Introduction

A taxonomically and ecologically diverse group of free-living bacteria are magnetotactic: their cytoplasm contains one or more chains of protein-bound, biomineralized magnetite or greigite (‘magnetosomes’), producing a magnetic dipole within the cell (Lefèvre and Bazylinski, 2013). Magnetotactic bacteria can align with the Earth’s magnetic field and move along geomagnetic field lines using their flagella. Many magnetotactic bacteria are chemolithoautotrophic microaerophiles that use magnetotaxis to locate the oxic–anoxic interface (OAI) within sediments, a favorable environment for them due to the proximity of reduced compounds (for example, hydrogen sulfide) and oxygen (Lefèvre and Bazylinski, 2013).

Although most chemolithoautotrophic bacteria have a free-living existence, some form obligate or facultative symbioses with marine invertebrates (Dubilier et al., 2008). The sulfur-oxidizing symbionts of thyasirid bivalves are among the few examples of chemosymbionts that are likely facultative (that is, capable of living both in hosts and in the outside environment), given that they are located outside rather than inside host gill cells (Southward, 1986; Dufour, 2005) and that, in at least one species (Thyasira n. sp. Guiness), host individuals can associate with different symbiont phylotypes, suggesting acquisition from surrounding sediments (Duperron et al., 2012a). Nevertheless, bacteria that associate with thyasirids are considered to be symbionts and not merely microbes trapped on gill epithelia as water circulates within the pallial cavity, because individual hosts associate with a single dominant phylotype of sulfur-oxidizing bacteria (Duperron et al., 2012b). Intriguingly, transmission electron micrographs of the gills of at least four thyasirid species reveal that their symbionts contain abundant electron-dense inclusions, interpreted as being viruses (Southward and Southward, 1991; Brissac et al., 2011; Duperron et al., 2012b).

Thyasira cf. gouldi from Bonne Bay were recently described as forming a cryptic complex in which there are two symbiotic operational taxonomic units (OTUs 1 and 2) and an asymbiotic OTU (3) of bivalves, which vary slightly in shell shape (Batstone et al., 2014). Similar to other thyasirids (Southward, 1986; Dufour, 2005), the symbionts of T. cf. gouldi OTUs 1 and 2 are maintained at the surface of gill epithelial cells within extracellular ‘pockets’ bounded by cytoplasmic extensions of host cells that often bear microvilli (Batstone et al., 2014). Symbiotic thyasirids meet part of their nutritional requirements by assimilating carbon fixed by symbionts (Dando and Spiro, 1993), and electron micrographs show that symbionts are periodically engulfed within gill epithelial cells, a process that results in whorls of lysed bacterial products in host cells (Southward, 1986; Dufour, 2005). As thyasirid symbionts are held among gill epithelial cells, they likely benefit from host bioirrigation and sulfur-mining behaviors, which give them access to oxygen and reduced sulfur (Dufour and Felbeck, 2003; Dando et al., 2004).

Here, we use multiple lines of evidence to show that inclusions in the symbionts of Thyasira cf. gouldi from Bonne Bay, Canada contain iron (most likely iron sulfide), can show magnetic contrast and, in some cases, are arranged in distinct chains; this is the first report of invertebrates forming associations with bacteria that contain magnetosomes.

Materials and methods

Sample collection and dissection

Sediments were collected from three sites (Deer Arm: 30 m depth; Southeast Arm: 30 m depth; and Neddy's Harbour: 15 m depth) in the fjord of Bonne Bay, Newfoundland, Canada, during multiple collection trips between September 2009 and November 2013. Sediments were sieved on a 1 mm mesh and symbiotic individuals of Thyasira cf. gouldi were retained; symbiotic and asymbiotic T. cf. gouldi were sorted on the basis of shell shape and gill morphology upon dissection (Batstone et al., 2014). We could not distinguish between the two symbiotic OTUs (which differ in 18S, 28S and CO1 gene sequences but not in morphological characters), so all specimens analyzed herein consist of either OTU 1 (the most common type) or OTU 2.

Upon dissection, the color of the gills (which varied from pale-pink to black) was noted. Gills of dissected individuals underwent different treatments: some were rinsed in distilled water and individually placed in vials of 95% ethanol for DNA extraction and analysis (see below), while others were fixed in 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (24 h) for either transmission electron microscopy (TEM), selected area electron diffraction (SAED), histological staining or environmental scanning electron microscopy (ESEM) and elemental analysis. Symbionts from the gill of one individual were isolated immediately for atomic force microscopy/magnetic force microscopy (AFM/MFM), as described below.

TEM

To observe symbionts, gills were post-fixed in 1% osmium tetroxide in 0.1 M sodium cacodylate buffer (1 h), dehydrated in an ascending ethanol series and embedded in TAAB 812 resin (Canemco, Canton de Gore, QC, Canada). Ultra-thin (60 nm) transverse gill sections were post-stained with uranyl acetate and lead citrate, and imaged using a Philips 300 transmission electron microscope (Tokyo, Japan).

Histological staining for iron

A prussian blue histological staining protocol (Sheehan and Hrapchak, 1980) was used to localize iron in the gills of 11 Thyasira cf. gouldi specimens. Gills were dehydrated in an ascending ethanol series (without post-fixation in osmium), embedded in paraffin and sectioned (5 μm thick). Deparaffinized and hydrated transverse sections of gill filaments were immersed for 20 min in a freshly prepared solution of 10% aqueous hydrochloric acid and 5% aqueous potassium ferrocyanide, rinsed three times in distilled water and counterstained with Hematoxylin (1 min) before dehydration, cover slipping and imaging with a Zeiss light microscope (Munich, Germany).

Elemental analysis

To investigate whether symbiont inclusions contained iron, we mounted thin (1 μm) resin sections of one gill (post-fixed in 1% osmium tetroxide) on stubs for ESEM and elemental analysis. We used a standard (solid state) backscattered electron detector in an FEI Quanta 650F ESEM (Eindhoven, The Netherlands) to observe and image symbiont inclusions in combination with a Bruker XFlash SSD 5030 X-ray detector (Berlin, Germany) for elemental analysis.

AFM/MFM

We used AFM/MFM to determine whether inclusions are magnetic, as in (Proksch et al., 1995). Symbionts were isolated from a single gill using a Percoll cushion (Distel and Felbeck, 1988), smeared on a glass slide, air-dried and magnetized by placing a magnet perpendicular to the plane of the slide for tens of seconds. We used an Asylum Research MFP-3D Atomic Force Microscope (Santa Barbara, CA, USA), operating in an interleaved AC/lift mode, for all imaging. A Co-Cr coated tip (Mikromasch NSC36, Wetzlar, Germany) was magnetized perpendicular to the cantilever plane before imaging, and samples were magnetized in the plane of the slide. Lift heights of 35–150 nm were tested for magnetic contrast imaging.

SAED

The crystal structure of inclusions was studied using SAED. Resin-embedded gill sections were coated with carbon and then examined in a FEI Titan low base TEM (Eindhoven, The Netherlands) operated at 300 kV. Electron diffraction patterns were obtained by focusing the beam on either one or two inclusions at a time, and d-spacings were determined from these patterns.

DNA extraction and amplification

To identify symbionts, DNA was extracted from the gills of three Thyasira cf. gouldi specimens (following confirmation of symbiont inclusions in the other gill using TEM) using the QIAgen DNeasy Blood and Tissue kit (Hilden, Germany), following the spin-column protocol for animal tissues. We performed PCRs using a universal primer set (27F and 1492R; Lane, 1991), the Promega Master Mix (Promega, Madison, WI, USA) and the following thermocycler conditions: initial denaturation at 94 °C for 4 min, 35 cycles of denaturation at 94 °C for 30 s, annealing (30 s) at 50 °C and elongation at 72 °C for 2 min, with a final elongation at 72 °C for 5 min. We then filtered all amplified products using Acro-Pro 100K-Omega filters (Pall Life Sciences, Port Washington, NY, USA), performed sequencing reactions using BigDye Terminator v3.1 Cycle Sequencing Ready Reaction Mix (Applied Biosystems, Foster City, CA, USA) and electrophoresed on an Applied Biosystems 3730 DNA Analyzer automated capillary sequencer running Sequencing Analysis v. 5.2 Software (Applied Biosystems).

We also sought to determine which form of the Calvin–Benson–Bassham enzyme RuBisCO was present in Thyasira cf. gouldi symbionts, as several magnetotactic bacteria had previously been shown to contain RuBisCO form II (cbbM) rather than form I (cbbL) (Bazylinski and Williams, 2006). To do so, we performed PCR with primer sets and conditions specific for RuBisCO forms I and II (Elsaied and Naganuma, 2001).

Molecular phylogenetic analysis and reconstruction

We used Sequencher (v. 5.2.3, Gene Codes, Ann Arbor, MI, USA) to edit the forward and reverse 16S sequences and used BLASTn (Altschul et al., 1990) to identify the most closely related sequences in GenBank. We then aligned the consensus sequence with various other published sequences confirmed as magnetotactic bacteria or close relatives using Clustal W in MEGA5 (Tamura et al., 2011). We constructed a phylogenetic tree using MrBayes (v. 3.2.2; Huelsenbeck and Ronquist, 2001) with GTR as the substitution model, rate variation set to gamma distribution with four discrete categories and allowance for invariant sites. Data sets were run twice, with initial settings of 10 000 000 generations each and a sampling frequency of once every 100 generations; however, once the convergence diagnostics hit the target value of 0.01, the analysis stopped resulting in a total of 442 trees read. The consensus tree was based on the 332 trees that were actually sampled. Support values at each node represent posterior probabilities calculated in MrBayes.

Detection of free-living symbionts in sediments

To determine whether free-living symbionts exist in sediments from the Thyasira cf. gouldi habitat, we first used fluorescence in situ hybridization (FISH) to label bacteria extracted from sediment samples, as in (Gros et al., 2003). Based on the 16S rRNA sequence obtained, a 20-base oligonucleotide probe specific for T. cf. gouldi symbionts (5′-GCTCACCAAGGCAGCGATCC-3′, melting T=57.97 °C) was designed using the Arb 5.5 probe design tool (Wolfgang et al., 2004) and the gammaproteobacterial 16S rRNA Silva database as of 10 January 2013. Oligonucleotide probes were labeled with fluorescein at the 5' end. Although other sequences in GenBank were identical to the symbiont probe sequence, closest matches belonged to bacteria from terrestrial or hypersaline environments; therefore, the probe was considered to be specific to the T. cf. gouldi symbiont, at least within its native sediments (out of several possible probes, we selected the one that produced the least non-specific hits). To reduce the likelihood of unspecific binding, an unlabeled competitive probe with one-base difference (5′-GCTCACCAAGGCAACGATCC-3′, melting T=55.19 °C) was used concurrently with the labeled probe in FISH assays (this approach was shown by Manz et al., 1992 to drastically reduce non-specific binding). To test probe specificity and determine appropriate hybridization temperatures and formamide concentrations, symbionts extracted from a homogenized gill were filtered on a 0.22-μm black polycarbonate filter and hybridization was attempted under different conditions; binding was successful at 46 °C and with 40% formamide. Labeling with a fluorescein-labeled version of the competitive probe did not occur under these conditions.

Sediment cores (1 cm diameter) were collected from grab samples at the study sites, and subdivided into 0–2-, 2–5- and 5–10-cm-depth fractions. Samples were stored at −20 °C in equal volumes of fixative (4% formaldehyde in the following buffer: 40 g l−1 Borax in sterile 0.22 μm filtered seawater, pH 8.8), in 15-ml centrifuge tubes. Bacteria were extracted from 1.5 ml of sediments by deflocculating using 250 μl of 100 mM sodium pyrophosphate and 3.25 ml of the above-mentioned buffer, followed by three cycles of sonication (10 s), and vortexing (30 s), and a final resting period of 15 min with periodic vortexing. Centrifuge tubes were then placed upright until most sediments appeared to have settled (1–3 min), and 1 ml of supernatant was filtered onto black 0.22 μm polycarbonate filters. Bacterial cells were permeabilized with 95% ethanol (1 min), placed on grease-covered slides and incubated with 50 ng probe in 38 μl hybridization buffer (40% formamide, 0.9 M NaCl, 20 mM Tris/HCl pH 7.4, 0.01% sodium dodecyl sulfate) at 46 °C for 2 h. Filters were then placed into 1.5 ml pre-warmed wash buffer (70 mM NaCl, 20 mM Tris/HCl pH 7.4, 5 mM EDTA, 0.01% sodium dodecyl sulfate) for 10 min, then rinsed with distilled water, dried and mounted on a slide using Permafluor for observation using a Zeiss Axio Imager A2 (Munich, Germany) with appropriate filter set. A sample of bacteria filtered from overlying seawater from the sampling site was used as a negative control; no fluorescent labeling occurred on these filters. Other filters containing sediment-extracted bacteria that had not gone through the hybridization process were tested for autofluorescence using the same filter set used for fluorescein detection; only very faint background fluorescence was noted on these filters.

Further confirmation of the presence of free-living forms of Thyasira cf. gouldi symbionts was obtained by searching for matching 16S sequence fragments among environmental DNA amplicons. Sediment samples from Deer Arm and Southeast Arm were collected in August 2013. Subsamples underwent total DNA extraction using the Xpedition Soil/Fecal DNA Miniprep kit (Zymo Research, Irvine, CA, USA) following the manufacturer’s protocol for soil samples. PCR was then performed using barcoded (IonXpress, Life Technologies, South San Francisco, CA, USA) universal microbial primers targeting the hypervariable V6 region of the 16S rRNA gene developed by the Lang Lab of Memorial University, St. John’s, NL, Canada (modified from Huber et al., 2007). Reverse primers (5′-CGACRRCCATGCANCACCT-3′) contained the barcode while forward primers (an equal mix of: 5′-CTAACCGANGAACCTYACC-3′, 5′-ATACGCGARGAACCTTACC-3′, 5′-CNACGCGAAGAACCTTANC-3′ and 5′-CAACGCGMARAACCTTACC-3′) contained the Ion Torrent adapter sequences (for simplicity, the barcode and adapter sequences are not shown). PCRs were run using Phusion High Fidelity Polymerase and High Fidelity Buffer (New England Biolabs, Ipswich, MA, USA), and dNTPs used in standard concentrations. Thermocycler settings were as follows: 98 °C for 30 s, 30 cycles of 98 °C for 10 s, 72 °C for 30 s, followed by 72 °C for 2 min.

The PCR product was cleaned and prepared for sequencing on the Ion Torrent PGM (Life Technologies) following the protocol provided in the Ion PGM Template OT2 200 Kit and Ion PGM Sequencing 200 Kit v2. An Ion 314 v2 chip was used for sequencing. Runs were cleaned and filtered using the FastQC suite. Raw reads were trimmed at both ends to obtain a quality score >20. These were then aligned to the known 16S sequence of the Thyasira cf. gouldi symbiont as well as close relatives using the Torrent Mapping Alignment Program mapall command. Changes to the default settings included only the best alignments being accepted (-a 0) and an increase from the default 10% misalignment allowance to a 2% misalignment allowance (map1 —max-misalignment 0). The mapping results were then visualized using Integrative Genomics Viewer (Robinson et al., 2011) and manually checked for misalignments.

Results

TEM and histology

Transmission electron micrographs of all symbiotic Thyasira cf. gouldi gills examined (N>80 individuals from various dates and from the three sampling sites) revealed numerous electron-dense octahedral inclusions within the cytoplasm of symbionts; in some cases, these inclusions formed the characteristic chains of magnetotactic bacteria (Figure 1a). In most sectioned bacteria, inclusions did not form a chain but were more or less aggregated in the bacterial cytoplasm (Figure 1b). Inclusions in chains measured 72.9±28.8 nm (N=20 particles) and were more electron dense than inclusions not forming chains (55.7±6.0 nm; N=20 particles).

Figure 1
figure 1

Transmission electron micrographs of symbionts associated with gill epithelial cells of Thyasira cf. gouldi. (a) Symbiont with a chain of magnetosomes, located close to the microvillar boundary (mv) of the bacteriocyte. Scale bar, 500 nm. (b) Multiple symbionts, with magnetosomes not organized in a chain. Scale bar, 500 nm. (c) View of symbionts in an extracellular ‘pocket’ (p), with some undergoing endocytosis (e). A large aggregate of lysed products of symbiont digestion (l), including magnetosomes, is visible in the host cytoplasm. Dark inclusions in the host cytoplasm (cy) may be the end products of lysosomal breakdown (that is, resistant particles of iron). Scale bar, 1 μm. (d) Magnified view of lysed bacterial whorls in the host cytoplasm; some magnetosomes appear in chains (m), others are in less dense aggregations. Scale bar, 500 nm.

Full size image

Evidence of the endocytosis and digestion of symbionts by host cells (with the formation of characteristic whorls; Figure 1c) was visible in many gills. Symbiont inclusions resist host digestion and accumulate in whorls (size: 56.0±6.3 nm; N=20 particles) and in host cytoplasm (size: 77.4±11.1 nm; N=20 particles; Figure 1d). The abundance of bacterial inclusions varied widely among host specimens. Gill color was related to the overall abundance of inclusions, particularly in host cytoplasm: pink gills contained fewer inclusions than darker gills (based on general observations made on all the gills examined).

The cytoplasm of the bacteriocytes of five specimens (out of 11) stained intensely with prussian blue, revealing the presence of iron (Figure 2a). All specimens that were prussian blue positive had either dark or black gills.

Figure 2
figure 2

Iron determination in bacteriocytes and in the magnetosomes of Thyasira cf. gouldi symbionts. (a) Transverse section of a gill of T. cf. gouldi, stained with prussian blue and hematoxylin. The cytoplasm of bacteriocytes is intensely stained, indicating the presence of ferric iron. Scale bar, 20 μm. (b) Backscatter electron image of a symbiont, showing a chain of magnetosomes. Scale bar, 100 nm. (c) Energy dispersive X-ray spectrum after analysis of a particle within the magnetosome chain. Peaks for iron are visible.

Full size image

Elemental analysis of inclusions

With ESEM, cellular structure could be observed in backscatter mode, and bacterial inclusions, sometimes forming chains, were visible (Figure 2b). X-ray spectra revealed the presence of iron in those structures (Figure 2c), but not in adjacent tissue. Sulfur was also identified within the inclusions. Osmium in the sample was an artifact of the postfixation procedure.

Magnetic contrast in isolated symbionts

AFM topography images showed clumps of cellular material, with a height range of approximately 120 nm (Figure 3a), and MFM imaging revealed clear magnetic contrast in certain regions, measuring <200 nm (Figure 3b). The magnetic contrast remained evident at MFM tip fly heights of 100 nm, which is far beyond the distance over which van der Waals interactions would be significant.

Figure 3
figure 3

Simultaneously acquired AFM and MFM images of isolated Thyasira cf. gouldi symbionts. Scale bar, 5 μm. (a) AFM topography image of dried symbionts, where taller features are shown in white. Magnetosomes are not evident. Inset shows AFM image of outlined area. (b) MFM phase-contrast image of the area shown in a at a fly height of 100 nm. White and black regions indicate magnetic interactions between the sample and the magnetic tip. Inset shows MFM image of outlined area.

Full size image

Crystalline structure of inclusions

Five diffraction patterns were obtained using SAED, and the following d-spacings were determined: 2.94, 2.5 (two times), 1.56 (three times) and 1.8 Å. Comparisons with reference materials (greigite, magnetite, mackinawite, hematite and pyrite; Lennie et al., 1995; Downs, 2006) indicate that inclusions are most likely a combination of greigite and mackinawite (Supplementary Table 1).

Symbiont identification and phylogeny

We amplified a single 16S rRNA sequence from each host specimen, with sequences from each host individual being exactly the same (length=1430 bp, accession number: KJ658209), and used it to construct a phylogenetic tree (Figure 4). Symbionts belong to the gammaproteobacteria and are closely related to the symbionts of other invertebrates, including Thyasira flexuosa (98.5% identity, accession number L01575.1) and the hydrothermal vent tubeworm Riftia pachyptila (94% identity), as well as free-living bacteria from marine sediments (Figure 4). The magnetotactic bacterium that is most similar to the T. cf. gouldi symbiont (93% similarity) is gammaproteobacterium SS-5 isolated from the Salton Sea (Lefèvre et al., 2012). Other more closely related bacteria are not known to have magnetosomes.

Figure 4
figure 4

Bayesian phylogenetic tree constructed using 16S rRNA gene sequences (1627 positions in the alignment, see text for method used), showing the relationships between the Thyasira cf. gouldi symbiont, other gammaproteobacteria (symbiotic or free living) and known magnetotactic bacteria (in bold). Numbers represent posterior probabilities calculated in MrBayes (v. 3.2.2). Accession numbers are in parentheses.

Full size image

From the same DNA extracts, we amplified the gene for RuBisCO form II (cbbM, length=397 bp, accession number: KJ658208), but could not amplify the gene for form I RuBisCO (cbbL). The most closely matched sequence (89% identity) was from a symbiont of a deep-sea tubeworm (Lamellibranchia sp., accession no. FM165442). There are no other thyasirid symbiont RuBisCO sequences available in GenBank.

FISH with the symbiont-specific 16S probe resulted in positive labeling of bacteria extracted from sediments collected at the three study sites, at all depth fractions examined. Following extraction of DNA from Bonne Bay sediments and analysis of PCR product sequences, we identified reads matching the known Thyasira cf. gouldi 16S sequence with 98–100% similarity in both the samples. This level of specificity allows for natural mutations within the population as well as sequencing errors; other authors have considered that sequences >97% identical belong to the same microbial species (Huber et al., 2007). Reads that aligned represented about 4% of the total data, suggesting that the symbiont represents a small portion of the bacterial community within these sediments. A blastn search revealed that the aligned reads were more similar to the T. cf. gouldi sequence than to any sequences available in the NCBI non-redundant database, with the top blast hits being 96% similar and originating from uncultured marine environmental samples. Therefore, there are free-living forms of the T. cf. gouldi symbionts in native sediments.

Discussion

Together, our results provide evidence for the presence of magnetosomes in the symbionts of Thyasira cf. gouldi, and allow us to propose a mechanistic pathway for symbiont acquisition in this species. We identified a single bacterial phylotype belonging to the sulfur-oxidizing gammaproteobacteria in T. cf. gouldi, as is typical of other thyasirids and many other chemosymbiotic bivalves (Dubilier et al., 2008; Duperron et al., 2012b). The single bacterial phylotype obtained from DNA extractions of T. cf. gouldi gills represents a symbiont and not simply an environmental contaminant trapped on epithelial cells, based on the following evidence: (1) numerous (N>80) symbiotic T. cf. gouldi specimens observed using TEM contain morphologically similar bacteria; (2) different host individuals (N=3 in this study) associate with the same, unique symbiont phylotype, a situation that would be highly unlikely if these bacteria were simply filtered from seawater; and (3) host epithelial cells clearly endocytose these bacteria similar to other bivalves where this process was demonstrated to be a pathway for nutrient transfer (Streams et al., 1997).

TEM observations show that inclusions are common among Thyasira cf. gouldi symbionts and fall within the general range of morphologies and sizes (tens of nanometers) reported for magnetosomes (Lefèvre et al., 2013a). The occasional chains of inclusions, along with the demonstration of iron (with magnetic properties) in both inclusions and host bacteriocytes show that those structures are not viruses, but rather magnetosomes. The diffraction patterns obtained from these inclusions suggest that particles are a mixture of the iron sulfides greigite and its precursor mackinawite as in other marine and estuarine species of magnetotactic bacteria (Pósfai et al., 1998), including gammaproteobacteria (Simmons et al., 2004). Mackinawite is non magnetic and is converted to greigite in magnetotactic bacteria (this transformation was observed after sample preparation and exposure to air; Pósfai et al., 1998), therefore, the presence of both forms of these minerals within a population of bacteria with magnetosomes is not surprising. The presence of those minerals is consistent with the identification of sulfur within inclusions using elemental analysis, and the shape of inclusions resembles that of other described greigite and mackinawite particles (Pósfai et al., 1998; Spring and Bazylinski, 2006). Further, we noted less MFM signal than we expected in our bacterial cell preparations, which may indicate that an important fraction of the magnetosomes contained mackinawite in our preparations. Recently, a strain of sulfate-reducing magnetotactic deltaproteobacteria (BW-1) was shown to biomineralize both greigite and magnetite, depending on environmental conditions (Lefèvre et al., 2011); another freshwater strain of deltaproteobacteria simultaneously produces magnetite and greigite (Wang et al., 2013). We have no evidence for similar physiological capabilities in the symbiont of T. cf. gouldi, but given that only five diffraction patterns were obtained herein, we cannot affirm that this species is unable to biomineralize magnetite.

Although the presence of magnetotactic bacteria within a host organism appears counterintuitive as symbionts have no need for magnetotaxis, a consideration of the entire ecological spectrum of these bacteria, along with host behaviors, helps to elucidate their presence in host organisms. The existence of free-living symbionts in sediments from the Thyasira cf. gouldi habitat suggests horizontal symbiont transmission in these thyasirids. In their free-living state, the symbionts of T. cf. gouldi would benefit from magnetotaxis (or magneto-aerotaxis, Frankel et al., 2006) as a means of tracking OAI in sediments, thereby facilitating sulfur oxidation for chemolithoautotrophy (as in other marine magnetotactic bacteria, Bazylinski and Williams, 2006). However, magnetotaxis would require the presence of flagella, which are not apparent on TEM images of gill-associated symbionts. The T. cf. gouldi symbiont is likely able to produce a flagellum when free living, and to lose it once associated with its host: the closely related, facultative symbiont of Riftia pachyptila (Cad. Endoriftia persephone) contains genes required for a functional flagellum and for chemotaxis, although flagella have never been observed in symbionts (Millikan et al., 1999, Robidart et al., 2008). Proteomic studies suggest physiological differences between host-associated and free-living states of Cad. Endoriftia persephone (Markert et al., 2007), and physiological differences probably also exist between free living and associated T. cf. gouldi symbionts. In magnetotactic bacteria, the physiological state was shown to affect magnetosome formation (Bazylinski and Williams, 2006).

Thyasirid bioirrigation can produce OAI along burrow walls (Dando et al., 2004; Hakonen et al., 2010), thereby enhancing the concentration of magnetotactic bacteria in near-burrow sediments. Thyasira cf. gouldi could collect symbionts from burrow walls on the mucociliary surface of their elongated feet (Allen, 1958). Upon retraction of the foot within the mantle cavity, symbionts trapped in mucus could collect on the gill. Once associated with host cells, symbionts would have no requirement for a flagellum or a functional magnetosome, as the host acquires oxygen and sulfide through bioirrigation and sulfide-mining behaviors (Dufour and Felbeck, 2003; Dando et al., 2004). Interestingly, the biomagnetic crystals are rarely organized in a chain in T. cf. gouldi symbionts, suggesting that proteins responsible for chain formation are not maintained in hospite. Other magnetotactic bacteria with clusters of magnetosome crystals rather than chains retain magnetotactic abilities (Cox et al., 2002; Zhang et al., 2013); however, the crystals in those species are aggregated at one end of the bacterial cell and not scattered throughout the cytoplasm as in T. cf. gouldi symbionts. Further, the apparent presence of both mackinawite and greigite (with the former possibly predominating) in the gill-associated symbionts of T. cf. gouldi begs the question of whether mechanisms for transforming mackinawite to greigite remain functional in associated symbionts. The few T. cf. gouldi symbionts with intact chains of magnetosomes may be recently acquired individuals, suggesting that symbiont uptake in thyasirids can occur more than once in their adult life, as in lucinids (Gros et al., 2012).

The whorls in the bacteriocytes of Thyasira cf. gouldi indicate that host cells can engulf and digest symbionts, a likely means of nutrient uptake in thyasirids (Southward, 1986; Dufour, 2005). Through this process, magnetic particles appear resistant to host degradation, concentrating in whorls and in the cytoplasm of epithelial cells. However, particles within host cytoplasm are slightly larger and appear more diffuse than those within intact bacteria. Similarly, ciliates having engulfed magnetotactic bacteria accumulated enlarged magnetosomes in their cytoplasm (Martins et al., 2007), and the size increase was interpreted as showing magnetosome dissolution. Occasionally, host gills contain such dense concentrations of particles that they appear black; such gills stained intensely with prussian blue and were therefore iron-rich. T. cf. gouldi may eliminate magnetic particles through the renewal of epithelial cells, an active process in bivalve gills (Hanselmann et al., 2000).

Although much remains to be learned about the phylogeny and metabolic capabilities of Thyasira cf. gouldi (and other thyasirid) symbionts, 16S rRNA sequencing supports their phylogenetic placement among other sulfur-oxidizing gammaproteobacteria, and the amplification of type II RuBisCO provides evidence for autotrophic metabolism in these bacteria. The symbionts of T. cf. gouldi represent one of the few described magnetotactic gammaproteobacteria (Simmons et al., 2004; Lefèvre et al., 2012; Wang et al., 2013). Genomic studies have shown that genes responsible for magnetosome formation are often organized within ‘magnetosome islands’ that contain mobile elements (Ullrich et al., 2005; Jogler et al., 2009), and recent work suggests a monophyletic origin for magnetotaxis within the proteobacteria (Lefèvre et al., 2013b). Inclusions resembling magnetosomes, but interpreted as viruses, have been noted in other thyasirid symbionts, notably in ‘symbiont of Thyasira flexuosa 1’ from the Mediterranean, in Figure 4; (Brissac et al., 2011), in symbionts of T. gouldi from Scotland (Southward and Southward, 1991) and of T. flexuosa from Long Beach, USA (Dufour, 2005) and Brest, France (J. Laurich, personal communication). Further phylogenetic and ecological studies of thyasirids and their symbionts should clarify why hosts from such different locations associate with closely related, magnetosome-forming symbionts.

The discovery of magnetotactic bacteria as symbionts of bivalves underscores their vast adaptability (Lefèvre and Bazylinski, 2013) and provides new opportunities to study magnetosome formation in an ecological context. The association of such bacteria with thyasirids is likely facilitated by the host’s sulfide mining and bioirrigation activities, which produce OAI along burrow linings. In thyasirids, the combination of host and symbiont behaviors appears to have been key to chemosymbiosis establishment, enhancing the success of this family in a variety of sedimentary environments.