Cold-water Coral Microbiome and Environmental Microbial Communities in a Remote NE Atlantic Submarine Canyon Setting: Microbial Diversity, Coral Health and Prospects

In the Porcupine Bank Canyon, Lophelia pertusa and Madrepora oculata are the main framework-forming corals producing three dimensional structures which provide a home for a range of benthic fauna and microbial communities. To understand the roles and functions that microbes perform in coral health in the Porcupine Bank Canyon, three groups of samples (corals, sediment and water) were collected between 600–800 m depth. DNA was extracted from these samples and metabarcoding was performed on the V3-V4 region of the 16S RNA gene using Illumina technology. The coral microbiome showed greater microbial diversity than both the surrounding sediment and water communities. The genera Pseudomonas, Pseudoalteramonas and Photobacterium were the bacterial communities conserved at 100% coverage of coral samples whereas at the order-level classication Clostridiales, Bacteroidales, Flavobacteriales, Rhodobacterales and Rickettsiales were in high abundance in all the coral samples. A disproportionate distribution of probiotic and pathogenic bacterial groups at the different levels of classication was observed on the corals. Corals do not appear, at present, to be stressed by climate induced changing environmental conditions in the upper Porcupine Bank Canyon. Overall, the corals in the Porcupine Bank Canyon are in a healthy state despite the detection of pathogenic bacterial groups. However, the current trend of climate change and subsequent deep-sea warming could shift the bacterial composition towards a more dominant pathogenic bacterial community, with serious implications for coral health and stability of this important ecosystem.


Introduction
Scleractinian corals are globally distributed and include cold-water and tropical corals 1 . Cold-water corals (CWCs) are greater in abundance in the northeast Atlantic 2 , including in submarine canyons, and provide several niches inhabited by diverse microbes such as archaea, bacteria as well as protists 3 .
Microbes live in the mucus, the polyp tissues and on the coral skeleton as well as in the ambient environment of the corals 4 . Microbial diversity and its interaction with the host and the environment facilitates identi cation of important linkages existing between the microbial communities and macroecological change 5,6,7 . Climate change, sedimentation, nutrients, pollution, salinity and temperature 8 disrupt the normal functioning of bene cial microbes 9 by altering the microbial community during stressful conditions 10,11,12 . Changes in coral physiology affect the rate of mucus production and its biochemical composition 13,5 , thereby favouring invading pathogens over resident microbes 14 . Also, while microbial assemblages support the growth and distribution of CWCs 15 , some are involved in biogeochemical processes, nutrients recycling and the breakdown of organic and waste products 16,17,18 .
The primary framework-forming coral in the Porcupine Bank Canyon (PBC) in the northeast Atlantic is the scleractinian coral Lophelia pertusa 19 . The distribution of L. pertusa (syn. Desmophllum pertusum 20 ) colonies in the northeast Atlantic is in uenced by microbes that live on them 21 as research suggests that microbes living on Lophelia and other CWCs potentially bestow on them probiotic properties 22,23 . While there are differences in the coral-associated microbial communities of living, dead and fossilized Lophelia and the ambient environment 24,25,26,27 , no divergence has been observed between the different colour morphotypes of L. pertusa 6 . Also, species-speci c distribution of cold-water coral-associated microbes has been reported 6,18 , whilst Lawler et al. 17 has noted that microbial diversity is conserved across different species. Cold-water coral-associated microbial communities are involved in the formation of carbonate mounds 28 . According to Templer et al. 28 , these microbes bind mound sediment together through carbonate precipitation, which eventually help preserve the mound structure. Generally, research on the contributions of microbial communities on and/or within CWCs is limited. As a result, our understanding of the roles that microbes play in the ecology of CWCs are insu cient, making it imperative that we assess the essential roles that microbes play in the processes that drive changes in coral distributions in the deep-sea. Different coral sampling strategies yield different results in microbial diversity 29,30 . Microbial gene sequencing is relatively fast and e cient and can be used to screen microbes associated with corals. The 16S rRNA gene is the most widely used sequence for microbial community pro ling, although the use of the internal transcribed spacer (ITS) region has also been reported 31 . The challenge with molecular techniques is that non-target genes can be ampli ed even when species-speci c primers are used. Several culture-independent techniques have been used to characterise cold-water coral-associated microbial communities 21,28,6,14,17 . Culture-based methods have also been applied to evaluate coralassociated microbes 32,33,34 although this approach of microbial community pro ling is in uenced by the type of medium used 29 . Also, marine microbes are generally di cult to grow on a medium. As such, a combination of both culture and culture-independent techniques has been advised as the use of only one of the methods in microbial studies can be limiting 29 , 34 .
Next generation sequencing (NGS) is an advanced deep-sequencing technique and its application in coldwater (speci cally in Lophelia pertusa and Madrepora oculata) coral-associated microbial community analysis is rather limited 22,35,16,36 . NGS is a high throughput technique which can sequence in parallel, millions of fragmented DNA from a single sample 37 . The NGS technique can detect microbes without prior knowledge of the target organism and discover novel organisms 38 , making its application suitable for environmental DNA samples 39 . Therefore, the present study aims to: 1) characterise the composition and diversity of microbial community in the PBC and, 2) determine the distribution of microbial communities in the corals, sediment and the ambient water in the PBC.

Results
In all, thirteen samples were successfully sequenced (seven coral and three each of water and sediment) ( Tables 1 and Supplementary Table S1). A total of 2,586,928 raw sequences yielded 529,513 effective tags (across 13 samples) for sequence analysis after processing and quality ltering of raw reads (Table   1). A total of 4,494 sequence variants with mean length of 190 were observed. Also, the least ASVs was observed in HF5 (20,588 sequences) whilst the highest ASVs was observed in FF3 (75,473 sequences) after quality ltering with Dada2 (Table 1). A total of 522,608 ASVs were observed across the thirteen samples (Table 1). Among the three groups of samples, corals contained 272,411 ASVs, sediment 161,726 ASVs whereas water yielded 88,471 ASVs. Five ASVs were commonly observed across thirteen samples (Supplementary Table S2: Feature detail, frequency and number of samples observed in). One ASV occurred twice in eleven samples whilst 236 ASVs occurred twice in only one sample (Supplementary Table S2: Feature detail, frequency and number of samples observed in). Species distribution curves showed that all thirteen samples were su cient to capture the different patterns of bacterial diversity (Supplementary Fig. S1). Raw reads = original paired end reads after sequencing; Input reads = tags merged from trimmed raw reads before Dada2 processing; Dada2 process = output reads after Dada2 processing; Feature counts = amplicon sequence variants (ASVs) after ltering out unassigned.
The bacterial groups Rhizobiales, Rhodospiralles, Spirochaetales, Legionalles, Enterobacteriales, Actinomycetales occur in six of the seven coral samples albeit in low abundances and a couple of the non-coral samples (Fig. 3). Members belonging to the genera Pseudomonas (0.24 ± 0.08%), Pseudoalteromonas (0.59 ± 0.14%) and Photobacterium (0.53 ± 0.13%) were relatively abundant among the corals and even more abundant in the non-coral samples.
Generally, archaea are only found in the total sample in low abundances (<5%) with the phylum Thaumarchaeota (25.92 -100%) dominant among the archaea communities, although not observed in all samples. Parvarchaea was identi ed in three samples in relatively low abundance (Fig. 4).

Discussion
Seabed morphology is known to in uence the dynamics of the physical and chemical properties (e.g. dissolved and particulate organic matter, oxygen saturation and current regimes) of the deep-sea 40 including the upper PBC 41 . These properties can potentially in uence the composition of microbial communities on CWCs, sediment and the ambient water 26,6 . Coral microbiologists have recognised that there is divergence among microbial communities inhabiting corals, sediments and the ambient water 24 .
In the present study, divergence in the microbial composition associated with the scleractinian corals, water and sediment samples were observed in the upper PBC (Fig. 1C). However, there was no divergence in the microbial communities in relation to site locations and coral species (Figs. 1A and B), even though geographical distribution can cause differences in host-bacterial composition 36 . In this context, observations in relation to CWCs in the present study should be interpreted with caution as only two samples of Madrepora oculata were used.
Shannon's diversity observed for CWCs in the present study seems to be higher than what has been reported in some microbial studies 3, 16 yet similar to other research 17,36 . Röthig et al. 18 and Hansson et al. 24 observed species-speci c differences in coral-associated bacterial composition as well as divergence in bacterial composition across corals and water samples. Meistertzheim et al. 16 reason that differences in the feeding strategies and thermal tolerance between the framework-forming corals L. pertusa and M. oculata are responsible for the structure observed in the host-bacterial community composition. In line with Neulinger et al. 21 , this study suggests that variations in water mass does not play a major role in structuring the bacterial communities across the different samples as these were collected between 600-800 m, where the only prevailing water mass is the Eastern North Atlantic Water 19 . Nevertheless, inherent local physical conditions (e.g., salinity, pH, temperature and oxygen saturation 14,10 ) might have contributed to the observed divergence in microbial composition across the different samples in the present study.
In the present study, Alteromonadales were abundant in non-coral (water and sediment) samples while unclassi ed Gammaproteobacteria were abundant in corals. Also, between the two coral species, we observed that unclassi ed Gammaproteobacteria were more abundant in M. oculata than in L. pertusa. In general, we observed that CWCs showed high microbial diversity compared to the surrounding environments. Similarly, Schöttner et al. 26 noticed that bacterial communities associated with coral habitats were signi cantly more biodiverse than those associated with non-coral habitats due to mucus release of corals which dissolves in the water and can fuel the growth of microbes. Scleractinian corals can produce and release dissolved and particulate organic matter in the form of mucus 42,43 , which microbes use as a source of carbon for growth 44 .
Coral associated microbial communities perform different functions that are important to the coral host and ensure the general health and existence of the corals 45 . The order-level bacterial communities including Clostridiales, Flavobacteriales, Rhodobacterales and Rickettsiales were consistently represented in relatively high abundance. However, Clostridiales, although detected in great quantity in this study, are generally highly abundant in diseased corals 46 . Bacteroidales are often associated with diseased corals 9, 46 while some members are recognised for their antimicrobial protein production 47 . Some members of Flavobacteriales contain genes that can perform nitrogen cycling while others have been used for contaminant removal 48,49 . That said, pathogenic forms have also been identi ed 50 as being overrepresented in stressed corals 18 . They have been observed in both azooxanthallate cold-water and zooxanthallate tropical corals 51 .
Neulinger et al. 21 , associated the distribution of Lophelia pertusa phenotypes in the northeastern Atlantic with the high Rhodobacterales abundance found on the species. According to Neulinger et al. 21 , the e ciency of the Rhodobacterales on white Lophelia allows it to adopt to areas of low organic materials in the deep-seas of the northeastern Atlantic and hence their dominance in this part of the ocean.
Rhodobacterales are generally widespread in marine environments including on CWCs 25, 17 . They have been described as sulfur-and metal-reducing bacterial taxon, are involved in carbon cycling and can be applied as probiotics 52 . As this bacterial group occur in high abundance in the present study, it is likely they aid in growth and nutrition of the corals. Also, white Lophelia seems to be the dominant phenotype in the PBC, suggesting that Rhodobacterales may indeed be involved in Lophelia distribution in the northeast Atlantic Members belonging to the order Rickettsiales have been described as opportunistic, facultative and pathogenic 53,54 , can be found in CWCs 34 and have been associated with hydrocarbon contamination 55 .
Rickettsiales have also been observed in tropical corals 56 and found in high abundance in healthy corals microbiome in the present study have members which are pathogenic, only a few pathogenic members of Myxococcales have been characterised 58 . This bacterial order is ubiquitous and can tolerate extreme environments. They can prey on bacteria and so have been applied as antibiotic, cytotoxic and antiviral compounds 59 . Myxococcales are also involved in breakdown of organic carbon and nutrient cycling 60 and have been observed in coral species 61 . Many of the bacterial groups identi ed in this study including Burkholderiales, Rhizobiales, Sphingobacterales, Flavobacteriales and Oceanospirillales have been described to be associated with petroleum hydrocarbons 55 . However, there is no data from the upper PBC that suggest that petroleum exploration has been carried out in and around the canyon or that seabed pockmarks formed by hydrocarbon seepage are present. As such their presence may be related to a different function other than oil-degradation. Furthermore, the fact that many of these bacterial order occur in relatively low abundance seem not to be a petroleum degradation characteristic of corals from the upper PBC. It is probable that the dominant bacterial groups detected in the corals in this study are rather involved in the nutrition and growth of these species as suggested by other researchers 21 16 . Also, Novosphingobium which was characterised to be part of L. pertusa core microbiome 36 was observed in only two sediment samples (SS2 and SS3) in this study while other studies on the same species failed to discover it, either on the coral or the surrounding environment 21,35 . Interestingly, we observed TM7, which was rst recorded by Neulinger et al. 21 in CWCs from the Trondheimsfjord and later by Van Bleijswijk et al. 35 from the Rockall Bank. Neulinger et al. 21 noted the possible similarities in cross-taxa bacterial composition, as TM7 had previously been observed in the sponge Chondrilla nucula 62 . Also, there seem to be a regional effect in the distribution of TM7 as it was observed in the northeastern Atlantic (this study; 21,35 ), western Atlantic 25 but not the Mediterranean 16 . That said, it is di cult to draw regional trends and patterns from hostbacterial composition. Meistertzheim et al. 16 and Kellogg et al. 36  The interaction between corals and sponges through mucus release demonstrates the importance of corals to the sponge-bacterial community and possibly other reef-associated organisms. For instance, the bacterium Tenacibaculum maritimum was represented in the corals in low abundance. This species can cause diseases in reef shes 67 .
Generally in the corals, relatively high abundances of bacteria belonging to the genera Pseudomonas, Pseudoalteromonas and Photobacterium were observed. They form part of the core microbiome (100% coral samples coverage). Species of Pseudoalteromonas are clinically relevant and perform probiotic functions 68 . P. porphyrae has the potential to increase the growth of some marine species in a stressful environment while P. luteoviolacea is shown to produce antibacterial compounds 69,70 . P. luteoviolacea was observed in the scleractinian corals Porites. lutea 63 , L. pertusa 33  Also in this study, the genus Vibrio including V. shilonii (0.22%) were observed in a couple of coral samples, which contradicts the ndings of researchers who have described Vibrio usually as part of the core microbiome 76 . That said, both genera Vibrio and Photobacterium belong to the family Vibronaceae and have been characterised as opportunistic pathogens whose virulence is in uenced by the environment 77,21 . For example P. damselae and V. shilonii can cause harm to their hosts in warm environments 77 , suggesting that their pathogenic activities to the CWCs are somewhat being suppressed by the cold temperatures of the deep. Several members of Vibrio also have probiotic properties, digestive enzymes (e.g. amylase, lipase, cellulase and chitinase) that aid in digestion, x nitrogen in anoxic or limiting oxygen conditions 73 , and have been observed in healthy and/or diseased tropical corals 33 as well as CWCs 35,17 . It is thus more likely that the bacterial species P. damselae, Tenacibaculum maritimum and V. shilonii observed in the present study (where only healthy corals were collected) are opportunistic pathogens that only contribute to the overall health of the corals. Other recognised important pathogenic species represented in a few samples in low abundance include Flexispiras rappini, Serratia mercens, Helicobacter cinaedi and Salmonella enterica, Cardiobacterium valvarum. Nevetheless, the relatively low abundant pathogenic bacterial species render them less important to the general health and functioning of the corals compared to the more abundant and dominant probiotic bacterial groups which play a major role in disease-prevention, nutition, and growth of the corals in the PBC.
Many archaeal taxa have been identi ed in CWCs (this study; 35 ). Thaumarchaeota was represented in high abundance in all the coral samples except HF5 while Thermoplasmata was the only taxon present in sample FF3 (Madrepora sample). Both Thaumarchaeota and Thermoplasmata were identi ed in Lophelia samples from the Rockall Bank 35 . Members of this taxon can extract and metabolise amino acids from the water column as well as reduce sulfur 78 . Lophelia as an opportunistic feeder can take up dissolved amino acids from the water column and incorporate them into its cells 79,80 . Thus, the dominance of Thaumarchaeota on the coral samples could be a selective mechanism which parallels with the feeding strategy of Lophelia. Our work corroborates with Van Bleijswijk et al. 35 who identi ed Thaumarchaeota as the single most important dominant archaeal taxa of CWCs (e.g. L. pertusa). Also, Parvarchaea has been described to be associated with petroleum hydrocarbon 55 . Interestingly, we note that all archaea in FF3 (Madrepora coral sample) from the ank are Thermoplasmata which is absent from SF2 (Madrepora coral sample), but highly abundant in only SS2 (sediment sample) which is also from the ank, suggesting that Thermoplasmata could be more associated with location in the canyon rather than the corals.
The current trends of deep-sea warming due to climate change and anthropogenic activities (e.g. deepsea mining and bottom-shing 81 ) put the corals in a vulnerable state, considering that most of the pathogenic microbes observed can cause disease in warmer and polluted environments. Although the coral microbiome suggests the coral are healthy with low abundances of pathogen, higher abundances of pathogens observed in the surrounding water and sediment suggest that this may change if corals become stressed. Appropriate measures ought to be implemented to reverse the current trends of climate change to protect the teeming corals of the deep or risk losing them, a situation that can arise from a shift from a generally benign and useful host-associated microbial community to pathogenic and disease-causing microbes as have been described for tropical corals elsewhere 9,46 .

Conclusions
Our study is the rst to describe the coral-associated microbiome from the Porcupine Bank Canyon. We characterised the pattern of distribution of microbial communities associated with coral and non-coral samples and their composition based on distances among samples in the upper PBC. We observed that the host-bacterial composition exhibited signi cant divergence among the different groups of samples, with coral samples showing high bacterial diversity. Comparing coral and non-coral assemblages, opportunistic pathogens are common in the water column and sediment. The dominance of pathogens in the water or sediment put the healthy corals at risk. Also, there was no structure in the bacterial composition associated with Lophelia pertusa and Madrepora oculata as well as in relation to site location in the canyon.
It is recognised that opportunistic pathogenic bacterial communities can pose a greater threat in warmer environments. Although it is di cult to predict the direction of the microbial composition inhabiting these corals in the short term, we believe that current trends of climate change and resulting deep-sea warming may shift the bacterial composition towards a more dominant virulent and pathogenic community. As such, it is important to protect the vulnerable coral fauna through policies that would reverse current trends in climate change to ensure the future safety of this fragile coral ecosystem.

Methodology
Study site, sample collection and storage  (Supplementary Table S1 and Supplementary Fig. S2). The PBC is tectonically controlled and terrestrially disconnected. The canyon head is sediment dominated and exhibits a gentle slope while the ank reveals a steep slope with exposed bedrock. The south branch is a smaller canyon system feeding into the main canyon and contains both live and dead coral communities 82,19 . DNA extraction of coral, sediment, and seawater samples All procedures were performed in a laminar ow hood which, along with the tools used, was decontaminated with a 10% bleach solution and DNA AWAY® before and between each sample preparation and each set of extractions. Coral samples were crushed to a ne powder using liquid nitrogen and aliquots of sediment were weighed to the required amount (200-250 mg) for DNA extraction. Extractions were performed using the DNeasy PowerSoil Pro kit (Qiagen) following manufacturer's protocol. DNA from seawater samples was extracted using DNeasy® Blood and Tissue kit (Qiagen) following the methodology outlined in Spens et al. 83 . Speci cally, seawater samples were ltered over sterile 0.22 µm Sterivex TM -GP lters (Merck Millipore, Germany) using 60 ml syringes (Soft-Ject®, HSW, Tuttlingen, Germany). After ltering, one opening of the Sterivex TM -GP lters was sealed with para lm, 720 µl of ATL-buffer and 80 µl of proteinase K were added, the opening sealed, and the mixture was incubated at 56°C for 2 h with agitation. The buffer mix was transferred with a 3 ml leuer-lock™ syringe into a 2 ml Eppendorf tube. DNA extraction was continued following the manufacturer's protocol, with buffer volumes adjusted for the amounts of lysate recovered from the sterivex lter column. All extractions were run alongside negative controls using ddH 2 O so that any laboratory contamination would be detected.
To test whether all the DNA yielded during the extraction process was entirely from the samples (and not from laboratory contamination), DNA extractions and negative controls were subjected to routine PCR of the 16s rRNA gene using the universal primers 8 F (5'-AGAGTTTGATCCTGGCTCAG-3') and 1492 R (5'-GGTTACCTTGTTACGACTT-3'). Negative controls did not yield any product indicating that sample DNA was suitable for further analysis. Genomic DNA samples were sent to Novogene, Cambridge, UK for microbial amplicon-based metagenomics sequencing 16S (V3-V4). The purity and concentration of genomic DNA samples were assessed on a 1% agarose gel and Qubit Fluorometer (Invitrogen, USA) respectively. The 341 F (5'-CCTAYGGGRBGCASCAG-3') and 806 R (5'-GGACTACNNGGGTATCTAAT-3') primers with barcodes were used to amplify the V3-V4 region (466 bp) of the 16S rRNA gene and sequenced using the Illumina NovaSeq 6000 SP with paired ends (2 x 250 bp) strategy (Novogene, Cambridge, UK). PCR reactions were performed in a 20 µl reaction volume using the Phusion® High-Fidelity PCR Master Mix (New England Biolabs) with ~ 1-2 ng template DNA. The nal primer concentration was adjusted to 0.5 µM. Amplicons were run on a 2% agarose gel for detection, desired bands were excised and puri ed using Qiagen Gel Extraction Kit (Qiagen, Germany).

Bioinformatic analysis and statistics
Raw reads were quality controlled with FastQC 0.11.9, trimmed (to remove reads shorter than 200 bp, mean quality score of 30 in a sliding window of 10 bp and a maximum of two primer mismatch) with Trimmomatic 0.39 and merged into contigs using FLASH 1.2.11. Quality ltering of the sequence data to remove chimeras was performed using Dada 2 as implemented in qualitative insights into microbial ecology (QIIME 2) v2020.8 84 . Sequencing data were rare ed to a sample depth of 20,000 corresponding to the least number of sequences available for all thirteen samples. Each Dada 2 amplicon sequence variants (ASVs) were utilized for taxonomic classi cation via the QIIME 2 feature-classi er with the sklearn technique. Taxonomic assignment was made from the Greengenes 13_8 database (DeSantis et al. 85 ; based on 99% similarity of sequence data to the Greengenes database). All unassigned, mitochondrial and chloroplast ASVs were ltered out. Also, archaeal ASVs were ltered out and analysed separately for taxonomic classi cation.
Diversity indices of the samples were assessed using the QIIME 2 pipeline (https://docs.qiime2.org). The sequences were aligned using Mafft and FastTree to construct a phylogenetic tree. Alpha diversity indices (Shannon, Pielou's evenness and Faith phylogenetic index) were assessed to determine the microbial diversity within each sample. Beta diversity based on Bray-Curtis distance and unit fraction analyses were performed. Principal coordinates analysis (PCoA) plots were made in QIIME 2 and using the Emperor tool, the differences in microbial composition among groups of samples, the different species and site locations were visualised. Signi cant differences among groups of samples, types of species and site locations were evaluated using PERMANOVA. Also, Kruskal-Wallis test was used to compare alpha diversity indices among groups of samples, the different species and site locations.