The coral Oculina patagonica holobiont and its response to confinement, temperature, and Vibrio infections

Abstract

Background

Extensive research on the diversity and functional roles of the microorganisms associated with reef-building corals has been promoted as a consequence of the rapid global decline of coral reefs attributed to climate change. Several studies have highlighted the importance of coral‐associated algae (Symbiodinium) and bacteria and their potential roles in promoting coral host fitness and survival. However, the complex coral holobiont extends beyond these components to encompass other entities such as protists, fungi, and viruses. While each constituent has been individually investigated in corals, a comprehensive understanding of their collective roles is imperative for a holistic comprehension of coral health and resilience.

Results

The metagenomic analysis of the microbiome of the coral Oculina patagonica has revealed that fungi of the genera Aspergillus, Fusarium, and Rhizofagus together with the prokaryotic genera Streptomyces, Pseudomonas, and Bacillus were abundant members of the coral holobiont. This study also assessed changes in microeukaryotic, prokaryotic, and viral communities under three stress conditions: aquaria confinement, heat stress, and Vibrio infections. In general, stress conditions led to an increase in Rhodobacteraceae, Flavobacteraceae, and Vibrionaceae families, accompanied by a decrease in Streptomycetaceae. Concurrently, there was a significant decline in both the abundance and richness of microeukaryotic species and a reduction in genes associated with antimicrobial compound production by the coral itself, as well as by Symbiodinium and fungi.

Conclusion

Our findings suggest that the interplay between microeukaryotic and prokaryotic components of the coral holobiont may be disrupted by stress conditions, such as confinement, increase of seawater temperature, or Vibrio infection, leading to a dysbiosis in the global microbial community that may increase coral susceptibility to diseases. Further, microeukaryotic community seems to exert influence on the prokaryotic community dynamics, possibly through predation or the production of secondary metabolites with anti-bacterial activity.

Video Abstract

Introduction

Corals form a dynamic meta-organism known as the coral holobiont, which involves a multipartite relationship between the cnidarian host, its endosymbiotic dinoflagellate algae (family Symbiodiniaceae; [1]), a diverse array of prokaryotes (Archaea and Bacteria), viruses, and eukaryotes (fungi and non-Symbiodiniaceae protists) [2,3,4]. Symbiodiniaceae, the primary photosymbiont in corals, has long been recognized for their crucial role in fulfilling most of the host energy needs, facilitating effective calcification and the formation of modern reefs [5]. Prokaryotes also play a pivotal role in the health and fitness of coral holobionts, and, although most of their specific roles and functions remain unknown, they are involved in nutrient acquisition, production of bioactive secondary metabolites, and probiotic mechanisms such as competitive exclusion of pathogenic bacteria [6].

All these microorganisms coexist and interact with their host, maintaining the coral’s health and facilitating its ability to adapt to climate change. When environmental stressors, such as rising temperatures, occur, coral species often lose Symbiodiniaceae algae, leading to coral bleaching due to an energy deficit [7,8,9]. Additional changes can range from the replacement of mutualistic species by commensalism or parasitic species, as well as an increase in potential pathogens such as Vibrio species [10,11,12]. The simultaneous occurrence of these changes makes it difficult to determine whether the variations in microbial composition in thermally stressed corals are a consequence or cause of bleaching.

The variability observed in coral microbial composition across seasons and spatial scales poses a significant challenge in unraveling the specific roles of microbial species within the coral holobiont under natural conditions. Consequently, many studies aiming to reduce complexity conduct experiments under controlled conditions, such as responses to pH gradients, nutrient fluxes, or thermal stress [13,14,15,16]. However, the applicability of data obtained from these controlled settings to natural conditions is still not well-understood. For example, in coral maintained in aquaria, some taxa rapidly decline, or even disappear, while other microorganisms remain unaffected for extended periods [17, 18]. This phenomenon has also been observed in incubations of seawater samples, where a reduction in microbial diversity occurs, accompanied by the replacement of autotrophic microbes with heterotrophs [19,20,21]. Therefore, within confined corals, additional factors, not well-understood and beyond nutrient imbalances or temperature variations, may contribute to the observed variations.

Omics tools have revolutionized coral research, allowing for a deeper understanding of the various partnerships within the coral holobiont. However, metagenomic approaches aiming to describe the entire holobiont face technical challenges, particularly due to the limited availability of genomes for individual holobiont components, especially microeukaryotes. In fact, only a few coral metagenomes have been sequenced, and most of these studies have analyzed specific partnerships separately, such as focusing on only the prokaryotic [22, 23] or the viral community [24].

In this work, we conducted a metagenomic analysis of the entire Oculina patagonica microbiome. This coral, known as an opportunistic colonizer, is currently expanding throughout the Mediterranean Sea coast [25,26,27]. Its expansion is attributed in part to its broad tolerance for varying light levels and trophic conditions [28, 29], as well as its heightened thermal resilience compared to other scleractinian corals [27, 30]. However, O. patagonica is still susceptible to high temperature, which can lead to changes in the coral microbiome, coral bleaching, and, in some cases, irreversible mortality [28, 31,32,33]. In the Western Mediterranean Sea, these coral bleaching episodes have been linked to the presence of two Vibrio species (Vibrio coralliilyticus and Vibrio mediterranei), whose pathogenicity increases when both bacteria infect the coral simultaneously [28].

The main objective of this study is to provide the first characterization of the O. patagonica microbiome and to investigate its changes in response to three different types of stress (confinement, thermal stress, and the presence of two Vibrio pathogens). For this purpose, we first described the O. patagonica holobiont in its natural environment and then compared it with that of corals maintained under controlled aquarium conditions. Our study sheds light on how alterations in the microeukaryotic community affect the prokaryotic community and whether these microbiome shifts result in a dysbiosis that may increase disease susceptibility in corals.

Material and methods

O. patagonica samples

Twenty-five fragments, each approximately 5 cm in diameter, of apparently healthy specimens of the coral O. patagonica, were collected in June 2016 from the Marine Reserve of Tabarca (Mediterranean Sea, 38° 09′ 60.00″ N 0° 27′ 59.99″ E, Spain) and transported to the laboratory. One coral fragment was used to assess the natural holobiont and will be referred to as the Ocu-2016 dataset. The remaining 24 coral fragments were utilized in the aquaria experiment, as described in Rubio-Portillo et al. (2020) [34]. In brief, corals were distributed in two sets of four tanks, with one set incubated at 20 °C and another set at 28 °C. Each set of tanks contained the following treatments: (I) control corals with no Vibrio added; (II) corals inoculated with the Vibrio co-culture; (III) corals with a Vibrio co-culture growing inside a dialysis membrane, and (IV) corals in the presence of monoculture from each of the two Vibrio species growing independently inside a dialysis membrane. Dialysis membranes prevented contact between the coral and the Vibrio cells but allowed the free diffusion of molecules with a molecular weight lower than 100 kDa. Vibrios used in this experiment were V. mediterranei (strain Vic-Oc-097) and V. corallilyticus (strain Vic-Oc-068).

DNA extraction and metagenomes sequencing

Coral fragments were gently washed three times with 50 ml of sterile-filtered seawater (SFSW) to remove any adventitious microbes. Coral fragments were broken into small pieces and centrifuged for 3 min at 2900 g (Labofuge 400R, Heraeus instruments) in order to remove the mucus from the coral. After centrifugation, the coral pieces were crushed in SFSW using a mortar and pestle, the CaCO₃ skeleton was allowed to settle for 15 min, and the supernatant (i.e., the crushed tissue) was collected and used for DNA extraction. DNA from the crushed coral tissue was extracted using the UltraClean Soil DNA Kit (Mo Bio; Carlsbad, CA, USA) following the manufacturer’s instructions for maximum yield.

The O. patagonica sample (Ocu-2016) was sequenced using paired-end reads on an Illumina MiSeq (Fisabio, Spain) and also a HiSeq4000 platform (Macrogen, South Corea), with a total of 69 Gb of sequence data. Reads from metagenomes obtained from the aquaria experiment were published previously in [34] to assess Vibrio abundance in the corals, but the metagenome assemblages reported herein have not been previously published. These metagenomes samples were sequenced using 150 bp single-end reads on an Illumina HiSeq 2000 platform.

Metagenomic analysis

Raw reads were processed by trimming adaptor sequences and low-quality ends (quality score < 30) using Prinseq [35]. Trimmed reads were used to assess the coverage of metagenomic sequencing with Nonpareil [36]. Subsets of ten million reads from each dataset were randomly separated to perform a taxonomic affiliation using Kaiju [37]. “All versus all” comparison was performed using subsets of one million metagenomic reads and BLASTn (cut-off: 95% nucleotide identity and 70% of query cover alignment length). When a match with the same score was found among the different datasets, it was equally counted for all.

Reads were assembled using MEGAHIT [38] for Ocu-2016 metagenome and meta-SPADES [39] for the aquaria experiments. Due to computational reasons, MEGAHIT was chosen to assemble Ocu-2016 libraries and meta-SPADES for the aquaria metagenomes. However, the rest of the bioinformatic pipeline, including the binning of the contigs, was the same for all the assembled metagenomes. Contigs larger than 5 Kb were selected for metagenome binning using two programs, MetaBAT [40] and MaxBin 2.0 [41]. The final optimized bins were obtained with DAS Tool 1.0 [42]. Genome completeness and putative contamination within the MAGs were determined with Anvi’o v.2.1.0 [43]. According to the standards suggested by Konstantinidis et al. [44], the quality of the MAGs was considered acceptable when completeness was higher than 80% and contamination lower than 5%. MAGs were manually checked for consistent coverage and taxonomy across contigs. The post-curated MAGs were subjected to fast pairwise comparison via a Mash algorithm [45] and grouped into primary clusters at 95% average nucleotide identity (ANI). After this de-replication step, taxonomic affiliation of MAGs was obtained using the GTDB-Tk pipeline, which uses a marker gene set to place genomes in the GTDB reference tree [46]. Gene predictions on the individual reads and also from the contigs of the MAGs were carried out using Prodigal [47]. CDS from the contigs bigger than 1 Kb were annotated with EggNOG v.5.0 [48]. Functional annotation of the predicted ORFs from contigs and MAGs was also performed by BLASTp comparisons against the NCBI nr database, Pfam [49], COG [50], and TIGRFAM [51]. To reconstruct the metabolic potential for the MAGs, GhostKOALA was used [52]. The search for genes encoding biosynthesis of secondary metabolites was performed with antiSMASH v.3.0 [53]. For the identification of putative enzymes involved in the breakdown, biosynthesis or modification of carbohydrates, the CDSs from the contigs bigger than 1 Kb were compared against the CAZy database (dbCAN) [54].

Recruitment analyses

To track the relative abundance of each MAG in the different samples, BBMap was used to map metagenomic short reads to MAG contigs [55]. Matches were filtered for single best alignments, using a minimum of 90% query cover alignment length and 95% nucleotide identity of reads mapping against the reference genome (ANIr). In order to remove biases from highly conserved regions and contig edges, the 80% central truncated average of sequencing depth of all bases (TAD80) was used as described previously [56]. MAG abundance in each metagenomic dataset (as percentage of total community) was calculated as the quotient of the MAG’s TAD80 value and the genome equivalents (GE) from MicrobeCensus [57].

In the case of the recruitment analysis of the dinoflagellate genomes against each of the metagenomes, a database containing the following 20 dinoflagellate assemblies was constructed: Breviolum minutum Mf 1.05b.01 (GCA_000507305.1), Symbiodinium microadriaticum (GCA_001939145.1), Symbiodinium microadriaticum (GCA_905231925.1), Symbiodinium microadriaticum AJIS2-C2 (GCA_018327485.1), Symbiodinium sp. clade A Y106 (GCA_003297005.1), Symbiodinium sp. clade C Y103 (GCA_003297045.1), Symbiodinium kawagutii (GCA_009767595.1), Symbiodinium natans (GCA_905221605.1), Symbiodinium tridacnidorum CCMP2592 (GCA_905221615.1), Symbiodinium sp. KB8 (GCA_905221625.1), Symbiodinium linuchaea CCMP2456 (GCA_905221635.1), Symbiodinium pilosum (GCA_905231905.1), Symbiodinium necroappetens (GCA_905231915.1), Amoebophrya sp. AT5.2 (GCA_005223375.1), Amoebophrya sp. A120 (GCA_905178155.14), Amoebophrya sp. A25 (GCA_905178165.1), Polarella glacialis (GCA_905237085.1), Polarella glacialis (GCA_905237095.1), Cladocopium goreaui (GCA_947184155.1), and Amphidinium carterae (GCA_019702695.1). BlastN comparisons of subsets of 10 million reads were performed and all the reads that hit over 90% identity with a coverage over 70% were considered. When a read matched several dinoflagellate genomes equally well, it was considered a match for all genomes using a modification of the BlastTab.best_hit_sorted of Enveomics Toolbox. To calculate the frequency of the dinoflagellates in the metagenomes, the number of hits was normalized against the dinoflagellate size dataset.

Comparison with other environmental datasets

For the MetaFast comparison [58], a subset of ten million reads randomly extracted from each collection was first obtained. The metagenome collections used came from Ocu-2016, samples from the aquaria experiments, together with other metagenomes obtained from different corals (Acropora palmata (SRR8789034), Cyphastrea (Australia, PRJNA364879), Diploira (Mexico, PRJNA364874), Siderastrea (Mexico, PRJNA364873), a sponge (Ircinia ramosa, PRJNA555144), a metagenome from soil (Soil Forest Indian, JGI-Ga0247671), several seawater metagenomes (two from the TARA collections: TARA-009, 5 m, ERR594288 and TARA-009, 55 m, ERR594315) [59]; 4 metagenomes from the MED collection [60]: Med-OCT2015-15 m (SRR5007106); Med-OCT2015-45 m (SRR5007115) and two seawater metagenomes affected by a “bottle-effect,” Med-OCT2015-15 m-7 h (SRR8503606) and Med-OCT2015-15 m-14 h (SRR8503605).

Viral sequence analyses

Viral contigs were included if their length was over ≥ 10 kb, and they were identified as a viral sequence of category 1 or 2 by VIRSorter v1.03 [61]. Also, contigs which were identified as prophages by VIRSorter v1.03 were also annotated with PHASTER [62], and only those contigs identified as intact or questionable prophages were used. To cluster viral sequences into viral operational taxonomic units (vOTUs), CD-HIT was used with a cut-off of 95% average nucleotide identity and 80% alignment fraction threshold for the smallest contig [63]. vConTACT2 0.9.09 was used to compare the viral contigs using the Prokaryotic Viral RefSeq version 201 (with ICTV and NCBI taxonomies). Only those connections with a score > 1 were considered. The nets were visualized with Cytoscape [64].

Statistical analysis

Given the absence of replicates to determine the statistical significance of the observed differences between samples, we employed a resampling approach [65]. For this purpose, each pair of samples was combined into a single sample, and this combined sample was then repeatedly split into two random samples of the same size as the originals, a process that was performed 1000 times. The Bray–Curtis (B-C) distance of the two original samples was then compared with the vector of 1000 distances obtained with the resampling. The fraction of distances greater than or equal to the original was taken as an estimate of the p-value for the hypothesis test with the null hypothesis “both samples come from the same community,” and the alternative “the two samples do not come from the same community.” This fraction thus estimates the probability that, if the two samples came from the same community, the B-C distance would be greater than or equal to that obtained with the two original samples. In all cases, this fraction was 0%, which gives an estimation of p-value < 0.001 that would remain significative after adjusting the whole set of p-values in order to account for multiple testing.

For each pair of samples of interest and for each genus, a proportion comparison test was performed. Depending on the conditions, two types of tests were employed: the parametric chi-square test when the expected frequencies were greater than or equal to 5, and a Monte Carlo chi-square test when any expected frequency was less than 5. Subsequently, p-values for each pair of samples were adjusted using the Benjamini-Yekutieli method, which is ideal for controlling the false discovery rate when the variables are not independent. A genus was considered to have significantly different proportions between samples when adjusted p-value < 0.05.

To evaluate whether the genes encoding the biosynthesis of secondary metabolites were predominantly produced by eukaryotes or prokaryotes, both in Ocu-2016 and in the aquarias, the proportions of genes were compared using chi-square tests and again adjusting the p-values using the Benjamini-Yekutieli method.

All the scripts used for the statistical analysis are available in Supplementary File 1.

Results and discussion

To characterize the coral holobiont community, encompassing eukaryotes, prokaryotes, and viruses, a metagenome of approximately 70 Gb of sequence from a sample of O. patagonica collected during the summer of 2016 in Tabarca (Alicante, Spain) was sequenced. This is referred to as the Ocu-2016 dataset. Additionally, to investigate alterations in these communities in response to heat stress or the presence of Vibrio coral pathogens, we conducted complementary analyses using published metagenomes from Rubio-Portillo et al. [34]. The published metagenomes came from the same coral sample that had been split into several experimental regimes, though the published analysis primarily focused on characterizing the Vibrio assemblages and did not encompass the complete holobiont. For all nine metagenomes, Nonpareil analysis indicated a coverage of approximately 94% for Ocu-2016 and between 50 and 68% for the metagenomes from the aquaria (Supplementary Table 1), indicating a relatively good representation of the biodiversity contained within them.

Oculina patagonica holobiont description

First, to obtain a general overview, a comparison of all O. patagonica metagenomes (including the natural and the stressed samples) was conducted against other published coral and Mediterranean seawater datasets. As expected, samples from seawater clustered separately from the coral samples, which formed a cluster comprising metagenomes from O. patagonica and other corals (Supplementary Fig. 1).

The reads annotated within the Ocu-2016 coral metagenome were distributed as follows: 10.86% belonged to Bacteria, 3.85% to Eukarya, 0.43% to Archaea, and 0.07% to Viruses, with 84.78% remaining unclassified (Supplementary Table 2). To assess whether this pattern was specific to our sample, a similar analysis was conducted using a dataset from the coral Porites lutea (SRR9182857) [23]. The results were comparable, with only 1.4% of the reads identified as eukaryotes and 7.7% classified as prokaryotes (Supplementary Table 2). The low percentage of eukaryotes found suggests that, due to the relatively lower availability of genomes from eukaryotes compared to prokaryotes, some of the unclassified reads may belong to the coral and other microeukaryotic organisms.

O. patagonica associated eukaryotes

Among coral symbionts, Symbiodoniaceae were the first and most important to be recognized, and the mutual transport of nutrients between both taxa has been well described [1]. O. patagonica harbors Symbiodinium species belonging to three different clades (A, B, and C) from which representative genomes are available (Supplementary Fig. 2). Among them, the genome of Breviolum minutum Mf1.05b (also known as Symbiodinium minutum), which belongs to clade B, exhibited the highest read recruitment rates. This result was consistent with previous findings that identified Symbiodinium type B1 as the primary clade in O. patagonica [25]. It is also consistent with prior investigations in other coral species, demonstrating the concurrent association of corals with multiple Symbiodiniaceae, with a prevailing clone [66].

Besides Symbiodinium, other protists were also abundant in the O. patagonica holobiont, with the Evosea, Euglenozoa, and Apicomplexa phyla being particularly prevalent (Fig. 1A). Previous studies have already documented the presence of other protists in the coral holobiont, suggesting that they may play a role in assisting coral hosts in obtaining sufficient nutrients or serve as an additional food source during recovery from stress, such as tissue loss and bleaching events [67]. Among the detected protists were two Choanoflagellate species, Monosiga brevicollis and Salpingoeca rosetta, which were not previously known to be associated with corals. The unexpected occurrence of reads matching the parasitic protist Plasmodium may have been attributed to the presence of the “ap1icoplast,” an organelle in Plasmodium and other apicomplexans [68] that is derived from endosymbiotic cyanobacteria. In other coral reef samples, sequences from apicomplexan-related plastids from Chromera and Vitrella have been consistently detected [69, 70], but given the low recruitment of these genera in Ocu-2016, it is possible that many of the reads matching Plasmodium belong to a yet unknown microalgae plastid(s) present in O. patagonica.

Fig. 1

Microbial taxonomic classification of O. patagonica microbiome inferred from metagenomic reads using Kaiju. Taxonomic classification at the phylum (A), family (B), and genera (C) levels

Numerous non-photosynthetic microeukaryotes associated with corals have been identified, exhibiting diverse roles within the holobiont, ranging from beneficial symbiosis to parasitic relationships, and even acting as primary pathogens [71,72,73]. In this study, most of the non-photosynthetic microeukaryotic reads were classified as Fungi (Fig. 1A). Among them, the most prominently represented phyla were Ascomycota, Basidiomycota, and Mucoromycota. Aspergillaceae and Glomeraceae were the two major families (Fig. 1B), in good agreement with previous results reporting Ascomycetes and Basidiomycetes as the two major groups of fungi associated with corals [72, 74]. Within the Ascomycota, the genera Aspergillus and Fusarium were the most abundant (Fig. 1C). A significant abundance of reads assigned to the genus Rhizophagus (Glomeraceae family) was also detected. This is a beneficial mycorrhizal fungus commonly used as a soil inoculant in agriculture and forest ecosystems to enhance phosphorus uptake [75]. Recruitment analysis of Ocu-2016 reads against the reference genome of Rhizophagus irregularis (DAOM 181602) revealed the presence of reads with similarity to their ribosomal operons and housekeeping genes. Similar results were also obtained using the microbial metagenomes of the corals P. lutea and Acropora palmata (results not shown). These findings suggest that although Rhizophagus species have not been previously described in aquatic environments, related fungi may contribute beneficial traits to corals.

O. patagonica associated prokaryotes

Among the prokaryotic community, bacterial reads greatly outnumbered those assigned to archaea, maintaining a ratio of approximately 15:1 (Supplementary Table 2). The archaeal community (Fig. 1A) was composed of Euryarchaeota (0.2% of the total reads) and Thaumarchaeota (0.1%). The genus Nitrosopumilus exhibited the highest abundance within this family (Fig. 1C). This genus has been previously linked to the crucial ammonium oxidation process occurring within the coral mucus layer [76]. Notably, Nitrosopumilus has been previously detected in O. patagonica, and its presence has been exclusively associated with healthy colonies [77]. Also in accordance with previous findings obtained by 16S rRNA gene analyses [77], the bacterial reads were primarily composed of Proteobacteria (4.3%), Actinobacteria (1.6%), Firmicutes (1.2%), and Bacteroidetes (1.1%), with Streptomyces, Pseudomonas, and Bacillus as the most abundant genera (Fig. 1).

Binning of the assembled contigs from the Ocu-2016 metagenome resulted in five metagenome-assembled genomes (MAGs), which belonged to Desulfobacterales (MAG1-Ocu2016), Flavobacteriales (MAG2-Ocu2016), Holosporales (MAG3-Ocu2016), Parvularculales (MAG4-Ocu2016), and Rhizobiales (MAG5-Ocu2016) (Supplementary Table 3). These MAGs recruited less than 0.001% of the metagenomics reads, with ANIr values over 99% (Supplementary Table 4), suggesting that only MAGs from low-abundance bacteria displaying a very low intra-population diversity could be retrieved. Excluding MAG1, the average genome size for the other four MAGs from the Ocu-2016 dataset was 1.7 ± 0.5 Mb, a size small enough to be consistent with a host-associated lifestyle, reflecting the loss of non-essential genes [78]. For example, MAG3-Ocu2016 (Candidatus Hepatobacter penai; 87.3% complete) lacks key biosynthetic pathways for essential amino acids and shows incomplete synthesis of purines and pyrimidines (Supplementary Table 5). However, it harbors genes for biotin (vitamin B7) synthesis. This could be crucial for the coral holobiont, as both corals and Symbiodinium are suggested to be auxotrophs for different B vitamins, potentially obtaining them from associated bacteria [79, 80].

Furthermore, within the MAGs, genes encoding mechanisms for stable symbiosis with the host, such as ankyrin repeats proteins (ARPs), were identified (Supplementary Table 3). ARPs are common protein interaction motifs that modulate intracellular processes, promoting stable symbiotic or pathogenic associations [81]. Previous analysis of ARP distribution in microbial genomes showed that species dedicating more than 0.2% of their protein-coding genes to ARPs are typically obligate intracellular or facultative host-associated species [82]. In this case, MAG3-Ocu2016 and MAG4-Ocu2016, both from the Alphaproteobacteria class, surpass this percentage. Moreover, in the case of MAG4-Ocu2016 (unknown Parvularculales), these proteins were affiliated with other corals and sponges, suggesting its potential as a symbiont for marine invertebrates.

MAG1-Ocu2016 represents a putatively diazotrophic dissimilatory sulfate-reducing bacterium (SRB) belonging to the Desulfobacter genus (Supplementary Table 5). The identification of a N₂-fixing SRB, not previously described in corals, carries substantial implications for coral communities that commonly inhabit nutrient-depleted environments. These bacteria may enhance the coral’s ability to efficiently convert gaseous nitrogen into a usable form such as ammonia, a vital process for sustaining a steady nitrogen supply for Symbiodinium-based primary production within corals [83].

O. patagonica associated virus

A total of 275 different viral operational taxonomic units (vOTUs) were retrieved from Ocu-2016 metagenome, which accounted for 0.08% of the metagenomics reads. Only 14% of these vOTUs exhibited similarity to known reference genomes, and most of them (86%) were classified as dsDNA phages belonging to the class Caudoviricetes (Supplementary Table 6). A phage phylogenetic network using phage reference genomes (Fig. 2) showed that most of the phages identified in the O. patagonica holobiont were new. Others grouped mainly with phages infecting Pseudoalteromonas, Pseudomonas, and Roseobacter, three bacterial genera consistently associated with corals. Three vOTUs presented a partial similarity to Suoliviridae sequences (Crassvirales), with 30 to 63% similarity to a major head protein, DNA polymerase or hypothetical proteins. Although marine crassviruses have been previously described [84], O. patagonica assembled crassviruses-like clustered separately from them and may thus represent novel lineages.

Fig. 2

vConTACT clustering of O. patagonica phages and prophages and related prokaryotic Viral RefSeq genomes (version 201). Color nodes represent the Oculina predicted viral sequences and black ones the RefSeq virus. Edges represent the vConTACT-generated similarity score between each pair of viruses (only similarity scores of ≥ 1 are included in the network). Highly similar viruses are positioned close together. Only reference viruses that are connected to ≥ 1 predicted phage are included in the network

Five of the identified vOTUs were assigned as eukaryotic viruses belonging to the Phycodnaviridae family within the Nucleocytoplasmic Large DNA Viruses (NCLDVs) (Supplementary Table 6). The family Phycodnaviridae infects phytoplankton and has been previously detected in both heat-stressed corals and cultures of Symbiodinium spp., suggesting that these viruses may have a role in the destruction of algal symbionts or the dysfunction of symbiont–host mutualism, although the extent of such infections is unknown [85,86,87].

O. patagonica holobiont changes under stress conditions

When the Ocu-2016 dataset (the natural holobiont metagenome) was compared with a collection of the metagenomes derived from the stress experiments, it clustered with the coral samples maintained at 20 °C, and apart from the 28 °C samples (Supplementary Fig. 1). Read-level analysis further supported these findings, with Ocu-2016 sharing 46.9% of reads with C20, only 15% with C28, and lower similarities were observed when compared to the Vibrio infection datasets (Fig. 3). Genus-level analysis using a resampling approach showed that coral microbiome composition significantly changed when corals were exposed to the different stress tested in the experiment (Supplementary Table 7). Further, Shannon diversity index was significantly higher in corals maintained in the aquaria compared to the Ocu-2016, mainly in corals under heat stresses and in corals in the presence of vibrios compared to the control ones (pairwise Hutcheson t-tests, all adjusted p-values 0, Supplementary Table 1). All these results highlight significant shifts in the microbial community of O. patagonica under heat stress, particularly in the presence of Vibrio species. These community shifts in the aquaria confinements were mainly enrichments of heterotrophic microbes. This phenomenon, known as the “bottle effect,” has been previously observed when seawater samples or corals are kept under prolonged confinement conditions [18, 88].

Fig. 3

Sequences shared between metagenomes determined using an “all versus all” comparison of metagenomic reads. Numbers indicate the percentage of shared sequences among datasets

O. patagonica holobiont changes due to aquarium confinement

One of the most extensive effects of confinement was the decrease in the proportion of reads assigned to Eukarya within the metagenome from the coral maintained at 20 °C (from 3.85 to 2.2% compared to Ocu-2016), and the increase of bacterial reads (from 10.85 to 17%). A differential abundance test was used to identify genera with different proportions in Ocu-2016 compared to the coral maintained at 20 °C (Supplementary Table 8). This analysis showed that in the aquaria a decline in Symbiodinium, as well as other microeukaryotes and prokaryotes compared to Ocu-2016. Meanwhile, an expansion of the (“natural”) rare biosphere was observed under confinement conditions (Fig. 4). Microeukaryotes that proliferated in the aquarium experiments were mainly members of the Saccharomycodes genus, which were not detected in Ocu-2016 (Fig. 4). Most of these reads showed similarity to the yeast Saccharomycodes ludwigii, with some matching Hanseniaspora. The presence of these sugar-consuming microorganisms may be attributed to the increased carbohydrate content in the mucus of stressed corals [89].

Fig. 4

Prokaryotic genera that showed significant different proportions among treatments. C, control corals; I, infected corals; MS, corals in the presence of Vibrio monocultures; MX, corals in the presence of Vibrio co-culture. The numerical annotations 20 and 28 refer to corals maintained at 20 °C and 28 °C, respectively

Also under confinement conditions, there was a notable increase in certain prokaryotic genera, including potential human pathogens like Acinetobacter, Bordetella, Neisseria, Klebsiella, and Salmonella, which were not detected in Ocu-2016 (Fig. 4). This shift could be linked to changes in the microeukaryotic community associated with the coral. It has been suggested that microeukaryotes play a role in regulating microbial communities within corals through processes like phagocytosis and the production of antimicrobial compounds [2, 70, 72]. Thus, a search for genes encoding the synthesis of secondary metabolites was conducted to investigate the potential antimicrobial activities of the O. patagonica microbiome. Genes encoding polyketide synthases (PKSs) and multienzymatic nonribosomal peptide synthetases (NRPSs) were detected. To determine whether they were predominantly produced by eukaryotes or prokaryotes, we compared the proportions of these genes in both the natural sample and the aquarium samples using the chi-square test. In the natural sample (Ocu-2016), these genes were found almost exclusively in eukaryotes (e.g., Symbiodinium, Scleractinia, Blastocladiomycota, and Dictyosteliales), with no bacterial PKSs detected (Fig. 5). Conversely, in the aquarium-maintained corals, these genes were primarily produced by prokaryotic microorganisms. Statistically significant differences were observed between the proportions of eukaryotic and prokaryotic genes (adjusted-p-value < 0.00005), except for the samples C20 and MS20, where the differences were not statistically significant (adjusted-p-value > 0.05). The absence or reduction of these natural eukaryotic PKS could potentially lead to an increase in the proliferation of fast-growing organisms, as discussed below. This would agree with previous results indicating that the coral-associated fungi that decreased under confinement conditions, such as species from Aspergillus or Fusarium, displayed antimicrobial activities against human pathogenic bacteria [90, 91].

Fig. 5

The taxonomic origins of the predicted polyketide synthase (PKS) clusters in O. patagonica. A Eukaryotic clusters and B bacterial clusters detected. C, control corals; I, infected corals; MS, corals in the presence of Vibrio monocultures; MX, corals in the presence of Vibrio co-culture. The numerical annotations 20 and 28 refer to corals maintained at 20 °C and 28 °C, respectively

O. patagonica holobiont changes due to thermal stress under experimental conditions

The proportion of annotated eukaryal reads in the metagenome from the coral maintained at 28 °C remained similar to that at 20 °C. However, there was a substantial increase in the bacterial reads, from 17 to 40%. The number of detected bacterial families also increased, from 112 at 20 °C to 171 at 28 °C (Supplementary Table 2). The differential abundance approach was used to identify genera with different proportions in corals maintained at 20 °C and 28 °C (Supplementary Table 8). This analysis identified various genera from the Rhodobacteraceae family, including Marivitia, Ruegeria, Loktanella, and Yoonia, as bacterial bloomers under heat-stress conditions (Fig. 4). In line with our results, the Rhodobacteraceae family has been recently suggested as an indicator species for thermal stress in corals [92]. Furthermore, a significant number of Rhodobacteraceae genera detected in our study are known to be involved in the breakdown of organic sulfur compounds like dimethylsulfoniopropionate (DMSP) and dimethylsulfide (DMS). The production of these compounds increases in corals under thermal stress [93,94,95]. In good agreement, genes responsible for DMSP catabolism (dmdA, dmdB, dmdC, dddD, dddP, and dddL) increased in abundance in corals kept in aquaria, particularly in corals maintained at 28 °C (from 0.00001% in Ocu-2016 to 0.0001–0.001% of reads). Furthermore, two MAGs recovered from the aquaria metagenomes (MAG9-MS28 and MAG10-MX28) corresponded to two Rhodobacteraceae genera potentially involved in DMSP metabolism, Pelagibaca and Yoonia, respectively (Supplementary Table 3). In fact, an orthologue gene encoding the DddL enzyme, responsible for DMSP catabolism, was detected in a Yoonia-related MAG, with a more pronounced increase in coral samples maintained at 28 °C (Fig. 6). This suggests that sulfur compounds produced in response to stressors, like confinement or thermal stress, may contribute to shaping coral-associated bacterial communities. This supports the hypothesis proposed by Raina et al. [96] that DMSP and DMS play a pivotal role in structuring coral-associated bacterial communities.

Fig. 6

Relative abundances of MAGs. The relative abundance of each MAG was estimated using fragment recruitment analyses carried out by BLASTn comparisons. Only reads that matched with over 95% identity and 70% coverage were considered. The fraction of nucleotides mapping to the respective MAG was normalized by the length of that MAG and size of the metagenome

Other genera that increased under confinement conditions and particularly under heat stress were Marinifilum, a member of the Bacteroidetes genus, and Halodesulfovibrio, originally affiliated with Desulfovibrio genus (Fig. 4). These genera have previously been observed in coral samples, and their proliferation under aquarium conditions has been documented by [97]. Marinifilum may play a role in the sulfur and carbon cycle, as well as lipid catabolism. On the other hand, Halodesulfovibrio has been identified as a secondary pathogen responsible for initiating and progressing black band disease in coral hosts, producing sulfide as a product of dissimilatory sulfate reduction [98].

The increase of bacterial bloomers under heat stress may be partially influenced by the recycling of metabolic waste products within the holobiont [99] and subsequent increase of available nutrients. Additionally, corals regularly release mucus into the surrounding seawater, which carries elevated concentrations of nutrients like organic carbon [100]. In heat-stressed corals, an increase in carbohydrate content in mucus has been detected [89], thus serving as a carbon source for heterotrophic microbes. To explore this idea, variation in the abundance of genes encoding carbohydrate-active enzymes (CAZymes) was examined. Compared to Ocu-2016, there was an increase in the proportion of glycosylhydrolases, glycosyltransferases, and carbohydrate-binding modules in corals maintained under experimental conditions, particularly in the thermally stressed ones (Supplementary Fig. 3). This suggests that mucus released under stressful conditions could provide newly accessible nutrients, serving as sustenance for fast-growing microbes. This process could potentially be enhanced by the decrease of other bacterial taxa and microeukaryotes under stress conditions, which may act as microbial regulators within the coral holobiont, as explained below.

Regarding viruses, it is well established that virus-like particles (VLPs) increase in corals during stressful conditions or bleaching events [85, 101]. In this case, recruitment analysis showed that one of the viruses, identified as Phycodnaviridae in the Ocu-2016 metagenome, clearly increased in abundance under thermal stress (Supplementary Fig. 4) when Symbiodinium abundances decreased. This fact suggests that this virus could be involved in the destruction of algal symbionts or the dysfunction of symbiont–host mutualism as mentioned above.

O. patagonica holobiont changes due to Vibrio infection under experimental conditions

All the above-described changes observed in confined or heat stress were enhanced in corals exposed to Vibrio cells (V. coralliilyticus and V. mediterranei), either directly or indirectly. Specifically, corals at 28 °C in the presence of Vibrio pathogens experienced a more significant increase in Vibrio species abundance compared to controls. Accordingly, a previous transcriptomic study carried out with these samples [34] revealed that pathogenic Vibrio release quorum sensing molecules, triggering alterations in coral-associated bacteria and an increase in other potential pathogens already present in the coral sample, thus explaining the increase of Vibrio spp. detected here. Furthermore, the increase in genera related to DMSP metabolism, such as Marivitia, Ruegeria, Loktanella, and Yoonia, was also more pronounced under the presence of Vibrio pathogens, which also suggests vibrios play a key role in coral microbiome modulation (Figs. 4 and 6).

Analysis of the viral contigs in these coral metagenomes revealed a significant decrease in the proportion of genomes classified as prophages in corals experiencing heat stress (0.002–0.0004%) and in those exposed to Vibrio coral pathogens (0.001–0.0005%) compared to the natural sample (0.03%) (Fig. 7A). To delve into the lysogenic dynamics within the coral holobiont, recruitment analyses were performed using only prophage genomes identified within bacterial contigs; this approach confirmed their integration into their respective bacterial hosts. Our data unveiled the induction of three distinct prophages in corals subjected to Vibrio infection (I20) or heat-induced stress (C28), and two of these prophages were found to be integrated within contigs belonging to the Rhodobacterales order (see Fig. 7B). These findings align well with recent research suggesting that coral pathogens, such as V. coralliilyticus, produce hydrogen peroxide to initiate the lytic cycle of prophages in their competitors, thereby providing the coral pathogen with an advantage by reducing competition during coral colonization [102]. This mechanism further contributes to host dysbiosis by shifting the balance from symbionts to pathobionts as observed in the human gut [103].

Fig. 7

Fragment recruitment plots for metagenome sequence reads on bacterial contigs with prophages (marked in grey in each contig) from each coral. The vertical axis indicates the sequence identity of an alignment between a metagenomic sequence and the reference contig using BLASTn, the identity ranges from 100% (top) to 95% (bottom). *Prophages induced from bacterial host upon heat or Vibrio infection stress

Conclusions

Taxonomic characterization of the O. patagonica holobiont showed a highly diverse microeukaryotic community. This, in addition to the well-known Symbiodinium, includes fungi from the genera Aspergillus, Fusarium, and Rhizophagus. Among prokaryotes, the most abundant genera found were the archaea Nitrosopumilus together with the bacteria Streptomyces, Pseudomonas, and Bacillus.

Our aquaria experiments were designed to address changes on coral microbiome related to different stressors. Importantly, many of these microeukaryotic symbionts suffered a significant decline (below detectable levels) in O. patagonica, particularly under heat stress. Concurrently, a notable increase in the abundance of the prokaryotic community was observed, shedding light on the complex interplay between these two communities. Our data suggest that the microeukaryotic community potentially exerts influence on the prokaryotic community dynamics, possibly through predation or the production of secondary metabolites with anti-bacterial activity.

The increase in the contribution of the prokaryotic community to the holobiont appears to coincide with an increase of CAZymes and genes responsible for DMSP catabolism. This correlation suggests the existence of a potential recycling mechanism for organic products generated by the holobiont under stress conditions. Thus, the alteration of conditions (such as temperature, confinement, and pathogen presence) within the coral holobiont appears to enhance the production of enzymes dedicated to carbohydrate degradation. This discovery underscores the dynamic and adaptive nature of microbial communities within coral ecosystems and offers exciting prospects for further research into the exploitation of these enzymatic resources in various industrial and environmental contexts.