The phylogenetic diversity of the gut microbial population of the hydrothermal shrimp Rimicaris exoculata was determined. The presence of micro-organisms in the hindgut of the shrimp was determined, and their morphology illustrated for the first time by transmission electron microscopy. DNA was extracted from the fore-, mid- and hindgut of shrimps collected at the Rainbow site, at the Mid-Atlantic Ridge, and the sequences of the small-subunit rRNA (16S rDNA) were analyzed. Seven different bacterial phylotypes were identified from the 90 clones screened. The sequences were affiliated to three major groups: ε-Proteobacteria (48,8%), Entomoplasmatales (23.3%) and representatives of the new phylum Deferribacteres (26.6%). These results show that the diversity in the shrimp gut is very low compared to that of the surrounding medium. Furthermore, the presence of groups that are not found in the external medium (Spiroplasma sp. and Geovibrio sp.) suggests the existence of a local microflora. The potential roles of these bacteria are discussed, involving the shrimp diet and metal bioremediation.
Indigenous gut microflora
Caridean shrimps belonging to the family Alvinocarididae are the dominant megafauna at the Mid-Atlantic Ridge (MAR) vent sites . Rimicaris exoculata is particularly abundant, forming dense, motile swarms around the chimney walls [3,4]. A striking characteristic of these shrimps is that they have an unusually enlarged gill chamber, housing epibiotic bacteria which are most abundant on mouthparts, on bacteriophore structures and on the inner surface of the gill chamber . Although three different morphotypes were observed for these bacteria (two filamentous and one rod-shaped ), they consist entirely of one single bacterial phylotype, forming a distinct branch within the ε-Proteobacteria.
As for the East Pacific Rise polychaete Alvinella pompejana, which exhibits similar (i.e. filamentous ε-Proteobacteria) bacterial epibiosis , a trophic role has been discussed extensively for the shrimp epibiotic bacteria, since bacteria have also been found in the gut [5,9]. The observation of large quantities of sulfide particles in the stomach and gut contents of R. exoculata first led Van Dover and collaborators  to propose that the primary food source of the shrimp consisted of bacteria living on the surface of the chimney, with a secondary contribution of the epibiotic bacteria to the diet. An alternative hypothesis proposed that the diet of R. exoculata might rely totally on its epibiotic bacteria [4,10–12]. A third hypothesis has emerged from different studies [13,14]: a local, highly active autotrophic bacterial population in the shrimp's gut might serve as a source of nutrition.
However, knowledge of the bacterial microflora associated with the gut is still poor. Polz et al.  have reported that about 20% of the gut bacteria (cardiac stomach and combined mid- and hindgut) belong to the phylotype found in the gill chamber epibionts. The aim of this study was to determine the shrimp gut bacterial diversity, based on the 16S rRNA distribution, in order to assess the 80% of unknown microflora.
2 Materials and methods
Specimens of R. exoculata were obtained during the French ATOS cruise (June 2001) on the Rainbow vent site (36°14.0′ N, MAR, 2300 m depth). Animals were collected with the slurp-gun of the ROV Victor, operated from the RV Atalante. The bowls of the suction sampler were aseptically washed with ethanol (95%) and then filled with distilled water before dives. Immediately after recovery, live specimens were dissected under sterile conditions to sample stomachs and digestive tracts. These tissue samples were either fixed in 3% glutaraldehyde-sodium cacodylate buffer solution, and later post-fixed in osmium tetroxide for microscopy, or frozen for DNA extraction.
2.2 Light and transmission electron microscopy (TEM)
For TEM observations, fixed samples were dehydrated in ethanol and propylene oxide series, and further embedded in epoxy resin (Serlabo). Semi-thin and thin sections were obtained from a Reichert-Jung Ultramicrotome (Ultracut E). Semi-thin sections were stained with Toluidine blue for observations by light microscopy (using a Nikon Optiphot-pol microscope). Thin sections were laid on 200-mesh copper grids and stained with uranyl acetate and lead citrate. Observations were carried out on a Philips 201 TEM, operating at 80 kV.
2.3 DNA extraction
Samples of the digestive tract (fore-, mid- and hindgut see Fig. 1) of two specimens were centrifuged (4000×g, 40 min, 4°C) and resuspended in TE-Na-1X lysis buffer (100 mM Tris–HCl, 100 mM NaCl, 0.5 mM EDTA, pH 8). Five hundred μl of 10% lauryl sarkosyl and 500 μl of 10% sodium dodecyl sulfate were added after homogenization. The samples were treated with proteinase K (20 mg ml−1, Eurobio, France) at 40°C for 3 h. Digested samples were centrifuged for 5 min (6000×g) and supernatants were extracted three times with cold phenol-chloroform-isoamyl alcohol (24:24:1 vol/vol/vol) and once with chloroform (mixed in a 1:1 ratio with lysates). Nucleic acids corresponding to the microbial community were obtained by overnight precipitation in −20°C absolute ethanol (2:1 ratio with lysates) and centrifugation for 1.5 h (12 000×g). They were dried, resuspended in 500 μl TE-1X (10 mM Tris–HCl, 2 mM EDTA, pH 7.5) and absorbance was determined at 230, 260, 280 and 320 nm, on a GenQuantII spectrophotometer (Pharmacia Biotech Inc., Uppsala, Sweden) to check the purity of the nucleic acids and to determine their concentration.
Morphology of R. exoculata and anatomy of its digestive tract. a: On board just after recovery. b: Schematic drawing of the shrimp (adapted from Williams and Rona ) on which the digestive tract is represented. The complete tract was sampled for the analyses, i.e. the foregut, midgut and hindgut.
2.4 Polymerase chain reaction (PCR) and cloning
Clone libraries were produced from community 16S rDNA by amplifying approximately 100 ng of bulk DNA in 50 μl reaction mixtures containing (final concentrations) 1×PCR buffer (QBIOgene, France), 200 μM of each deoxyribonucleoside triphosphate, 100 μM of each Bacteria specific forward and reverse primer, and 0.5 μl of Taq polymerase (QBIOgene). Reaction mixtures were incubated in a Robocycler Gradient 96 (Stratagene, La Jolla, CA, USA). For Bacteria, the cycling program was as follows: 94°C for 3 min, 30 cycles of 94°C for 1 min, 48°C for 1.5 min, 72°C for 2 min and a final extension period of 6 min at 72°C. The forward primer specific for the bacterial 16S rDNA was 27F (5′-AGAGTTTGATCATGGCTCAG-3′) and the reverse primer was the universal 1492R (5′-GTTACCTTGTTACGACTT-3′). The PCR products were cloned directly to separate them from each other, with the TA cloning method, using the pCR2.1 vector according to the manufacturer's instructions (Invitrogen Corp., San Diego, CA, USA). Clone libraries were constructed by transforming E. coli TOP10F’.
2.5 Screening of rDNA clones by amplified ribosomal DNA restriction analysis (ARDRA)
Each recombinant colony was suspended in 100 μl of sterile deionized water. Ribosomal DNA inserts from recombinant clones were amplified by PCR from 5 μl of colony suspension, using the reaction mixtures described above. The PCR program was the same as that for the initial amplification of 16S rDNA (see above), except that the initial step lasted 10 min (for cell lysis). Three to 5 μl of the PCR products was digested with 3 U of the 4-base-specific restriction endonuclease HhaI (New England Biolabs) in 1×NEB buffer 4 supplemented with 100 μg ml−1 of bovine serum albumin, in a final volume of 20 μl for 3 h at 37°C. The digested fragments were separated by electrophoresis in a 3% agarose gel (Sigma) stained with ethidium bromide and the bands were visualized by UV transillumination (Fluor-S-multi-imager, Biorad, Hercules, CA, USA). ARDRA patterns for each library were normalized and grouped with ‘Molecular Analyst’ software (Biorad). One gene representative of each pattern was selected for sequencing. Where patterns indicated dominance (more than 12 clones per pattern), three clones were partially sequenced to check for the homogeneity of the group. ARDRA screening was stopped when no new pattern was observed after 30 successive clones.
2.6 Sequencing of rDNA clones
Representative rDNA inserts were sequenced by Genome Express (Grenoble, France) with automatic sequencers (Perkin Elmer, Sanger method ). Two sequence reactions per insert with specific primers were required to determine the bacterial 16S rDNA sequence (about 1300 bp). The primers used for sequencing included the universal primer 27F (see above), and an internal bacterial primer EUBINT (5′-GCGCCAGCAGCCGCGGTAA-3′) (position 520–539 on E. coli 16S rDNA sequence). Sequences were then assembled using DNASTAR software (Madison, WI, USA).
2.7 Phylogenetic analysis and chimera detection
Sequences were compared with those available in databases using the BLAST (Basic Local Alignment Search Tool) network service , in order to determine approximate phylogenetic affiliations. Chimeric sequences were detected by using the BLAST program (with the 5′ or 3′ partial sequences) and confirmed by phylogenetic reconstruction on partial sequences (unstable sequences were removed from the study). Sequences were compiled in MEGALIGN (DNASTAR) and aligned with the rDNA sequences from the Ribosomal Database Project, using the CLUSTALW program . Alignments were refined manually, using the SEAVIEW program . All the trees presented were constructed by using the PHYLO-WIN program , on the basis of evolutionary distance neighbor-joining  and maximum likelihood methods . The robustness of inferred topologies was tested by the bootstrap resampling  of trees calculated on the basis of evolutionary distance (neighbor-joining algorithm with Kimura two-parameters correction ) and maximum likelihood. If ARDRA patterns were not identical, while related sequences displayed over 98% sequence similarity, then only one of the sequences was retained for phylogenetic analysis.
2.8 Nucleotide sequence accession numbers
The EMBL accession numbers of the 13 sequences used in this study are AJ515712 to AJ515724.
3.1 Light and TEM
Semi-thin sections of the stomach and the gut were observed. Both of these parts of the digestive tract were full of black and brown (respectively iron sulfides and iron oxides, Zbinden, unpubl. data) mineral particles (Fig. 2). TEM observations (Fig. 2) revealed the presence of bacteria in the gut, whereas none were observed in the stomach. Morphotypes observed in the hindgut indicate that bacteria are either cocci or rods, 0.35 μm in diameter, and at least 1.5 μm in length. They seem to be undamaged (i.e. not degraded by a digestion process), since the cell wall, composed of two membranes (as in Gram-negative bacteria), is readily visible and intact (see Fig. 2c).
R. exoculata hindgut content. a: Photonic observation of a semi-thin section of the digestive tract (hindgut). The gut is full of mineral particles, generating stria on the sections. Scale bar=100 μm. b: Bacteria observed in the hindgut in TEM. Scale bar=0.5 μm. c: Higher magnification of the membrane of the bacteria. Scale bar=10 nm.
When one ARDRA pattern contained more than 12 clones, three of them were partially sequenced. Most of the clones from the same ARDRA pattern displayed more than 99% sequence similarity, showing that ARDRA screening with HhaI was efficient. One exception was Entomoplasmatales, for which clone 2 and clone 69 displayed the same ARDRA pattern but had different sequences.
3.3 Phylogenetic analysis
Among the 100 recombinant colonies, 92 contained an insert and eight were false positive clones (no insert). The ARDRA screening for these 92 clones evidenced 16 different patterns, with seven patterns containing only one clone. Among these seven patterns, two were removed from the study as they were probably chimeric. The 14 ‘positive’ patterns represented 90 clones. Sequences of these clones were associated with three phylogenetic groups: ε-Proteobacteria, Spiroplasma and Geovibrio, from a new phylum described in 1997 .
Forty-four clones (47.8% of the study) were related to ε-Proteobacteria (see Fig. 3a; Table 1) and could be divided in two subgroups . Twenty-four clones (representing four ARDRA patterns) were related to the epibionts of R. exoculata group (called hydrothermal vent ε-groupII) while 20 clones were related to cold seep environmental clones (cold seep ε-groupII), 11 (two patterns) being related to BD21  and nine (two patterns) related to NKB11 . Twenty-one clones (three patterns) were related to the Entomoplasmatales and can be divided in two lineages, both related to Spiroplasma spp. and Mycoplasma spp. clusters (see Fig. 3b). Twenty-four other clones (representing two ARDRA patterns) showed 97% sequence similarity and are related to the genus Geovibrio of the new phylum Deferribacteres (Fig. 3c). One clone was related to γ-Proteobacteria, closely related to a Red sea clone  (Fig. 3a).
Phylogenetic trees obtained using maximum likelihood analysis with bootstrap resampling. Topologies were confirmed with maximum parsimony and neighbor-joining methods. Bootstraps values are indicated on nodes above 70%. Accession numbers of the sequences used are indicated on trees. a: This tree was based on 27 species and 673 sites were analyzed. Bacillus subtilis was chosen as outgroup. b: This tree was based on 13 species and 1060 sites were analyzed. Achoplasma spp. were chosen as outgroup. c: This tree was achieved with 11 species and 1121 sites were analyzed. Deinococcus radiodurans and Thermus aquaticus were chosen as outgroups.
The presence of bacteria in the digestive tracts (particularly the hindgut) of Crustacea is widespread, occurring across the taxa, feeding types and habitats [28,29]. These associations range from pathogenic to mutualistic, and from facultative to obligate , but in general, the roles that these bacteria play in the nutrition of their hosts are unknown.
Our TEM observations showed for the first time the presence of micro-organisms in the hindgut of the shrimp, and illustrate their morphology and integrity. As we were unable to observe bacteria in the foregut on our thin sections, they are probably much less numerous, if present, in the stomach than in the mid- and hindgut. The presence of large amounts of bacterial cell-material in the cardiac stomach of R. exoculata has been evidenced by lipopolysaccharide assays . However, their presence could not be confirmed by scanning electron microscopic observations, suggesting that only bacterial cell-wall remnants were present in this part of the gut. The bacteria observed in the hindgut had intact membranes, which suggested that they were resident and not submitted to a digestion process. This hypothesis is reinforced by the ‘high molecular mass’ of the extracted DNA, which was not lysed, indicating bacterial activity in the gut, or at least suggesting that the DNA (and thus the bacteria) was not degraded. The physiological state of the bacteria should nevertheless be tested on fresh samples to confirm this. These assumptions are consistent with the hypothesis of the occurrence of a local microflora [13,14].
The phylogenetic diversity of microbial communities of the MAR was reported for the Snake Pit and ‘Les Ruches’ sites [31,32]. Bacterial diversity revealed the presence of numerous phylotypes (27) related to the ε and β subclasses of the Proteobacteria, the Aquificales, the genus Desulfurobacterium and the Cytophagales. Although these analyses were not performed on samples from the same origin site as our samples, comparison of the results from both studies indicates that the microbial populations present in the hydrothermal fluid and in the shrimp gut are different. Two main differences are to be noticed: (1) the hydrothermal fluid displayed an important phylogenetic diversity, which did not exist in the R. exoculata gut, where only three major groups were present; (2) occurrence of Deferribacterales and Entomoplasmatales was observed in the gut, which was not found in the outer medium.
Deferribacterales: Species from the order Deferribacterales was first described from submarine petroleum reservoirs  and more recently from hydrothermal vents (from MAR: R. Miroshnichenko, pers. comm., and from a seamount in the Japanese sea ). To our knowledge, no Geovibrio species has yet been described from the Atlantic hydrothermal sites. If they are present, they are probably very rare. As they represent 25% of our clones, they are unlikely to come from the outer medium, and most probably represent a part of the local microflora of the gut. Most of these bacteria are iron reducing. Furthermore, although a new species of the Deferribacteres phylum (under description, R. Miroshnichenko, pers. comm.) is reported to be lithoautotrophic, the rest of the group is strictly heterotroph. Physiological assumptions based only on the phylogenetic position of a sequence must be made with caution. However, it is reasonable to think that the Deferribacteres occurring in the R. exoculata gut are heterotrophs and iron-reducing bacteria, the presence of iron oxides in the shrimp gut (, Zbinden, unpubl. data) suggesting the existence of metal-reducing micro-organisms. They would most probably be involved in detoxification processes or bioremediation of the mineral compounds found in the environment, rather than in primary production used as food source by the host.
Entomoplasmatales: Almost all members of the genus Spiroplasma that have been described are associated with arthropods, but are mainly found in terrestrial insects [34,35]. Most Spiroplasma spp. are known or suspected to be parasitic, although the degrees of pathogenicity may be extremely diverse between different species, ranging from evidently detrimental to sometimes slightly beneficial [36,37]. From available data, it is difficult to speculate on the roles of Spiroplasma species in the gut of R. exoculata. Nevertheless, as they were recovered from apparently healthy animals, they are probably non-toxic for the shrimp.
ε-Proteobacteria: In all deep-sea vent microbial communities studied, ε-Proteobacteria have been found to be dominant, accounting for 40% to 98% of the bacterial clones libraries . Among these bacteria, several are capable of autotrophic growth [14,38,39].
Genetic analysis of the bacterial populations associated with R. exoculata[6,14] showed that the population of bacterial epibionts consisted entirely of a single phylotype, which was also found to contribute to 60% of the detectable 16S rRNA amplified from the sulfide chimney samples, and to more than 20% of the 16S rRNA amplified from the shrimp gut (which is consistent with our results). Furthermore, very high carbon dioxide fixation rates were reported  in homogenates of total shrimp guts. From these results, Polz and collaborators  suggested that there should be additional, highly active bacterial populations in the shrimp gut that could serve as a previously unsuspected source of nutrition. Considering that most Deferribacterales and Entomoplasmatales are heterotrophs, it is reasonable to think that the high CO2 fixation rates measured in the shrimp  are due to the ε-Proteobacteria activity. This suggests that the putative nutritional symbiosis with a specific gut microflora, if existing, occurs through bacteria from this group.
In conclusion, it appears from our results that most of the bacteria from the surrounding medium are not found in the gut. Geovibrio sp. and Entomoplasmatales sp., which are found in the gut, do not seem to be present in the surrounding medium ([31,32], Cambon-Bonavita, unpubl. data). These results lead us to propose the following hypotheses: the foregut of R. exoculata is highly active and the ingested bacteria are rapidly degraded. Extracted DNA from the mid- and hindgut belongs to a local microflora. Some of these bacteria (the ε-Proteobacteria) could provide the shrimp with their metabolic by-products as a source of nutrition. A trans-epidermal transfer of dissolved organic matter from the bacteria has already been suggested for the filamentous bacteria present on the mouthparts . Gebruk and collaborators  rejected this hypothesis based on studies on other Crustacea, which show the active transport of dissolved organic matter to be unlikely. But recently, absorption of hydrolysates resulting from gut bacterial activity has been suggested to occur through hindgut cuticle of a deposit-feeding shrimp . Another explanation for the low bacterial diversity found in the gut is that the physico-chemical conditions existing in the gut are very specific and not suitable for numerous micro-organisms. This could explain why most of the bacteria found in the external medium are not able to develop in the gut environment.
The authors wish to thank P.M. Sarradin, chief scientist of the ATOS cruise, as well as the captain and crew of the Atalante and the Victor team. The authors are grateful to F. Gaill and J. Querellou for their constructive comments on this manuscript. Electron microscopy was performed at the Service de Microscopie Electronique, IFR 83 de Biologie Integrative-CNRS/Paris VI. This work was funded with the help of INSU, CNRS, the Region Bretagne, the European Ventox and CNRS (Geomex) programs.
(1992) Les peuplements associés à l’hydrothermalisme océanique du Snake Pit (dorsale médio-Atlantique, 23°N, 3480 m): composition et microdistribution de la mégafaune. C. R. Acad. Sci. III 314, 593–600.
(1997) Occurrence of 16(n-4) and 18:2 (n-4) fatty acids in the lipids of the hydrothermal vent shrimps Rimicaris exoculata and Alvinocaris markensis: nutritional and trophic implications. Mar. Ecol. Prog. Ser 1562167–174.
(1994) CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res 22, 4673–4680.
(2001) Growth and phylogenetic properties of novel bacteria belonging to the epsilon subdivision of the Proteobacteria enriched from Alvinella pompejana and deep-sea hydrothermal vents. Appl. Environ. Microbiol 67, 4566–4572.