Dominance of Endozoicomonas bacteria throughout coral bleaching and mortality suggests structural inflexibility of the Pocillopora verrucosa microbiome

Abstract The importance of Symbiodinium algal endosymbionts and a diverse suite of bacteria for coral holobiont health and functioning are widely acknowledged. Yet, we know surprisingly little about microbial community dynamics and the stability of host‐microbe associations under adverse environmental conditions. To gain insight into the stability of coral host‐microbe associations and holobiont structure, we assessed changes in the community structure of Symbiodinium and bacteria associated with the coral Pocillopora verrucosa under excess organic nutrient conditions. Pocillopora‐associated microbial communities were monitored over 14 days in two independent experiments. We assessed the effect of excess dissolved organic nitrogen (DON) and excess dissolved organic carbon (DOC). Exposure to excess nutrients rapidly affected coral health, resulting in two distinct stress phenotypes: coral bleaching under excess DOC and severe tissue sloughing (>90% tissue loss resulting in host mortality) under excess DON. These phenotypes were accompanied by structural changes in the Symbiodinium community. In contrast, the associated bacterial community remained remarkably stable and was dominated by two Endozoicomonas phylotypes, comprising on average 90% of 16S rRNA gene sequences. This dominance of Endozoicomonas even under conditions of coral bleaching and mortality suggests the bacterial community of P. verrucosa may be rather inflexible and thereby unable to respond or acclimatize to rapid changes in the environment, contrary to what was previously observed in other corals. In this light, our results suggest that coral holobionts might occupy structural landscapes ranging from a highly flexible to a rather inflexible composition with consequences for their ability to respond to environmental change.

Given the likely critical contribution of bacteria to coral holobiont function, changes in the identity and abundance of associated bacteria may allow for rapid adaptation to environmental change (Reshef, Koren, Loya, Zilber-Rosenberg, & Rosenberg, 2006;Rosenberg et al., 2007;Theis et al., 2016;Torda et al., 2017). Consequently, a dynamic relationship between associated microorganisms and environmental conditions is assumed that selects for the most beneficial host microbiome (Reshef et al., 2006). Thereby, diversity and function of microbes need to be considered when assessing acclimatization and adaptation of corals and the ecosystems they shape (Ainsworth & Gates, 2016;Jessen et al., 2013;Röthig et al., 2016;Ziegler, Seneca, et al., 2017).
Accordingly, changes in bacterial community structure or activity may be either beneficial or deleterious, depending on the environmental context (Ainsworth & Gates, 2016;Rädecker et al., 2015Rädecker et al., , 2017Santos et al., 2014).
Although putative core members (Ainsworth et al., 2015; Hernandez-Agreda, Gates, & Ainsworth, 2017;Hernandez-Agreda, Leggat, Bongaerts, & Ainsworth, 2016;Neave, Rachmawati, et al., 2017) and temporally stable versus more sporadic bacterial taxa (Hester, Barott, Nulton, Vermeij, & Rohwer, 2016) of coral microbiomes were previously identified, the common denominator of coralassociated microbiomes may ultimately be its flexibility. The notion of flexible bacterial associations, however, has never been systematically assessed. Further, it remains unclear whether all corals possess flexible bacterial microbiomes. Here we sought to investigate the flexibility of microbial association of the common Indo-Pacific coral Pocillopora verrucosa, previously suggested to display a limited acclimatization potential Sawall et al., 2015;Ziegler, Roder, Büchel, & Voolstra, 2014). We approached this by exposing the P. verrucosa holobiont to different excess dissolved organic nutrient treatments. Specifically, we integrated microbial community data from the here-conducted experiment assessing the effects of excess dissolved organic nitrogen (DON; 40-to 50-fold enrichment compared to ambient conditions) and complemented these data with a previously published companion experiment exposing P. verrucosa holobionts to excess labile dissolved organic carbon (DOC; >10-fold enrichment; . In these two independent 14day experimental treatments, corals exhibited bleaching and mortality reflected by progressed tissue sloughing, respectively. The composition and concentration of the excess DOC treatment were selected to mimic sewage input (Huang, Li, & Gu, 2010) and exudates of the reef macroalga Halimeda (Nelson et al., 2013), which share a similar pool of the most abundant oligosaccharides as used here. The excess DOC and excess DON conditions in the present study constitute severe stress scenarios not representative of natural, healthy reef conditions , but rather reflect conditions of degraded coastal ecosystems impacted by anthropogenic activity (Kline, Kuntz, Breitbart, Knowlton, & Rohwer, 2006;Peña-García, Ladwig, Turki, & Mudarris, 2014). Excess organic matter can cause rapid and dramatic compositional and metabolic changes in bacterial communities Haas et al., 2016) and can affect the physiology of coral holobionts (Kline et al., 2006;Vega Thurber et al., 2009. As the uptake and cycling of organic nutrients involve all holobiont members, choosing an approach to assay Symbiodinium and bacterial community dynamics allows to identify rapid responses associated with and potentially involved in holobiont resilience and/or breakdown. We assessed such responses in algal and bacterial symbionts by assessing community composition dynamics via ITS2 and 16S rRNA gene-typing, respectively, alongside algal symbiont density and chlorophyll a content measurements.

| MATERIALS AND METHODS
The data presented in this manuscript were collected in two independent companion experiments during November/December 2014 (excess DOC experiment) and January/February 2015 (excess DON experiment) at the wet laboratory facilities of the Coastal and Marine Resources Core Lab (CMOR) at the King Abdullah University of Science and Technology (KAUST). Part of the data presented for the excess DOC experiment were published previously in a companion paper ; this applies to the bacterial 16S rRNA gene amplicon sequencing data, Symbiodinium population density data, and seawater DOC concentrations for the excess DOC experiment presented in this study. This is complemented by new data on Symbiodinium chlorophyll a content and ITS2 sequencing for the excess DOC and DON experiments, as well as bacterial 16S rRNA gene amplicon sequencing data for the DON experiment. For the present study, the specified data previously published in  along with the new data from both experiments were jointly analyzed.

| Coral collection and husbandry
Three colonies of the brown color morph of P. verrucosa were collected for each of the excess nutrient experiments (i.e., a total of n = 6 coral colonies) from a Central Red Sea midshore reef ("Al-Fahal" reef, N22°18′19.98″, E38°57′46.08″; Kingdom of Saudi Arabia). The colonies had an average diameter of 40 cm and were collected from 7 to 8-m water depth. Care was taken to sample corals at least 5 m apart from each other to avoid collection of clonal colonies of P. verrucosa (Robitzch, Banguera-Hinestroza, Sawall, Al-Sofyani, & Voolstra, 2015).
Immediately after coral collection for each of the experiments (i.e., excess DOC and excess DON), coral colonies were fragmented and acclimated for a 4-week period in aquaria at the wet laboratory facilities. For each of the experiments, the aquarium system consisted of two separate identical units, each consisting of three closed replicate experimental tanks (100 L) connected to reservoir bins (100 L) containing filtration and heating equipment. For each of the experiments, one of the units was maintained at ambient conditions (control), while the second unit was used for the respective nutrient manipulation (excess DOC or DON); the two experimental nutrient manipulations were conducted independently from each other and consecutively.
Red Sea reef water was circulated in each tank, and 30% of the water was replaced on a daily basis, maintaining close to natural water parameters. Maintenance conditions were kept constant (temperature 26.9 ± 0.4°C, salinity 41.0 ± 0.8 PSU, photosynthetic active radiation ~100 quanta μmol m −2 s −1 , on a 12:12-hr daylight cycle; dissolved oxygen levels remained >6 mg/L at all times). Individual colonies were fragmented and glued to 40 × 40 mm stone tiles with a two-part epoxy putty. After the acclimation period, replicate fragments of each of the 3 colonies in each excess nutrient experiment were redistributed among the six aquarium tanks (i.e., 100 L each and each tank contained replicates of each colony) of the two separate identical units ( Figure S1). The sampling design was fully paired, except for day 14 in the excess DON treatment. Here, due to mortality of individual fragments, colony replication was n = 2 for day 14.

| DOC and DON enrichment experiments
For each of the experiments (i.e., excess DOC and excess DON), three tanks per unit were used for nutrient manipulations and three tanks of the remaining unit were used as controls, that is, maintained at ambient levels (see above). In the first experiment, elevated DOC conditions were achieved in three aquarium tanks by daily additions of a monosaccharide mixture (10 mg/L; composition: (in mg/L; (D+) xylose: 3.82; (D+) glucose: 2.56; (D+) mannose: 1.39; (D+) galactose: 2.22; . The respective contribution of each monosaccharide was based on the carbohydrate composition of sewage (Huang et al., 2010) and released exudates of the coral reef macroalga Halimeda after chemical hydrolysis (Nelson et al., 2013).
In the second experiment, dissolved organic nitrogen (DON) levels were increased in three of the aquarium tanks of one unit by the addi- Peña- García et al., 2014). Urea rapidly photodissociates into carbon dioxide and ammonium in aqueous solutions (Glibert, Harrison, Heil, & Seitzinger, 2006), thereby allowing a continuous co-enrichment of dissolved organic and inorganic nitrogen. Notably, while urea is predominantly taken up by the coral animal itself (Grover, Maguer, Allemand, & Ferrier-Pagès, 2006), ammonium is preferably taken up by algal symbionts (Grover, Reynaud-Vaganay, & Ferrier-Pagès, 2002). Thereby, the here applied excess nitrogen conditions ensured the co-enrichment of both the coral animal and the associated algal symbionts.

| Sampling and measurements
All sampling procedures and measurements were identical for both experiments. Fragments for all response parameters were sampled on day 0, 7, and 14 for the respective experiments. For the assessment of Symbiodinium cell density and chlorophyll a content, Symbiodinium typing, and bacterial community analyses, single fragments originating from all mother colonies were freshly collected for each treatment condition and time point and rinsed with filter-sterilized seawater (FSW; 0.22 μm), flash-frozen in liquid nitrogen, and stored at −80°C until further processing. Seawater samples for bacterial community analyses were collected in triplicates (1 L each) for each treatment and time point. The seawater samples were filtered through 0.22 μm, and the filters immediately frozen and stored at −80°C until further processing. Coral tissue loss was assessed by visual estimates based on photographs.

| Seawater nutrient analysis
Water samples for the analysis of nutrients were collected at all sampling points in 30 ml triplicates for each condition. Each water sample was filtered (GFF 0.7 μm, Fisher Scientific, USA). Water samples for DOC measurements were additionally filtered through 0.45μm GFF filters to remove particulate organic carbon and subsequently acidified with 100 μl of 35% phosphoric acid to remove inorganic carbon content. Subsequently, all water samples were frozen at −20°C.

| Symbiodinium cell density and chlorophyll analysis
For the assessment of Symbiodinium population densities and chlorophyll a content in hospite, Symbiodinium cells were freshly isolated from coral tissue by NaOH extraction (Zamoum & Furla, 2012). Subsamples of individual coral fragments were incubated in 1M NaOH. After 1 hr, the skeleton was removed. Suspended Symbiodinium cells were spun down in a bench-top centrifuge for 5 min at 3,000 RCF, the supernatant discarded, and the Symbiodinium pellet resuspended in 1 ml 1 × PBS. After a second centrifugation step, the pellet was resuspended in a 10% PBS-buffered formaldehyde solution and stored at 4°C until further processing. Symbiodinium density was determined using flow cytometry (BD LSRFortessa, BD Biosciences, USA). The relative cell chlorophyll a content of Symbiodinium per sample (i.e., for each coral fragment) was calculated based on the relative chlorophyll autofluorescence of Symbiodinium cells. For this, cells were excited at a wavelength of 488 nm and fluorescence emission was recorded at 695/40 nm.

| DNA extraction from seawater and coral samples, PCR conditions, and sequencing
To address changes in the Symbiodinium community (coral-associated) and bacterial community (seawater and coral-associated) composition, we sequenced the ribosomal internal transcribed spacer 2 (ITS2) and the hypervariable regions v5 and v6 of the 16S rRNA gene, respectively.
Before DNA extraction, each frozen coral fragment was transferred into a sterile zip-lock bag. While thawing, fragments were doused with 5-ml Qiagen AP-1 tissue lysis buffer (Qiagen, Germany) and tissue was subsequently removed by air-blasting on ice. The coral tissue slurry in AP-1 was stored frozen at −80°C until further process- Individual PCRs were run using 10μl Qiagen Mix, 0.5 μl of each 10-

| Statistical analysis
All statistical analyses of Symbiodinium responses and microbial alpha diversity were conducted in R v3.3.0 (R Development Core Team, 2015). Symbiodinium population density as well as relative abundance of clade A and D symbionts was tested for significant individual and interactive effects of treatment and time in generalized linear models (GLMs). The models were based on a Gamma distribution with best fitting link function to account for skewing of data. To illustrate significant differences between manipulations, treatment effects of individual time points were compared using unpaired Welch's unequal variances t test. Bacterial community compositions were compared between treatment and time points using analysis of molecular variance (AMOVA) as implemented in mothur. Alpha diversity indices of bacterial communities were compared between treatment and time points using two-way analysis of variance (ANOVA), and homoscedasticity of models was confirmed using the Breush-Pagan test as implemented in the "lmtest" package (Hothorn et al., 2017).

| Coral phenotypic response to excess DOC and DON
The enrichment with excess labile dissolved organic carbon (DOC; >10-fold enrichment) and nitrogen (DON; 40-to 50-fold enrichment compared to ambient conditions) during two independent 14-day experimental treatments resulted in severe coral stress phenotypes (for details, refer to Table S1). While the control fragments remained in a visibly healthy state, coral fragments under excess DOC exhibited a bleaching response (Figure 1; . Conversely, all coral fragments subjected to DON initially (within 7 days of treatment) displayed visibly darkened tissue.
However, minor tissue loss (<10%) became apparent after 13 days of F I G U R E 1 Phenotypic response of Pocillopora verrucosa subjected to excess dissolved organic carbon (DOC) and excess dissolved organic nitrogen (DON). (a and b) Corals exposed to excess DOC exhibited pronounced bleaching over the course of the 14-day treatment compared to control coral colonies. (c and d) Corals exposed to excess DON exhibited a marked darkening of the tissues, even when partial mortality (>90% tissue loss) was visible (day 14)   (Figure 2a). The remaining ITS2 sequences belonged to Symbiodinium of clades B and C, which consistently comprised <1% of all sequences. While excess DOC addition had no significant effect on Symbiodinium community composition over time (GLM, χ 2 (5,n=18) = 2.241, p = .815), observed decreases in the relative abundance of Clade D symbionts in controls might suggest that Symbiodinium communities were still adjusting to aquaria conditions. However, we did not observe such a shift in control fragments in the DON treatment. Control coral colonies in the excess DON experiment were stable, while under the excess DON treatment, they exhibited a significant shift in the relative clade abundance (GLM, χ 2 (5,n=18) = 18.703, p < .002) and Symbiodinium of Clade D almost entirely disappeared over time (Figure 2b).

| Bacterial community composition under excess DOC and DON
To assess changes in seawater and coral-associated bacterial commu-   found in seawater), 2,126 were associated with corals (860 of those were only found in corals), and 1,266 OTUs were present in seawater and corals (Table S4) Table S3).
Bacterial community composition was highly uneven and dominated by only two bacterial OTUs of the genus Endozoicomonas (Oceanospirillales: Hahellaceae, recently put into the proposed family Endozoicomonadaceae (Bartz, Blom, Busse, Mvie, Hardt et al., 2017)) with a combined relative average abundance of 90% across all coral fragments. These two OTUs also comprised the "core" microbiome, defined as OTUs present in 100% of all coral samples (for a complete presentation of coral-associated OTU abundances, see Table S4). Notably, after omitting all Endozoicomonas sequences from analysis, no treatment effects on the remaining coral-associated bacterial communities were apparent under excess DOC (AMOVA, p = .843, F = −0.006) or excess DON (AMOVA, p = .894, F = −0.998). However, despite the overall stable bacterial community and the dominant contribution of Endozoicomonas OTUs to the overall bacterial community composition, changes in the relative abundance of the two Endozoicomonas OTUs became apparent. Endozoicomonas OTU 2 decreased substantially under excess DOC from 22% to 1% within 14 days of treatment (corresponding to an increase in relative contribution of Endozoicomonas OTU 1 from 67% to 89%). In contrast, Endozoicomonas OTU 1 and OTU 2 remained proportionally stable under excess DON over 14 days (about 56% for OTU 1 vs. 29% for OTU 2 average relative contribution). The reasons for this are unclear at present, although metabolic subfunctionalization/specialization of Endozoicomonas taxa residing in corals was suggested previously Neave, Rachmawati, et al., 2017).

| DISCUSSION
To assess the stability and flexibility of coral holobiont structure in the common reef-building coral P. verrucosa, we monitored community composition of Symbiodinium and bacteria to severely altered nutrient conditions. Excess DOC and DON clearly impaired coral health and functioning within days as indicated by coral bleaching and tissue loss (sloughing) and eventually mortality, respectively. Despite the severe phenotypic responses, coral-associated bacterial communities, in contrast to Symbiodinium communities, remained remarkably stable and were dominated by two OTUs of the genus Endozoicomonas, comprising on average 90% of the coral-associated 16S rRNA gene amplicon sequences. This outcome has implications for our understanding of coral holobiont structure and microbiome flexibility and stability, as discussed in the following.

| Breakdown of the coral-algae symbiosis
Both DOC and DON treatments affected the coral-algae symbiosis, although the responses of Symbiodinium and coral stress phenotypes differed between both experimental treatments. While under excess DOC corals experienced a rapid loss of Symbiodinium (for a detailed discussion see , the Symbiodinium population density and chlorophyll a content rapidly increased under elevated DON as suggested by our measurements and the visibly darkened tissue, which constitutes a common response of coral holobionts to elevated nitrogen levels (Ezzat, Maguer, Grover, & Ferrier-Pagès, 2015;Falkowski et al., 1984Falkowski et al., , 1993. The rapid proliferation of algal symbionts under excess DON was associated with fluctuations in the abundance of the two dominant symbiont clades A and D (as opposed to a stable algal community composition under excess DOC). No symbiont shuffling was observed (Mieog, Van Oppen, Cantin, Stam, & Olsen, 2007). Instead, P. verrucosa exhibited strong clade fidelity concomitant with an increase in clade A Symbiodinium. Hence, clade A symbionts seem to perform better in high nitrogen environments when compared to symbionts from clade D. In this regard, it is interesting to note that clade D Symbiodinium exhibited an inferior capacity for nitrogen acquisition compared to other clades at ambient temperatures in previous work (Baker, Andras, Jordán-Garza, & Fogel, 2013;Pernice et al., 2015).
At large, our findings may explain the prevalence of clade A symbionts in the highly phototrophic P. verrucosa along most of the Red Sea basin . Importantly, clade A symbionts are rare in most scleractinian corals (LaJeunesse et al., 2004) and are often considered opportunistic (Stat, Morris, & Gates, 2008). In the oligotrophic conditions of the Red Sea, however, the association with clade F I G U R E 4 Bacterial community response in the coral Pocillopora verrucosa subjected to excess dissolved organic carbon (DOC) and excess dissolved organic nitrogen (DON) over time. (a) Excess DOC, (b) excess DON. Community composition is presented at the class and OTU level (inner vs. outer circle of pie charts, respectively). While seawater bacteria showed significant differences indicating effects of excess DOC and DON on community structure, coral-associated bacterial communities remained remarkably stable over the course of the experiments (each pie plot represents means for n = 3 A Symbiodinium may increase holobiont productivity due to its assumingly efficient nitrogen uptake capability (Aranda et al., 2016). Under excess DON, on the other hand, we argue that clade A Symbiodinium may be detrimental to P. verrucosa holobiont health due to opportunistic growth and reduced carbon translocation.

| Stable dominance of Endozoicomonas during bleaching and mortality
Despite the severe coral stress responses, the bacterial community structure of P. verrucosa remained dominated by Endozoicomonas under excess DOC and DON over the 14-day experimental treatments. This observation does not rule out that community shifts occurred within the rare fraction of the microbiome, but the bacterial community at large was consistent. Importantly, we did not observe an intrusion and propagation of putative opportunistic or pathogenic bacteria from the surrounding seawater, which harbored bacterial communities that were highly distinct from bacteria associated with P. verrucosa (Figure 4).
While the absence of bacterial community changes in coral holobionts counters previous work (Vega Thurber et al., 2009Ziegler et al., 2016), recent work from the Red Sea reports on similarly stable bacterial communities in P. verrucosa across sites subject to differential anthropogenic impact (sewage, municipal waste water, and sediment input). Notably, this bacterial community "stability" in P. verrucosa as reported by Ziegler et al. (2016) could largely be attributed to the high abundance of taxa in the family "Endozoicimonaceae". Similarly, in our current study the stable bacterial community of P. verrucosa was driven by the prevalence and dominance of two Endozoicomonas OTUs. Importantly, the same two Endozoicomonas phylotypes were previously shown to prevail in the microbiome of P. verrucosa across its entire global distribution range with little geographic partitioning, suggesting a particularly intimate and conserved host-microbe relationship (Neave, Apprill, Ferrier-Pagès, & Voolstra, 2016).

| Structural stability of bacterial communities: implications for holobiont functioning
The remarkable stability of the association between Endozoicomonas Consequently, coral holobionts with adaptable bacterial communities may respond rather dynamically and readily to environmental change, as observed in Fungiidae (Roder et al., 2015;Röthig et al., 2016), Acroporidae (Ziegler et al., 2016;Ziegler, Seneca, et al., 2017), or Dendrophyllidae (Röthig et al., 2017). By comparison, coral hosts with structurally "stable" (and presumably strongly selected) microbiomes may host highly uneven bacterial communities (as observed in P. verrucosa) with a potentially very specialized set of (metabolic) functions (Ley, Peterson, & Gordon, 2006;Sawall et al., 2015;Ziegler et al., 2014Ziegler et al., , 2016. Such strongly structured bacterial communities may provide an advantage under highly stable conditions, but at the same time may implicitly come with the restriction to a comparatively narrow ecological niche space, as known for heritable obligate host-microbe symbioses (Bennett & Moran, 2015). Further, the strong reliance on few selected bacterial symbionts may come at the cost of low stress resistance under adverse environmental conditions (Wittebolle et al., 2009). Indeed, Red Sea P. verrucosa holobionts can be considered highly specialized as indicated by their fairly limited bathymetric range and restriction to shallow sunlit and nutrient-poor waters, where fairly stable environmental conditions prevail Sawall et al., 2015;Ziegler et al., 2014).
Future studies should assess coral microbial community association and flexibility under environmental stress or across environmental gradients including coral species covering different ecological traits, for example, autotrophy versus heterotrophy or spawning versus brooding corals. This will further help to understand whether coral species indeed represent different holobiont structural landscapes (i.e., specialized and stable vs. functionally redundant and flexible) and how bacterial community stability aligns with environmental resilience.

DATA ACCESSIBILITY
Microbial data as well as aquaria seawater conditions are available as

CONFLICT OF INTEREST
None declared.

AUTHORS CONTRIBUTION
CP, NR, CW, CRV designed and conceived the study; CP, NR, AC generated data; CP, NR, CRV analyzed data; AG, CW, CRV contributed reagents/tools/materials; CP, NR, CRV wrote the manuscript; all authors read and approved the final manuscript.