Microbiomes of gall-inducing copepod crustaceans from the corals Stylophora pistillata (Scleractinia) and Gorgonia ventalina (Alcyonacea)

Corals harbor complex and diverse microbial communities that strongly impact host fitness and resistance to diseases, but these microbes themselves can be influenced by stresses, like those caused by the presence of macroscopic symbionts. In addition to directly influencing the host, symbionts may transmit pathogenic microbial communities. We analyzed two coral gall-forming copepod systems by using 16S rRNA gene metagenomic sequencing: (1) the sea fan Gorgonia ventalina with copepods of the genus Sphaerippe from the Caribbean and (2) the scleractinian coral Stylophora pistillata with copepods of the genus Spaniomolgus from the Saudi Arabian part of the Red Sea. We show that bacterial communities in these two systems were substantially different with Actinobacteria, Alphaproteobacteria, and Betaproteobacteria more prevalent in samples from Gorgonia ventalina, and Gammaproteobacteria in Stylophora pistillata. In Stylophora pistillata, normal coral microbiomes were enriched with the common coral symbiont Endozoicomonas and some unclassified bacteria, while copepod and gall-tissue microbiomes were highly enriched with the family ME2 (Oceanospirillales) or Rhodobacteraceae. In Gorgonia ventalina, no bacterial group had significantly different prevalence in the normal coral tissues, copepods, and injured tissues. The total microbiome composition of polyps injured by copepods was different. Contrary to our expectations, the microbial community composition of the injured gall tissues was not directly affected by the microbiome of the gall-forming symbiont copepods.


Results
Overall, about 150,000 reads were obtained per sample (standard deviation 84,500) after chimaera and error checking (Table 1), comprising 54,329 OTUs at the threshold similarity level of 0.987 after removal of singletons and normalization of read numbers by the analysis of rarefaction curves ( Figure S1).
The diversity of the prokaryotic communities expressed as three different metrics of alpha diversity, namely the number of OTUs per sample (ranging from 4,732 to 17,730), the Shannon-Wiener index, and the Simpson index, was not statistically different between the healthy and diseased tissues, and symbiotic copepods for both coral species, nor between the species ( Table 2).
The community composition of the samples was significantly different between the coral host species (ADONIS on OTU-based Bray-Curtis distances: F = 3.8, p = 0.001, R 2 = 0.21), with OTU-based Bray-Curtis distances between the species ranging from 0.64 to 0.96 and within the species from 0.48 to 0.93. No significant differences were observed between the types of substrate for the microbiome (healthy tissues, gall tissues, and symbiotic copepods) (F = 1.4, p = 0.073, R 2 = 0.15), with distances between the substrates ranging from 0.48 to 0.93, and within from 0.51 to 0.84. The level of similarity between the microbiome types was the same for both coral host species (ADONIS interaction term: F = 1.4, p = 0.074, R 2 = 0.15).
Upon considering in detail individual coral colonies, the results were confirmed. The microbial composition of the diseased tissues in Gorgonia ventalina was neither significantly different from the healthy tissues in the same coral colony (ADONIS on Bray-Curtis distances: F = 0.9, p = 0.7, R 2 = 0.31), nor from the microbiome of the symbiotic copepod (F = 1.2, p = 0.3, R 2 = 0.28). Similarly, the microbial composition of the diseased tissues in Stylophora pistillata was neither significantly different from the healthy tissues in the same coral colony (F = 1.2, p = 0.3, R 2 = 0.29), nor from the microbiome of the symbiotic copepod (F = 1.3, p = 0.2, R 2 = 0.30). Yet, a plot of the differences between the samples at the genus level suggests some potential effect of the microbiome of the symbiotic copepod on the gall tissues, at least for one gall tissue of Stylophora pistillata that clusters with the microbiomes of the symbiotic copepods (Fig. 2). Among three copepod samples of Gorgonia ventalina, one clustered with coral samples, both gall and healthy, and two formed a separate cluster. Thus, the microbiomes were highly variable, with an uncertain effect of the symbiotic copepods on the gall tissues. The main significant difference between the microbiomes was between the coral species systems in two oceans, dominated by different bacterial phyla (Fig. 3). Actinobacteria, Alphaproteobacteria, and Betaproteobacteria were more prevalent in the samples from Gorgonia ventalina in the Caribbean, while Gammaproteobacteria dominated in the samples from Stylophora pistillata in the Red Sea. At the genus level, the PCA visualization confirmed the difference between the coral systems but not by the substrate type (Fig. 4). The main difference between the samples was in the prevalence of Algicola and some unclassified bacteria and Gammaproteobacteria in all samples from Stylophora pistillata with additional very abundant Oceanospirillales family ME2 in the gall and copepod samples, while the samples from Gorgonia ventalina were rich in Propionibacterium and unclassified Microbacteriaceae.
The difference in the microbial composition among the copepod samples and the coral samples was mainly generated by Endozoicomonas, which was present in all coral samples, and was the dominant taxon in most normal coral samples from the Red Sea, while it was absent or a minority in all copepod samples. Instead, the copepod samples from Gorgonia ventalina were enriched in Propionibacterium and unclassified Microbacteriaceae. On the other hand, the copepod samples from Stylophora pistillata together with the gall tissues were enriched in ME2 of Oceanospirillales, and the latter was the only taxon that clearly distinguished the copepod and gall samples from the healthy coral samples. No such predominant taxa were observed in the samples from Gorgonia ventalina.

Discussion
Microbiomes of corals and copepods are objects of numerous studies; however, the interactions between copepods and coral microbial communities are poorly understood and have not been studied in detail. The definition of a core healthy coral microbiome meets numerous challenges, the most critical ones being that coral microbial communities are temporally and spatially dynamic 8 , with different coral species possessing different microbiomes 36   Here we studied bacterial communities of two different copepod-coral associations, focusing on the identification of bacteria possibly involved in the gall formation. These two systems have different microbiomes. The microbiomes of normal corals, galls, and copepods within these systems differ less than between the systems. While we could not distinguish between the influence of geographical location and species-specific host-microbiome interactions due, in particular, to a small number of samples and studied systems, we found that in all metrics, and contrary to initial expectations, the gall samples had a microbiome biodiversity similar to that of the regular coral samples, but with a different microbiome structure. The microbiomes of the normal coral samples from the Red Sea were found to be similar to those previously described in literature 35 and showed enrichment with widespread coral-associated Endozoicomonas bacteria, while the gall samples were enriched with bacteria unusual for regular Stylophora, like the family ME2 of Oceanospirillales, or by potential pathogens like Rhodobacteraceae. The former is normally absent in Stylophora, but is one of the dominant taxa in the octocoral Corallium rubrum 69 . Similarly, the gall samples from the Caribbean were enriched with potential pathogens like Arcobacter and Pseudoalteromonasas known to be associated with injured tissues of corals and algae 71,72 . Oceanospirillales family ME2 in the Red Sea was not only present in galls but also in all copepods, while absent in the regular coral tissue, which may indicate that these bacteria can be transmitted by copepods and expand in gall tissues. This is reminiscent of the Roseobacter prevalence in the microbiomes of the Cryptochiridae crabs and white plague coral microbiomes 60 .
Regardless of the large variability in the analyzed microbiomes, no clear evidence of a role of the microbiome associated to the symbiont copepod was found in affecting the microbiome of the gall tissue of the coral. We cannot rule out an influence of the microbiome of the symbiont in the induction of the gall tissue, but its effect is not visible in the microbiome of fully formed coral galls. We acknowledge that our study involved a limited number of samples and of analysed systems, and further studies could still provide evidence of a role of symbiotic copepods in causing or facilitating the spread of disease to corals.  Table 1 and Fig. 2.  Table 1  Each coral was photographed underwater, placed in a separate plastic bag and brought to the surface. The parts of corals with galls were dissected with sterilized needles in sterilized Petri dishes using dissecting microscope Olympus SZX 7, then rinsed several times and preserved in a 95% solution of ethanol. One copepod individual found in the gall was selected per gall of the coral colony. The copepods have been rinsed in ethanol, some of the copepods present in the samples have been inspected by scanning electron microscopy (SEM) in order to detect the presence of microbes.
Scanning electron microscopy. For scanning electron microscopy (SEM) analyses, copepods were dehydrated through graded ethanol concentrations, critical point dried, mounted on aluminum stubs, coated with gold, and examined in a JEOL scanning electron microscope at the Laboratory of Electron Microscopy (Biological Faculty of Lomonosov Moscow State University) 75 .
DNA extraction and 16S rRNA gene sequencing. DNA from the ethanol-preserved copepods, galls, or a normal coral tissue was extracted simultaneously using a standard silica-based DNA extraction kit (Diatom DNAprep 100, Isogene, Moscow, Russia). The DNA extraction was conducted according to the manufacturer's protocol for the fresh blood samples.
Negative controls with PCR grade water occurred without amplification. PCR products were quantified in the QIAxcel (QIAGEN). PCR products were pooled in an equimolar concentration. The pool was cleaned using AMPure magnetic bead-based purification system (Beckman Coulter). The clean pool was quantified using the Bioanalyser (Agilent). The amplicon library was sequenced using an Ion 314 Chip by an Ion Torrent Personal Genome Machine (Life Technologies) at the Naturalis Biodiversity Center following manufacturer protocol. Sequence analysis. The quality of reads was analyzed with FastQC 77 . Long reads were trimmed to 300 bp, first 10 bp and low quality (Phred <20) ends of reads were trimmed, and then reads shorter than 140 bp were removed with Trimmomatic 78 . On average 8.5% of reads were removed. For OTU definition we used CD-HIT-EST 79 with similarity level 0.987 that was more restrictive than the commonly used 0.97 threshold and was more likely to group together only reads from the same species 80,81 . All single-read OTUs were filtered out (approximately 97500 OTUs accounting for 4.3% of reads). We used two common approaches for accounting for different sequencing depth between samples: (1) normalizing OTU sizes by dividing them by the total number of reads in each sample, and (2) construction of rarefied samples that contained equal numbers of reads by random sub-sampling of the reads. Both methods produced similar results and following 82 we used normalized OTUs for the beta-diversity analysis. Since some methods for estimating alpha-diversity require absolute numbers, we used rarefaction for all such analyses. Representative sequences from each OTU were scanned for possible chimeras with DECIPHER 83 and 3056 minor OTUs (that contained less than 2.5% of all reads) were marked as such. We assigned taxonomy to each read the Mothur software package 84 with standard parameters and SILVA (version 128) as the reference database 85 . If at least 75% of reads from an OTU shared the same taxonomy, it was transferred to the whole OTU. All OTUs classified as chloroplast or eukaryote-related were removed. Rarefaction curves, alpha and beta diversity were calculated using the package vegan for R 86 . To estimate the alpha diversity, we used the Shannon-Wiener index, the Simpson index, and the observed number of species. For the beta diversity analysis, we used the Brain-Curtis dissimilarity and the Hellinger distance, the latter better accounting to low-abundant OTUs. The hierarchical clustering and principal component analyses were performed with built-in R functions based on the OTU or taxon distribution between the samples. Significance of differences between the microbiome compositions was tested with ADONIS implemented in the package vegan for R. The models included the coral holobiont species, the type of substrate for the microbiome (healthy tissue, gall tissue, or symbiont copepod), and the statistical interaction between the coral species and the substrate type.
Data availability. Sequence data determined in this study are available at NCBI under BioProject Accession PRJNA433804 (http://www.ncbi.nlm.nih.gov/bioproject/433804). Other data are available in the Supplementary Data file.