Introduction

The association between microorganisms and reef–building corals, also called the coral holobiont, has been studied due to its influence on coral physiology and health. The microorganisms associated with corals can play a role in the host's health by providing a food source1, protecting the host against pathogenic bacteria by the production of antibacterial compounds and occupying specific niches2,3,4. Studies have shown that environmental changes, such as a decrease in pH or a rise in temperature and organic matter, can impact the composition of the coral holobiont5,6,7. Based on these observations, Reshef et al.8 suggested the probiotic hypothesis: changes in microbial communities under different environmental conditions allow a rapid and versatile adaptation of the coral holobiont.

Despite the importance of microbial communities to coral health, there are few studies that have focused on the discovering of species-specific associations between corals and bacteria. Species-specific bacterial associations have been found in other animals, such as sponges, hydra and colonial hydroids9,10,11, but the existence of these associations in corals is still controversial. Rohwer et al.12 found a ribotype of the Gammaproteobacteria group, closely associated to the coral Porites astreoides. Reis et al.13 and de Castro et al.14 suggested bacterial species-specific associations in the mucus of Mussismilia braziliensis and Mussismilia hispida. However, neither of these studies used the microbiota of sediment or of other coral species to confirm the possibility. Kvennefors et al.15 found species- and site-specific associations in the bacterial communities of two coral species from the Australian Great Barrier Reef. Ceh et al.16 analyzed the microbial communities associated with the tissues of three coral species and compared them with the surrounding water and sediment, finding no species-specific associations. Lema et al.17 assessed the diversity of the nifH gene, which encodes a bacterial dinitrogenase, in the tissue and mucus of three coral species from the Australian Great Barrier Reef. The sequences found in the mucus showed a great diversity and no specific associations with either coral species or sampling site. The authors suggested that this finding was due to the ephemeral nature of the mucus. However, they also found that the dominant nifH sequence differed among the tissues of the coral species studied.

The existence of species-specific interactions between bacteria and corals is still uncertain. Few studies have characterized the coral bacterial community composition in a sufficiently statistically robust manner to support it (which means, with replicates, multiple coral species, sampling of the surrounding environment and a sample size large enough to find rare bacterial species). The aim of this work was to investigate the existence of species-specific associations between Brazilian coral species and bacteria. Pyrosequencing of the V3 region of the 16S rRNA gene was used to analyze the taxonomic composition of bacterial communities associated with the mucus of four coral species (Madracis decactis, Mussismilia hispida, Palythoa caribaeorum and Tubastraea coccinea) found along the southeast Brazilian coast, during two seasons (winter and summer). The results were compared with the surrounding environment (water and sediment). Our hypothesis is that different coral species harbor distinct mucus microbial communities.

Results

Taxonomic composition and community analyses

A total of 25,035 OTUs were found in all the samples, of which more than 43% were singletons (Table 1). The Good's coverage ratio ranged between 83 and 98%. A greater diversity (H') of OTUs was found in the sediment samples (5.7 ± 0.9). The water samples presented the lowest diversity (3.5 ± 0.5), while the mucus samples were more diverse, with values ranging from 1.2 to 5.8.

Table 1 Sample information
Full size table

In all the samples, most of the OTUs were classified as Proteobacteria (Figure 1). A higher proportion of Cyanobacteria sequences were found in water samples and Actinobacteria and Acidobacteria sequences were more frequent in sediment samples. However, no substantial differences of phylum distribution were observed among the habitats, seasons, sampling sites, or coral species. Among the Proteobacteria, Alphaproteobacteria dominated in the water samples, while Gammaproteobacteria dominated in the samples of sediment and the mucus of P. caribaeorum, T. coccinea and M. hispida. A higher proportion of Betaproteobacteria was found in the mucus samples of the scleractinian corals M. hispida (1.2%), M. decactis (1.6%) and T. coccinea (2.8%) than in the other environments (Water = 0.53%, Sediment = 0.51% and P. caribaeorum = 0.53%).

Figure 1
figure 1

Average of phylum relative abundance in each habitat.

Full size image

The analysis of similarities (ANOSIM) indicated that the OTU composition was significantly different among the communities found in the three habitats (mucus, water and sediment, P = 0.0001).

The n-MDS was performed using Bray-Curtis similarities between the samples (Figure 2; stress value = 0.1554). In Figure 2, there is a clear grouping of the samples from water and of the samples from sediment. According to Figures 2 and 3 and the ANOSIM analysis (Table 3), samples of P. caribaeorum mucus were the most dissimilar and there was no clear differentiation between the mucus samples from the Scleractinian coral species.

Table 3 P-values derived from ANOSIM pairwise comparisons of bacterial community composition values using Bray-Curtis values. The values highlighted in bold are statistically significant (P < 0.05)
Full size table
Figure 2
figure 2

Nonmetric multidimensional scaling (nMDS) plots of the first 2 dimensions based on Bray-Curtis dissimilarities for water (dark blue), sediment (dark green), M. hispida mucus (light blue), M. decactis (light green), T. coccinea (red) and P. caribaeorum (pink).

The stress value is 0.1554.

Full size image
Figure 3
figure 3

UPGMA cluster analysis of the mucus communities of the four coral species studied.

The UPGMA clustering was based on Bray-Curtis dissimilarities using the mean of the relative abundance of OTUs for each coral species. Bootstrap values (in%) for 1000 replicates are given at the nodes.

Full size image

The ANOSIM analysis showed that there were no significant differences between the mucus communities in the two seasons (comparison of the summer and winter mucus samples gave P-values of 0.329 (M. hispida), 0.3303 (M. decactis), 0.3345 (P. caribaeorum) and 0.6665 (T. coccinea)). In the case of the water and sediment samples, the differences between the summer and winter were significant (P = 0.0004 and 0.0011, respectively). In our study, the communities of water and sediment did not differ significantly between the two sampling sites (P = 0.1749 and 0.1723, respectively). The difference between the sampling sites was significant for the samples of mucus (P = 0.0124), however the difference was not significant (P = 0.0808) when the samples of Palythoa were excluded from the analysis.

Most abundant OTUs for each habitat

The SIMPER analysis highlighted the OTUs primarily responsible for the observed differences between the habitats studied, using the relative abundances of the OTUs in each habitat. The five main results for each habitat are presented in Table S1 and their relative abundances are given in Table S2 (supplementary material). The OTU 23763, which was mainly associated with the water samples, was classified to the GPIIA group of Cyanobacteria, which includes the genera Prochlorococcus and Synechococcus that contribute most to the primary production in oligotrophic oceans18.

Some of the OTUs most related with M. decactis presented similarity with bacterial genera often associated with the degradation of organic pollutants. These included OTU 6997, which showed similarity with Rhodococcus erythropolis19, OTU 23270, related to Sphingomonas20 and OTU 19915, classified as Brachybacterium21. In addition, OTU 6481, attributed to the Acinetobacter genus, has been found in great abundance in bleached corals22.

The OTUs of M. hispida mucus, identified by SIMPER as important for distinction of the microbial community, were only found in the colonies of this species, but in low abundance. None showed similarity with previously described microorganisms. However, four of them (14290, 20636, 25133 and 23022) showed greater similarity in databases with sequences of 16S rDNA libraries of Brazilian coral.

The OTU 7634 dominated in the communities of P. caribaeorum mucus, representing around 40% of all reads found in these samples. This OTU and another two, also identified by SIMPER analysis, were classified as Gammaproteobacteria and showed similarity to sequences of 16S rDNA libraries of the soft coral Sinularia sp subjected to thermal stress.

One of the OTUs of T. coccinea, 17001, was classified in the genus Tenacibaculum and showed similarity with the sequence of 16S libraries of corals affected by black-band disease. The OTU 9209, classified as Alphaproteobacteria, showed similarity with a sequence of 16S from the Scleractian cold-water coral Lophelia pertusa. The OTU 16320 was classified in the genus Haliagium, a marine myxobacterium isolated in Japan that has known antifungal and antibiotic properties23.

The OTU 16689, associated with sediment, was classified to the nitrogen-fixing genus Mesorhizobium24 and the OTU 20185 was classified as a Gammaproteobacterium, which showed 100% similarity with a sequence from a 16S library of bleached Siderastrea stellata.

In addition to these results, genera present in at least two coral species, identified from mucus libraries, included Achromobacter (Betaproteobacteria), Acidovorax (Betaproteobacteria), Agarivorans (Gammaproteobacteria), Faecalibacterium (Firmicutes), Fibrobacter (Fibrobacter), Flammeovirga (Bacteroidetes), Marinobacter (Gammaproteobacteria), Marmoricola (Actinobacteria), Psychrobacter (Gammaproteobacteria), Serratia (Gammaproteobacteria) and Zunongwangia (Bacteroidetes). Marmoricola and Psychrobacter were found in the four coral species.

Species-specific OTUs

We found species-specific OTUs in P. caribaeorum, T. coccinea and M. hispida. All these OTUs were present in low abundance (between 20 and 190 reads, around 0.02% of the total reads for each coral).

The OTUs 23022, 14290 and 21841 were only found in three of the four colonies of M. hispida. The OTUs 6336, 11074 and 15129 were only found in the four colonies of T. coccinea. The OTUs 9266 and 3573 were found in the four colonies of P. caribaeorum. Table 2 presents information about these OTUs.

Table 2 Information about species-specific OTUs
Full size table

Most of the genera that were found associated with mucus are aerobic and heterotrophic and are commonly found in a wide range of hosts. These include Acinetobacter, Brachybacterium, Escherichia, Janthinobacterium, Pseudomonas, Sphingomonas and Tenacibaculum. However, some of the species-specific OTUs related to the Scleractinian corals were classified to autotrophic bacterial groups, such as Alphaproteobacteria and Chromatiales (purple sulfur bacteria).

Discussion

The bacterial communities found in the three habitats (mucus, water and sediment) were significantly different, which is in agreement with most of the studies published to date25,26. Some bacterial groups and OTUs were most related to a determined habitat. For example, the OTUs attributed to Betaproteobacteria, a group that includes the Janthinobacterium genus, were found to be most associated with the mucus of M. hispida, M. decactis and P. caribaeorum. Janthinobacterium, previously belonging to the Chromobacterium genus, is a gram-negative violacein-producing Betaproteobacterium found in soil, water and some animal surfaces27,28. Members of this family have already been found in the mucus of corals and Kooperman et al.29 suggested that they can be an obligate symbiont of Fungia granulosa.

Our results indicate not only that the microbial community of the mucus was distinct from the surrounding environment, but also that it was more stable throughout the seasons than the communities of the water column and the sediment. We suggest that the community of the mucus is less influenced by physical factors, such as temperature, compared to the communities in the surrounding environment. Our results differ from those obtained by Ceh et al.16, who studied the microbial communities of three coral species in Australia and concluded that seasonal changes were more important in determining the microbiota than the coral species and the spatial separation of coral. We believe that the stability found in the mucus community in our study can be attributed to the antimicrobial properties of the mucus. Studies have demonstrated the bacterial growth inhibitory activity of the mucus of several coral species2,30 and we have observed this property for the mucus of the coral species studied in this work (unpublished data). As suggested by Kvennefors et al.15, host and microbial factors can be responsible for the selection and maintenance of the bacteria that can inhabit the coral tissues.

The results also indicated that host phylogeny and genetic factors might be more important in determining the bacterial community than the presence of the zooxantellae (found in association with P. caribaeorum, M. hispida and M. decactis, but not T. coccinea), since we observed that the bacterial community of the zoanthid P. caribaeorum was different to the communities of the other species studied. This supports the notion of the existence of Scleractinian-specific associations31.

We have found OTUs that seem to be specific for each coral species. The species-specific OTUs of P. caribaeorum are related to the Endozoicomonas genus, which belongs to the order Oceanospirillales. Endozoicomonas and Oceanospirillales bacteria have been frequently found associated with corals, especially of the Porites spp. genus32. They dominate the community of Porites astreoides33 and the species Endozoicomonas montiporae was first isolated from the coral Montipora aequituberculata34. The role of these bacteria and the characteristics that allow their widespread association with marine organisms are still unknown32.

One of the most interesting species-specific OTUs was 23022, only found in the mucus of M. hispida and classified to the order Chromatiales, a group of microaerophilic or anaerobic, photosynthetic and sulfur-oxidizing bacteria35. The major hits of this OTU in the databases are with sequences of 16S libraries of Mussismilia species, which is strong evidence for the specificity of this OTU to M. hispida. In corals, phototrophic sulfur-oxidizing bacteria are usually found in the microbial consortia that cause black band disease. This is a major disease affecting corals and is characterized by degradation of the tissue of the animal by the action of hydrogen sulfide (H2S) produced by sulfate-reducing bacteria36, such as those found in sediment. Childress & Girguis37 suggested that the H2S detoxification function of sulfur-oxidizing bacteria in the symbiotic relationships with marine animals is not as important as the inorganic carbon fixing function. Nevertheless, we believe that H2S detoxification may be the main function required in a relationship between zooxanthellate corals and sulfur-oxidizing bacteria, since algae endosymbionts are responsible for production of most of the energy used by the corals. Thus, our hypothesis is that the presence (at low density) of phototrophic sulfur-oxidizing bacteria, such as those characterized by the OTU 23022, could assist in the balance of sulfur compounds in the coral tissues, protecting the host against the action of H2S. Raina et al.38,39 have suggested the role of sulfur compounds in determining the bacterial communities associated with corals. Based on these observations, we believe that more studies are needed to investigate the role of sulfur compounds in the coral holobiont, in order to elucidate their importance in species-specific associations between bacteria and corals.

We emphasize the importance of including coral reef sediments in future studies of coral microbiota. As observed in this study as well as by Schöttner et al.25,26, sediment presented the most diverse bacterial community among the coral reef habitats studied. We observed that mucus shared more OTUs with sediment than with water, suggesting that sediments might act as a seed bank of bacteria that could colonize the coral surfaces. Furthermore, sediment can also provide a reservoir of opportunistic coral pathogens, so that the study of this habitat could aid in understanding bacterial dynamics in coral reef environments.

We are aware that it is not possible to affirm that the species-specific OTUs found in this study represent mucus residents, because (as discussed by Lema et al.16) the ephemeral nature of the coral mucus could preclude strong bacterial associations. Hence, many of these bacteria may actually inhabit the coral tissues, which could explain their low density in the mucus samples.

Methods

Sample collection

The samples were obtained during the winter of 2010 (June) and the summer of 2011 (January), at two locations: São Sebastião Channel (S 23°51'467 W 45°25'867) and Búzios Island (S 23°48'157 W 45°07'181) in São Paulo State, Brazil. The sea surface temperature was around 20°C in the winter and 28°C in the summer at the two locations. The water depth was 9 m at São Sebastião Channel and 11 m at Búzios Island. The samples of Mussismilia hispida (Scleractinia) and Palythoa caribaeorum (Zoantharia) were collected from the São Sebastião Channel and the samples of Tubastraea coccinea (Scleractinia) and Madracis decactis (Scleractinia) were collected from Búzios Island.

From each coral colony (apparently healthy) we collected 10 ml of mucus, 1 liter of surrounding water directly above the colony and 50 ml of sediment directly below the colony. Four colonies of each coral species were sampled, spaced around 10 m from each other. Water samples were collected in sterile plastic bottles and first filtered through sterile filter paper to remove debris and then vacuum filtered through a 0.22 micrometer Millipore membrane, which was used for the DNA isolation. Sediment samples were collected in sterile 50 ml centrifugation tubes. The mucus samples were collected using sterile syringes. All the samples were stored on ice during transport to the laboratory and frozen at −20°C until DNA isolation was performed.

DNA isolation and 16S library construction and pyrosequencing

Total DNA of all samples was isolated using the Wizard Genomic DNA Purification Kit (Promega), according to the manufacturer's instructions for Gram-positive bacteria, using a pellet of the environmental sample (or a membrane in the case of water samples, instead of the pellet of cells). The average yield was around 10 ng/μL of DNA for each sample.

The amplification of the V3 region of the 16S rDNA was performed according to Clingenpeel et al.40, using the primers 338F (ACTCCTACGGGAGGCAGCAG) and 533R (TTACCGCGGCTGCTGGCAC)41. The amplification reactions (final volume 25 μL) employed 3–10 ng of DNA, 1U AccuPrime Pfx® DNA polymerase (Invitrogen), AccuPrime Pfx Reaction Mix® 1X (Invitrogen) and 0.4 mM of each primer. Amplification was performed under the following conditions: initial denaturation at 94°C for 1 min and 25 cycles at 94°C for 30 seconds, 62°C for 1 min and 72°C for 1 min. After these 25 cycles, a further five cycles were performed, using the same conditions, in order to add the specific barcodes from each sample and the adapters A and B (reaction with final volume of 50 μL). The amplification products were purified with the GFX PCR and DNA Gel Band Purification Kit (GE Healthcare), quantified using the Qubit Kit (Invitrogen) and the quality (integrity and presence of a unique band) was confirmed by 1% agarose gel electrophoresis. An equimolar mixture of the samples from each library (summer and winter) was employed in order to obtain 5000 ng of PCR product in total, with a minimum concentration of 20 ng/μl. The winter library was composed of 20 samples: 6 of water, 8 of mucus and 6 of sediment. The summer library was composed of 21 samples: 7 of water, 7 of mucus and 7 of sediment. The amplicons were sequenced with the 454 technology, a GS FLX Roche Company platform.

Sequence processing and statistical analysis

The initial processing of the sequences was carried out with the package QIIME (http://qiime.org/). The sequences were screened according to quality (minimum of 25) and size (between 150–240 bp). No ambiguous bases or mismatches were admitted in the sequence of the primers and we checked for the existence of chimeras. The sequences with similarity greater than or equal to 97% were grouped into operational taxonomic units (OTUs) and the taxonomic classification was performed using the RDP (Ribosomal Database Project: http://rdp.cme.msu.edu/classifier/classifier.jsp) classification threshold of 80%. Rarefaction curves were also constructed with the QIIME package. The Good's coverage ratio (C) was calculated according to the formula C = 1 − ni/N, where ni is the number of OTUs that were sampled only one time (singletons) and N is the total number of sequences in the sample. The Shannon index (H) was calculated according to the formula H = − Σ pi ln pi, where pi is the proportion of OTU i relative to the total number of OTUs.

The community analysis was performed according to Laverock et al.42 and Theis et al.43. Multidimensional scaling (n-MDS) analysis was used to obtain the community similarities (Bray-Curtis values), employing the relative abundances of OTUs in each sample. Analysis of similarities (ANOSIM) was used to test the hypothesis that bacterial communities from the same habitat were more similar to each other than to communities from different habitats. Similarity percentage (SIMPER) analysis was used to identify the most abundant OTUs for each habitat, employing the relative abundances of OTUs in each sample. The UPGMA (Unweighted Pair Group Method with Arithmetic Mean) clustering was based on Bray-Curtis dissimilarities using the average between the relative abundances of OTUs of all the samples from each coral species. Bootstrapping with 1000 resamplings was performed to determine the robustness of the clustering. All these analyses were performed with the statistical software PAST (http://folk.uio.no/ohammer/past/).

We have considered as species-specific OTUs those found in only one coral species and in at least three of the four colonies studied, since they are not necessarily a symbiont found in all the individuals of the species.