Transient Shifts in Bacterial Communities Associated with the Temperate Gorgonian Paramuricea clavata in the Northwestern Mediterranean Sea

Background Bacterial communities that are associated with tropical reef-forming corals are being increasingly recognized for their role in host physiology and health. However, little is known about the microbial diversity of the communities associated with temperate gorgonian corals, even though these communities are key structural components of the ecosystem. In the Northwestern Mediterranean Sea, gorgonians undergo recurrent mass mortalities, but the potential relationship between these events and the structure of the associated bacterial communities remains unexplored. Because microbial assemblages may contribute to the overall health and disease resistance of their host, a detailed baseline of the associated bacterial diversity is required to better understand the functioning of the gorgonian holobiont. Methodology/Principal Findings The bacterial diversity associated with the gorgonian Paramuricea clavata was determined using denaturing gradient gel electrophoresis, terminal-restriction fragment length polymorphism and the construction of clone libraries of the bacterial 16S ribosomal DNA. Three study sites were monitored for 4 years to assess the variability of communities associated with healthy colonies. Bacterial assemblages were highly dominated by one Hahellaceae-related ribotype and exhibited low diversity. While this pattern was mostly conserved through space and time, in summer 2007, a deep shift in microbiota structure toward increased bacterial diversity and the transient disappearance of Hahellaceae was observed. Conclusion/Significance This is the first spatiotemporal study to investigate the bacterial diversity associated with a temperate shallow gorgonian. Our data revealed an established relationship between P. clavata and a specific bacterial group within the Oceanospirillales. These results suggest a potential symbiotic role of Hahellaceae in the host-microbe association, as recently suggested for tropical corals. However, a transient imbalance in bacterial associations can be tolerated by the holobiont without apparent symptoms of disease. The subsequent restoration of the Hahellaceae-dominated community is indicative of the specificity and resilience of the bacteria associated with the gorgonian host.


Introduction
The interactions between microbial communities and sessile marine invertebrates such as sponges or corals are increasingly recognized as a critical component of the overall biology of these organisms. Recent research on the microbiology of scleractinian corals has demonstrated close relationships between the hexacoral animal and numerous prokaryotic organisms, and studies on coral reef ecosystems should consider the coral host with its associated microbiota as a unique evolutionary unit termed the ''holobiont'' [1][2][3]. Corals harbor highly diverse bacterial communities, and several reports have highlighted the existence of specific coralbacteria associations that are mostly maintained among colonies from the same locality or even across distinct geographical locations [2,[4][5][6][7]. These bacterial assemblages may play important roles in the host's physiology, mainly through their functions in nutrient cycling and health status [8]. In particular, there is increasing evidence for a relationship between coral-associated bacteria and disease resistance, which is likely the result of antibiotics produced by the resident microbiota and/or niche competition with potential pathogens [1,9,10].
In contrast to scleractinian reef-forming corals, information on the microbial diversity associated with gorgonians (octocorals) is limited. A few studies have investigated bacterial associations with tropical or cold-water species [11][12][13], but to our knowledge, the bacterial diversity of shallow temperate octocorals has not yet been explored. In the Northwestern (NW) Mediterranean Sea, gorgonians play an important ecological role in highly diverse coralligenous outcrops [14]. During the last few decades, coralligenous benthic invertebrates have suffered from large-scale disease and mortality [15][16][17][18]. These events are linked to unusual positive anomalies of seawater temperature in the summer and have affected a wide range of macro-benthic species over hundreds of kilometers, from the Italian to Spanish coasts [19]. Gorgonians populations have undergone extensive damage, although differences in the degree of impact have been observed in various geographic areas.
One of the most affected species is the red gorgonian Paramuricea clavata (Risso, 1826), a long-lived, aposymbiotic colonial octocoral that is considered a key species within coralligenous assemblages [20]. During the summer mortality outbreaks, P. clavata colonies exhibited symptoms of necrosis that may have been caused by multiple factors acting in synergy with thermal stress, including food limitation, metabolic constraints and microbial virulence [15,21]. Colonies that were affected during mass mortality events in 2003 and 2008 harbored culturable isolates of Vibrio coralliilyticus, a pathogenic bacterium that can induce host tissue damage at elevated temperatures [22,23]. Members of this Vibrio clade were previously identified as the etiologic agents of bleaching in the Indo-Pacific coral Pocillopora damicornis and White Syndrome disease in other coral species, thus satisfying Koch's postulate [24,25]. This finding suggests that at least some mechanisms underlying the disease process in P. clavata are comparable to the mechanisms involved in tropical coral outbreaks. Several studies have concluded that coral pathogens take advantage of changes in the bacterial community structure during stressful conditions to proliferate and cause tissue damage [4,26,27]. In support of this hypothesis, compositional shifts in bacterial assemblages have been observed in diseased coral colonies [28,29]. Similarly, the recurrent disease and mortality of P. clavata may be promoted by a disturbance of the resident microbiota in response to anomalous high-temperature conditions during the summer.
To evaluate the contribution of microbial communities to temperate gorgonian health and disease, a detailed knowledge of the structure of the bacterial assemblages in healthy colonies is required. The objective of the present study was to provide a baseline of the bacterial communities associated with P. clavata populations in the NW Mediterranean basin. Given the predicted increase in extreme climatic events in this region, new mortality outbreaks are expected [30,31]. Thus, tracking potential modifications of this baseline during future high-temperature stress could greatly aid our understanding of the role of bacterial associations in maintaining gorgonian health.
In this work, we conducted a seasonal sampling over 4 years (2007-2010) in 3 distinct geographical locations separated by hundreds of kilometers (Provence, Corsican and Catalan coasts), which allowed us to characterize the bacterial communities associated with P. clavata and the spatiotemporal variability of their structure. Three different molecular culture-independent approaches (denaturing gradient gel electrophoresis (DGGE), terminal-restriction fragment length polymorphism (T-RFLP) and 16S RNA gene clones library construction) were used to determine which bacteria might be conserved across geographically remote P. clavata populations. Our findings reveal the presence of host-specific bacterial associations in the gorgonian holobiont, although transient and reversible variations in microbiota composition were also recorded during our seasonal survey.

Ethics Statement
The sampling in the Mediterranean Sea was performed in accordance with French professional diving rules and did not involve endangered or protected species. Annual sampling permits for the site located in Riou (France) were delivered by the Préfet de la region Provence-Alpes-Côte d'Azur. Specific permits delivered by the DIRM Méditerranée, advised by the Conseil Scientifique de la Réserve Naturelle de Scandola and the Departament d'Agricultura, Ramaderia, Pesca, Alimentació i Medi Natural de la Generalitat de Catalunya, were obtained to sample gorgonian tissues in the marine protected areas of Scandola (Corsica) and Medes Islands (Spain). Gorgonian sampling entailed the collection of only a few centimeters of the apical part of the branches. Because gorgonians are colonial organisms, the sampling caused unnoticeable injuries that healed rapidly after the collection.

Sample collection and processing
Samples of P. clavata were collected from 3 sites in the NW Mediterranean Sea, Riou Island (France) (43u10.3459 N, 05u23.3199 E), Scandola (France) (42u22.7939 N, 8u33.0009 E) and Medes Island (Spain) (42u2.8829 N, 3u13.5799 E) (Figure 1), at an approximately 20 m depth in winter (January to March) and summer (June to September) from 2007 to 2010. Apical branch fragments (2 cm length) of randomly chosen, apparently healthy colonies (i.e., with no visible signs of necrosis) (n = 3) were collected using shears and placed in plastic bags underwater. Three liters of surrounding seawater were also collected in plastic bottles. The collected samples were then transferred to the Mediterranean Institute of Oceanography within 2 h (Riou site) or frozen on dry ice for subsequent transportation to the laboratory (Scandola and Medes sites).
The P. clavata samples were rinsed 3 times with sterile 0.22 mmfiltered seawater to remove loosely associated bacteria. Tissues were detached from the central axis and gently crushed using sterile forceps and a scalpel blade in 3 ml of sterile seawater. The tissue slurry was homogenized and aliquoted into 3 microtubes. Samples were then spun at 13,000 g, and the supernatant was removed before storing the tissue pellets at 280 uC. The surrounding seawater samples were immediately filtered through 0.22 mm SterivexH filter units (Merck Millipore, Billerica, MA, USA). The filters were stored at 280 uC for subsequent DNA extraction.

DNA extraction
Bacterial DNA in gorgonian tissue samples was extracted following the protocol of Bourne et al. [28]. Briefly, tissue slurries were resuspended in 0.5 ml extraction buffer (50 mM Tris-HCl pH 8.0, 40 mM EDTA, 0.75 M sucrose) with 1 mg of salmon sperm DNA and incubated for 5 min at room temperature. Next, a lysozyme solution (75 ml of 100 mg?ml 21 ) was added. The samples were incubated for 1 h at 37 uC with slow agitation and submitted to 3 freeze-thaw cycles (ranging from 280 uC to 70 uC); these steps were followed by the addition of 100 ml 25% sodium dodecyl sulfate (SDS) and a further incubation of 10 min at 70 uC. The samples were digested by proteinase K (20 ml of 20 mg?ml 21 ) for 1 h at 37 uC with slow agitation. The tissue lysates were again submitted to 3 freeze-thaw cycles and extracted by standard phenol-chloroform DNA extraction procedures. After precipitation with 50 ml 3 M sodium acetate and an equal volume of isopropanol, the DNA was pelleted by centrifugation (16,000 g for 30 min at 4 uC), washed with cold 70% ethanol and suspended in 30 ml sterile ultra-pure water. The DNA concentration was estimated by spectrophotometry using a biophotometer (Eppendorf, Hamburg, Germany), and the DNA samples were stored at 220 uC until further processing.
After an overnight incubation at 55 uC, the cell lysate was recovered from the filter unit, and total DNA was extracted by standard phenol-chloroform extraction procedures. The DNA in the liquid phase was then precipitated with ethanol and concentrated by centrifugation at 16,000 g for 10 min at 4 uC. The DNA pellets were washed with 70% cold ethanol, resuspended in 100 ml sterile ultra-pure water and stored at 220 uC until analysis.
PCR (35 cycles) was performed on a MastercyclerH thermocycler (Eppendorf) in a total volume of 50 ml containing 1 X TaqH Flexi Buffer (Promega, Fitchburg, WI, USA), along with 1.5 mM MgCl 2 , 200 mM of each deoxynucleoside triphosphate (dNTP), 1 mM forward and reverse primer, 0.3-3 ml (approximately 100-300 ng) of bacterial DNA template, 1 U of GoTaqH Flexi DNA Polymerase (Promega), and sterile ultra-pure water. After an initial denaturation of 1 min at 96 uC, a touchdown PCR protocol was performed in which the primer annealing temperature was initially set at 53 uC (27F/536R), 62 uC (EUB f933/EUB r1387) or 58 uC (341F/907R and 341F/907RA) and was decreased by 0.5 uC with each cycle of the next 20 cycles. The final annealing temperature was then maintained for the remaining 15 cycles of the reaction. For each cycle, denaturation was performed at 94 uC for 1 min, followed by annealing for 1 min and extension at 72 uC for 1 min. The PCR products were evaluated by electrophoresis on 1% agarose gels stained with ethidium bromide. To identify the 16S rDNA bacterial sequences, selected DGGE bands were excised from the gels and eluted overnight at 4 uC in 50 ml sterile ultra-pure water. PCR re-amplification was performed with the corresponding primer set. The PCR products were purified with WizardH PCR Clean-Up minicolumns (Promega), and the recovered 16S rDNA gene fragments were sequenced (Eurofins MWG Operon, Ebersberg, Germany).

T-RFLP analysis
PCR amplification, enzymatic digestion and T-RFLP. For T-RFLP analysis, the bacterial 16S rRNA genes were amplified by PCR with the extracted DNA using the universal primers 63F [36] and 786R-Eub (a reverse-complement version of 768F-Eub; [37]). This primer pair target the V1-V4 hypervariable regions of the 16S rRNA gene. The forward primer was labeled at the 59 end with phosphoramidite fluorochrome 6carboxyfluorescein (Applied Biosystems, Carlsbad, CA, USA). PCR amplifications were performed in a total volume of 25 ml containing 1 X Taq Buffer with (NH 4 ) 2 SO 4 (Fermentas, Burlington, Canada), 200 mM each dNTP, 1 mM forward and reverse primer, 2 mM MgCl 2 , 0.1 mg ml 21 bovine serum albumin (BSA), 0.3-3 ml (approximately 100-300 ng) bacterial DNA template, 1 U native Taq DNA Polymerase (Fermentas) and sterile ultra-pure water. After an initial denaturation step (5 min at 94 uC), 35 amplification cycles (94 uC for 1 min, 48 uC for 1 min, 72 uC for 1 min) were performed on a MastercyclerH thermocycler (Eppendorf). Amplification was confirmed by migration on a 1% agarose gel in TAE and ethidium bromide staining. The labeled PCR products were purified using the WizardH PCR Clean-Up System (Promega) following the manufacturer's instructions and eluted in 30 ml of DNase-free distilled water (Promega).
The restriction reactions were performed for 16 h at 37 uC using a digestion mixture containing 6 ml of the purified PCR products, 1 X reaction buffer (Promega), 0.1 mg ml 21 BSA and 20 U CfoI or MspI (Promega) in a final volume of 20 ml. Each digested sample (3 ml) was mixed with 0.3 ml GeneScan 600-LIZH (Applied Biosystems) size standard and 10 ml Hi-DiH Formamide (Applied Biosystems) in a 96-well plate and denatured at 90 uC for 3 min. The labeled terminal restriction fragments (TRFs) were then separated on an ABI 3130 Genetic Analyzer (Applied Biosystems) with default fragment analysis parameters. Each sample was run twice as replicates to account for migration discrepancies. All T-RFLP electropherograms were visually inspected for possible artifacts and incorrect peak determination and were tabulated in GeneMapperH version 4.0 software (Applied Biosystems) using the Local Southern method as the size-calling algorithm [38]. Only fragments longer than 80 bp and smaller than 600 bp were considered.

Statistical
analysis of the spatiotemporal variability. The raw data sets (peak sizes in base pairs and peak areas in fluorescence units) for the 2 restriction enzymes were exported to Excel (Microsoft, Redmond, WA, USA). To produce a single composite profile for each sample, the average area and size for each peak were calculated between duplicates (TRFs with sizes within 0.5 bp of each other were considered identical, while TRFs that were present in only one of the duplicates were removed). Because the amount of DNA template loaded for T-RFLP analysis could not be controlled accurately, T-RFLP profiles were normalized by applying the variable percentage threshold, as described by Osborne et al. [39]. To eliminate background fluorescence, only peaks contributing to more than 1.2% of total fluorescence (according to the variable percentage threshold result) were considered. Data were binned using the interactive and automatic binning algorithms [40] implemented in the free R programming language [41], applying custom R binning scripts (WS 1; Sh 0.1) to account for size calling imprecision. Based on the ''best binned frame'' identified by the algorithm, the data were transformed for further graphical and statistical representations. The relative abundances in the different analyzed samples were calculated for each TRF. The data were transformed to a binary matrix (presence/absence), and the distance matrix between samples was computed from the Dice similarity index using R (R package ade4; [42]). Patterns of bacterial community similarities were represented on a 2-D scatter plot using multidimensional scaling (MDS) ordination. The K-means clustering method was then performed to determine the center coordinates of the clusters (R package fpc; [43]). Confidence ellipses were superimposed on the MDS ordination plot for each previously determined cluster, with 2 confidence levels of 50% and 95% (R package car; [44]).

Bacterial clone library construction
The universal bacterial primers 63F and 1389R [36] were used to amplify nearly full-length 16S rRNA genes from the gorgonian tissue samples. The PCR mixtures contained 1 X of TaqH Flexi Buffer (Promega), 1.5 mM MgCl 2, 200 mM each dNTP, 1 mM each primer, 1 U GoTaqH Flexi DNA polymerase (Promega) and 0.3-3 ml (approximately 100-300 ng) DNA template in a total volume of 25 ml. PCR was performed with an initial 5 min denaturation step at 94 uC, followed by 35 amplification cycles consisting of 94 uC for 1 min, 55 uC for 1 min and 72 uC for 1 min. The final elongation step was performed at 72 uC for 10 min. For each library, the DNA extracted from 3 P. clavata colonies was PCR-amplified in 3 separate reactions, each performed in triplicate. The amplified products were pooled and purified with WizardH PCR Clean-Up minicolumns (Promega), followed by cloning into the pGEMH-T Easy vector (Promega) according to the manufacturer's instructions. After transformation into E. coli JM109 competent cells, the orientation of the inserted 16S rRNA gene of each resulting clone was determined by PCR re-amplification using 63F and M13 reverse primers. Clones producing a PCR product of the expected size (approximately 1,500 bp) were sequenced with a plasmid forward primer (LGC Genomics GmbH, Berlin, Germany). A total of 700 partial 16S rRNA gene sequences were obtained.

Phylogenetic analysis
All bacterial 16S rDNA clone sequences were trimmed to 750 bp starting at the 59 end of the 63F primer, and the production of chimeras during PCR was evaluated with the Pintail program [45]. The resulting libraries were dereplicated using FastGroupII ( [46]; http://biome.sdsu.edu/fastgroup/index.htm) with a similarity cutoff value of 97% for sequence group definition. The unique sequence groups were classified taxonomically at the family level by the SeqMatch tool of the Ribosomal Database Project (RDP; http://rdp.cme.msu.edu). Within the dominant family-level groups, the sequence affiliations and closest relatives were identified by comparison with the GenBank database using the BLAST algorithm ( [47]; http://blast.ncbi.nlm.nih.gov/Blast. cgi). To describe the compositional differences between the 3 summer 2007 libraries, the Shannon-Weaver index for diversity [48] and the Chao1 richness index [49] were calculated using the RDP Pipeline tools, with sequence groups defined with 97% similarity.
The consistency between the identified ribotypes in the libraries and the major TRFs in each community's peak profiles was confirmed by T-RFLP analysis of representative 16S rDNA clones of dominant ribotypes. Partial sequence reads obtained from the excised DGGE bands were taxonomically assigned by BLAST analysis against the GenBank database.

Nucleotide sequences accession numbers
The nucleotide sequence data for the partial 16S rRNA genes reported in this paper were deposited in GenBank under the accession numbers JX874190-JX874360 (Medes; summer

DGGE fingerprinting of P. clavata-associated bacterial communities
To estimate the intra-and inter-colonial variability of bacterial community composition, apical branches from P. clavata colonies sampled in Riou in winter 2007 were analyzed by 16S rDNA-DGGE fingerprinting. PCR amplification with 27F-GC/536R universal primers followed by denaturing electrophoresis yielded repeatable banding patterns for all 3 fragments sampled from each colony ( Figure 2, lanes 1-9). The consistency of the banding patterns from different P. clavata colonies was clearly apparent, indicating the presence of similar bacterial diversity. By contrast, the DGGE profile of seawater collected in close vicinity of the colonies (Figure 2, lane 10) was markedly different and did not share detectable bands with the P. clavata tissue samples. The dominant 16S rDNA band in the fingerprint of P. clavata was excised from the gel and sequenced for identification of the corresponding ribotype. This band was related to 16S rDNA sequences from marine bacteria belonging to the order Oceanospirillales. The closest cultivated relative was Spongiobacter nickelotolerans (GenBank accession number AB205011), with 94% similarity along the 479 bp sequence length.
The dominant DGGE bands obtained with 3 other universal bacterial primer pairs were also sequenced. The band obtained with the EUB f933/EUB r1387 primer pair (Figure 2, lane 11) was related to Spongiobacter nickelotolerans 16S rDNA (94% similarity among 417 bp). The sequences recovered from the 341F-GC/ 907RA and 341F-GC/907R amplifications (Figure 2, lanes 12-13) shared 91 to 92% similarity with the 16S rDNA of Endozoicomonas elysicola (AB196667), a close relative of Spongiobacter spp. This similarity indicated that the predominance of single bands observed in all the P. clavata DGGE profiles was not caused by amplification biases but likely resulted from an abundant population of ribotypes belonging to the Oceanospirillales order.

T-RFLP analysis of bacterial community variation
Temporal and spatial changes in bacterial diversity associated with P. clavata were investigated by T-RFLP analysis. A total of 186 profiles, including run duplicates (96 and 90 profiles for CfoI and MspI digestion, respectively), were obtained from tissues of 48 P. clavata colonies that were seasonally sampled between March 2007 and September 2010 at the Riou, Medes and Scandola sites (3 colonies per site for each sampling). The profiles generated by run duplicates were consistent. The diversity and abundance of the detected TRFs are shown in Figures 3 and 4 for each restriction enzyme. In total, 57 different TRFs were observed upon CfoI digestion (Figure 3), and 50 TRFs were observed for MspI digests (Figure 4). None of the TRFs were shared by all samples, independent of the endonuclease. The T-RFLP profiles obtained for replicate colonies of each sampling were consistent, although variations in peak intensity were observed, likely due to the nonquantitative PCR amplification step. However, the dominant TRFs between replicates from the same seasonal sampling were broadly conserved between the CfoI and MspI data sets.
Two dominant TRFs were consistently recovered throughout most of the survey, indicating the presence of conserved ribotypes. The The library constructed from colonies sampled in winter 2007 in Riou was largely dominated by a single ribotype that accounted for 98.9% of the sequenced clones ( Figure 6A). The winter 2008 library was exclusively composed of the same ribotype. This uncommon dominance is unlikely to be the consequence of a contamination bias because bacterial DNA extractions were performed separately, in March 2007 and February 2008, for the construction of the corresponding 16S rDNA libraries. The sequence divergence was very low (,1%) between clones belonging to this dominant ribotype in the 2 winter libraries, suggesting that it may represent microheterogeneity of 16S rRNA gene copies within a unique bacterial species. According to pairwise comparisons, this ribotype is related to members of the Hahellaceae family within the Oceanospirillales order (class Gammaproteobacteria). The closest cultivated strains were Endozoicomonas elysicola (AB196667), Endozoicomonas montiporae CL-33 (FJ347758) and Spongiobacter nickelotolerans (AB205011), with 92-93% 16S rDNA sequence similarity (see Figure S1 for a 16S rDNA-based phylogenetic tree). Although the Hahellaceae-related ribotype is undoubtedly a highly abundant component in the P. clavataassociated assemblage, a better description of diversity within these libraries would certainly require the analysis of a larger number of clones.
By contrast, the 16S rDNA sequences obtained from the summer 2007 clone libraries exhibited greater diversity and were distributed among 5 bacterial phyla, namely Firmicutes, Actinobacteria, Proteobacteria, Bacteroidetes and Cyanobacteria ( Figure 6B-D). Firmicutes was dominant in each of the libraries, constituting 76.1% of the sequences from Riou, 69.6% from Medes and 43.8% from Scandola. The second most abundant phylum was Actinobacteria, which comprised 13.4% of the cloned sequences in the Medes library and 25.0% in the Scandola library. The Gammaproteobacteria were represented to a lesser degree in the bacterial community in each of the 3 summer 2007 libraries (2.3% to 6.8%), and none of these ribotypes were related to the Hahellaceae family. Analysis at lower taxonomic levels revealed the presence of a high proportion of almost identical sequences affiliated with the genus Paenibacillus, accounting for 84.4% to 98.6% of the Firmicutes species (36.9% to 75.0% of the clone libraries). According to BLAST and SeqMatch searches, the closest relative is Paenibacillus validus (AB073203), which shares .98% sequence similarity with this ribotype. A second dominant ribotype was identified within Actinobacteria, which represented 52.4% to 83.5% of this phylum in each of the libraries (9.4% to 14.2% of the total number of clones). The nearest relative of this group of sequences is Propionibacterium acnes (AE017283; .98% similarity for 16S rDNA). While Paenibacillusand Propionibacterium-related sequences were predominant in the 3 libraries from summer 2007, the communities exhibited differing diversities. The Chao1 and Shannon-Weaver indices were highest in the Scandola library, although there is considerable overlap of the estimated total number of ribotypes at the 95% confidence interval within these 3 libraries (Table 1).
To identify the TRFs of the dominant ribotypes, the T-RFLP profiles of representative clones that were recovered from the 5 libraries were analyzed. The observed TRFs of the winter libraries clones carrying the Hahellaceae 16S rDNA sequences matched CfoI TRF-171.4 and MspI TRF-105.1, which correspond to the dominant peaks in all the P. clavata community profiles, with the exception of the summer 2007 samples (Figures 3 and 4). Clones corresponding to the Paenibacillus sequences in the summer 2007 libraries generated CfoI and MspI TRFs (TRF-193.4 and TRF-100.1, respectively) that matched the dominant peaks in the corresponding T-RFLP summer profiles. Propionibacterium 16S rDNA cloned sequences only generated an MspI peak (TRF-126.1), and no signal was detected following CfoI hydrolysis. In silico digestions of these Propionibacterium-related sequences showed that the size (636 bp) of the predicted CfoI TRF was beyond our detection range.

Discussion
This study provides the first description of the spatial and temporal patterns of the structure of the microbial communities associated with a temperate gorgonian in the Mediterranean Sea. The analysis of the DGGE fingerprint data indicated that the bacterial diversity hosted by P. clavata is broadly similar within and between individual colonies from a population and differs from the community composition within the surrounding water. In addition, the MDS ordination of the T-RFLP diversity profiles revealed a lack of clustering by location for the gorgonian populations sampled among our 3 study sites, which are separated by hundreds of kilometers. These results clearly suggest that the structure of bacterial communities that inhabit P. clavata does not rely on local environmental drivers or discrete gorgonian populations. Most of the T-RFLP profiles from the 48 colonies sampled between March 2007 and September 2010 grouped together, thus supporting the notion of specific bacterial-host associations. Taken together, data from the T-RFLP analysis and 16S rRNA gene clone libraries indicate that P. clavata colonies are associated with an almost permanent microbial population that is highly dominated by a unique gammaproteobacterial ribotype. However, we detected a transient compositional shift between this relatively stable bacterial assemblage that was observed throughout 7 of the 8 investigated seasons, with a strikingly different diversity pattern in summer 2007.
The dominant identified ribotype is affiliated with the family Hahellaceae within the order Oceanospirillales and is related to Endozoicomonas and Spongiobacter spp. Interestingly, a meta-analysis of coral-associated bacterial assemblages revealed that Oceanospirillales-affiliated sequences were among the most frequent ribotypes found in healthy hexacorals [50], and several recent reports indicated that Hahellaceae members may represent a common group of coral-associated Oceanospirillales. Spongiobacter sequences have been retrieved from the hexacoral Acropora millepora, in which they represent the largest proportion of bacterial clone libraries [28,51]. In 2 other hard coral species, Acropora hyacinthus and Stylophora pistillata, Kvennefors et al. [29] identified a cluster of ribotypes that are closely related to the genus Endozoicomonas. This so-named ''Type A Associates'' cluster includes several ribotypes that were previously observed in different tropical coral species or specifically associated with benthic marine invertebrates. Spongiobacter-related 16S ribosomal sequences were also detected in the cold-water coral Madrepora oculata and dominated the DGGE banding pattern of the associated bacterial community in this species [52]. Very recently, Morrow et al. [53] found that Endozoicomonas represented the most abundant bacterial genus in communities of the reef-building coral Porites astreoides, representing up to 99% of the 16S rRNA sequences recovered from colonies in one of the studied sites. The presence of Hahellaceae as a major component of the bacterial communities in both hexacorals and gorgonians (octocorals) suggests that these associates have differentiated to form a stable symbiotic complex specific to a cnidarian taxon. Further research focusing on the patterns of diversity of this bacterial group will certainly help to understand the evolutionary forces that may drive host-associated community composition in divergent phylogenetic lineages of cnidarians. Figure 3. Relative abundances of CfoI TRFs in P. clavata samples analyzed by T-RFLP. Each row represents a TRF generated by CfoI digestion of PCR amplicons from samples that were collected from the different sites (Riou (Ri), Medes (Me) and Scandola (Sc)) during the seasons indicated in columns. TRFs were designated by their size (in bp) after binning. The filling colors represent the relative fluorescence of the TRFs according to the range presented in the color key. doi:10.1371/journal.pone.0057385.g003 Figure 4. Relative abundances of MspI TRFs in P. clavata samples analyzed by T-RFLP. As described in Figure 3, with the restriction enzyme MspI. doi:10.1371/journal.pone.0057385.g004  In contrast to the unique Hahellaceae ribotype revealed by our study in P. clavata, several distinct Endozoicomonas-Spongiobacter sequences coexist in hexacoral species. This difference suggests that distinct hexa-and octocoral hosts may exert different selective controls on their microbial partners, resulting in various diversity patterns of Hahellaceae associates. Consistent with this hypothesis of host-driven control on associated Hahellaceae bacteria, the ribotype identified in P. clavata did not closely match the Spongiobacter-Endozoicomonas sequences found in hexacorals (,93% similar) but was more closely related (.96% similar) to bacterial sequences retrieved from the tropical gorgonian Gorgonia ventalina (GenBank accession number GU118518; [13]), suggesting that the species associated with P. clavata may belong to a new Hahellaceae genus that is adapted to gorgonian hosts.
To date, very little is known about the potential contribution of Hahellaceae bacteria to the functioning of the holobiont. One study has shown that Spongiobacter-related organisms isolated from A. millepora tissues possess the ability to metabolize the organosulfur compound dimethylsulfoniopropionate (DMSP), suggesting a role in sulfur cycling in scleractinian corals [51]. However, because DMSP production in corals is believed to rely on the presence of endosymbiotic zooxanthellae [51], a similar hypothesis for Hahellaceae-affiliated bacteria in P. clavata seems unlikely because this gorgonian species is devoid of photosynthetic symbionts [20]. Despite multiple trials, we failed to isolate culturable Hahellaceae bacteria from P. clavata tissues; consequently, their metabolic potential could not be investigated. Another interesting property of the members of the Oceanospirillales clade is the production of extracellular hydrolytic enzymes that are involved in the degradation of various complex organic substrates, providing a potential nutritional integration between bacterial associates and their host, as suggested for the endosymbiotic Oceanospirillales found in the deep-sea worm Osedax [54]. In the case of coral-associated bacteria, one might speculate that heterotrophic Hahellaceae help supply the host with macromolecular nutrients that are not directly assimilated under a complex form. The previous identification of Gammaproteobacteria aggregates within the gastrodermis tissue layer in the digestive cavity of several reef-building corals suggests that these bacteria may play such a role in the coral diet [55]. However, the available data on the 16S rRNA gene sequences of these intracellular bacteria are restricted to the Gammaproteobacteria FISH probe regions, thus precluding phylogenetic affiliation at the genus or family levels and hindering comparisons with Hahellaceae sequences. Further investigations, including in situ localization of bacteria in P. clavata, are clearly required to provide information on their possible role in the functioning of the holobiont. Hopefully, the 16S rRNA gene sequence retrieved from the dominant Hahellaceae ribotype in our study will allow us to design oligonucleotide probes for their specific detection in gorgonian tissues.
While our T-RFLP and clone libraries analysis indicated the existence of a relatively stable Hahellaceae-dominated microbiota in P. clavata populations across a large geographic range and throughout the 4-year survey, we identified a simultaneous and deep compositional shift of communities in summer 2007 at a regional scale. During this summer, the Hahellaceae sequences were not detected by either technique and were transiently replaced by more diverse assemblages with prominent Paenibacillusand Propionibacterium-related sequences, 2 bacterial genera that respectively belong to Firmicutes and Actinobacteria. Members of Firmicutes and Actinobacteria have previously been detected in cold-water and tropical scleractinian corals [4,29,56,57], where they are generally found in low abundance in healthy and/or bleached colonies [50] although representatives of Paenibacillus and Propionibacterium spp. were among the most commonly recovered bacteria from a Brazilian coral species [58]. Because we did not observe a similar transition in any of the other summers that were analyzed, we suggest that an abnormal and transient disruption of the host-Hahellaceae association occurred in summer 2007. However, ordination of the T-RFLP MspI data set revealed that the profiles of all summer 2007 samples grouped with profiles from several colonies sampled in other seasons ( Figure 5B). Significantly, the Paenibacillus TRF (TRF-100.1) was observed together with the dominant Hahellaceae TRF (TRF-105.1) in the latter profiles ( Figure 4). These observations indicate a possible overlap between the 2 communities and suggest that bacteria related to Paenibacillus could be normal but infrequently detected associates, except during periods of anomalous abundance. Therefore, the shift in dominant ribotypes from Hahellaceae to Paenibacillus may result from a transient imbalance of endogenous bacterial populations that naturally reside in the holobiont rather than opportunistic colonization by Paenibacillus from the surrounding water concomitant with a decrease in the Hahellaceae community. The subsequent shift back toward the initial Hahellaceae-dominated community further supports the hypothesis that the holobiont strongly regulates its microbial diversity and indicates that Hahellaceae most likely represent host-specific bacterial associates of P. clavata.
The causes of the transition in P. clavata bacterial diversity from Hahellaceaeto Paenibacillus-dominant ribotypes during summer 2007 are not clear. As mentioned above, the shift occurred at the 3 studied areas in an almost synchronous manner, suggesting the involvement of environmental and/or biological factors acting on a large geographical scale. Several previous studies have identified shifts in coral-associated bacterial communities during bleaching and disease outbreaks or under conditions of environmental stress, such as increased temperature and organic matter enrichment [55,59,60]. For instance, changes in microbial associates were observed in the stony coral A. millepora during a mass bleaching event, and occurred prior to visual signs of bleaching on the sampled colonies [28]. Major shifts in bacterial communities were also observed in diseased colonies of 2 coral species that were affected by White Syndrome disease [29]. Interestingly, the Hahellaceae-related bacteria found in healthy colonies were replaced by other bacterial groups in bleached or diseased individuals. Whether the shift in bacterial assemblages is the cause or the effect of the disease is unclear, but these observations indicate that changes in Hahellaceae-host associations are related to an altered physiological state of the holobiont during stress conditions. Regarding the potential causes of stress within the gorgonian populations in summer 2007, the available temperature time series data recorded at the study areas during the survey period (T-MedNet network; http://t-mednet.org) did not reveal any temperature anomaly. In addition, surveys of P. clavata populations conducted during and after summer 2007 in the 3 study areas did not reveal any symptoms of disease (J. Garrabou et al., unpublished data). Notably, we did not detect the presence of Vibrio coralliilyticus 16S rDNA in the Paenibacillus-dominated summer 2007 clone libraries, although this Vibrio has been implicated in recent disease outbreaks in P. clavata populations during climatic anomalies [22,23]. Several other Vibrio species were also involved in P. clavata tissue necrosis in Mediterranean [17,61], or recovered from healthy and diseased cold-water gorgonians [62]. Altogether, this suggested that unhealthy conditions of the host are allowing colonization by potentially pathogenic Vibrio strains that can also be a natural component of the holobiont and may exploit the disturbance of the normal microbiota [63]. By contrast, we detected only 2 Vibrionaceae sequences among the 523 clones that were analyzed from the summer 2007 libraries (data not shown). This result suggests a very moderate abundance of Vibrio and indicates that the shift observed in P. clavata was likely not related to the onset of a disease. However we cannot rule out that other bacterial groups among the taxon diversity in summer 2007 may represent potential pathogens or detrimental associates that overcome the Hahellaceae-dominated natural assemblage. For instance, although Paenibacillus and Propionibacterium spp. are not considered as coral pathogens and can be normal components of the associated microbiota, their overwhelming dominance in summer 2007 might have resulted in transiently compromised holobiont homeostasis without expression of disease symptoms.
Besides climate-related stresses and disease, regional oceanographic perturbations including land run-off and anthropogenic disturbance have also been shown to cause changes in microbial communities associated with corals [56,64,65]. Degraded ecological conditions may facilitate the existence of alternate states of microbial community structure and overgrowth of opportunistic bacteria [59]. However, an exposure of P. clavata colonies to a pollutant or effluent in summer 2007 is very unlikely considering the large geographical area in which this bacterial shift was observed. In addition, our studies sites are variably impacted by anthropogenic effects as the 3 regions are subjected to different levels of protection. Riou site was located in a non-protected region, a few kilometers away from Marseille, the second major French city. Medes and Scandola sites are both located in marine protected areas, Scandola being a no-take area submitted to very limited human pressure. Therefore, local anthropogenic impacts do not seem to be the cause for the bacterial shifts simultaneously observed in the 3 sites in summer 2007, although a monitoring of environmental parameters and microbial community structure in the water column would be required to better evaluate differences between the sites. Finally, the factors that could have directly or indirectly caused a transient change in the natural bacterial community remain unclear and several other hypotheses cannot be ruled out, such as a bacteriophage infection targeting Hahellaceae and causing microbial mortality [66], or the occurrence of subtle alterations in gorgonian physiology in summer [67] that would disrupt the host-bacteria relationship without macroscopic disease signs.
In conclusion, the present work provides a reliable evaluation of the structure of and variations in gorgonian-associated bacterial communities. To our knowledge, this is the first spatiotemporal study demonstrating that transient but dramatic shifts in the natural baseline of these assemblages can arise in apparently healthy octocorals. In light of our results, it is important to encourage multi-year monitoring to avoid erroneous or incomplete descriptions of the bacterial assemblages, as we might have drawn by investigating the P. clavata microbiota during the phase shift. Although this shift was only observed once in the 4-year survey and thus presumably corresponds to an abnormal or stressful event, it indicates that transient bacterial populations may replace the natural Hahellaceae-dominated community without visible evidence of deleterious effects. Further studies are now required to address the question of whether the phase shift in the microbial community may represent a potential monitoring tool for determining the health state of gorgonians and the ecological condition of their environment. This research may aid in understanding the causes of recurrent mass mortalities and designing effective management strategies to preserve one of the most emblematic Mediterranean species. Figure S1 Phylogenetic relationships between the dominant P. clavata-associated ribotype and its closest relatives. Neighbor-joining tree based on 16S rRNA gene sequences filtered to ,720 aligned nucleotide positions. Representative clones retrieved from P. clavata libraries in winter 2007 and winter 2008 are marked in red, and the GenBank accession numbers of reference strains are shown in parentheses. Branch points supported by boostrap values .50% or .95% (based on 1000 resamplings) are indicated by open and filled circles, respectively. The Gammaproteobacteria Pseudomonas aeruginosa ATCC 10145 was used as an outgroup. Scale bar represents 0.02 changes per nucleotide. (TIF)