Introduction

Mostly described as endosymbioses (Dubilier et al., 2008), associations between invertebrates and chemoautotrophic bacteria are regularly encountered in reducing marine habitats (Cavanaugh et al., 2006). Among the Crustacea, chemosynthetic epibioses are reported in only a few hydrothermal decapods: the shrimp Rimicaris exoculata (Segonzac et al., 1993), the galatheid crabs Kiwa hirsuta (Macpherson et al., 2005; Goffredi et al., 2008), Kiwa puravida (Thurber et al., 2011), and Shinkaia crosnieri (Miyake et al., 2007), and two amphipods from littoral sediments (Gillan and Dubilier, 2004; Gillan et al., 2004) and sulphide-rich caves (Dattagupta et al., 2009). Most of these associations are regarded as nutritional ectosymbioses, but to date, there is little direct evidence of bacterial autotrophy and no direct demonstration of mutualistic nutritional bacteria–host interactions.

The shrimp R. exoculata (Williams and Rona, 1986) dominates the megafauna of several deep-sea hydrothermal vent sites of the Mid-Atlantic Ridge, forms dense aggregates around active chimneys (Segonzac, 1992) and harbours a luxuriant bacterial community in its enlarged gill chamber (Van Dover et al., 1988; Casanova et al., 1993; Segonzac et al., 1993; Polz and Cavanaugh, 1995; Zbinden et al., 2004) (Figure 1). The integument surfaces are periodically re-colonised during the moult cycle (Corbari et al., 2008b), each ecdysis causing shedding of the old cuticle with its bacterial mats. The shrimp R. exoculata, listed as a model organism of an extreme deep-sea environment (CAREX, 2010), is exceptional among crustaceans for its association with bacteria. Recent studies suggest that R. exoculata gill chamber epibionts constitute a diversified community with various morphotypes, phylotypes and metabolisms (Zbinden et al., 2004, 2008; Petersen et al., 2009; Hügler et al., 2011; Guri et al., 2012), but no study has directly demonstrated which substrate(s) the bacteria oxidise to fuel their ‘chemosynthetic’ metabolism(s). This shrimp also harbours a digestive bacterial community (Zbinden and Cambon-Bonavita, 2003; Durand et al., 2010) whose role remains hypothetical (detoxification, nutrition and/or pathogen control).

Figure 1
figure 1

Illustration of the shrimp Rimicaris exoculata gill chamber with the associated bacterial community. (a) Whole shrimp with dilated gill chamber visible on both sides of the cephalothorax. (b) Left gill chamber opened by removal of the branchiostegite (lying below, Brs) and showing the very large mouthparts (scaphognathite, Sca, and exopodite, Exo) as well as the posterior gill set (Gi)—both the mouthparts and the inner side the branchiostegite are colonised by bacterial mats encrusted with particles of red iron oxide. (c) Isolated mouthparts bearing bacteriophore setae sheathed by a thick iron-oxide-encrusted biofilm. (d) Semi-thin cross-section (toluidine blue stained) of the branchiostegite (anterior part), showing the inner and outer integument leaflets (IB and OB) and the bacterial biofilm growing on the inner side. (e, f) Transmission electron microscopy and scanning electron microscopy images of the branchiostegite biofilm, showing several bacterial morphotypes including rods, nested methanotrophic-like bacteria, thick and thin filaments (pictures taken by Dr L Corbari). Scale bars: (a) 1 cm; (b) and (c) 500 mm; (d) 100 μm; (e) 5 μm; (f) 10 μm. C, cuticle; Epi, epidermis; FB, filamentous bacteria; IB, inner branchiostegite; MB, methanotrophic-like bacteria; Mi, mineral particles; RB, rod-shaped bacteria; S, blood sinuses.

Full size image

There is strong evidence, based on stable isotope signatures and essential fatty acid composition, suggesting that the shrimp gets organic matter mainly from its epibionts, rather than from grazing free-living bacteria associated with chimney walls (Gebruk et al., 1993; Pond et al., 1997a, 2000; Rieley et al., 1999; Colaço et al., 2007). Yet this evidence of symbiont–host nutritional relationships is indirect, and how carbon is transferred from bacteria to shrimp remains an open question. The hypothesis of direct transfer of dissolved molecules across the shrimp integument, especially the gill chamber lining, has been proposed on the basis of morphological observations (features dedicated to bacteria farming, absence of scrape marks on the biofilm) (Zbinden et al., 2004; Corbari et al., 2008b), but the nature of the transferred molecules and the integument areas involved remain to be discovered.

The major aims of this study were to test (1) inorganic carbon fixation by active bacterial chemosynthetic metabolisms and (2) the hypothesis of bacterial organic carbon transfer (soluble molecules) to the shrimp tissues. For this, we incubated live shrimps in a pressurised vessel with different isotope-labelled molecules and electron donors, to trace carbon assimilation by the epibionts and shrimp tissues.

Materials and methods

Shrimp collection

R. exoculata specimens were collected with the slurp-gun operated from the manned submersible Nautile at the Rainbow hydrothermal vent site (36°14′N–33°54′W, 2320 m depth) during the ‘MoMARDREAM-Naut’ cruise, July 2007. Healthy specimens were maintained alive in a pressurised IPOCAMP incubator (Shillito et al., 2001) at in-situ pressure (230 bars) and temperature (15 °C) for in-vivo experiments (Zbinden et al., 2008). Control specimens were directly frozen at −80 °C after collection.

Experiments

Experiments were carried out only on red-coloured pre-ecdysial specimens (that is, having an established and still-growing bacterial community) (Corbari et al., 2008b). In the IPOCAMP, incubations with isotopic tracers were carried out in flasks filled till to the top with seawater (SW), whose volume (50 or 250 ml) was chosen according to the treatment duration and number of shrimps to ensure shrimp survival in good physiological condition at the end of the experiment (Shillito et al., 2001; Ravaux et al., 2003). Incorporation of 13C- and 14C-bicarbonate by the chemoautotrophic bacteria was tested with different chemical energy sources. 14C-acetate and 3H-lysine were chosen as representative microbial substrates and/or metabolic by-products liable to mimic transfer of small biomolecules from epibionts to host.

After each experiment, tracer excess was washed off in 0.22-μm-filtered SW: 15 min for 13C and 1, 5 and 30 s, 5 and 15 min for the other experiments. The shrimps were then either directly frozen at −80 °C or dissected into body parts and tissues of interest: (1) regions colonised by the bacteria: the branchiostegites (integument folds enclosing the gill chamber), later separated into the inner bacterial biofilm (BB) and the outer branchiostegite (OB) or shrimp tissues, and the mouthparts (MP, scaphognathites and exopodites); (2) exchange or absorption organs: gills (Gi), digestive tract (DT) and hepatopancreas (HP); (3) abdominal muscles (Mu). In some specimens, the old abdominal cuticle (AC) was separated from the inner tissues (new cuticle and epidermis). In the laboratory, the samples were rinsed once again (see below) and treated to remove unfixed tracers and to preserve the labelled organic molecules before quantification.

13C-bicarbonate incorporation

Four and two shrimps were incubated for 4 and 10 h, respectively, in 250-ml flasks containing 1 g/L NaH13CO3 (98%; Sigma 372382; St Louis, MO, USA) in filtered SW enriched or not (control) with iron (100 μM FeCl2; Sigma 372870) or thiosulphate (100 μM Na2S2O3; Sigma 72049) as chemical energy source. Before use, the dissected tissue samples were rinsed three times in distilled water before dehydration overnight at 60 °C. Dried samples were finely ground to obtain a powder to vaporise in the mass spectrometer.

14C-bicarbonate incorporation

Two specimens were incubated for 6 h in 50-ml flasks containing 0.25 g/L NaH12CO3 labelled with 80 μM NaH14CO3 (4 μCi/ml; GE Healthcare CFA3-5MCI; Buckinghamshire, UK) as inorganic carbon source and dissolved in SW enriched or not with an electron donor (Fe2+ or Na2S2O3) at the same concentration as above. The first shrimp for each condition was directly frozen at −80 °C for autoradiography and scintillation analysis. The second shrimp was dissected into body parts as described above before storage in cryotubes at −80 °C. In unlabelled shrimps (control), the radioactivity level was undetectable to very low in both bacterial mats and shrimp tissues.

In the laboratory, the samples were rinsed four times in distilled water and their fresh weight was determined. After overnight digestion at 50–55 °C in glass scintillation vials containing 1 ml quaternary ammonium hydroxide (0.5 N) in toluene (BTS-450 Tissue Solubilizer; Beckman Coulter 580691; Fullerton, CA, USA), they were acidified for 2 h at room temperature by adding 250 μl acetic acid to remove any remaining 14C-bicarbonate and the calcium carbonate present in the cuticle or HP. Scintillation liquid (Ready Organic; Beckman Coulter 586600) (10 ml) was added to all samples just before reading.

14C-acetate and 3H-lysine incorporation

Two sets of three to four shrimps were incubated for 1 h in 50 ml SW containing either 20 μCi/ml 3H-lysine (GE Healthcare TRK520-1MCI) (that is, 0.2 μM) or 10 μCi/ml sodium 14C-acetate (GE Healthcare CFA14-1MCI) (that is, 0.2 mM). In both experiments, one entire specimen was immediately frozen after rinsing (for autoradiography and analysis). The others were dissected and the pieces stored at −80 °C. Samples were treated as in the 14C-bicarbonate experiment.

Stable isotope and elemental analysis

The atomic percentage of stable isotope (%at 13C) of each sample and its carbon content expressed as a percentage of dry weight (%Cdw) were obtained with an isotope ratio spectrometer (Optima, Isoprime, UK) coupled to an elemental analyser (Carlo Erba, Milan, Italy) (Lepoint et al., 2004).

The percentage of natural 13C in control shrimps was first subtracted from the measured 13C percentage in each sample to obtain the percentage of 13C (or total carbon) incorporated during the experiment (%at 13Cinc). Carbon incorporation rates (Cinc), expressed in moles of carbon incorporated per time unit and weight unit of total carbon in the sample (μmolCinc × gCsam−1 × h−1), were calculated according to the formulas:

where %at 13Cinc is atomic percentage of incorporated 13C; DWsam is the dry weight of the sample, %Cdw is the carbon content expressed as a percentage of dry weight, DWsam.%Cdw is the weight of carbon in the sample, W13Cinc is the weight of 13C incorporated into the sample during the experiment, MMC is the molar mass of carbon, t is the duration of the experiment, and Cinc is the carbon incorporation rate expressed in moles of incorporated carbon per time unit and weight unit of total carbon in the sample.

Measurement of incorporated radioactivity

The beta-radioactivity of the samples was measured for 15 min in liquid phase in a Beckman LS 6500 Scintillation Counter (effective resolution: 0.06 keV).

Moles of incorporated radioelement are known from the measured number of disintegrations per second (disintegrations per second × 2 half-life in seconds/Avogadro’s number). The total incorporation rate, expressed in moles of incorporated molecules per time unit and weight unit of total carbon in the sample (μmolRinc × gCsam−1 × h−1), was calculated according to the formula:

where xEinc is the number of moles of incorporated radioelement, f is the ratio of radioactive to cold molecules in the incubation medium, FWsam is the fresh weight of the sample, p is the dry weight/fresh weight ratio measured on control shrimps, %Cdw is the carbon content expressed as a percentage of sample dry weight, and Rinc is the total incorporation rate expressed in moles of incorporated molecules per time unit and weight of total carbon in the sample.

Autoradiography

Histological localisation of radiotracers was done by autoradiography on frozen shrimps incubated with a radiotracer. To isolate shrimp segments, each frozen shrimp was embedded in Tissue-Tek gel (Sakura 4583; Torrance, CA, USA), directly frozen in liquid nitrogen, and cut into segments 1 cm long with a hacksaw in a cryostat at −20 °C. Segments of the cephalothorax (including MP and gills) and the abdomen were directly thawed overnight in fixative solution (2.5% glutaraldehyde), decalcified in EDTA (0.2 M, pH 8), post-fixed in 1% OsO4 for 2 h, rinsed in distilled water (4 × 10 min), dehydrated with ethanol, and embedded in Steedman’s wax (polyethylene glycol distearate; Sigma 305413—hexadecanol 9:1; 99%; Sigma 258741). Twelve-micrometre cross-sections were obtained with a microtome (Reichert, Austria) at three levels (MP, gills and abdominal muscles) and deposited on gelatinised glass slides. They were rehydrated, washed in distilled water and then coated with a photographic emulsion film (Hypercoat LM-1; GE Healthcare RPN40) in the dark according to the classical wet method (Ullberg et al., 1982). The slides were developed with KODAK products (KODAK D19-Revelator and KODAK-Fixator; Rochester, NY, USA) after an exposure time of 6–9 months for 14C-bicarbonate or 6 h to 3 months for 14C-acetate and 3H-lysine. The slides were mounted in epoxy resin before being observed and photographed with an Olympus AX70 light microscope (Shinjuku, Tokyo, Japan) equipped with a Visicam 5.0 camera (Hilden, Germany).

Statistical analysis

All data are reported as mean values±s.d. Outliers (possibly due to experimental contamination or to the absence of labelling) were eliminated with Dixon’s Q test (P<0.05). One-way analysis of variance (Kruskall–Wallis) was used and all pairwise multiple comparison procedures (Dunn’s Method) performed with Sigma Plot 11.0 software (Systat Software Inc., San Jose, CA, USA) were used for sample comparisons. P<0.05 was taken as fiducial limit for statistical significance.

Dunn’s multiple comparisons vs control group method were used to test the real incorporation of 14C.

Results

13C-bicarbonate experiments

The data for 13C-bicarbonate incorporation are shown in Figure 2a. Bacterial mats of branchiostegites and MP displayed a high rate of inorganic carbon incorporation from 13C-bicarbonate. Other body parts (OB, DT and Mu) showed lower rates. All the results appeared to be statistically significant.

Figure 2
figure 2

Rates of inorganic carbon incorporation into the bacterial mats of the mouthparts and branchiostegite and into shrimp tissues, as calculated from 13C fixation (a) or 14C fixation (b) data. The shrimps were incubated: (a) for 4 or 10 h in NaH13CO3 in pure seawater (dark grey), supplemented with reduced iron (light grey) or thiosulphate (white); (b) for 6 h in pure SW with NaH14CO3 (dark grey), supplemented with iron (light grey) or thiosulphate (white). The radioactivity measured in control shrimp tissues is shown (black) for comparison. Stars indicate significant differences between groups of tissues; ( × 10, × 100) indicates tissues for which the values were multiplied by 10 or 100 for graph presentation; ‘N’ indicates the number of samples for each data bar.

Full size image

After a 10-h incubation, the 13C uptake measured in the bacterial mats was approximately twice as high as after 4 h (Table 1). In the BB, it was higher in thiosulphate-incubated shrimps (78.8 μmolCinc × gCsam−1) than in SW-incubated specimens (52.4 μmolCinc × gCsam−1).

Table 1 Total amounts of inorganic carbon (μmolCinc × gCsam−1±s.d.) incorporated after 4 and 10 h into the body parts of shrimps incubated with 13C-bicarbonate in the presence of seawater, or seawater supplemented with iron or thiosulphate
Full size table

The incorporation rate was higher in the MP (12.1–16.7 μmolCinc × gCsam−1 × h−1) than in the BB (3.9–7.3 μmolCinc × gCsam−1 × h−1). In shrimp tissues, it was higher in the OB (1.1 μmolCinc × gCsam−1 × h−1, in the presence of thiosulphate) than in the DT (0.3 μmolCinc × gCsam−1 × h−1) or the Mu (0.2 μmolCinc × gCsam−1 × h−1). The incorporation rates for MP and BB were significantly different from those of shrimp tissues (P<0.05).

14C-bicarbonate experiments

The radioactivity levels measured in the bacteria and tissues of control shrimps appeared negligible as compared with radiotracer uptake (P⩽0.001) (Figure 2b).

High 14C incorporation rates were observed in the bacterial mats of the MP and BB (Figure 2b), of the same order of magnitude as the corresponding 13C-bicarbonate incorporation rates. The rate was much higher in the presence of iron or thiosulphate. In the MP and BB, respectively, it reached 10.4 and 5.4 μmolCinc × gCsam−1 × h−1 in the presence of thiosulphate and only 3.9 and 2.8 μmolCinc × gCsam−1 × h−1 for shrimps incubated in SW alone.

Most of the shrimp tissues analysed showed significant 14C incorporation, whatever the incubation conditions (Figure 2b). The highest carbon incorporation rates were recorded in the tegumental tissues of the gill chambers, that is, the OB (2.7 μmolCinc × gCsam−1 × h−1) and the Gi (0.2 μmolCinc × gCsam−1 × h−1). The rates recorded for the internal tissues of all shrimps (Mu: 0.01; DT: 0.04; HP: 0.01 μmolCinc × gCsam−1 × h−1) were low to very low, but nevertheless significant (P⩽0.001), the digestive gland displaying the lowest values. The AC showed a surprisingly high incorporation rate (0.6 μmolCinc × gCsam−1 × h−1 in the presence of iron), to be explained below.

14C-acetate experiments

The results for 14C incorporation from acetate are shown in Figure 3a. Bacterial mats (MP and BB) showed high rates, indicating that acetate is a suitable carbon source for these bacteria. The rate recorded in SW (61.3 μmolAcinc × gCsam−1 × h−1) was 15 times the rate of carbon incorporation from 14C-bicarbonate (3.9 μmolCinc × gCsam−1 × h−1 in MP).

Figure 3
figure 3

Rates of incorporation of 14C (a) and 3H (b) into the bacterial mats and inner tissues of R. exoculata shrimps incubated for 1 h with 14C-acetate or 3H-lysine in pure SW. Stars indicate significantly different groups of tissues; ( × 10, × 100, × 1000) indicates tissues for which the values were multiplied by 10, 100 or 1000 for graph presentation; ‘N’ indicates the number of samples for each data bar. U, uropods.

Full size image

High rates of 14C incorporation were measured in the integument, especially in areas lining the gill chamber, that is, OB and Gi (7.9 and 7.3 μmolAcinc × gCsam−1 × h−1, respectively). The level remained low but measurable in the internal tissues. In the Mu and HP, the rate did not exceed 0.1 μmolAcinc × gCsam−1 × h−1. It was slightly higher in the DT (0.4 μmolAcinc × gCsam−1 × h−1). In the AC, a distinction was made between the old, outer cuticle (ACo) and the inner tissues (ACn). The 14C incorporation rate for the whole AC was mainly due to high 14C uptake into the inner, living tissues (ACn, 4.8 μmolAcinc × gCsam−1 × h−1), that is, the epidermis secreting the new cuticle, while the inert old cuticle showed low labelling (ACo, 0.7 μmolAcinc × gCsam−1 × h−1).

3 H-lysine experiments

3H-lysine incorporation rates were about 100 times lower than the 14C-acetate incorporation rates (0.7 μmolLysinc × gCsam−1 × h−1 vs 61.7 μmolAcinc × gCsam−1 × h−1 in the BB) (Figure 3b). The highest rates were again found in the bacterial mats (MP and BB), suggesting that these bacteria can use lysine, probably for protein synthesis. Significant rates were measured in the gill chamber integument (OB and Gi) (34.1 and 16.1 nmolLysinc × gCsam−1 × h−1, respectively), while the internal tissues (except the Mu: 0.8 nmolLysinc × gCsam−1 × h−1) showed low labelling.

Autoradiography

Our autoradiographic observations were in good agreement with the analytical data, confirming the distribution of the radiolabelled molecules among the bacterial mats and shrimp tissues (see Supplementary Information 2—Figures 7a and b for negative controls).

In the bacterial mats associated with the MP and branchiostegites, results for the three radiotracer incubations were quite similar. The labelling intensity depended on the specific radioactivity of the isotope, the exposure time and the amount of radiotracer incorporated. 3H-lysine gave the best result with the shortest exposure time, followed by 14C-acetate and 14C-bicarbonate. Silver grain precipitation (in the photographic emulsion) always appeared first along the bacterial filaments (Figures 4a and c), eventually covering the whole length of the filaments as a dense black deposit (Figures 4b and d). By comparing the pictures obtained after short and long exposure times we were able to distinguish the labelled filamentous bacteria from the mineral crust of iron oxides: the latter also appeared as dark masses, but they were not labelled, whereas the bacteria located inside or below the mineral crust were moderately labelled (Figure 5a).

Figure 4
figure 4

Autoradiographs of histological sections: cross-sectioned bacteriophore setae of the mouthparts of R. exoculata shrimps incubated for 1 h with 3H-lysine (a, b) or for 6 h with 14C-bicarbonate (c, d). (a) Filamentous bacteria are distinguished from mineral deposits and labelled on short segments only. (b) The filamentous bacteria are labelled over their entire length. (c) Filamentous bacteria are poorly or not labelled. (d) Discontinuous labelling appears on the distal part of the filamentous bacteria but does not stain them close to the setae. Exposure times: (a) 12 h, (b) 11 days, (c) 10 weeks and (d) 34 weeks. Scale bars: 100 μm. FB, filamentous bacteria; Mi, mineral.

Full size image
Figure 5
figure 5

Autoradiographs of different cross-sectioned body parts of R. exoculata shrimps incubated for 1 h with 14C-acetate (a, c) or 3H-lysine (b): (a) branchiostegite (3-day exposure) showing intense labelling of emerging filamentous bacteria and much weaker labelling below the mineral crust, a dense silver deposit covering the living tissue, that is, both the inner and the outer epidermis and their respective new cuticles, while the old outer cuticle remains almost free of silver particles; (b) gills (20-days exposure); (c) abdominal muscles (29-days exposure) labelled moderately but more intensely than the background. Scale bar: 100 μm. AS, sectioning artificial space; FB, filamentous bacteria; Mi, mineral crust; Mu, muscles; OB, outer branchiostegite (old cuticle separated from the new one); RB, rod-shaped bacteria; T, tissues including the epidermis bordered by the new cuticle being deposited.

Full size image

The shrimp tegumental tissues appeared densely labelled after incubation with 3H-lysine or 14C-acetate, but sections of shrimps incubated with 14C-bicarbonate required very long exposure times before the label appeared. After 20–30 days of exposure, gills and muscles also appeared moderately labelled (Figures 5b and c), but neither the DT nor the HP showed any measurable labelling. Detailed views of the branchiostegite (Figure 5a) confirmed and explained the analytical data obtained for the abdominal integument: living tissues synthesising new cuticle and the new cuticle itself showed the most intense labelling. In contrast, the overlying old cuticle remained free of silver grains except in its lower part, which was probably infused with moulting fluid enzymes secreted by the epidermis (Compère et al., 2004).

Discussion

The in-vivo experiments presented here provide direct evidence that R. exoculata epibionts are autotrophic. These are the first experiments to have been carried out on live shrimps together with their epibionts, and the first to demonstrate a mutualistic relationship in this bacteria–shrimp symbiosis under in-situ pressure conditions.

Carbon fixation by bacterial chemosynthesis

Once thought to be a monoculture of a single, pleomorphic phylotype of epsilonproteobacteria, the gill chamber epibiont population of R. exoculata is now regarded as a functional consortium or syntrophic community (Stams and Plugge, 2009) dominated by gamma- and epsilonproteobacteria. By combining 16S rRNA and functional gene analysis with in-situ hybridisation and transmission electron microscopy observations on R. exoculata epibionts and their free-living or symbiotic relatives from vent habitats, investigators (Campbell et al., 2006; Zbinden et al., 2008; Petersen et al., 2009; Goffredi, 2010; Hügler et al., 2010, 2011; Guri et al., 2012) have evidenced epsilonproteobacteria of Marine Groups 1 (the new family Thiovulgaceae) and 2, as thick (3-μm wide) and thin (1-μm wide) filaments, respectively. Both should oxidise sulphur compounds through the Sox pathway and fix carbon through the reverse tricarboxylic acid cycle (SoxB and aclA gene sequences). Authors have also identified gamma-proteobacterial epibionts of the Thiotrichaceae family (Leucothrix group), together with some free-living members and some other invertebrate ectosymbionts, mainly of crustaceans (Goffredi, 2010). These gammaproteobacteria appearing at least as filamentous and rod-shaped morphotypes can probably grow as sulphide oxidisers, most likely using the APS pathway for energy generation and the Calvin-Benson-Bassham cycle for carbon fixation (aprA and cbbM gene sequences, respectively) (Zbinden et al., 2008; Hügler et al., 2010, 2011; Guri et al., 2012). In addition, coccoid-shaped cells with intracytoplasmic membrane stacks have been identified as type-I methanotrophic gammaproteobacteria (pmoA gene sequence) (Corbari et al., 2008a; Zbinden et al., 2008; Guri et al., 2012), and deltaproteobacteria related to Desulfocapsa have been identified thanks to their aprA and hydrogenase gene sequences (Hügler et al., 2011), suggesting they can grow lithotrophically using hydrogen as electron donor for sulphate reduction. On this basis, the authors hypothesise that an internal sulphur cycle between sulphur-oxidising epsilon- and gammaproteobacteria and sulphate-reducing deltaproteobacteria could take place in the shrimp gill chamber, as proposed for oligochete symbiosis (Dubilier et al., 2001).

The following results of the present study support the view that shrimp-associated bacteria carry out chemosynthetic inorganic carbon fixation: (1) the carbon incorporation rates for 13C- and 14C-bicarbonate are in good agreement, (2) they increase in the presence of electron donors (S2O32−, Fe2+), (3) they are stable over time, as confirmed by the time-related increase in 13C incorporation and (4) autoradiographs of histological sections show labelling of bacterial mats.

The carbon fixation rates observed here are similar to but somewhat higher than those observed by Polz et al. (1998): 3.9 vs 2 μmolCinc × gCsam−1 × h−1 for the MP of shrimp incubated in SW. This comparison is somewhat uncertain, as we had to convert Polz’s data, expressed per weight unit of protein, into values given per weight unit of total carbon, the carbon content of proteins being estimated at 45% (Rouwenhorst et al., 1991). Yet, the consistently higher rates found here might be due to the incubation of live shrimps under in-situ pressure (230 bars in this study), as opposed to the use, in previous experiments, of pieces of dead shrimp under atmospheric pressure. Active chemoautotrophy might be promoted by preserving shrimps and bacteria in good physiological condition (Ravaux et al., 2003). Accordingly, Zbinden et al. (2008) observed <30% degraded epibionts in re-pressurised shrimps. The carbon fixation rates determined here are much lower, however, than those obtained for other chemosynthetic symbionts (bacterial homogenates or endobacteria of bivalve gills), for example, those of Bathymodiolus thermophilus: 150 μmolCinc × gCsam−1 × h−1 in SW (Nelson et al., 1995), or Solemya velum: 2500 μmolCinc × gCsam−1 × h−1 in the presence of thiosulphate (Scott and Cavanaugh, 2007). Moreover, owing to the relatively low energy level of thiosulphate as compared with sulphide, the latter authors observed threefold higher carbon fixation rates with 0.2 mM Na2S (70 μmolC × gprot−1 × min−1) than with 1 mM Na2S2O3 (20 μmolC × gprot−1 × min−1).

That the presence of an electron donor increases the carbon fixation rate indicates that Na2S2O3, and Fe2+ at least contribute to fuelling R. exoculata epibionts, and our autoradiography data suggest that chemosynthesis occurs at least in the large filamentous bacteria (epsilonproteobacteria of Marine Group 1), which appeared more intensely labelled than the thin filamentous and rod-shaped ones (epsilonproteobacteria of Marine Group 2 and gammaproteobacteria). The view that R. exoculata epibionts have sulphide-oxidising activity is consistent, moreover, with the fact epsilon- and gammaproteobacteria are assumed to have such activity, on the basis of the group affiliation of the former (Petersen et al., 2009; Goffredi, 2010; Guri et al., 2012) and because the latter possess functional SoxB and aprA genes (Hügler et al., 2010, 2011). We cannot say, however, what the epibiotic bacteria of R. exoculata use as predominant energy source. Methane and hydrogen, found in hydrothermal fluids from the Rainbow vent field along with sulphide and iron (Charlou et al., 2002; Douville et al., 2002; Schmidt et al., 2008), are also potential drivers of chemoautotrophy. In particular, the type-I methanotrophic gammaproteobacteria may use methane as both electron and carbon source (Zbinden et al., 2008; Guri et al., 2012), deltaproteobacteria are believed to use H2 for sulphate reduction (Hügler et al., 2011), and epsilonproteobacteria (of Marine Groups I and II) may use H2 instead of sulphide for chemosynthesis, as proposed for Bathymodiolus gill endosymbionts (Petersen et al., 2011). The possibility that R. exoculata epibionts might use H2 is strongly supported by the identification of a hupL gene sequence from Group-2 epsilonproteobacteria (Hügler et al., 2011) among the R. exoculata epibionts and by the metabolic capabilities reported for Campylobacterales, especially of the family Thiovulgaceae (Campbell et al., 2006; Goffredi, 2010). We have also tested methane and hydrogen (see Supplementary Information 1), but we observed rapid death of the incubated shrimps, so they were discarded. Furthermore, whether CH4- or H2-metabolic genes are actively expressed or not has not been determined. Biotic iron oxidation by R. exoculata epibionts has been proposed (Zbinden et al., 2004, 2008; Corbari et al., 2008a) on the basis of the close association of these bacteria with heavy iron-oxyhydroxide deposits. Our results support this hypothesis, but evidence is lacking as to the ability of the bacterial phylotypes identified so far to oxidise iron. Furthermore, abiotic iron-oxide precipitation can compete strongly with biotic iron oxidation (Schmidt et al., 2009), bacterial cell walls can induce it (Corbari et al., 2008a).

We also show here that R. exoculata epibionts can incorporate 14C-acetate and 3H-lysine much faster than they incorporate 14C from bicarbonate. In the branchiostegite bacterial mats, the rate of organic carbon incorporation from acetate (61.7 μmolAcinc × gCsam−1 × h−1) was 10–20 times the rate observed with 14C-bicarbonate, depending on whether thiosulphate was present (5.4 μmolCinc × gCsam−1 × h−1) or not (2.8 μmolCinc × gCsam−1 × h−1). This may be due to the presence of heterotrophic bacteria, but to date, this has not been investigated. Our observation that the thick bacterial filaments were intensely labelled with all three radiotracers, and that the thin ones were also labelled to a lesser extent, suggests that at least the epsilonproteobacteria might switch to a mixotrophic metabolism and use small organic molecules, when available, as energy and/or carbon sources. Acetate and amino-acid assimilation is thus unsurprising and might reflect a capacity of certain chemolithoautotrophic bacterial epibionts to adjust their metabolism to changes in their geochemical environment. A versatile (autotrophic/mixotrophic/heterotrophic) energy metabolism using several electron donors (including hydrogen and organic compounds) and acceptors (including oxygen and nitrate) has indeed been reported for free and symbiotic epsilonproteobacteria of the Thiovulgaceae family (Campbell et al., 2006; Goffredi, 2010) and suggested for R. exoculata ectosymbionts (Hügler et al., 2011). Mixotrophy/heterotrophy is also commonly accepted as a possible feature of gammaproteobacteria of the Thiotrichaceae family (Goffredi, 2010), as most Beggiatoa species can grow on acetate as sole added carbon source (Larkin and Strohl, 1983). It has been proposed but not investigated for R. exoculata epibionts (Hügler et al., 2011). In support of this hypothesis, a metagenomic study has revealed in Riftia pachyptila endosymbionts the presence of all genes required for heterotrophic metabolism (Robidart et al., 2008), and a metaproteomic study has pinpointed acetate as a substrate for bacterial endosymbionts of the gutless marine worm Olavius algarvensis (Kleiner et al., 2012). Schulz and Jørgensen (2001) interpret the metabolic flexibility of filamentous sulphur-oxidising bacteria, especially those of the Thiothrix group, and their ability to grow heterotrophically or autotrophically on different sulphur species, as a need to overcome disadvantages of their sessile life and to adapt their activity to changes in substrate availability (oxygen, sulphide). In R. exoculata, the obvious variability of bacterial assemblages according to the gill chamber area, the individual and the vent site (Zbinden et al., 2008; Petersen et al., 2009; Hügler et al., 2011) suggests a consortium composition plasticity as a response to varying vent fluid composition (Charlou et al., 2002; Schmidt et al., 2008) and to fluctuating geochemical or micro-environmental conditions in the zone of mixing with deep water.

The relatively high carbon fixation rate observed in SW alone suggests that R. exoculata epibionts have an internal energy store that supplies autotrophic metabolisms in the absence of an external electron donor, for example, when the shrimps swim from reduced-compound-rich hydrothermal fluid to oxic (reduced-compound-poor) deep water. Such carbon fixation has previously been evidenced in Bathymodiolus thermophilius and R. exoculata symbionts (Nelson et al., 1995; Polz et al., 1998). Energy could be stored as iron polyphosphate and sulphur globules like those observed ultrastructurally by Wirsen et al. (1993) and Zbinden et al. (2008) in thin filamentous bacteria. These could be identified as Leucothrix-like gamma-ectosymbionts because their APS pathway allows production of intermediate sulphur compounds and because the storage role of similar globules is well documented in their shallow-water relatives (Larkin and Strohl, 1983; Schulz and Jørgensen, 2001). Furthermore, these globules were shown to be empty in specimens maintained in a pressurised aquarium without electron donors (Zbinden et al., 2008). This suggests that the enhanced carbon fixation observed in the presence of thiosulphate as compared with pure SW might reflect a switch of gamma- and/or epsilon-epibionts to an external energy source supporting autotrophic metabolism.

Bacteria–host transfers

Several authors propose that organic nutrients might be transferred from epibionts to R. exoculata (Casanova et al., 1993; Zbinden et al., 2004; Corbari et al., 2008b). Others, arguing that the cuticle is impermeable, reject this hypothesis (Segonzac et al., 1993; Gebruk et al., 2000). The latter authors refer to an experiment (Anderson and Stephens, 1969) or to subsequent review papers (Stephens, 1988; Preston, 1993; Gebruk et al., 2000) suggesting that crustaceans, in contrast to many other marine invertebrates, cannot take up dissolved organic matter or dissolved organic carbon. According to Anderson and Stephens (1969), the apparent incorporation of radiotracers by crustaceans was due to uptake by epibionts alone and not by the crustaceans themselves. Yet these authors worked with whole organisms, and never separated the internal tissues from the integument and the epibionts as we have done here.

The present results demonstrate the transfer of carbon fixed by bacteria to the host shrimp tissues and the ability of shrimps to take up dissolved organic molecules such as acetate or lysine across their integument. Together with our autoradiographs, the higher radiotracer concentrations in the shrimp tissues than in the incubation media (Table 2) and the incorporation of 3H-lysine and 14C-acetate into the Mu (proteins) and the HP (storage lipids) confirm their active uptake and their accumulation in shrimp insoluble organic matter. Our separate analysis of the old and new cuticles of the abdominal integument and the corresponding autoradiographs also suggest that the new cuticle and the secreting epithelium incorporate more labelled molecules, probably because of the intense biosynthetic activity at this level in pre-ecdysial crustaceans.

Table 2 Initial and final concentrations (in μM) of labelled compounds (13C, 14C-bicarbonate, 14C-acetate, 3H-lysine) in the incubation media and the shrimp body parts
Full size table

These results thus constitute in-vivo experimental evidence of nutritional transfer of bacteria-fixed carbon to shrimps. They are in good agreement with previous stable isotope studies (Gebruk et al., 1993; Pond et al., 1997a, 1997b, 2000; Rieley et al., 1999; Colaço et al., 2007) and confirm that R. exoculata depends on its epibionts for its nutrition. They thus strengthen the view that the relationship between bacteria and shrimp is mutualistic: the epibionts supply carbon compounds to the shrimp and the shrimp, thanks to its gill chamber flow and its swimming behaviour, offers them protection and a supply of chemical compounds, maintaining their position at the oxic/anoxic interface around active chimneys (Casanova et al., 1993; Segonzac et al., 1993). The R. exoculata bacterial epibiosis can thus be regarded as a true mutualistic trophic ectosymbiosis, similar to what is known about the chemosynthetic endosymbiosis of organisms such as R. pachyptila (Felbeck, 1981; Fisher et al., 1989), Calyptogena magnifica (Childress et al., 1991) and S. velum (Stewart and Cavanaugh, 2006; Scott and Cavanaugh, 2007).

Finally, the view that bacterial products are assimilated across the shrimp integument rather than via the DT is strongly supported by: (1) the significant incorporation of both 14C-acetate and 3H-lysine after an incubation time as short as 1 h. This is insufficient for uptake by ingestion assimilation, which commonly takes several hours (ingestion assimilation of carbon from bicarbonate, for example, requires prior incorporation by the bacteria, followed by grazing of bacteria from the MP) (Chipps, 1998; Hoyt et al., 2000); (2) the high incorporation levels recorded in the gill chamber integument lining (OB, Gi), as opposed to (3) the much lower levels recorded in the DT (Figure 6). These results are in agreement with the view that the shrimps do not graze on their epibionts (Zbinden et al., 2004; Corbari et al., 2008b), but rather farm them (Segonzac et al., 1993; Polz and Cavanaugh, 1996), although they might also ingest some chimney bacteria along with sulphides (Van Dover et al., 1988; Segonzac, 1992). The low but nevertheless significant incorporation of 14C from bicarbonate in the DT corroborates the data of Polz et al. (1998), although they found higher rates than ours (upon incubating isolated gut segments). This incorporation might be due to chemosynthetic activity of the midgut microbial community (Polz et al., 1998; Zbinden and Cambon-Bonavita, 2003; Durand et al., 2010), but this community would be less active than the gill chamber community.

Figure 6
figure 6

Schematic view of R. exoculata (a) and its gill chamber—transversal view (b) showing the level of 14C-tracers incorporation in the internal organs by colour range from the highest levels in red to the lowest in yellow. U, uropods.

Full size image

The view that soluble carbon-containing compounds are transferred from bacteria to shrimp by crossing the gill chamber integument tallies with the ultrastructural characteristics of this integument, which resemble those of permeable and/or transporting tissues (Martinez et al., 2005): a very thin cuticle (30 μm), reduced blood-water distance, mitochondria-rich cells with apical infolding in the branchiostegite epithelium. In R. exoculata, the uptake of bacterial nutrients across the integument could be the major source of nourishment, along with complementary sources such as grazing on chimney minerals (Van Dover et al., 1988; Segonzac et al., 1993), ingestion of bacteria-colonised exuviae, and midgut microflora activities (Zbinden and Cambon-Bonavita, 2003; Zbinden et al., 2004; Corbari et al., 2008b; Durand et al., 2010). It could be regarded as a peculiar adaptation of the shrimp. Although some authors have discarded this hypothesis, the present results may bring it back into favour. By extension, they suggest that this process might exist in many crustacean species, which would derive their nourishment at least partly from dissolved organic matter or dissolved organic carbon, as is commonly accepted for many marine invertebrates: annelids (Richards and Arme, 1982; Ahearn et al., 2000; Peppler and Ahearn, 2003), molluscs (Wright and Secomb, 1986; Wright, 1988; Wright and Pajor, 1989; De Eguileor et al., 2000) and cuticle-lined parasitic organisms such as nematodes (Fleming and Fetterer, 1984; Gordon and Burford, 1984; Masood, 1984). Arbitrarily rejected for crustaceans (Stephens, 1988) and especially decapods, this idea is surprisingly accepted without demonstration for parasitic rhizocephalan barnacles such as Sacculina carcini (Bresciani and Høeg, 2001) and for parasitic copepods (Kannupandi, 1976; Poquet et al., 1994) with a modified cuticle. It is well known, furthermore, that in pre-ecdysis, crustaceans (and other arthropods) absorb soluble molecules (for example, sugars and amino acids) from their degrading old cuticle through the new cuticle and re-utilise them for its synthesis (Compère et al., 2004). These conflicting views regarding the capacity of marine crustaceans to assimilate dissolved organic carbon will be tested in further studies.