Chemosynthetic symbiont with a drastically reduced genome serves as primary energy storage in the marine flatworm Paracatenula

Significance

Animals typically store their primary energy reserves in specialized cells. Here, we show that in the small marine flatworm Paracatenula, this function is performed by its bacterial chemosynthetic symbiont. The intracellular symbiont occupies half of the biomass in the symbiosis and has a highly reduced genome but efficiently stocks up and maintains carbon and energy, particularly sugars. The host rarely digests the symbiont cells to access these stocks. Instead, the symbionts appear to provide the bulk nutrition by secreting outer-membrane vesicles. This is in contrast to all other described chemosynthetic symbioses, where the hosts continuously digest full cells of a small and ideally growing symbiont population that cannot provide a long-term buffering capacity during nutrient limitation.

Abstract

Hosts of chemoautotrophic bacteria typically have much higher biomass than their symbionts and consume symbiont cells for nutrition. In contrast to this, chemoautotrophic Candidatus Riegeria symbionts in mouthless Paracatenula flatworms comprise up to half of the biomass of the consortium. Each species of Paracatenula harbors a specific Ca. Riegeria, and the endosymbionts have been vertically transmitted for at least 500 million years. Such prolonged strict vertical transmission leads to streamlining of symbiont genomes, and the retained physiological capacities reveal the functions the symbionts provide to their hosts. Here, we studied a species of Paracatenula from Sant’Andrea, Elba, Italy, using genomics, gene expression, imaging analyses, as well as targeted and untargeted MS. We show that its symbiont, Ca. R. santandreae has a drastically smaller genome (1.34 Mb) than the symbiont´s free-living relatives (4.29–4.97 Mb) but retains a versatile and energy-efficient metabolism. It encodes and expresses a complete intermediary carbon metabolism and enhanced carbon fixation through anaplerosis and accumulates massive intracellular inclusions such as sulfur, polyhydroxyalkanoates, and carbohydrates. Compared with symbiotic and free-living chemoautotrophs, Ca. R. santandreae’s versatility in energy storage is unparalleled in chemoautotrophs with such compact genomes. Transmission EM as well as host and symbiont expression data suggest that Ca. R. santandreae largely provisions its host via outer-membrane vesicle secretion. With its high share of biomass in the symbiosis and large standing stocks of carbon and energy reserves, it has a unique role for bacterial symbionts—serving as the primary energy storage for its animal host.

The discovery of dense animal communities at deep-sea hydrothermal vents in the late 1970s led to paradigm shifts in deep-sea ecology and animal physiology. Large macrofauna was shown to be independent of photoautotrophic production and instead lives off primary production performed by chemosynthetic bacterial symbionts (1–3). In the most extreme forms, animals such as the giant tubeworms have lost their mouth and digestive system and instead are nutritionally dependent on their symbionts (4).

Since then, chemosynthetic symbioses have been documented from diverse habitats and in various lineages of marine invertebrates and protists (5, 6). The symbionts are all Proteobacteria, typically Gamma- or Epsilonproteobacteria, that are either vertically transmitted or horizontally acquired from the environment (5–12). Genomes from chemosynthetic symbionts can range from 4.88 to 1.02 Mb, reflecting varying metabolic and ecological strategies (13–15). Surprisingly, some mouth- and gutless hosts take up their symbionts anew in every generation, although the lack of a digestive system suggests extreme dependence on their symbionts (12, 16). As these symbionts have to thrive in both free-living environmental and symbiotic states, it is difficult to attribute their genomic features to either functions they provide to their host, or traits that are necessary for environmental survival or to both.

The smallest genomes of chemoautotrophic symbionts have been observed for the gammaproteobacterial symbionts of vesicomyid clams that are directly transmitted between host generations (13, 14). Such strict vertical transmission leads to substantial and ongoing genome reduction. Reduced genomes of intracellular endosymbionts reflect the essential set of functions that the symbionts provide to their invertebrate hosts (17–23). Host provisioning by chemosynthetic bacteria is accomplished through a conveyor belt-like turnover of symbiont cells and compared with their hosts, they only have a small share of the total biomass in the consortium (24). The patterns of genome reduction in the symbionts of vesicomyid clams mirror this mode of nutrition since they only retained limited metabolic and storage capacities and primarily serve as biosynthesis factories (13, 14, 18).

A second host group that lives in obligate symbiosis with vertically transmitted chemosynthetic endosymbionts is Paracatenula, a genus of marine flatworms that can be a dominant member of the meiofauna in shallow-water sediments (9). Unlike the vesicomyid clams that can still filter feed (3, 25), Paracatenula lack both a mouth and gut and are completely dependent on intracellular endosymbionts for nutrition (9, 26). All Paracatenula host species harbor a species-specific symbiont phylotype (9). These symbionts, called Candidatus Riegeria, comprise a third to a half of the total animal volume and are housed in vacuoles of specialized cells, the bacteriocytes, that make up the nutritive trophosome organ (9, 26, 27). Ca. Riegeria are the only known alphaproteobacterial chemosynthetic symbionts (9). They are the oldest extant clade of symbiotic chemoautotrophic bacteria, having been passed on in strict vertical transmission for more than 500 million years since the last common ancestor of all Paracatenula species (9).

In this study, we focus on a Paracatenula species that is highly abundant in the coarse shallow-water sediments of the bay Sant’Andrea on Elba, Italy (SI Appendix, Supplementary Note 1 and Fig. S1). We propose the name Ca. Riegeria santandreae for its endosymbionts, in reference to the collection site. Based on the genome of Ca. R. santandreae that we generated using shotgun metagenomics, we analyzed its metabolism, ecophysiology, and evolution. Compared with free-living alphaproteobacterial relatives from the Rhodospirillaceae, the genome is highly reduced. We reconstructed the physiology of Ca. R. santandreae by integrating transcriptomics and correlative imaging approaches with highly sensitive MS-based methods for proteomics and metabolomics. Despite the reduced genome, the symbiont performs energy-efficient sulfide oxidation and couples it to versatile means of carbon fixation and carbon storage. The symbionts likely provision their host via outer-membrane vesicle (OMV) secretion. With its large share of the biomass in the consortium, the symbiont has a unique role in animal-bacteria associations—serving as the bulk storage of carbon and energy reserves for its animal host.

Results and Discussion

Ca. R. santandreae Belongs to Rhodospirillaceae and Has a Reduced Genome.

The complete Ca. R. santandreae genome was assembled into one contig from the sequenced metagenome of a single Paracatenula sp. santandrea specimen. The genome had a size of 1.34 Mb, encoding 1,344 protein-coding and 56 RNA genes at a coding density of 83.9% and a genomic GC of 51.8% (SI Appendix, Fig. S2). The Ca. R. santandreae genome encoded pathways for sulfur oxidation based on reverse-acting dissimilatory sulfate reductase (rDSR) and for carbon fixation using the Calvin–Benson–Bassham (CBB) cycle, indicating that Ca. R. santandreae could acquire energy and carbon via thioautotrophy (SI Appendix, Supplementary Note 2).

Ca. R. santandreae was classified as a member of the family Rhodospirillaceae within the class Alphaproteobacteria, based on a phylogenetic analysis of 43 conserved single-copy marker genes (Fig. 1). It was phylogenetically nested among Rhodospirillaceae taxa that have genome sizes of 4.29–4.97 Mb, indicating that massive genome reduction has occurred in the lineage leading to Ca. R. santandreae (Fig. 1 and SI Appendix, Table S1). The Rhodospirillaceae are remarkably diverse and include photoheterotrophs and chemoheterotrophs but also chemoautotrophs, diazotrophs, and even magnetotactic genera (28–31). The taxa most closely related to the Ca. R. santandreae endosymbiont included Rhodospirillum rubrum [56.19% amino acid identity (AAI)], Novispirillum itersonii (56.22% AAI), Caenispirillum salinarum (56.99% AAI), and Magnetospirillum magneticum (57.01% AAI).

Fig. 1.

The endosymbiont Ca. R. santandreae clusters deeply within Rhodospirillaceae. Phylogenetic reconstruction was based on a protein alignment of 43 conserved marker genes calculated using FastTree and rooted with Rickettsiales as the outgroup. Scale bar indicates substitutions per site.

Ca. R. santandreae Is a Chemoautotroph with a Versatile Carbon Metabolism Despite Its Reduced Genome.

To investigate the physiology of Ca. R. santandreae, we applied a combination of genomics, transcriptomics, and imaging analyses with sensitive MS protocols that allow proteomic and metabolomic measurements from single microscopic animals. The expression data from specimens sampled directly from the environment covered up to 1,288 out of 1,400 predicted genes (n = 3 transcriptomes), of which 407 were also identified in the proteomes generated from single Paracatenula individuals as well as pools of up to three specimens (n = 8) (Fig. 2 and Dataset S1 A–E).

Fig. 2.

The chemoautotroph Ca. R. santandreae expresses a versatile metabolism. Metabolic reconstruction of Ca. R. santandreae, pathways, or enzymes are colored by expression levels in the transcriptome (mean of three samples). Only the highest 10% of expressed genes were colored relative to DNA gyrase A (gyrA) mean expression. Expressed proteins detected in the proteome samples are indicated. Not all genes of certain pathways were transcribed, and not all transcribed genes were found in the proteomes (Dataset S1 B–E). Dotted arrows correspond to indirect synthesis of metabolites. Apr, sulfite reductase; bc₁, cytochrome C oxidase type bc₁; Cyt. a₃, cytochrome c oxidase type a₃; Cyt. C, cytochrome c; FccAB, flavocytochrome c sulfide dehydrogenase; Mce, mammalian cell entry proteins Mla; NDH, NADH-quinone oxidoreductase/dehydrogenase; PckG, phosphoenolpyruvate carboxykinase; PDH, pyruvate dehydrogenase complex; Pst, high-affinity phosphate transporter; Sat, ATP sulfurylase; SDH, succinate dehydrogenase; Sec, secretion pathway; Tat, twin-arginine translocation; TPP, trehalose-P phosphatase; TPS, trehalose-P-synthase; TST, thiosulfate sulfur transferase (rhodanese).

Genes of the rDSR sulfur oxidation pathway and CBB carbon fixation pathways were highly expressed and underline the thioautotrophic role of Ca. R. santandreae [sum of 8 CBB cycle proteins: 4.88 ± 1.17% (mean ± SD) normalized spectral abundance factor (NSAF); sum of 14 rDSR pathway proteins: 7.48 ± 0.95% NSAF; n = 8 proteomes] (Fig. 2 and SI Appendix, Figs. S3 and S4 and Supplementary Note 2). CBB cycle genes including the ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO) subunits cbbL and cbbS were among the top 10% of most highly represented genes in transcriptomes as well as proteomes (SI Appendix, Figs. S3 and S4).

Microautoradiography corroborated the gene expression data, as signals for carbon fixation were limited to the symbiont-bearing trophosome region. Without addition of an external electron donor supply, the Ca. R. santandreae symbionts fixed 9.48 ± 10.2 µmol C g⁻¹ h⁻¹ (maximum: 33 µmol C g⁻¹ h⁻¹) (SI Appendix, Figs. S5–S7 and Dataset S1F). These rates were in the range determined for other sulfur-oxidizing symbionts of, e.g., Ifremeria nautilei (0.7 µmol C g⁻¹ h⁻¹) and Riftia pachyptila (27 µmol C g⁻¹ h⁻¹) (32). The calculated fixation rates would be sufficient to sustain a doubling of the length of a Paracatenula sp. santandrea specimen in 10–34 d (Dataset S1F). Such growth rates, together with frequent asexual reproduction (33, 34), could explain their high abundances at the sampling site where they occurred in densities of 233–5,400 individuals per m³ of sediment (1,809 ± 1,483; n = 40).

The CBB cycle encoded in Ca. R. santandreae was missing fructose-1,6-bisphosphatase and sedoheptulose-1,7-bisphosphatase, which could be replaced by a reversible and PP_i-dependent 6-phosphofructokinase (PFK) (35). Ca. R. santandreae encoded two PFKs, an ATP-dependent ATP-PFK and a PP_i-PFK (SI Appendix, Fig. S8). The PP_i-PFK had a 10-fold higher expression [transcripts per million (TPM): 5,073 ± 780; n = 3] than the ATP-PFK (TPM: 504 ± 128; n = 3) and also was among the top 10% of highest expressed genes (SI Appendix, Figs. S3 and S4). Together with the PP_i-energized proton pump (PP_i-H⁺-pump) (SI Appendix, Fig. S9), which was among the 10% most expressed genes (TPM: 1,686 ± 163; n = 3), the PP_i-dependent variant of the CBB cycle can save up to 31.5% of the ATP invested in carbon fixation (35). We tested for a possible lateral gene transfer of the PP_i-PFK and PP_i-H⁺-pump genes to Ca. R. santandreae as the PP_i-dependent CBB cycle has also been found in gammaproteobacterial chemosynthetic symbionts, including the symbionts of gutless oligochaetes, vestimentiferans tubeworms, and the bivalve Solemya (35–37). In contrast to the Gammaproteobacteria, the PP_i-PFK and PP_i-H⁺-pump genes in Ca. R. santandreae did not form an operon (35). Both genes were of alphaproteobacterial origin and clustered with gene sequences of related Rhodospirillaceae (SI Appendix, Supplementary Note 2 and Figs. S8 and S9). Other Rhodospirillaceae such as Rhodospirillum rubrum only lacked a sedoheptulose-1,7-bisphosphatase but still encoded a canonical fructose-1,6-bisphosphatase for the last step in the CBB cycle that has the highest energy-saving potential (35) (Dataset S1A). This suggests that the most energy-efficient variant of PP_i-dependent CBB cycle evolved independently in Ca. R. santandreae, possibly due to the same energy constraints that led to the selection of the most efficient sulfur oxidation pathway across the diversity of thiotrophic symbionts (38) (SI Appendix, Supplementary Note 2).

The Ca. R. santandreae symbiont encoded form IA RuBisCO for carbon fixation rather than the form II RuBisCO that was previously identified by PCR amplification in another Ca. Riegeria species, Ca. R. galateiae (9) (SI Appendix, Fig. S10). The presence of two different RuBisCO forms in the Ca. Riegeria clade that both clustered with sequences from Alphaproteobacteria suggests that the last common ancestor encoded both forms and that descendants differentially retained only one. Several members of the family Rhodospirillaceae are known to encode more than one RuBisCO form per genome, including forms IA and II (39). As form IA is adapted to higher oxygen and lower carbon dioxide levels compared with form II, the retention of form IA in Ca. R. santandreae could indicate a preference for habitats with higher oxygen concentrations for Paracatenula sp. santandrea compared with Paracatenula galateia (39). Niche differentiation via different forms of RuBisCO has been observed in free-living thiotrophic bacteria, e.g., in strains of Ca. Thiomargarita nelsonii (40), but not within a single clade of chemoautotrophic symbionts (15, 36, 41, 42).

Characteristic of obligate autotrophs such as the vesicomyid clam endosymbionts is an incomplete TCA cycle that lacks the alpha-ketoglutarate dehydrogenase complex (13, 14). In contrast, Ca. R. santandreae encoded a complete TCA cycle, which in combination with glycolysis, allows it to use sugars and other organic substrates such as fatty acids as carbon and energy sources (Fig. 2 and SI Appendix, Fig. S11). While these pathways might indicate a heterotrophic lifestyle, the results of our analysis of the transporters suggest that Ca. R. santandreae uses these pathways for internal cycling of carbon stocks (SI Appendix, Supplementary Note 2). We identified seven import transporters in the symbiont’s genome with transmembrane domains and found only a very low number of importers of organic compounds (SI Appendix, Fig. S12 and Dataset S1 G and H). The Ca. R. santandreae symbiont can only take up selected peptides, but no sugars or fermentation end products that might be present in the sediments or originate from the host metabolism.

Still, TCA cycle genes were highly transcribed, and the corresponding proteins were abundant, constituting a total of 3 ± 0.49% NSAF (sum of seven proteins; n = 8 proteomes) of all measured proteins in the proteomes (Fig. 2 and SI Appendix, Figs. S3 and S4). We hypothesize that carbon fixation and the TCA cycle do not operate simultaneously in a single cell but are expressed in separate symbiont populations to avoid futile cycling. To replenish intermediates of the TCA cycle and to synthesize and convert acetyl-CoA into biomass, Ca. R. santandreae expresses the ethylmalonyl-CoA pathway and an incomplete 3-hydroxypropionate bicycle (3-HPB) pathway. The ethylmalonyl-CoA pathway (sum of five proteins, 2.25 ± 0.34% NSAF; n = 8 proteomes) is typical for Alphaproteobacteria (43–45) and has a similar function to the glyoxylate shunt that operates in many Gammaproteobacteria but additionally allows the coassimilation of carbon dioxide (43) (Fig. 2). This type of carbon fixation coupled to the replenishment of TCA cycle intermediates is called anaplerotic carbon fixation. For three molecules of acetyl-CoA that enter the ethylmalonyl-CoA pathway, two molecules of CO₂ can be fixed at the expense of only one ATP (46). A similarly highly expressed but incomplete 3-HPB pathway (sum of five proteins, 1.94 ± 0.42% NSAF; n = 8 proteomes) could allow the assimilation of organic compounds such as acetate, propionate, succinate, and malate, even though several key reactions that would permit it to function autotrophically were not predicted in Ca. R. santandreae (35, 47). The elements of the 3-HPB pathway that are encoded represent a second anaplerotic pathway, similar to the ethylmalonyl-CoA pathway, and consume one ATP for the fixation of one molecule of CO₂ (35, 47). Compared with the 4 ¹/₉ ATP that would be needed in the PP_i-dependent CBB cycle for the fixation of two molecules CO₂ (35), both the ethylmalonyl-CoA pathway and the 3-HPB pathway would allow for a cheaper maintenance and expansion of carbon stocks in Ca. R. santandreae. The total proteomic investment in both anaplerotic pathways (i.e., 3-HPB and ethylmalonyl-CoA) was similar to autotrophic carbon fixation via the PP_i-dependent CBB cycle (SI Appendix, Fig. S4). This suggests, in combination with the highly expressed TCA cycle, that the two pathways contribute substantially to carbon assimilation in Ca. R. santandreae. While this has been shown for other mixotrophs (46) or apparent heterotrophs like SAR11, one of the most successful clades of marine bacteria (48), it has not been observed in chemosynthetic symbioses. How much the possible anaplerotic fixation in Ca. R. santandreae contributes to the total carbon fixation in the symbiosis remains to be shown. Both pathways are not essential to replenish intermediates of the TCA cycle they synthetize. It is therefore tempting to speculate that without an essential benefit to the efficiency of the symbiont’s carbon fixation or replenishment of TCA cycle intermediates, one or both pathways would have been lost in the process of reductive genome evolution.

The exact same gene set of the partial 3-HPB pathway described above has also been found in thiotrophic gammaproteobacterial symbionts from gutless oligochaetes and ciliates (35, 49). Our phylogenetic analyses showed that most of the Ca. R. santandreae homologs of genes characteristic for the 3-HPB pathway clustered with genes from these two gammaproteobacterial symbiont groups, except for one that clustered with sequences from other Rhodospirillaceae (SI Appendix, Fig. S13). For three out the five enzymes, the cluster of sequences from thiotrophic symbionts were the sister clade to homologs from Rhodospirillaceae, whereas two of them (Meh and Mct) had no homologs in Rhodospirillales. This extended phylogenetic distribution for several of the enzymes corroborates an origin of these genes for enzymes of the 3-HPB pathway outside of the Chloroflexi, as also suggested for the chemosynthetic symbionts of Kentrophoros (49), and points to the Alphaproteobacteria as one possible group of origin (50).

Paracatenula sp. santandrea shares the Sant’Andrea bay sediments with several groups of chemosynthetic meiofauna such as gutless oligochaetes, stilbonematinae nematodes, and the ciliate Kentrophoros (6, 51, 52). The most abundant taxon is the gutless oligochaete Olavius algarvensis, followed by Paracatenula sp. santandrea. They represent the most contrasting forms of these chemosynthetic symbioses. O. algarvensis hosts a mixotrophic consortium of five bacterial symbionts (53) with little signs of genome reduction in any of the symbiont groups (51), while Paracatenula relies on a single endosymbiont with a drastically reduced genome. These contrasting metabolic capabilities likely reflect their different ecological niches. The Olavius symbiosis can use several energy sources in addition to sulfur oxidation, such as carbon monoxide, hydrogen, and small organic compounds (35, 54), whereas the Paracatenula–Ca. Riegeria symbiosis is an energy-efficient specialist with sulfur oxidization as its main energy source.

Specialization and Convergence in Intracellular Thiotrophic Symbionts.

To test for deviating evolutionary patterns in the alphaproteobacterial Ca. R. santandreae symbiont, we compared its genome to those of 62 bacteria representing free-living relatives of Ca. Riegeria, parasites, nonthiotrophic mutualistic symbionts, as well as thiotrophic bacteria from other symbiotic and nonsymbiotic clades (for a list of genomes, see Dataset S1I). We analyzed the cellular processes encoded in each their genomes in a nonmetric multidimensional scaling (NMDS) ordination of the distribution of each their genes to clusters of orthologous genes (COGs) and to modules of the Kyoto Encyclopedia of Genes and Genomes (KEGG) (Fig. 3 and SI Appendix, Supplementary Note 3).

Fig. 3.

The unique functions in Ca. R. santandreae allow efficient maintenance of large carbon and energy stocks. (A and B) Functional diversity of thiotrophic symbionts and selected reference bacteria. The diameter of each circle represents the genome sizes. Free-living bacteria include Alpha- and Gammaproteobacteria and one Cyanobacterium. A list of genomes used and the raw plots can be found in Dataset S1I (n = 63). (A) NMDS plot of COG category distributions with a 2D stress of 0.152. (B) NMDS plot of KEGG module distributions with a 2D stress of 0.147. (C) Unique pathways in vertically transmitted thiotrophic symbionts with reduced genomes. Metabolic pathways specific for Ca. R. santandreae or shared between two or more symbionts are highlighted. See details in Dataset S1L.

Apart from Ca. R. santandreae, the thiotrophic symbiont genomes formed a spectrum that was dependent on their genome sizes, suggesting that COG- and KEGG-based profiles and genome contents in the other symbionts with small genomes were largely convergent. At the other end of this spectrum, the profiles of symbionts with nonreduced genomes overlapped with free-living thiotrophs such as Allochromatium vinosum and Thiocapsa marina (Dataset S1 J and K).

While Ca. R. santandreae had the most similar distribution of encoded functions to vesicomyid symbionts, another group of vertically transmitted chemosynthetic symbionts, its profile stood out in the observed NMDS patterns across the categories of COGs and KEGG modules (Fig. 3 A and B, SI Appendix, Supplementary Note 3 and Fig. S14, and Dataset S1 J and K; for a list of genomes, see Dataset S1I). Ca. R. santandreae deviated from the vesicomyid symbionts with an enrichment of genes coding for carbohydrate transport and metabolism (COG category G), lipid transport and metabolism (COG category I), and coenzyme transport and metabolism (COG category H) (Dataset S1J). We could corroborate this pattern in a comparison on the level of pathways as Ca. R. santandreae encoded 37 unique pathways that allow for flexible carbon fixation and the versatile storage of carbohydrates (Fig. 3C and SI Appendix, Supplementary Note 3). In particular, the capabilities to synthesize long-chain fatty acids and polyhydroxyalkanoates (PHAs) as well as trehalose and glycogen underline a possible dual role of the symbionts in the Paracatenula symbiosis, where the symbionts provide large storage capacities for the association (Fig. 3C and Dataset S1 L and M). This is in contrast to all other chemosynthetic symbioses and could be coupled to the high proportion of the total biomass that the Ca. Riegeria cells achieve in the Paracatenula consortium (9).

While Ca. R. santandreae has a substantial set of unique pathways, it also shared 107 pathways with the thiotrophic symbionts with the most reduced genomes, Ca. Ruthia magnifica, and Ca. Vesicomyosocius okutanii. Besides energy conservation through sulfur oxidation, these shared pathways were largely related to biosynthesis and include carbon fixation via the CBB cycle and pathways for amino acid, nucleoside, nucleotide, fatty acid, and lipid synthesis (SI Appendix, Supplementary Note 3 and Dataset S1 L and M). These pathways represent the core functions of a nutritional chemosynthetic symbiont and reflect the strong selection based on the conserved metabolic needs of their invertebrate hosts. In such nutritional symbioses, essential biosynthetic pathways must be retained because their loss would lead to a decrease in fitness or even the extinction of the affected host (13, 14, 17). In contrast, a loss of catabolic pathways might not severely impact the symbionts but might, in return, even benefit the consortium as it leaves resources accessible to the host.

Ca. R. santandreae Is a Major Carbon and Energy Buffer in the Paracatenula Symbiosis.

Our comparative analyses showed that Ca. R. santandreae possessed and expressed an unexpectedly high number of pathways for a chemoautotrophic symbiont with a reduced genome, many connected to carbon metabolism and storage. The pathways included the synthesis of PHAs, trehalose, and glycogen (SI Appendix, Fig. S11 and Dataset S1L). We observed a large number of intracellular inclusions in the symbionts in both light microscopy and EM that caused the orange-white appearance of the trophosome region and further highlighted the symbionts storage capabilities (Figs. 4 and 5E). By combining Raman spectroscopic imaging with sensitive MS protocols, we were able to account for most of these storage inclusions that make up large proportion of the total Ca. R. santandreae volume (SI Appendix, Figs. S15 and S16 and Supplementary Note 2).

Fig. 4.

Ca. R. santandreae are large endosymbiotic bacteria with massive storage inclusions. (A) Habitus of Paracatenula sp. santandrea. The white trophosome contains endosymbionts, while the anterior and transparent part of the worm (rostrum) is bacteria-free. (B) Differential interference contrast image of Ca. R. santandreae symbionts indicated multiple intracellular inclusions. (C) Confocal laser-scanning image of CARD-FISH combined with a lipophilic staining (Nile Red) on a transverse section of a Paracatenula specimen in the symbiont-bearing region. Overlay of DAPI signal (blue), Nile Red (red), and probe targeting the symbionts (green, EUB I–III) is shown.

Fig. 5.

Ca. R. santandreae is rarely digested and likely nourishes the host with OMVs. (A–E) TEM analysis of Paracatenula sp. santandrea. (A) Trace illustration of the trophosome region with the status of the intracellular bacterial symbiont Ca. R. santandreae indicated in dark blue (OMVs detected), light blue (no OMVs detected), and orange (digested—lysosomal bodies). See the raw overview image in SI Appendix, Fig. S22. (B–D) Representative image outcrops of the OMV-secreting symbiont population. CR, Ca. R. santandreae; hc, host cytosol; om, OMVs; ph, storage inclusion (PHA). (E) A symbiont cell that has undergone lysosomal digestion (ly). The enclosed structure in the lysosome resembled a PHA inclusion (ph) (Fig. 4C and SI Appendix, Fig. S21A).

Two conspicuous inclusions were detectable in differential interference contrast microscopy. One of them, identified using Raman spectroscopy, was elemental sulfur that appears as light-refractile granules of 1–2 µm in diameter (Fig. 4B and SI Appendix, Fig. S15). The second type of inclusions was larger, with sizes of up to 3.5 µm in diameter, apparent in most cells and occupied 13.1 ± 9.4% of the cell volumes (n = 40 symbiont cells) (Fig. 4C). These were likely composed of PHA, as they were selectively stained by the lipophilic dye Nile Red, and both the polymer polyhydroxyvalerate and its monomer hydroxyvalerate could be identified by gas chromatography–MS (GC-MS) (SI Appendix, Supplementary Note 2 and Fig. S16). Furthermore, genes for PHA synthesis and the phasin protein PhaP were highly transcribed, and PhaP was the most abundant protein in six out of eight proteomes (6.67 ± 1.83% NSAF; n = 8 proteomes) (SI Appendix, Figs. S3 and S4). Besides storing carbon, PHA might function as an electron sink under anoxic conditions (55–57). The host animal Paracatenula likely traverses the oxic and anoxic sediment layers on a regular basis to provide its symbionts with sulfide, a behavior shown for invertebrate meiofauna with thiotrophic symbionts (58, 59). The Ca. R. santandreae symbiont had the necessary genes for aerobic respiration coupled with ATP synthesis but lacked terminal oxidases for utilization of nitrate or alternative inorganic electron acceptors (Fig. 2 and SI Appendix, Supplementary Note 2). A possible electron sink for the oxidation of sulfide to elemental sulfur when oxygen is absent could be the synthesis of PHA from acetyl-CoA, which would simultaneously function as both a store of energy and of carbon. The accumulation of PHA under anaerobic conditions has also been shown in other Alphaproteobacteria that live in changing redox conditions, such as the free-living Ca. Defluviicoccus tetraformis (60). A similar mechanism was shown for Ca. Accumulibacter phosphatis (Betaproteobacteria) and was also proposed for the unrelated gammaproteobacterial thiotrophic symbionts of Olavius algarvensis (55, 56, 61).

Electron-dense particles with the typical size and shape of glycogen inclusions were identified in both host and symbiont tissue (SI Appendix, Supplementary Note 2 and Fig. S17). Glycogen is widely used to store energy and carbon in both bacteria and animals and is considered to be a long-term energy reserve under anaerobic conditions (35, 62–64). Proteins for glycogen metabolism were also detected in the symbiont proteomes, although their genes were not highly expressed (Fig. 2 and SI Appendix, Figs. S3 and S4). We detected proteins for both biosynthesis and degradation in the same samples, suggesting a dynamic switching between synthesis and utilization, possibly distributed in subpopulations of the symbionts (Dataset S1 D and E).

Unexpectedly, the disaccharide trehalose was by far the most abundant soluble metabolite measured in Paracatenula holobionts (SI Appendix, Supplementary Note 2 and Figs. S16 and S18). We could link the trehalose pool to the symbionts as it was not detectable in host tissue without symbionts (anterior rostrum fragments; n = 3; SI Appendix, Fig. S18A). The Ca. R. santandreae symbiont uses the OtsAB pathway for trehalose synthesis from UDP-glucose and glucose-6-phosphate. The trehalose-P-synthase was among the 10% most highly expressed genes and was detectable in the proteome (SI Appendix, Figs. S3 and S4). Trehalose can be found in a wide spectrum of eukaryotes and bacteria and has multiple biological functions including carbon storage but also the protection from osmotic stress, e.g., due to varying salinities (65). Paracatenula sp. santandrea could experience lower salinities due to freshwater input into Sant’Andrea bay in periods of prolonged precipitation. Changes in salinity during salinity tolerance tests however had no effect on trehalose levels, indicating a storage rather than osmolyte function. Specimens survived 5-h treatments in seawater of a salinity ranging from 25 to 45‰, but no significant differences in trehalose concentrations could be measured (salinities from 20 to 50‰; n = 5; two-sided t test; SI Appendix, Fig. S18B).

Based on expression data, chemical measurements, and visualizations, Ca. R. santandreae has a unique ability to accumulate massive amounts of multiple energy and carbon storage compounds. Our chemical characterization and quantification showed that all of these carbon and energy stocks are available with large standing stocks. The carbon stock is dominated by PHA and trehalose (each a few micrograms of C per worm), whereas glycogen is underrepresented (a few nanograms of C per worm) (SI Appendix, Supplementary Note 2). Elemental sulfur, PHA, and trehalose can be solely attributed to the symbionts. Taken together, our observations imply that the Ca. R. santandreae symbiont performs the primary energy storage for the whole animal–microbe symbiosis.

OMVs Secreted by Ca. R. santandreae Likely Form the Basis of Host Nutrition.

Like many marine invertebrates associated with thiotrophic symbionts, Paracatenula lacks a mouth and a gut and depends on its endosymbionts for essential nutrients. This was reflected in the broad genomic repertoire and the expression of amino acid, vitamin, and cofactor biosynthesis pathways in the symbiont (SI Appendix, Supplementary Note 2 and Figs. S19 and S20).

Active export across the symbiont membrane appears to be limited as we only identified 15 export related gene products in Ca. R. santandreae, 8 of which had at least 1 transmembrane (TM) domain based on the identification using the Transporter Classification Database (TCDB) (SI Appendix, Supplementary Note 2 and Fig. S12 and Dataset S1 G–I). The main identified transporters were the Tat and Sec secretion systems that could translocate folded and unfolded proteins to the periplasm. A Type I secretion system could export unfolded proteins across the outer membrane and in the hosting vacuoles but no secretion system for the transfer of proteins into the host bacteriocyte could be detected. Although several transporters such as secretion systems have essential functions conserved across bacteria (66) and might be involved in the nutrition of the host, unexpectedly only the Tat secretion system for the translocation of folded proteins was highly expressed (Fig. 2 and Dataset S1B).

Given that we could not find expression based evidence of substantial export via transporters, we were surprised that we did not observe Ca. Riegeria cells undergoing lysosomal digestion in our transmission EM (TEM) analyses (n = 5 host specimens, 682 symbiont cells, 54,895 µm² total analyzed area) (Fig. 5 and SI Appendix, Fig. S21 and Table S2). We also only identified nine lamellar bodies containing inclusions of symbiont cells in the same dataset. The absence of active bacterial degradation and the presence of few lamellar bodies suggests a low level of symbiont digestion. In contrast, lysosomal digestion of symbiont cells is readily observed in bacteriocytes of other chemosynthetic hosts, e.g., members of the deep-sea mussel genus Bathymodiolus as well as in vesicomyid clams (67, 68). Host gene expression indicated lysosome formation and trafficking as well as proteolysis, all likely involved in the digestion of symbiont biomass (Dataset S1N). One of the most important enzymes in the lysosomal digestion of bacterial cells are lysozymes. They are glycoside hydrolases that cleave the 1,4-beta-linkages in peptidoglycan of bacterial cell walls and lead to cell lysis. Unexpectedly, we could not detect lysozyme transcripts in the host transcriptome [79.1% completeness identified by Benchmarking Universal Single-Copy Orthologs (69)], suggesting that lysozyme expression was below detection limit or absent. Several other types of digestive enzymes were expressed, such as the cysteine protease cathepsin B that has been shown to play a major role in the digestion of gut-bearing flatworms and is also expressed in the symbiotic gutless annelid Olavius algarvensis (70, 71) (Dataset S1N). Because the symbiont very likely has a peptidoglycan cell wall (Fig. 2 and SI Appendix, Supplementary Note 4), we conclude that complete symbiont cells are not the main target for the lysosomal digestion expressed in Paracatenula. This corroborates the ultrastructural observations and is in contrast to other animals with chemosynthetic bacterial symbionts. Hosts such as the Bathymodiolus deep-sea mussels have been shown to constitutively express lysozymes that play a key role in symbiont digestion and the control of symbiont populations (67).

Our data rather suggest that the main mode of symbiont-to-host transfer is based on the release of OMVs and their digestion by the host. OMVs naturally package lipids, proteins, and sugars and can also carry nucleic acids, but despite increased attention into their role in symbiotic interactions, they have not been shown to be broadly involved in host nutrition in animal-microbe interactions (72). We imaged random tissue sections from five individuals using TEM and quantified the number of bacteria, bacteria with OMVs, and lysosomal structures in the bacteriocyte region. On average, of the 136 ± 35 Ca. R. santandreae cells per section, 62.85 ± 10.94% showed OMV formation (total endosymbiont cells: 682; n = 5 specimens; Fig. 5A and SI Appendix, Table S2). We typically observed several clusters of small to large OMVs in close proximity to Ca. R. santandreae cells (Fig. 5 A–D). Correspondingly, we identified a highly expressed intermembrane lipid transport system (ILTS) [also known as mammalian cell entry proteins (mlaED)] of the endosymbiont that has been shown to play an important role in the formation of OMVs (73). The two most highly expressed genes of the ILTS (mlaE and mlaD; TPM: 2,536 ± 682 and 2,097 ± 531, respectively; n = 3) were among the top 10% of expressed genes and represent the subunits that form the transport pore in the inner membrane (74) (Fig. 2). In contrast, the outer-membrane protein of the ILTS, VacJ, had 60× lower expression (VacJ; TPM: 39 ± 13; n = 3). VacJ maintains outer-membrane asymmetry by removing phospholipids from the outer leaflet and supports their trafficking from the outer membrane (75). The observed low expression of VacJ has been linked to increased OMV formation in VacJ-deficient Gammaproteobacteria (73). Although we could not resolve the exact mode of OMV uptake by the host, their channeling into host lysosomal digestion is likely (76) and would also complement the observed expressions of digestive enzymes by the host.

Based on (i) the limited expression of transporter-based export by the endosymbionts, (ii) the observation of OMV formation in the majority of symbiont cells and corresponding symbiont expression profiles, (iii) the detection of host expression of lysosomal digestion but no expression of lysozymes, and (iv) the rare occurrence of digestion of full symbiont cells, we propose that digestion of symbiont material in the form of OMVs as the main mode of nutrition in the Paracatenula symbiosis. In this model, the Ca. R. santandreae symbiont builds up and efficiently maintains the primary energy storage for the whole animal–microbe consortium. The host then directly accesses all these stockpiles via the digestion of symbiont-derived OMVs and gains lipids, amino acids, and sugars, as well as many other possible nutritive constituents. Only a very small subpopulation of its symbionts is digested at any given time, in contrast to all other chemoautotrophic symbioses where symbiont cell digestion drives biomass transfer.

Conclusion

After 500 million years of strict vertical transmission and reductive genome evolution, Ca. R. santandreae has a drastically reduced genome compared with its free-living relatives. Although the symbiont has been associated with its host much longer than, e.g., the oldest known insect symbionts (9, 77), it had to retain a much larger genome and metabolic toolkit, reflecting the multitude of biosynthetic functions that are essential for maintaining the symbiosis. It has maintained a versatile carbon metabolism and carbohydrate storage, in stark contrast to the similarly reduced vesicomyid clam symbionts that appear to be optimized only for autotrophic biomass production. The versatile function of Ca. R. santandreae might be connected to its ancestry in Rhodospirillaceae, with many of which they share a complete intermediary carbon metabolism as well as the ethylmalonyl-CoA pathway that is common in Alphaproteobacteria (43). In addition, Ca. R. santandreae has several unexpected features, even for a versatile alphaproteobacterial background that are shared with gammaproteobacterial symbionts—an incomplete 3-HPB pathway and a PP_i-dependent CBB cycle. In Ca. R. santandreae, the majority of the genes involved in these pathways are of alphaproteobacterial origin and not acquired via lateral gene transfer. Instead of showing an acquisition as proposed for the gammaproteobacterial symbionts, our analyses extended the phylogenetic distribution of these energy-efficient and versatile pathways and shed light on their origins and possible histories of transfer from the metabolically flexible Alphaproteobacteria.

The key elements in the physiology of Ca. R. santandreae such as enhanced modes of carbon fixation, intermediary carbon metabolism, and sulfur oxidation are optimized for high energy efficiency. The endosymbionts can create and maintain versatile carbon pools as they represent the bulk storage organ for the consortium and likely support its host via OMV secretion. This integration of storage and transfer into a single and streamlined chemoautotrophic organism with a reduced genome represents a unique type of bacteria–animal interaction. Ca. R. santandreae extends the spectrum of nutritional symbioses that range from single vitamin or amino acid supplementation to full nutrition (17–23) and adds the function of primary energy and carbon storage that in all other animal hosts is performed by the animal itself.

Materials and Methods

Sample Collection and Light Microscopy.

Paracatenula sp. santandrea specimens were collected between 2013 and 2017 from the bay off Sant’Andrea, Elba, Italy (9). Paracatenula specimens were fixed in RNAlater (Ambion) and stored at 4 °C or fixed in methanol and stored at −20 °C. Specimens for TEM were kept for up to 7 months in 12 mL miniaquaria. Live specimens were imaged using a Nikon SMZ25 stereomicroscope (Nikon) and differential interference contrast on a Zeiss LSM 780 confocal laser microscope (Carl Zeiss AG).

EM.

All specimens were fixed with PHEM (Pipes, Hepes, EGTA, MgCl₂)-buffered glutaraldehyde, processed with a high-pressure freezer, and freeze-substituted in acetone with 1% wt/vol osmium tetroxide. After low viscosity resin embedding and ultrathin sectioning (70 nm), mounted and contrasted sections were randomly chosen for quantitative image analysis of host and symbiont structures.

CARD-FISH and Nile Red Staining.

CARD-FISH, DAPI, and Nile Red staining were performed on paraffin sections of Paracatenula sp. santandrea, and relative volumes of Nile Red-stained inclusions were estimated using quantitative image analysis.

DNA Sequencing, Assembly, and Binning.

DNA was extracted using DNeasy Blood and Tissue Micro Kit (Qiagen). Illumina MiSeq reads were sequenced from paired-end libraries (Ovation Ultralow Library System V2; NuGEN), kmer-filtered for the high coverage symbionts, and assembled. The genome bin derived from the metagenomic assembly graph was improved using iterative read mapping and reassembly.

Metabolic Reconstructions and Comparison.

The Ca. R. santandreae genome assembly was annotated, and the automated annotations were manually curated (transporters, PHA synthesis, and incomplete 3-HPB). Protein sequences were assigned to Clusters of Orthologous Genes (COGs) and KEGG Orthologs (KOs), and metabolic features were predicted from KOs and our curated annotations. The distribution of genes across (i) COG categories was quantified, and (ii) the completeness per KO module and genome was encoded into a matrix and analyzed using NMDS plots. Pathways were compared between Ca. R. santandreae, Ca. R. magnifica, and Ca. V. okutanii based on Pathway Tools and represented in a Venn diagram.

RNA Extraction and Transcriptome Analyses.

Host and symbiont transcription was analyzed from total RNA extractions of single Paracatenula sp. santandrea specimens. Libraries (Ovation RNASeq System v2; NuGEN) were sequenced on the HiSeq 2500 platform, and transcript counts were generated using kallisto. The TPM were sorted based on the average for the three replicates.

For host transcriptomics, a de novo transcriptome assembly was generated from filtered mRNA reads using Trinity. After cleanup and deduplication, the final host transcriptome that consisted of 54,926 predicted genes was annotated with Trinotate.

Protein Extraction and Proteomic Analyses.

For 1D liquid chromatography–tandem MS (LC-MS/MS) proteomics, tryptic digests from eight samples (four from one animal and four from three animals pooled) were separated using a 260-min gradient and roughly 110,000 MS/MS spectra were acquired per sample using the Orbitrap. The final protein identification database consisted of all protein sequences predicted from the Ca. R. santandreae genome and the database for common laboratory contaminants. A total of 407 symbiont proteins were identified using the Sequest HT node in Proteome Discoverer with an overall false discovery rate of <5%.

Phylogenetic Analyses.

16S rRNA and dsrB sequence identities between Ca. R. santandreae and Ca. R. galateiae were determined using single gene alignments. The average AAIs of Ca. R. santandreae and free-living relatives were calculated with the AAI calculator (enve-omics.ce.gatech.edu/aai).

The phylogenomic tree of Ca. R. santandreae and representatives from five alphaproteobacterial orders (Rhodospirillales, Sphingomonadales, Rhodobacterales, Rhizobiales, and Rickettsiales) was calculated based on 43 conserved marker and estimated from the amino acid alignment using FastTree.

Single-gene phylogenetic analyses based on amino acid alignments were conducted for RuBisCO, PP_i-PFK, PP_i-H⁺-pump, and five genes of the incomplete 3-HPB pathway (Mch, Mcl, Smt, Mct, Meh).

Physiological Experiments.

All physiological experiments were performed with biological replicates; the numbers are indicated in each experiment. Individual specimens served as technical replicates. Size normalization was based on images of each specimen.

Pulse and pulse/chase incubations of Paracatenula specimens with ¹⁴C-labeled bicarbonate were performed in either glass beads and artificial seawater or washed Sant’Andrea sediment and natural Mediterranean seawater. Samples were fixed in paraformaldehyde or glutaraldehyde, and the anterior half was used for microautoradiography and the posterior half for bulk measurements. Bulk ¹⁴C incorporation was measured by liquid scintillation counting. Microautoradiography was performed on LR White (London Resin Company) sections.

For salinity tolerance experiments, specimens were incubated with filtered natural seawater of salinities ranging from 20 to 50‰ for 5 h and fixed in methanol. No specimens could be recovered from 20 and 50‰ setups.

Metabolite Measurements.

Untargeted and PHA-targeted GC-MS–based metabolite measurements were performed on extractions from single Paracatenula specimens (trehalose quantification), five Paracatenula specimens (PHA quantification), and 25 Paracatenula specimens (bulk metabolome analysis). Additionally, three Paracatenula specimens were sectioned to separately analyze equal-sized fragments of rostrum (symbiont-free, anterior) and trophosome (symbiont-hosting, posterior).

A kit-based fluorometric glycogen assay (MAK016; Sigma-Aldrich) was used to quantify glycogen contents in tissue extracts from single worm specimens. Raman spectroscopy was performed on live Paracatenula sp. santandrea specimens to identify highly refractile inclusion in symbiont cells.

Data Availability.

The assembled and annotated Ca. R. santandreae genome has been deposited in the European Nucleotide Archive (ENA) and is accessible under project no. PRJEB26644 and the assembly under accession no. LR026963. Raw total RNA library reads and the host assembly have been deposited in ENA under project no. PRJEB31702 via The German Federation for the Curation of Biological Data (GFBio) (78).

The MS proteomics data and the protein sequence databases have been deposited in the ProteomeXchange Consortium via the PRIDE partner repository (79) for the pure culture data with the dataset identifier PXD009856.

The raw data for the EM-based quantitative image analysis have been deposited in Figshare (https://doi.org/10.6084/m9.figshare.7746806) (80).

Code Availability.

The script that was used to classify the transporter families is available at https://github.com/kbseah/tcdbparse_sqlite.