Comparative genomics of a vertically transmitted thiotrophic bacterial ectosymbiont and its close free‐living relative

Abstract Thiotrophic symbioses between sulphur‐oxidizing bacteria and various unicellular and metazoan eukaryotes are widespread in reducing marine environments. The giant colonial ciliate Zoothamnium niveum, however, is the only host of thioautotrophic symbionts that has been cultivated along with its symbiont, the vertically transmitted ectosymbiont Candidatus Thiobius zoothamnicola (short Thiobius). Because theoretical predictions posit a smaller genome in vertically transmitted endosymbionts compared to free‐living relatives, we investigated whether this is true also for an ectosymbiont. We used metagenomics to recover the high‐quality draft genome of this bacterial symbiont. For comparison we have also sequenced a closely related free‐living cultured but not formally described strain Milos ODIII6 (short ODIII6). We then performed comparative genomics to assess the functional capabilities at gene, metabolic pathway and trait level. 16S rRNA gene trees and average amino acid identity confirmed the close phylogenetic relationship of both bacteria. Indeed, Thiobius has about a third smaller genome than its free‐living relative ODIII6, with reduced metabolic capabilities and fewer functional traits. The functional capabilities of Thiobius were a subset of those of the more versatile ODIII6, which possessed additional genes for oxygen, sulphur and hydrogen utilization and for the acquisition of phosphorus illustrating features that may be adaptive for the unstable environmental conditions at hydrothermal vents. In contrast, Thiobius possesses genes potentially enabling it to utilize lactate and acetate heterotrophically, compounds that may be provided as byproducts by the host. The present study illustrates the effect of strict host‐dependence of a bacterial ectosymbiont on genome evolution and host adaptation.

Here, we provide genomic information on two closely related SOBs with different lifestyles, that is, an obligate vertically transmitted ectosymbiont and a free-living chemolithoautotrophic bacterium.
Candidatus Thiobius zoothamnicola of the Chromatiaceae (Chromatiales, Gammaproteobacteria) (Oren, 2017, short Thiobius) -originally introduced as Candidatus Thiobios zoothamnicoli (Rinke et al., 2006) -covers the surface of the giant colonial ciliate Zoothamnium niveum (short Zoothamnium) that can typically be found in shallow-water environments from tropical to temperate waters with decaying organic material (Bauer-Nebelsick et al., 1996a, 1996b, Rinke et al., 2006, 2009, reviewed in Bright et al., 2014).Thiobius is transmitted vertically (Bauer-Nebelsick et al., 1996a, 1996b;Rinke et al., 2006) and has not been found free-living despite extensive searches using general bacterial primers as well as a symbiont-specific primer and direct Sanger sequencing (Monika Bright personal observation, 2022).To our best knowledge, the giant ciliate mutualism is the only thiotrophic symbiosis that has been cultivated in the laboratory over several host generations (Rinke et al., 2007).In contrast, the facultative endosymbiont Thiosocius teredinicola of the shipworm Kuphus polithalamius is the only thiotrophic symbiont that has been cultivated without its host (Distel et al., 2017).
The host Zoothamnium niveum grows fast and reproduces asexually developing specialized macrozooids that leave the colony as ectosymbionts-covered propagules called swarmers to found new colonies upon settlement (Bauer-Nebelsick et al., 1996a, 1996b).
Thiobius thriving on the host surface fixes inorganic carbon using sulphide and provides organic carbon to the host (Volland et al., 2018).
In return, benefits to the symbiont come through the host's peculiar contraction and expansion behaviour that ensures the supply of oxygen and sulphide (Rinke et al., 2007).The ciliate colonies can show unspecific overgrowth on the stalk, which connects them to the substrate surface (Bauer-Nebelsick et al., 1996a).
Strain Milos ODIII6 (short ODIII6) is a cultured, but not yet formally described bacterium, that has been isolated from the shallow-water hydrothermal vents in Paliochori Bay (Milos, Greece;Kuever et al., 2002;Sievert, 1999).ODIII6 is mesophilic (optimal growth at 34°C) and oxidizes reduced sulphur compounds under aerobic conditions.
Evolutionary theory predicts that the genome of obligate microbial symbionts, transmitted vertically from one to the next host generation, will be reduced in size compared to free-living relatives and will contain a lower proportion of mobile elements than facultative symbionts (Newton & Bordenstein, 2011;Sachs et al., 2011).Being obligately host-associated, these bacteria experience bottlenecks in population size during each transmission event leading to genome reduction (Bobay & Ochman, 2017;McCutcheon & Moran, 2012;Moran, 1996;Moran et al., 2009;Toft et al., 2009;Toft & Andersson, 2010;Wernegreen, 2015).Trait function compensation by the host or other symbionts can lead to further gene loss and reduced genome size (Ellers et al., 2012).Since obligate symbiotic bacteria are often isolated in the host as endosymbionts, they experience limited possibilities for horizontal gene transfer and access to foreign DNA.Some even lack the machinery for uptake and incorporation of DNA (Medina & Sachs, 2010).In contrast, obligate ectosymbionts have access to novel gene pools (Newton & Bordenstein, 2011).Nevertheless, some thermophilic ectosymbiotic archaea undergo genome reduction.This socalled thermophilic streamlining was explained by having fewer and shorter genes in hot environments compared to archaea from other environments (Nicks & Rahn-Lee, 2017).Much less studied are vertically transmitted bacterial ectosymbionts located on the surface of eukaryotic hosts.Mostly, these ectosymbiotic microbiomes are composed of complex microbial communities.Only a few cases with a single microbial partner are known and even fewer genomic studies are available (Fokin & Serra, 2022;Husnik et al., 2021).
Here, we show that the ectosymbiont Thiobius has a smaller genome, fewer genes, reduced GC content, and a smaller mobilome than the free-living bacterium ODIII6, matching theoretical predictions.The analyses of metabolic capabilities revealed that the functional traits of Thiobius are largely a subset of the repertoire of ODIII6, which has a higher functional versatility to cope with broader environmental conditions as a free-living bacterium from a highly unstable, fluctuating hydrothermal vent environment.In contrast, Thiobius shows a potential genetic capability to grow heterotrophically as an adaptation to its host.
A culture frozen at −80°C in dimethylsulphoxide was reactivated in 2015 and used to obtain DNA for genome sequencing.

| DNA preparation and sequencing
The lower parts of Zoothamnium niveum colonies are usually overgrown with diverse microbes (Bauer-Nebelsick et al., 1996a;Rinke et al., 2006), and were therefore cut off to minimize contamination using Schreiber micro scissors (Fridingen, Germany, European Union).
The upper parts, covered by a monolayer of Thiobius, were homogenized with Tris-EDTA buffer by vortexing and DNA was extracted according to Zhou et al. (1996).The DNA extraction yielded up to 7 ng DNA per μL in ~30 μL of final volume for G43.Nextera XT (Illumina) DNA library preparation was used for multiplexing the Zoothamnium samples, and they were sequenced in a paired-end mode with 125 nucleotides of read length using the Illumina HiSeqV4 platform at the Vienna Biocenter Core Facility (https:// www.vienn abioc enter.org/ vbcf/ ).The low amount of DNA yielded by a single Zoothamnium ciliate colony precluded the use of long-read sequencing technologies.
The ODIII6 isolate was regrown from a frozen stock culture in 2015 and DNA was extracted with an UltraClean® microbial DNA isolation kit (MoBio laboratories).The DNA was sequenced with a MiSeq (Illumina) sequencer by a commercial provider, MR DNA (Shallowater, TX, https:// www.mrdna lab.com/ ), using the 600 Cycles v3 Reagent Kit (Illumina).At MR DNA, the library of the sample was prepared with a Nextera DNA Sample Preparation Kit (Illumina).

| Quality refinement, de novo assembly and binning
The sequenced reads of the four Zoothamnium samples were quality filtered with the bbduk command of the suite bbmap v35.92 F I G U R E 1 Maximum likelihood (ML) tree based on the 16S rRNA gene.Bootstrap support of the internal nodes is indicated with the colour of the outgoing branches.Topology conflicts with two other techniques Maximum parsimony (MP, ☻) and Bayesian inference (BI, ★) indicated with symbols in the affected internal nodes.The two microbes analysed in this study are highlighted in bold.The tree was generated with 306 sequences (Table S1), but only 40 organisms that have publicly released genomes are shown.
(BBMap-Bushnell B.-sourc eforge.net/ proje cts/ bbmap /) under a more stringent quality threshold of 25 (Phred units).The filtered paired-end reads were assembled with SPAdes v3.7.1 in the metagenomic mode (Nurk et al., 2016). MetaBAT v0.26.3 (Kang et al., 2015) was then used for binning with default parameters.A nucleotide sequence homology BLAST search (Altschul et al., 1990) of the published 16S rRNA gene sequence of Thiobius (Rinke et al., 2006) identified the targeted bin.The genome of ODIII6 was de novo assembled by MR DNA using NGEN (DNAstar, https:// www.dnast ar.com).Completeness, heterogeneity and contamination of the ODIII6 and Thiobius assemblies were assessed with CheckM v1.0.5 (Parks et al., 2015).Presence of rRNAs and tRNAs for each amino acid was checked with tRNAscan-SE (Lowe & Eddy, 1997).Average amino acid identity (AAI; Konstantinidis & Tiedje, 2005) was calculated with CompareM (https:// github.com/ dpark s1134/ CompareM).The assembly of ODIII6 with SPAdes was performed to exclude contamination, using the same set of reads as the MR DNA assembly.The alignment between both assemblies was generated and visualized with Mauve (Darling et al., 2010, establishing the correspondence between the contigs.The fastg file containing paired-end linkage information of the de Bruijn graph was visually inspected with Bandage v0.8.0 (Wick et al., 2015), allowing the assessment of physical connectivity between contigs of the bacterial chromosome.

| Synteny analysis and localized gap filling
The four Thiobius draft genome assemblies were aligned with Mauve (Darling et al., 2010) and the ordering of the contigs was partly reconciled.A less stringent quality threshold of 20 (Phred units) was applied for the quality filtering and trimming of the G43 reads with bbduk resulting in the recovery of the full 16S rRNA gene.

| Phylogenetic analyses
A total of 306 16S rRNA gene sequences were chosen following the taxa selection of Petersen et al. (2016) and Distel et al. (2017).
16S rRNA gene sequences were directly obtained from GenBank and RefSeq when available or extracted from publicly available genomes using Metaxa v2.2 (Bengtsson-Palme et al., 2015).The alignment was done with MAFFT v7 (Katoh &Standley, 2013) andTrimAl v1.2 (Capella-Gutierrez et al., 2009) was employed for trimming and filtering the alignments with default parameters.The obtained sequences were allocated to 292 putative species according to a 98.65% identity threshold (Kim et al., 2014;Table S1).
Bootstrap support (Felsenstein, 1985) in ML and MP trees was calculated with 100 trees in each, and in BI posterior probability was used as metric of support.To facilitate visualization, 16S rRNA gene trees were reduced by pruning most of the branches (Rodriguez-Puente & Lazo-Cortes, 2013) retaining only the 40 organisms with genomes available.By comparing the tree topologies, shared internal nodes and conflicting topologies were manually identified.We used ML tree as the base of the representation to show the support of the three approaches.

| Functional inference of predicted genes and metabolic pathways
To predict and annotate genes the web version of RASTtk (https:// rast.nmpdr.org, Brettin et al., 2015) was employed with default settings.Functional categories were assigned to the predicted genes by the use of eggNOG mapper version 2 (Huerta-Cepas et al., 2017, 2019).Metabolic pathways were obtained using the predicted and annotated genes as an input by two independent methods, followed by a manual curation: (a) the online KEGG tool (www.kegg.jp, Kanehisa & Goto, 2000, Moriya et al., 2007) interrogating pathway complete modules; and (b) the PathoLogic module from Pathway Tools (v23.5, Karp et al., 2002Karp et al., , 2010) ) of MetaCyc (Caspi et al., 2008).
To these two levels of granularity (the fine grain of the genes and the coarser grain of the metabolic pathways; Vogt, 2010) we added the overarching level of functional traits (De Oliveira et al., 2022), defined as microbial characteristics that can be observed and are linked to fitness (Green et al., 2008).Additionally, secretion, motility and defence potential capabilities encoded in the genomes were assessed with MacSyFinder (Abby et al., 2016).The potential presence of traits was determined by putting together constituent genes following MetaCyc and KEGG when available, or MacSyFinder for motility and interaction traits (Abby et al., 2016).Some traits were not detected by MetaCyc and KEGG, and therefore, were manually inferred based on scientific literature (Table S2).In the manual curation logical operators (AND, OR) were employed to combine the gene presence/absence evidence into traits, as outlined in Karaoz and Brodie (2022) (Table S2).

| Orthology analysis
A total of 30 bacterial genomes, closest to Thiobius and ODIII6 based on the 16S rRNA gene tree phylogeny (Table S3), were selected excluding the genome of Bathymodiolus sp.SMAR symbiont due to poor quality of the assembly (52 scaffolds comprising 339 contigs with N 50 as low as 10,280 nt).Protein sequences were clustered into orthologous groups using OrthoFinder v2.5.4 (Emms & Kelly, 2019).

| Close phylogenetic relationship of Thiobius and ODIII6
In order to confirm the previously reported close phylogenetic placement of Thiobius and ODIII6 (Distel et al., 2017;Lenk et al., 2011;Nunoura et al., 2014;Rinke et al., 2006Rinke et al., , 2009;;Schuster & Bright, 2016), 16S rRNA gene ML, MP and BI phylogenies were constructed.A pruned visualization of the resulting ML tree with indication of the conflicts between the three approaches is shown in Figure 1 (full ML tree in Figure S1).While some of the shallower internal nodes (closer to the tips) showed good support (e.g.bootstrap values greater than 70%) and agreement between the three phylogenetic approaches, many of the deeper internal nodes were poorly supported and showed topology conflicts.While ODIII6 falls in a clade containing the endosymbionts of the hydrothermal vent snails Chrysomallon squamiferum and Alviniconcha sp.Lau Basin, and free-living bacteria, Thiobius can be assigned as sister taxon of this clade.The internal node connecting these clades obtained bootstrap supports of 70% (ML) and 63% (MP), while the posterior probability support of BI was 73%.The 16S rRNA gene sequences of Thiobius and ODIII6 were 95% identical (over 1396 aligned nucleotide positions).A relatively low average amino acid identity (AAI) of 67% was obtained for the whole genomes of Thiobius and ODIII6.

Thiobius and ODIII6
The best Thiobius metagenome assembled genome (MAG) in terms of assembly completeness and contiguity was obtained from a single ciliate colony collected in Belize (labelled G43, Figure S2; Tables 1   and S4).A total of 8.8 million of the initial 13.5 million read pairs passed a more stringent quality threshold filtering (phred 25).
SPAdes assembled these reads into 44,538 contigs.MetaBAT binning grouped these contigs into four bins, one of them containing two partial matches at contig ends to the published 16S rRNA gene of Thiobius according to a BLAST search (Rinke et al., 2006; Figure S3).
A less stringent quality threshold filtering (phred 20) led to a very similar SPAdes assembly that recovered the full 16S rRNA gene.The corresponding long contig containing the full 16S rRNA gene was used to replace three contigs from the stringent filtering assembly that covered the same span (Figure S4, Table S5).Overall, Thiobius G43 assembly resulted in 46 contigs with a total length of 2.38 Mb, a N 50 value of 98,217, a coverage of 216× and a GC content of 49.4% (Figure 2).CheckM estimated its completeness as 96.0% and its contamination as 0.1%, and found no heterogeneity (Table 1), meeting the MIMAG standard for high-quality MAGs (Bowers et al., 2017).
Synteny analysis with Mauve between the four Thiobius MAGs led to the identification of 26 clusters of contiguity that were utilized in the final ordering of the contigs (Figures 2 and S2, Table S5).
ODIII6's genome MR DNA assembly consisted of 26 contigs, and had no paired-end linkage associated information available (fastg file).Because RASTtk annotation revealed one contig with only phage related genes, we re-assembled the genome using SPAdes and found through the paired-end linkage information that the corresponding contig was circular and was thereby excluded (Figure S5).
Accordingly, the final ODIII6 assembly resulted in 25 contigs, with a total of 3.53 Mb, a N 50 value of 245,984, a coverage of 103×, and a GC content of 61.9%.CheckM estimated its completeness as 99.6% and its contamination as 0.8%, and found no heterogeneity either (Table 1).
A comparison between the high-quality draft genomes of Thiobius and ODIII6 (both appropriate for general assessment of gene content, Chain et al., 2009), revealed a complete set of tRNAs for translation of all 20 amino acids in both bacteria.In agreement with theoretical TA B L E 1 Statistics of the genomes assembled in this study.predictions (Sachs et al., 2011), the genome of the vertically transmitted ectosymbiont Thiobius was 33% smaller than the one of the free-living bacterium ODIII6, and showed a reduced relative GC content, consistent with the significant positive correlation between GC content and genome size in bacterial genomes (Almpanis et al., 2018).Whether the reduced genome size of the vertically transmitted ectosymbiont is due to any of the mechanisms that are known to affect vertically transmitted endosymbionts, such as those of thiotrophic vesicomyid clams (Kuwahara et al., 2007;Newton et al., 2007) and of marine flatworms Paracatenula (Jäckle et al., 2019), remains to be studied.

| Thiobius encodes fewer genes, pathways and functional traits compared to ODIII6
To characterize the potential functional capabilities of Thiobius and ODIII6, their gene complements were predicted with RASTtk (Brettin et al., 2015).The annotation of Thiobius yielded 2486 predicted protein-coding sequences, of which 1494 (60%) received a meaningful automatic functional prediction.In the manual curation process 21 hypothetical genes (18 non-redundant) obtained a putative annotation.ODIII6 had 3452 predicted protein-coding sequences including 2063 (60%) with meaningful functional predictions, and in the manual curation 21 hypothetical genes (19 non-redundant) were functionally annotated.Thiobius contains about 28% less protein-coding sequences than ODIII6.Accordingly, the lower number of proteincoding sequences in Thiobius compared to ODIII6 is consistent with their genome sizes.
We manually selected 46 relevant functional traits involved in key metabolism and additional functions like storage or interactions, of these two SOBs.These traits can range from a single gene to the combination of several metabolic pathways (Tables 2 and S2).Less than half of the traits ( 19) showed a compact localization of their constitutive genes within contigs (Figure 2).PathwayTools inferred obtained 31 traits and ODIII6 yielded 39, of which 24 traits were shared (Table S6).Accordingly, Thiobius showed less pathways and relevant traits compared to ODIII6, particularly in those related with oxygen, phosphorus and sulphur.
In order to assess the functional commonalities between Thiobius and ODIII6 we performed an orthology analysis over a background set of 28 additional genomes of species included in the phylogenetic analysis (Table S3).Thiobius and ODIII6 shared 1409 orthogroups (Figure 3).A higher proportion of functional pathways were shared between the two organisms (Figure 3).Most of the genes of both or- ganisms presented an orthologous one-to-one relationship (84% of genes in Thiobius and 90% of genes in ODIII6; Table S7).Overall, the relatively high percentage of shared orthogroups, pathways, modules and traits in Thiobius indicates that its functional capabilities are mainly a subset of those present in ODIII6 (Figure S6).
To further investigate the distribution of functional categories, eggNOG was applied to determine the attribution of COG categories (Tatusov et al., 2000).In general, the gene counts of the COG categories followed the genome sizes (Table S8), but there were some exceptions.ODIII6 and Thiobius COG categories yielded high count similarities in the categories 'lipid metabolism' and 'secondary structure'.Nevertheless, this did not imply a fully identical sets of genes, for example, of the orthogroups involved in the 'lipid metabolism' category only a 52% were shared, indicating that different 'non-orthologous' gene sets were recruited to perform same biological tasks in the two different bacteria.The least similar categories were 'cell motility' and 'signal transduction', for which the respective proportions were much lower in Thiobius in relation to its genome size (Table S8).Therefore, the COG functional categories composition reflects the free-living lifestyle of ODIII6, with increased motility representation and a more complex gene regulation that match the requirements of a free-living lifestyle.

| Potential for a mixotrophic lifestyle as an adaptation to the ciliate host
Both Thiobius and ODIII6 show the genomic potential to fix inorganic carbon through the Calvin-Benson-Bassham cycle (Cbb) with the capability to form carboxysomes to concentrate ribulose-1,5-bisphosphate carboxylase oxygenase (RuBisCO; Badger & Bek, 2008), the key enzyme of the Calvin-Benson cycle responsible for CO 2assimilation.According to the neighbouring genes the RuBisCO type in both Thiobius and ODIII6 is form IAc with a potential functional niche of low CO 2 and low to high oxygen concentration (Badger & Bek, 2008).This is in line with previous studies in Thiobius, in which RuBisCO was histochemically detected, carboxysomes were identified by TEM (Bauer-Nebelsick et al., 1996b) and a type IA RuBisCO large subunit sequence was retrieved (Rinke et al., 2009).
In addition, carbon fixation in Thiobius was confirmed through tissue autoradiography and NanoSIMS (Volland et al., 2018).Further, the contraction and expansion behaviour of the host creates a continuously changing abiotic environment for Thiobius ranging from oxygen rich to sulphidic, anoxic conditions (Bright et al., 2014).ODIII6 shows two putative bicarbonate transporter-encoding genes downstream of the carboxysome structural genes operon, consistent with previous reports on other chemoautotrophs (Axen et al., 2014;Scott et al., 2020).ODIII6 further possesses a gene for a beta class carbonic anhydrase not integrated in the carboxysome operon but elsewhere in the genome, which converts bicarbonate to CO 2 for carbon fixation (Supuran & Capasso, 2017).
Organic carbon is stored differently in Thiobius compared to ODIII6.While Thiobius has the genetic potential for using polyhydroxyalkanoates (Pha), genes for glycogen (Glg) and cyanophycin synthesis (Cph; storing carbon and nitrogen) were found in ODIII6.
In both organisms, the organic carbon is oxidized through the TCA cycle (TcaC), although.Thiobius encodes additionally genes for the glyoxylate cycle pathway (GlC; Cozzone & El-Mansi, 2005).Putative genes encoding for transporters to import acetate (ActP) and lactate (LakP) along with genes encoding for all three components of L-lactate dehydrogenase for lactate utilization (Lut) were also present.Both acetate and lactate could potentially be metabolized by the glyoxylate cycle (El-Mansi et al., 1986;Serafini et al., 2019).While in many organisms (including Bacillus subtilis) lutABC belongs to the same operon with a lactate permease (Chai et al., 2009), Thiobius possesses a different DctP-TRAP-like transporter (LakP), similar to the one described for Thermus thermophilus (Fischer et al., 2010).
The presence of these transporters indicates a potential for heterotrophic metabolism in Thiobius.As the ciliate host may potentially be able to switch to an anaerobic metabolism under sulphidic conditions and produce lactate and acetate, similarly to the rumen ciliate Entodinium caudatum (Park et al., 2021), it is tempting to speculate that these fermentation products may then be released from the host and taken up by the symbiont, but these anaerobic processes have yet to be studied in this mutualism.No equivalent capabilities were found in the genome of ODIII6 in line with its obligate autotrophic metabolism.

| Oxygen is the only electron acceptor in both bacteria, but ODIII6 is more versatile than Thiobius
According to the genomic potential of their draft genomes, oxygen is the only electron acceptor that Thiobius and ODIII6 can utilize.
Genes for nitrate respiration, known in many free-living and symbiotic thiotrophic bacteria living at oxic-anoxic interfaces (De Oliveira et al., 2022;Flood et al., 2015;König et al., 2016;Nunoura et al., 2014;Paredes et al., 2021), were not found.This indicates that both bacteria strongly rely on oxygen both for respiration and for the oxidation of reduced sulphur species to generate energy.
Indeed, both genomes encode for a cytochrome bc 1 electron transport complex (Pet), and two cytochrome terminal oxidases, that is,  Note: '+' indicates that the following gene is required, '-' indicates that the following gene is not required, ',' indicates interchangeability; brackets are used for grouping; 'ndhA_to_ndhN' stands for 14 genes.

TA B L E 2 (Continued)
in Bradyrhizobium japonicum (Pitcher & Watmough, 2004), and aa 3 (CytAA3), which belongs to the Class A of oxidases with lower apparent oxygen affinities (Han et al., 2011).This indicates the ability to utilize oxygen at various concentrations.Interestingly, the genome of ODIII6 contains two copies of the genes encoding cbb 3 , which is unusual and could be an additional adaptation to optimize oxygen utilization under varying concentrations, as was shown for Pseudomonas aeruginosa (Comolli & Donohue, 2004).Further, ODIII6 possesses two more high-affinity terminal oxidases, cytochrome bd (CytBD)-and ba 3 (CytBA3)-encoding genes, which points to a higher versatility under a broader range of oxygen regimes than Thiobius.
In contrast, Thiobius lives on a ciliate host that can position itself in microhabitats with optimal oxygen concentrations, which potentially renders the ability to express multiple terminal oxidases unnecessary.

| Both organisms use the oxidation of reduced sulphur compounds to generate energy
The energy fuelling processes for carbon fixation in Thiobius and ODIII6 comes from the oxidation of reduced sulphur species using oxygen as terminal electron acceptor.Sulphide oxidation to elemental sulphur occurs in both bacteria through two possible pathways, sulphide dehydrogenase (Fcsd; flavocytochrome C, Sorokin et al., 1998), and sulphide:quinone oxidoreductase (Sqr).ODIII6 has a type VI Sqr and Thiobius has type I and type VI Sqr (Dahl, 2017).
The elemental sulphur formed by sulphide oxidation is stored in sulphur globules that are enveloped by proteins (Dahl, 2017) identified in both genomes (Sgp).Indeed, transmission electron micrographs Bauer-Nebelsick et al., 1996b) and Raman microspectrometry previously revealed membrane bound elemental sulphur vesicles in Thiobius (Maurin et al., 2010) used to store sulphur under sulphidic conditions and oxidize it further during oxic conditions (Volland et al., 2018).In ODIII6 cultures, the formation of sulphur globules attached to the cell could also be observed if thiosulphate was provided (Stefan Sievert personal observation, 2022).
Thiosulphate is potentially oxidized in both organisms to sulphate and elemental sulphur through the truncated Sox pathway (tSox, Welte et al., 2009, Dahl, 2020), with the possible involvement of SoxL (Weissgerber et al., 2011).For both organisms, elemental sulphur is oxidized to sulphite in the cytoplasm by the reverse dissimilatory sulphate reductase pathway (rDsr; Dahl, 2015, Gregersen et al., 2011, Hensen et al., 2006), and sulphite is oxidized to sulphate either by the adenylylsulphate reductase and the sulphate adeny- Predicted and functionally annotated genes are grouped in metabolic pathways.Functional traits are features such as single metabolic pathways or composites of them that affect the organism fitness.Structural features such as transporters or secretion systems, and storage capabilities are also considered as traits.Dahl, 2017).Gene sequences for the alpha and beta subunits of the reverse-type dissimilatory sulphite reductase (dsrAB) and for the alpha subunit of the adenylylsulphate reductase (aprA) were previously reported in Thiobius (Rinke et al., 2009).Finally, the putative sulphate transporter CysZ exports the sulphate to the periplasm (Figure 4a; Hryniewicz et al., 1990).
In contrast to Thiobius, ODIII6 has additional sulphur-metab- Although many sulphur oxidation traits are shared, we observe a higher versatility in ODIII6 than in Thiobius.In addition, the presence of [NiFe] hygrogenase Hup genes (Vignais et al., 2001) in ODIII6 may indicate the potential of hydrogen oxidation.However, it was recently found in Candidatus Endoriftia persephone that hup genes present in its genome were not involved in hydrogen oxidation but may instead facilitate intracellular redox (Mitchell et al., 2019).The two subunits L and S in ODIII6 show 75% and 71% amino acid identity with their Candidatus Endoriftia persephone orthologues.

| Similarities in nitrogen and phosphorous metabolism
Both organisms show similar potential capabilities in nitrogen metabolism.Nitrate is imported into the cytoplasm through the nitrate transporter Nrt (Maeda et al., 2019), and is reduced to ammonium in the first step of assimilatory nitrate reduction, and from there assimilated as organic nitrogen in form of biomolecules (AsN; Takai, 2019).
In addition, the ammonium can be imported into the cytoplasm by the ammonium transporter AmtB (Wang et al., 2012) Oliveira et al., 2022).Overall, urea as well as ammonium are wellknown nitrogen waste products of ciliates (Caron & Goldman, 1990), and may serve as byproducts the host provides to the symbiont.
Urea was shown to be a source of incorporated nitrogen for bacteria in intertidal sediments (Veuger & Middelburg, 2007).Whether urea is utilized by ODIII6 in its natural environment remains to be studied.
Phosphorus is a limiting nutrient in many marine environments, present in its inorganic dissolved fraction generally as orthophosphate (Paytan & McLaughlin, 2007).Both organisms encode the genes for a high-affinity phosphate transporter (Pst), and additionally ODIII6 possesses genes for two other transporters: a high-affinity Na + /P i symporter (Hnp) and a low-affinity phosphate transporter (Pit).Once phosphate is incorporated into the cells, both organisms also show the potential for polymerizing it into polyphosphate and hydrolysing back to inorganic phosphates (PHKandPHX), using it as energy storage (Achbergerová & Nahálka, 2011).

| Interaction and motility in both organisms
Both Thiobius and ODIII6 have a repertoire of genes to interact with other organisms and with the environment.They share genes for the type II (T2SS) and type V (T5SS) secretion systems.Thiobius additionally has genes for the type VI secretion system (T6SS; Kapitein & Mogk, 2013) known to function in host interaction (Hachani et al., 2016) and CRISPR-Cas proteins (Cas), while ODIII6 has genes for the type I secretion system (T1SS).It remains to be studied how both bacteria use these traits in their natural environment and whether they may help Thiobius to interact with its host.
Both organisms further possess genes for type IV pilus potentially involved in twitching motility (T4P; Ayers et al., 2010), and additionally ODIII6 genome contains several loci for the biosynthetic genes encoding for a flagellum (Fla).Indeed, ODIII6 was observed to be motile in cultures, however, loses its motility after a longer period of cultivation (Sievert pers. obs.).This suggests that the flagellum may no longer be expressed in ODIII6 if constant‚ favourable conditions render motility unnecessary.ODIII6 additionally shows genes for photolyase DNA protection against UV radiation, and repair (Sancar et al., 1987).
3.9 | Mobile genetic elements are less abundant in Thiobius than in ODIII6 Bacteria can experience horizontal gene transfer through mobile genetic elements such as phages, plasmids, transposons and insertion sequences, also referred to as the mobilome (Frost et al., 2005).
Free-living bacteria like ODIII6 are more likely to be exposed to novel gene pools than symbionts (Newton & Bordenstein, 2011).
In addition, vertically transmitted symbionts, such as Thiobius experience population bottlenecks as each swarmer is covered with relatively few symbionts that grow to cover the new colony (Bauer-Nebelsick et al., 1996a, 1996b).Accordingly, we hypothesized a smaller mobilome in the obligate symbiont Thiobius than in the freeliving ODIII6.Indeed, RASTtk annotation of Thiobius revealed three genes attributed to phages and 18 genes attributed to other mobile genetic elements, while ODIII6 has two genes attributed to phages and 58 genes attributed to other mobile genetic elements (Table S9).
Roughly 60% of these mobile elements were located at the extremes of the contigs, consistent with their disruptive effect on the assembly processes.These results point to a smaller mobilome in the hostassociated Thiobius than in the free-living ODIII6.

| CON CLUS IONS
The phylogenetic relationship of the ectosymbiont Thiobius and the strain Milos ODIII6 is confirmed through 16S rRNA phylogeny and average amino acid identity and serves as baseline to compare the genomes of these two bacteria with very different lifestyles.In agreement with theoretical predictions, but hardly studied in ectosymbiotic bacteria, Thiobius' genome is smaller than that of its free-living relative ODIII6.The characterization at the levels of genes, metabolic pathways and traits reveals that Thiobius and ODIII6 share a large proportion of their genetic repertoire and metabolic capabilities.The lower number of lineage-specific metabolic pathways and relevant traits in Thiobius compared to ODIII6 may point to a more stable environment provided by the host, requiring less versatility.This may have led to a loss of genetic potential in Thiobius and/or gain in ODIII6.In comparison with Thiobius, ODIII6 shows a larger functional repertoire, in particular for its energy metabolism regarding the utilization of sulphur, oxygen and hydrogen, consistent with the requirements for a free-living bacterium to live under the fluctuating conditions of the hydrothermal vent environment.Thiobius, however, shows potential

F
Representation of the draft genomes of Thiobius strain G43 (left) and ODIII6 (right).The most outer grey solid arcs represent the contigs.ODIII6 contigs are ordered by decreasing length.In Thiobius, information of three additional MAGs was employed to retrieve putative clusters of contiguity that are presented in decreasing length order.The bridging contigs are shown as solid arcs in colours, the rest of their contigs are omitted.GC skew typically reflects the attribution of the contig to the leading or to the lagging DNA strand.The correct ordering of the contigs has not been fully resolved, and this is reflected in the inconsistent GC skew pattern between neighbouring contigs.GC content and coverage are shown in the most inside rings.The location of relevant traits that show contiguity of their constitutive genes is also shown.203MetaCyc metabolic pathways for Thiobius, while 39 complete KEGG modules were identified.The annotation of ODIII6 resulted in 233 MetaCyc pathways and 53 complete KEGG modules.Thiobius which was shown to have a high affinity for oxygen TA B L E 2 Selection of relevant functional traits.
lyltransferase (AprSat), or by the sulphite-oxidizing enzyme (Soe; F I G U R E 3 Euler diagrams of different levels of granularity showing shared and unique genetic potential capabilities of Thiobius and ODIII6.The finer granularity level is the genes from the orthogroups (a).Two metabolic pathway databases were employed: MetaCyc (b), and KEGG modules (c).The coarser granularity level are the traits (d).F I G U R E 4 Major metabolic capabilities found in the draft genomes of (a) Candidatus Thiobius zoothamnicola strain BelizeG43, and (b) strain Milos ODIII6.Chosen relevant potential functional capabilities are shown (for simplicity not all transporters are depicted in the figure).
olizing capabilities: the periplasmic disproportionation of thiosulphate by a rhodanese-like sulphurtransferase to sulphide and sulphite (Tst;Deckert et al., 1998); elemental sulphur reduction to sulphide in the periplasm by a putative polysulphide reductase (Psr;DeOliveira et al., 2022), that might also be involved in oxidation as suggested for Allochromatium vinosum(Weissgerber et al., 2013); cytoplasmic disproportionation of elemental sulphur to thiosulphate, sulphite and sulphide by the sulphur oxygenase reductase (Sor;Janosch et al., 2015), as suggested for Thioalkalivibrio paradoxus(Rühl et al., 2017); reduction of elemental sulphur to sulphide coupled to the oxidation of hydrogen by the (Hyd;Ng et al., 2000); and sulphate exportation to the periplasm by the sulphate permease SulP (Figure4b; Aguilar-Barajas et al., 2011).A thiosulphate transporter gene (YeeE;Tanaka et al., 2020) is also found in the genome of ODIII6.