Coming together—symbiont acquisition and early development in deep-sea bathymodioline mussels

(a) Sampling, fixation and sample preparation

Deep-sea mussels were collected from the sea floor with remotely operated vehicles (electronic supplementary material, table S1). Mytilus edulis were collected in the Baltic Sea at a site close to Kiel, Germany (electronic supplementary material, table S1). Samples were preserved for morphological, FISH and TEM analysis (electronic supplementary material, Methods a). All specimens were photographed and shell dimensions and shell margin limits were recorded as shown in electronic supplementary material, figure S1b–d (see also electronic supplementary material, Methods b, table S2). For histological analysis, all samples were post-fixed, embedded, serial sectioned and stained with a toluidine blue and sodium tetraborate solution. For TEM, semi-thin sections were re-sectioned according to [19] (electronic supplementary material, Methods c). Paraformaldehyde fixed samples were decalcified and DOPE-FISH was performed on embedded and sectioned samples using general and specific probes (electronic supplementary material, table S3, Methods d and e) [20]. Details of the light, fluorescence and electron microscopes used are found in electronic supplementary material, Methods f.

(b) SRμCT measurements

SRμCT datasets were recorded at the DESY using the P05 beamline of PETRA III, operated by the Helmholtz-Zentrum Hereon (Geesthacht, Germany [21]). The X-ray microtomography setup at 15–30 keV and 5× to 40× magnification was used to scan resin-embedded samples with attenuation contrast and uncontrasted samples in PBS-filled capillaries [22] with propagation-based phase contrast. Scan parameters are summarized in electronic supplementary material, table S4. The tomography data were processed with custom scripts implemented in the ASTRA toolbox [23–25] (electronic supplementary material, Methods g).

(c) Correlative workflow for SRµCT, light and electron microscopy, and FISH

Section series were screened by eye to identify individual host cells and predict symbiont colonization state based on three morphological characteristics: hypertrophy, loss of microvilli and loss of cilia (table 1, electronic supplementary material, Methods i). The location and symbiont colonization state of the analysed host cells was marked using Cell Counter in Fiji [26]. The analysed sections were re-sectioned and the same fields of view were recorded with TEM (electronic supplementary material, figure S2). To validate the LM-based predictions, symbionts were identified in TEM images based on their morphology (electronic supplementary material, video S5).

Table 1. Overview of number of individuals analysed for each species, developmental stage, and imaging method. The numbers correspond from left to right to B. puteoserpentis (black), B. azoricus (dark yellow) and B. childressi (blue).

Collapse

(d) Image processing and three-dimensional visualization

Microscopy images were adjusted and figures composed using Fiji, Adobe Photoshop and Adobe Illustrator 2021. LM-images were stitched and aligned with TrackEM2 [27] in Fiji. Amira 2020.2 (ThermoFisher Scientific) was used to generate three-dimensional models from LM and µCT datasets. Co-registration between µCT, LM and TEM datasets was carried out after [19] (electronic supplementary material, Methods j).

(a) Identification of developmental stages

We measured the shell lengths of 259 specimens: 129 B. puteoserpentis, 124 B. azoricus and 6 G. childressi individuals. We assume that these had already settled or were in the process of settling, as we collected them from mussel beds on the sea floor. The specimens ranged from 370 µm to 4556 µm shell length (electronic supplementary material, tables S2 and S5). The earliest developmental stages were pediveligers, with shell lengths of 366–465 µm. Developmental stages could only be identified through detailed analyses of the mussels' soft body anatomy (electronic supplementary material, figure S1a,b). These analyses revealed that the shell sizes of 58 pediveligers and mussels in metamorphosis overlapped with those of the smallest plantigrade stages (electronic supplementary material, figure S1f).

(b) Morphological characterization of Bathymodiolus and Gigantidas developmental stages

For our morphological analyses, we analysed 39 individuals (table 1). Bathymodiolus puteoserpentis specimens were best preserved and covered the widest range of developmental stages. We therefore focussed our detailed morphological analyses on B. puteoserpentis, and compared these with selected B. azoricus and G. childressi stages (electronic supplementary material, figures S3 and S4). In the following, we describe the shared morphological features of all three species unless specified otherwise.

The pediveliger larvae were characterized by the presence of a velum, a fully developed digestive system, a foot with two pairs of retractor muscles and two gill baskets (figure 1a; electronic supplementary material, table S6, figure S5a,d and video S1). The digestive system consisted of the mouth, oral labial palps, oesophagus, stomach, two digestive glands, gastric shield, the style sac with crystalline style, mid gut, s-shaped looped intestine and anal papillae (figure 1a; electronic supplementary material, figure S6). The diverticula of the digestive glands and the epithelia of the stomach contained membrane-bound lipid vesicles (figure 1; electronic supplementary material, figures S6 and S7). The gill baskets on each side of the foot (figure 1a; electronic supplementary material, figure S6) consisted of three to four single gill filaments in B. puteoserpentis and five in B. azoricus and G. childressi (electronic supplementary material, figure S3b,c). These filaments form the descending lamella of the inner demibranch in later life stages. For further details, see electronic supplementary material, Note 1.

In mussels undergoing metamorphosis, the first steps from a planktonic to a benthic lifestyle were visible in the degradation of the velum (electronic supplementary material, figure S5) and the appearance of byssus threads (electronic supplementary material, figure S8). Rearrangements of all organs occurred in this stage, for example, the alignment of the growth axis of the gill ‘basket’ with the length axis of the mussel (figure 1a–c; electronic supplementary material, figure S9 and video S2). The number of gill filaments increased by one in all species (electronic supplementary material, figure S6). Furthermore, gill filaments separated from each other, increasing the gaps between them from 47 µm to 120 µm (figure 1c). In the digestive gland and stomach epithelia, the number and volume of lipid vesicles decreased from 12.8% of the soft body volume in pediveligers, to 3.9% in metamorphosing mussels, 1.5% in plantigrades, and were no longer present in juveniles and adults (figure 1a–d; electronic supplementary material, table S6). The organs involved in filter feeding also changed. In pediveligers, the main feeding organ was the velum, which collects particles from the seawater and transports them to the oral labial palps and mouth. After the degradation of the velum, particle sorting was taken over by the highly ciliated foot (electronic supplementary material, figures S5, S8 and S9) in metamorphosing mussels, and by the gills in late plantigrades and juveniles. The plantigrade stage began once the mussels secreted the dissoconch and completed metamorphosis (electronic supplementary material, figures S1 and S10). As the mussels transitioned from the plantigrade to the juvenile stage, the digestive system straightened (figure 1c,d; electronic supplementary material, figure S11 and videos S3 and S4). This morphological change was most prominent in the intestine, which went from a looped to a straight shape and remained straight in all later developmental stages (figure 1c,d; electronic supplementary material, figure S11). For further details see electronic supplementary material, Note 2.

(c) Establishment of the symbiosis

Central to an accurate assessment of symbiont colonization and symbiont-mediated morphological changes was our correlative approach, which combined SRµCT, light and electron microscopy (electronic supplementary material, video S5 and figure S2) and was complemented with FISH. This allowed us to rapidly screen whole animals, yet achieve the resolution needed to identify the colonization of single eukaryotic cells by symbiotic bacteria. We first searched for a morphological characteristic that was visible using light microscopy and reliably revealed the presence of symbionts in host cells. Previous studies [30,31] showed that in juvenile and adult Bathymodiolus mussels, the morphology of epithelial cells colonized by symbionts is fundamentally altered: (i) the microvilli that cover all epithelial cells are lost (known as microvillar effacement) and (ii) epithelial cells become hypertrophic (swollen) compared to aposymbiotic cells. We identified a third characteristic change in epithelial cells colonized by symbionts that has not received much attention, namely the loss of cilia (figure 2c,f,i). We tested if these three morphological characteristics had predictive power for symbiont colonization by analysing 1965 epithelial cells from a subset of seven B. puteoserpentis individuals (table 1) and comparing LM images of these cells with their correlated TEM images. Our analyses revealed that all cells predicted to have symbionts in the LM dataset were indeed colonized in the TEM dataset, and likewise, all cells predicted to be aposymbiotic were free of symbionts (electronic supplementary material, figure S2). Our approach allowed us to identify host cells that were colonized by only a few SOX symbionts based on the absence of microvilli and cilia, indicating that these are lost immediately after the first symbionts colonize host cells (figure 2e). We next used our verified morphological characters to reveal the onset of symbiont colonization in B. puteoserpentis in this subset of seven mussels. All pediveliger cells were free of symbionts (n = 797 host cells in two pediveliger). In the metamorphosing mussels examined with our correlative approach 1–15% (n = 488 host cells in two metamorphosing mussels) and in plantigrades 21–26% (n = 680 host cells in three plantigrades) of all analysed gill, mantle, foot and retractor muscle epithelia cells were colonized by symbionts.

• Download figure
• Open in new tab
• Download PowerPoint

We then expanded our analyses to LM on another 10 B. puteoserpentis individuals (table 1). These analyses confirmed our results from the correlative dataset: All additional pediveligers (n = 3) and metamorphosing mussels were aposymbiotic (n = 3), all additional plantigrades (n = 2) and juveniles (n = 2) were colonized by symbionts. Finally, we performed FISH on another seven B. puteoserpentis individuals (table 1). These analyses corroborated our LM data on the timing of symbiont colonization, with all pediveliger aposymbiotic and symbiont colonization beginning at metamorphosis (figure 3; electronic supplementary material, figures S12 and S13).

• Download figure
• Open in new tab
• Download PowerPoint

Our correlative workflow revealed that pediveliger were aposymbiotic in all three mussel species (figure 2a,d,g; electronic supplementary material, figure S15a,c). Interestingly, two of the aposymbiotic B. puteoserpentis pediveligers had bacterial morphotypes similar to the SOX and MOX symbionts attached to the outside of their shell (electronic supplementary material, figure S16). Mussels that were undergoing metamorphosis were the earliest developmental stage in which we found symbionts in all three host species, with a shell length of 432 µm in the smallest B. puteoserpentis individual, 510 µm in B. azoricus, and 383 µm in G. childressi. In these metamorphosing mussels, we observed symbionts in epithelial cells of the gill filaments, mantle, foot and retractor muscle (figure 2b,e,h,j–l; electronic supplementary material, figure S15). Once symbiont colonization began, gill tissue morphology was similar to that of adult mussels [30–32]: the majority of gill cells were symbiont-containing bacteriocytes without microvilli or cilia, whereas the only gill cells without symbionts were those at the ventral ends of the gill filaments and at the frontal-to-lateral zones along the length of the filaments, as well as the intercalary cells (electronic supplementary material, figures S6, S9 and S10). In addition to the gills, the epithelial tissues of the mantle, foot and retractor muscles also had symbionts in all three host species (figure 2c,f,i; electronic supplementary material, figure S15). We never observed other bacteria besides the two symbionts in any of the developmental stages, including the intranuclear parasite that infects these mussels, based on FISH analyses with symbiont-specific and eubacterial probes of B. puteoserpentis, and TEM analyses of symbiont morphology of B. puteoserpentis, B. azoricus and G. childressi specimen (figures 2 and 3; electronic supplementary material, figures S12, S13 and S15).

The superior preservation of B. puteoserpentis specimens allowed us to analyse the process of symbiont colonization in this species in more detail. We first found evidence of symbiont colonization in metamorphosing mussels that had only a few bacteria in gill, mantle and foot epithelial (figure 2b,e,h,j–l). Bacterial density per gill cell was the lowest in metamorphosing mussels with 13.7% (±6.3) of the host cell area occupied by symbionts, and steadily increased in later developmental stages, reaching up to 29.0% (±5.0) in plantigrades and 32.1% (±6.1) in adult mussels (electronic supplementary material, table S7). With the onset of symbiont colonization, we observed phagolysosomal digestion of the symbionts in all epithelial cells, even those with only very few symbionts (figure 2b,c,e,f). The process of symbiont colonization appeared to be extremely rapid. Nearly all metamorphosing mussels had either no symbionts at all, or all of their epithelial tissues were colonized. In only two out of six individuals, we occasionally observed SOX and MOX symbionts that were not completely engulfed by the host's apical cell membrane, which we interpreted as ongoing colonization (figure 2j–l). These first steps in colonization were particularly common in mantle epithelial cells, while in the same specimen the gill epithelial cells were already fully colonized (figure 2b,e,f). Furthermore, in host cells where colonization was ongoing, we observed that these were colonized only by SOX (figure 2e), or by both SOX and MOX (figure 2h), but we never observed host cells only colonized by MOX symbionts.

(a) Post-metamorphosis development in Bathymodiolus and Gigantidas deviates from the mytilid blueprint

Our study shows that the use of shell characteristics alone to determine developmental stages of deep-sea mussels is not reliable (e.g. [16]), particularly for shell lengths of settling pediveligers and early plantigrades. Our analyses of shell lengths in developmental stages of the three mussel species revealed an overlap in size of 50 µm between pediveligers and plantigrades. This inconsistency in shell lengths between developmental stages indicates that metamorphosis begin is not dependent on size, and provides further evidence for the ability of mussels to delay metamorphosis while continuing to grow, as previously suggested for G. childressi [33]. Such a delay could be due to a lack of settlement cues or limited nutrition, similar to what is known from M. edulis [34]. Delaying metamorphosis would favour dispersal, potentially leading to an increase in geographical distribution and the colonization of new and remote habitats [35].

The pre-metamorphosis development of Bathymodiolus and Gigantidas mussels is similar to that of their close relatives from the genus Idas [7,15] and shallow water mytilids such as M. edulis [18,29,36]. The pediveligers of Bathymodiolus, Gigantidas, Idas and Mytilus have a large velum, foot, mantle epithelium, digestive system, central nerve system and two preliminary gill baskets consisting of three to five gill buds [7,15,18]. During metamorphosis, the velum is degraded, organs within the mantle cavity are rearranged and the lipid vesicles in the digestive diverticula and stomach are reduced. In early developmental stages of mussels from the Mytilidae, lipid vesicles serve as storage compounds to fuel the energy-demanding process of metamorphosis [18] and they probably have a similar function in Bathymodiolus and Gigantidas. Furthermore, these lipid vesicles could provide energy for movement of the pediveligers during their searches for sites to settle [37,38].

Although early development appears to be conserved across Bathymodiolus, Gigantidas, Idas and their shallow water relatives, marked differences occur as soon as symbiont colonization begins in the deep-sea mytilids. All colonized epithelial cells lost not only their microvilli but also their cilia, and developed a hypertrophic habitus, as previously shown for gill bacteriocytes [7,30,31]. If the symbionts of Bathymodiolus and Gigantidas actively induce these cellular changes or if these are a response by the host to symbiont colonization remains unresolved, but our data shows that these processes were tightly linked spatially and temporally. Furthermore, our correlative analyses demonstrate that these cell surface modifications serve as reliable markers for the state of symbiont colonization. Observations of effacement of cilia and microvilli have been reported for a wide range of bacteria that invade epithelia, particularly pathogens, and for these it also remains to be shown if the bacteria or the host drive these processes [39–41].

Although it has been known for several decades that the symbionts of Bathymodiolus and Gigantidas mussels supply their host with nutrition, and that adults possess only a rudimentary gut, nothing was known about the development of the digestive system in these mussels. We observed a straightening of the digestive system after completion of metamorphosis in all three species (figure 4). The stomach and the intestine straightened and the digestive system changed from the complex looped type found in Mytilus to the straight type seen in adults of bathymodioline mussels (electronic supplementary material, table S8). This transformation is striking, as in mytilids like M. edulis, such drastic morphological changes after metamorphosis are not known [42]. In Bathymodiolus and Gigantidas, the straightening of the digestive system did not coincide with the first stages of symbiont colonization or metamorphosis, but rather occurred during the transition from the plantigrade to the juvenile stage, well after metamorphosis and only when these hosts had become fully colonized by their symbionts (figure 4).

• Download figure
• Open in new tab
• Download PowerPoint

In general, the morphology of an animal's gastrointestinal tract reflects its food sources. Animals that digest complex foods possess enlarged compartments and lengthened gastrointestinal structures to slow down the flow of digested material and increase the breakdown of complex molecules [43]. We hypothesize that the straightening of the digestive system in the vent and seep mussels we analysed here was induced by their shift from filter feeding to gaining nutrition from their symbionts. The straightening of the digestive tract could be an evolutionary adaptation to minimize the energy needed for maintaining the digestive tract once the majority of nutrition is gained through intracellular digestion of symbionts in the bacteriocytes.

Not all bathymodiolins have a straightened digestive system. Idas intestines are complex with one or more loops [7,15,44]. These small mussels are commonly found at organic falls, which typically have higher inputs of organic matter than vents and seeps. Also, symbiont abundances are lower in Idas than in the large vent and seep bathymodiolins, indicating that Idas mussels may depend more on filter feeding than their larger relatives. Intriguingly, the relatively large adults of the bathymodioline genus Vulcanidas have a pronounced looped intestine despite high symbiont abundances [45]. Vulcanidas inhabits relatively shallow hydrothermal vents close to the photic zone (140 m compared to 1000–2500 m depth for the mussels in our study) where more organic material from the photic zone is available for their nutrition. It is thus likely that filter feeding plays a greater role in the nutrition of Idas and Vulcanidas than in the three species analysed in this study. These findings raise the question whether the straightening of the digestive tract is a conserved developmental trait in only some genera, or if the environment and availability of organic matter and other non-symbiotic food sources drives the morphology of the digestive tract in bathymodioline mussels.

(b) Symbiont colonization begins during the plantigrade stage as soon as the velum is degraded

As recently highlighted, how and when Bathymodiolus and Gigantidas mussels acquire their symbionts has remained, as yet, unclear [9]. Here, we used correlative imaging analyses to reveal that the early developmental stages of two Bathymodiolus and one Gigantidas species were aposymbiotic, and narrowed the window of symbiont acquisition to mussels undergoing metamorphosis (figure 4). Our analyses revealed that the symbionts colonize the gills only after the velum is lost. As long as the velum is present and active, particles from the surrounding seawater are either transported directly into the digestive system or expelled from the mussel. In M. edulis, once the velum is degraded the gill filaments separate further from each other and take over the task of sorting food particles and generating a water current [18,29]. Beginning in the late plantigrade stage of Bathymodiolus, we identified a similar increase of space between the gill filaments. Given that the developmental changes in the morphology of the bathymodiolin gills are so similar to those of M. edulis [7,15,18,29], it is likely that the functional roles of the gills in Bathymodiolus change in a similar manner as in M. edulis: The pediveligers' gills of Bathymodiolus are used for respiration only, and then become responsible for water current generation in late metamorphosing mussels, and the sorting of particles in plantigrades and juveniles. We hypothesize that this functional shift of the gills plays an important role in enabling the symbionts to adhere to the gill epithelium and initiate symbiont colonization.

The initial colonization of the host by its symbionts appears to be rapid, based on our observation that 22 out of 24 B. puteoserpentis individuals were either completely aposymbiotic or fully colonized, and the first stages of symbiont colonization were only visible in two metamorphosing mussels. As we were working with preserved samples that represent a snapshot of development, the chance of observing a process depends on how often it occurs and how long it takes. The less frequent or the faster a process happens, the smaller the chance of observing it. We therefore conclude that symbiont colonization occurs rapidly in B. puteoserpentis, given the small percentage of individuals in which we observed the first steps of colonization. Symbiont colonization seems to be even more rapid than in I. simpsoni and I. modiolaeformis [7,15]. This could reflect the importance for Bathymodiolus mussels to quickly acquire symbionts once they have nearly consumed their internal energy reserves and settled in an environment that lacks energy-rich planktonic nutrition [46].

We found symbionts in epithelial cells of the gills, mantle, foot and retractor muscle in the plantigrades and juveniles of all three bathymodioline species, similar to previous studies [16,47,48]. Previous work suggested that the symbionts first colonize the mantle epithelia, and from there colonize gill cells, as the first gill filaments are formed from mantle tissues [47]. Our data contradicts this assumption, as the gills had already begun to develop in pediveligers before the onset of symbiont colonization. Furthermore, in two B. puteoserpentis specimens we detected fully colonized gills, while the mantle tissue was still in the process of being colonized. Our findings indicate that symbionts first colonize gill cells before colonizing other epithelial tissues, and that the SOX symbionts colonize individual host cells first, before the MOX.

Our analyses revealed that B. puteoserpentis, B. azoricus and G. childressi larvae are aposymbiotic during their planktonic phase and first acquire their symbionts when they transition to a benthic lifestyle. The timing of symbiont acquisition and major developmental changes was similar in all three species, and, with the exception of the digestive system, corresponds to observations of early life stages of Idas [7,9,15]. These similarities in developmental biology and symbiont acquisition suggest that these traits are conserved in bathymodiolins. The acquisition of symbionts after settlement allows these mussels to recruit locally adapted symbionts. As the geochemistry of vent and seep environments varies strongly across spatial scales [49], recruiting locally adapted symbiont populations would confer a strong fitness advantage to these hosts. Indeed, recent studies have revealed that Bathymodiolus mussels host multiple strains of symbionts that vary in key functions, such as the use of energy and nutrient sources, electron acceptors and viral defence mechanisms [50,51]. By acquiring their symbionts from the sites where they settle, bathymodioline mussels can establish symbioses with those strains that are best adapted to the local environment.