Changes in microbial communities, photosynthesis and calcification of the coral Acropora gemmifera in response to ocean acidification

With the increasing anthropogenic CO2 concentration, ocean acidification (OA) can have dramatic effects on coral reefs. However, the effects of OA on coral physiology and the associated microbes remain largely unknown. In the present study, reef-building coral Acropora gemmifera collected from a reef flat with highly fluctuating environmental condition in the South China Sea were exposed to three levels of partial pressure of carbon dioxide (pCO2) (i.e., 421, 923, and 2070 μatm) for four weeks. The microbial community structures associated with A. gemmifera under these treatments were analyzed using 16S rRNA gene barcode sequencing. The results revealed that the microbial community associated with A. gemmifera was highly diverse at the genus level and dominated by Alphaproteobacteria. More importantly, the microbial community structure remained rather stable under different pCO2 treatments. Photosynthesis and calcification in A. gemmifera, as indicated by enrichment of δ18O and increased depletion of δ13C in the coral skeleton, were significantly impaired only at the high pCO2 (2070 μatm). These results suggest that A. gemmifera can maintain a high degree of stable microbial communities despite of significant physiological changes in response to extremely high pCO2.

Scientific RepoRts | 6:35971 | DOI: 10.1038/srep35971 response to OA, or whether theses changes also alter host physiology. Preliminary laboratory-based investigations have revealed a remarkable impact of increased pCO 2 or reduced pH on coral microbial communities 12,16,18 . In contrast, no significant changes were observed in the microbial communities of transplanted corals in natural CO 2 vents 19 and associated with two Pacific corals after 8 weeks of exposure to increased pCO 2 20 . These contradictory findings underscore the need for further research. Most recently, microbial communities associated with coral and sponge from natural CO 2 seeps have demonstrated species-specific acclimatization to their habitats 21 . Therefore, the potential of natural microbial communities in corals to acclimatize/adapt to OA cannot be overlooked.
Natural fluctuations in seawater pH/pCO 2 are common, especially diel pH/pCO 2 fluctuations in shallow water coral reefs 22 and these fluctuations affect the abundance and distribution of marine organisms 23 . The fauna have been suggested to be locally acclimatized/adapted to the variable pH environment as an evolutionary mechanism to cope with future acidification [24][25][26] . However, the adaptation of the coral holobionts to OA remains largely unexplored and is worth careful investigation.
The Luhuitou fringing reef (18°12′ N, 109°28′ E) is located in the southern Hainan Island, South China Sea (see Supplementary Fig. S1) and used to have a high coverage of living coral, which has declined by 80% since the 1960s 27 . The diurnal and seasonal variation of the reef flat seawater pH/pCO 2 is high [28][29][30] , and the recorded extreme level is even lower than the value currently predicted at the end of this century (see Supplementary Table S1). The rapidly-growing branching coral Acropora, which is distributed worldwide, is an ecologically important genus in this reef flat. In the present study, A. gemmifera colonies collected from this reef flat were exposed to three pCO 2 levels to test our hypothesis that both the coral physiology and the microbial communities associated with this coral species are stable and resistant to OA exposure.

Results and Discussion
Overview of the microbial communities. After quality filtering, 308,591 reads were used for the downstream analyses. The number of operational taxonomic units (OTUs), Chao1 estimation of species richness, and Shannon index were obtained at a dissimilarity of 3% (Table 1). The rarefaction analyses revealed that the sequencing effort for each sample was sufficient to reflect the microbial diversity, and the rank-abundance curve showed that most OTUs had an abundance lower than 0.1%, which demonstrated that the microbial communities were occupied by rare species (see Supplementary Fig. S2). The prevalence of rare species has been widely demonstrated elsewhere, yet the ecological and functional roles of these rare species remain unknown 31 . There were no significant differences in beta diversity among the pCO 2 treatments, which is in contrast to the findings of a previous report showing an increase in coral microbial diversity with decreasing pH, possibly caused by an intermediate disturbance 16 .
In total, 24 bacterial and 2 archaeal phyla were detected in the coral and seawater samples, including Proteobacteria, Actinobacteria, Bacteroidetes, Cyanobacteria, Firmicutes, Thaumarchaeota and Euryarchaeota (see Supplementary Fig. S3). The relative abundance of archaea made up less than 0.1% of the seawater samples, with most OTUs belonging to autotrophic ammonia-oxidizing archaea (AOA) within the phylum Thaumarchaeota, which was even less abundant in the coral samples. It has been suggested that archaea may play prominent roles in corals and reefs 5 , although their abundance in both corals and reef water is much lower than that of bacteria 32 . Notably, the bacterial communities in both A. gemmifera and seawater were dominated by Proteobacteria and the most abundant class was Alphaproteobacteria (56-80%), among which the majority were assigned to the family Brucellaceae in the order Rhizobiales ( Fig. 1) followed by Gammaproteobacteria (9-26%). Both Alphaproteobacteria and Gammaproteobacteria are commonly highly abundant in corals, but their relative abundance varies among species 33 .  Table 1. Sample information and summary of microbial communities in corals and seawaters. "H", "M" and "C" refer to high, medium and control pCO 2 treatment, respectively. "A" and "W" refer to coral and seawater samples, respectively. The number following the letter indicates a replicate. Chao1, Observed species and Shannon index were determined at 3% dissimilarity after normalizing the full 16S dataset (including bacterial and archaeal sequences) to 12,459 sequences per sample.

Sample ID Treatment
Scientific RepoRts | 6:35971 | DOI: 10.1038/srep35971 Taxonomic assignment at the genus level was summarized, and genera with an abundance of greater than 1% in at least one sample are shown in a heat map (Fig. 2). In the present study, the unclassified Brucellaceae (>24%), Acinetobacter (> 9%) and Pannonibacter (> 5%) were the most abundant genera in coral, regardless of the pCO 2 treatment. Diazotrophs within the order Rhizobiales have been found in other coral species and were considered to be important for coral holobiont in nitrogen-limited waters 5 . It has been shown that copiotrophic taxa including Brucellaceae were enriched in algal-dominated environment 34 . Diverse algal communities on the Luhuitou fringing reef 35 might contribute to the dominance of unclassified Brucellaceae in A. gemmifera. Acinetobacter spp. have also been commonly reported in bleached and healthy corals 36 . Therefore, it is reasonable to suggest that the dominant genera, including Acinetobacter and the unclassified Brucellaceae, play critical roles in A. gemmifera. Interestingly, the putatively endosymbiotic Endozoicomonas 37 was detected at a very low level in all coral samples (< 2%). The photosynthetic Cyanobacteria assigned to the genus Synechococcus have been reported in sponge and coral 21 and were also detected at a very low abundance (0.2%) in A. gemmifera. However, the functions of bacteria and archaea and their interactions in the coral holobiont remain largely unclear.

Stability of microbial communities in A. gemmifera. As estimated by Adonis analysis at all taxo-
nomic levels (Adonis test, p > 0.05) and nMDS ordination (see Supplementary Fig. S4), there were no significant differences in microbial community compositions in A. gemmifera among the different pCO 2 treatments even after a 4-week exposure. Additionally, results from the SIMPER analysis showed that the dissimilarity of microbial communities among pCO 2 treatments was very small (see Supplementary Fig. S5). Taken together, these findings suggest that the A. gemmifera microbiome was not significantly affected by elevated pCO 2 and could remain relatively stable (Fig. 2). This result is inconsistent with the findings of some previous studies in which the coral microbiome shifted under higher pCO 2 or lower pH treatments over treatment periods ranging from days to months 16,38 . However, our finding is consistent with some other studies. For example, there were no differences in the microbial community structure in coral between pH 7.7 (pCO 2 = 1187 μ atm) and 7.5 (pCO 2 = 1638 μ atm) whereas a significant difference was observed between pH 8.1 (pCO 2 = 464 μ atm) and 7.9 (pCO 2 = 822 μ atm) Figure 1. Coral and seawater microbial communities at the order (a) and family (b) level. The minor group represents the sum of all orders or families representing < 2% in all samples. "H", "M" and "C" refer to high, medium and control pCO 2 treatment, respectively. "A" and "W" refer to coral and seawater samples, respectively. The number following the letter indicates a replicate. after 6 weeks of CO 2 exposure 18 . Moreover, Meron et al. 19 observed no significant changes in microbial communities associated with two Mediterranean coral species that were transplanted along natural pH gradients. A recent study reported that the microbial communities of two Pacific coral species were tolerant to reduced pH 7.9 (pCO 2 738-835 μ atm) 20 . These inconsistent results might reflect that some coral-microbial associations are more resistant to increases in pCO 2 /decreases in pH than others, but these findings could also be partially attributed to differences in the experimental conditions (e.g., field vs laboratory, pCO 2 or pH level, among others), and the exposure duration.
In most cases, microbial communities are dynamic and can rapidly respond to OA in seawater 39 , biofilms 40 and other associated systems 41 . The genus Acropora is among the most sensitive coral to environmental change 42 . The potential for coral acclimatization or adaptation to climate change has been studied 43 , and the physiological and molecular mechanisms responsible for OA resistance have recently been proposed 8 . Although there is limited evidence for biological adaptation to climate change in coral microbial symbionts, the adaptive power to climate change has been well documented in the photosymbiotic Symbiodinium 10,11,44 . The shallow habitat of the coral A. gemmifera sampled in the present study has been experiencing regular large diurnal and seasonal variations in pH/pCO 2 (see Supplementary Table S1), which are mainly driven by biological activities of the reef [28][29][30] . Therefore, it is most likely that microbial communities harbored by the natural population of A. gemmifera are resistant to the increased pCO 2 , due to long-term acclimatization/adaptation to the highly dynamic pH conditions within the reef flat. Thus, there may be a resilient relationship between coral and microbial partners that can help corals overcome the fluctuations in seawater pH/pCO 2 . However, we note that the stability of the coral microbiome is based on only one species colleting from a fluctuating environment. The application of variable pCO 2 conditions and controls from stable pH/pCO 2 environments in highly replicated culture experiments with consideration of tank effects could further confirm this assumption in future studies.
A recent study supports this interpretation. Morrow et al. 21 found that microbial communities associated with coral and sponge originally from natural volcanic CO 2 seeps were distinct from the nearby control sites, reflecting the acclimatization of the host-symbiont to the high pCO 2 environment. Local acclimatization/adaptation to environmental variations in pCO 2 , temperature and nutrients, among others, has revealed the capabilities of marine organisms including reef-building corals and symbiotic algae, to adapt to future climate change 8  However, in general, the species-specific response of marine organisms to OA remains poorly understood 1,23,46 . Thus, it is rather premature to conclude whether we can extrapolate the adaptive power of coral and its associated microbes documented in the present study to other coral species living in highly fluctuating reef environments.
Skeletal isotopic response to ocean acidification. During our experiments, all blocks of A. gemmifera exposed to the different pCO 2 treatments grew, survived and formed new skeleton (see Supplementary Fig. S6), even at the high pCO 2 (pH reduced to 7.47). When the fast-growing coral A. gemmifera skeletal δ 13 C and δ 18 O were compared among the different pCO 2 treatments, the skeletal δ 13 C values in A. gemmifera were significantly different between any two pCO 2 treatments except between the control and the medium (Fig. 3). Skeletal δ 13 C values were depleted by 1.10‰ and 1.04‰ for the control vs. high pCO 2 and for the medium vs. high pCO 2 , respectively (one-way ANOVA, Tukey test, p < 0.05). Skeletal δ 18 O values in A. gemmifera were enriched with increased pCO 2 ; they were 0.55‰ and 0.38‰ heavier in response to high pCO 2 than those in the control and medium, respectively (one-way ANOVA, Tukey test, p < 0.05). Compared with the previous data 47 , skeletal δ 18 O values revealed greater depletion in fast-growing coral A. gemmifera, while the skeletal δ 13 C values remained within range. In addition, the relationship between skeletal δ 13 C and δ 18 O in the non-photosynthetic coral Tubastrea sp. deviated the most from those both in the photosynthetic corals A. gemmifera and Pavona sp. (Fig. 3).
The isotopic composition of the coral skeleton can be affected by metabolic isotope effects (e.g., photosynthesis and respiration) and kinetic isotope effects (e.g., the calcification process) 48 . The coral skeletal δ 13 C and δ 18 O were generally used as an effective proxy to study photosynthesis, respiration and calcification processes 48,49 . In general, photosynthesis and calcification can enrich skeletal δ 13 C but deplete skeletal δ 18 O due to isotope fractionation 47,49 . Compared non-photosynthetic (Tubastrea sp.) and photosynthetic (Pavona sp.) corals 47 , the relationship of δ 13 C vs. δ 18 O in Pavona sp. and A. gemmifera was different from that in Tubastrea sp. (non-photosynthetic coral) due to active photosynthesis. In addition, more δ 18 O deviation was observed in A. gemmifera than Tubastrea sp. and Pavona sp., mostly due to the highest growth rate in A. gemmifera (Fig. 3). Skeletal δ 13 C values in A. gemmifera were significantly depleted at the high pCO 2 , suggesting that the photosynthetic rates were much lower at the high pCO 2 than at the control pCO 2 . The variation in skeletal δ 18 O values of A. gemmifera was consistent with the findings of a previous study demonstrating an enrichment of δ 18 O in the coral skeleton in response to elevated pCO 2 49 . The coral calcification rate decreases under reduced pH conditions 9 , which corresponds to heavier skeletal δ 18 O, whereas low pCO 2 and higher pH lead to species with lighter δ 18 O because HCO 3 − is isotopically heavier than CO 3  . Consequently, the significantly enriched δ 18 O and more depleted δ 13 C in A. gemmifera observed herein may reflect slight reductions in photosynthesis and calcification at the high pCO 2 . It should be noted that A. gemmifera skeletal δ 13 C and δ 18 O values did not vary significantly at the medium pCO 2 , potentially because this stress level did not exceed its acclimatization range. These findings indicate that the coral A. gemmifera is able to acclimate to an acidifying ocean, even in the presence of a dramatically increasing atmospheric CO 2 concentration.
Although the mechanisms by which extremely high pCO 2 induces decreased photosynthesis and calcification efficiencies in A. gemmifera are unknown, several potential mechanisms have been proposed, such as photoinhibition and suppression of the carbon concentrating process 3,9 . Photosynthesis, calcification and other physiological processes in reef-building corals can be influenced by their microbial partners or vice versa under OA 17 . However, the microbial communities associated with A. gemmifera remained unchanged as a consequence of host physiological changes, further supporting our hypothesis that highly stable microbial associations are likely driven by local acclimatization/adaptation to the fluctuating environment. Alternatively, host physiological costs might result from a potentially increasing energy demand to maintain stable microbial assemblages at the extremely high pCO 2 that exceeds its tolerance level.
It has also been proposed that physiological differences among symbiotic algal phylotypes may influence the stable isotopic composition of coral skeleton 50 . Furthermore, the distinct mechanisms used to concentrate carbon by different Symbiodinium phylotypes and their physiological responses to OA are phylotype-specific 51 . For example, Symbiodnium community shifts may occur in response to environmental stresses 10,11 . In the present study, we did not investigate Symbiodinium phylotypes associated with A. gemmifera. A recent study found no changes in Symbiodinium phylotypes associated with corals among different pH conditions 19 , suggesting the presence of stable Symbiodinium assemblages in corals in response to OA. In general, a stable microbial partnership to maintain key metabolic functions can improve coral holobiont acclimatization or adaptation to environmental stresses 5 . However, the interactions between microbial communities and coral physiology remain far from clarified.

Conclusions
In this study, the tropical fast-growing coral A. gemmifera from a shallow habitat with natural pH/pCO 2 fluctuations was selected as a representative species and was exposed to a 4-week CO 2 treatment. The microbial communities and skeletal isotopic compositions were examined simultaneously. We found that the microbial communities in A. gemmifera remained remarkably stable. In contrast, neither photosynthesis nor calcification in the coral were impacted under medium pCO 2 but were both negatively affected under extremely high pCO 2 , as demonstrated by an enrichment of δ 18 O and increased depletion of δ 13 C in the skeleton under extremely high CO 2 stress. The present findings indicate that some reef-building corals may be more tolerant to OA in pH/pCO 2 fluctuating environments and have a high degree of host-symbiont fidelity, despite the observed impairment of host physiological processes in response to high CO 2 stress. This study also contributes to our understanding of the variability of OA resistance among coral-microbial associations. Because coral reefs are facing other environmental stresses in addition to OA, the synergistic effects of multiple stressors on the coral microbiome must be carefully examined to understand the persistence of the coral holobiont and coral reefs in the future ocean.

Methods
Experimental design and sample collection. The experiment was conducted in an outdoor seawater flow-through system at the Tropical Marine Biological Research Station in Hainan (TMBRS) near the Luhuitou fringing reef. Seawater was pumped directly from a depth of 6 m in the front of the TMBRS into three 2000-L header tanks in which the designated pCO 2 level was adjusted with CO 2 gas. Three pCO 2 treatments were projected current pCO 2 levels, those at the end of the present century, and double the end of the present century: 421 μ atm, 923 μ atm, and 2070 μ atm with pH values of 8.07, 7.76 and 7.47, respectively. The well-mixed seawater from the header tank continually flowed into three aquaria at a rate of 0.5 l min −1 . Each aquaria was equipped with a submerged pump to drive the water flow, and all aquaria were maintained under a natural light-dark cycle to mimic the field condition.
Six healthy colonies of A. gemmifera were collected from the Luhuitou fringing reef flat at a depth of ~2 m in May 2014 and divided into small pieces. After acclimation for two weeks in large aquaria with running water, one coral nubbin was randomly selected and was suspended using fine nylon strings in three aquaria for exposure to each of the three pCO 2 treatments. The coral nubbins from same colony were evenly distributed among the pCO 2 treatments to avoid any possible sampling bias. A total of 9 coral nubbins were maintained for further analysis during the experiment.
At the end of the experiment (i.e., 4-week exposure), 1 L of seawater from each treatment was filtered through 0.2 μ m polycarbonate (PC) membrane filters and stored at − 20 °C for further analyses. One coral nubbin from each treatment aquarium was sampled and divided into two aliquots. One was rinsed three times and then preserved in 70% ethanol and stored at − 20 °C for DNA extraction; the other was used for stable isotope analyses.
Determination of environmental parameters. Photosynthetically active radiation (PAR) in each aquarium was recorded every 30 min below the seawater surface using the Hobo ® logger (Onset, USA). The average diurnal variations in PAR during the 4-week period are shown in Fig. S7. The seawater pH and temperature were measured daily in each aquarium using a pH meter (Orion Star ™ ) and the total alkalinity was also determined weekly using an automated titration system (Metrohm 877 Titrino plus, Switzerland ], pCO 2 and aragonite saturation state (Ω A ), were calculated from the measured pH and total alkalinity values using the CO2SYS program 52 (Table 2). DNA extraction and 16S rDNA amplicon sequencing. A preserved piece of coral was homogenized in liquid nitrogen and then the total DNA from the resulting coral powder and filtered seawater samples was subsequently extracted using the Fast DNA ® SPIN Kit for Soil (MP Biomedicals, Irvine, CA) according to the company's protocol. The DNA samples were amplified by PCR using barcoded primers targeting the hypervariable region V3-V4 of the 16S rRNA gene of Bacteria and Archaea: 341F (5′ -CCTAYGGGRBGCASCAG-3′ ) and 802R (5′ -TACNVGGGTATCTAATCC-3′ ) 53 . The PCR amplification was performed using a thermocycle controller (MJ Research Inc., Bio-Rad) with the following program: an initial denaturation at 94 °C for 5 min, followed by 35 cycles at 94 °C for 30 s, 50 °C for 30 s, 72 °C for 30 s, and a final extension at 72 °C for 5 min. All PCR products were purified using the Qiagen Agarose Gel DNA Purification Kit (Qiagen, Germany) and quantified using a NanoDrop device (Thermo Scientific, USA). All amplicon products were mixed at equal concentrations and sequenced on an Illumina Miseq platform using 2 × 300 bp mode at Novogene (Beijing, China). The raw reads were submitted to the NCBI Sequence Read Archive under accession number SRP066229 (SRR2917919).
Sequence data processing. Overlapping paired-end reads were merged to obtain full-length 16S V3-V4 fragments using PEAR 54 . After de-multiplexing and quality control, the downstream bioinformatics analysis was performed with QIIME1.5.0 pipelines 55 . Briefly, OTUs with 97% similarity were defined after the qualified reads were clustered using Uclust 56 . Representative sequences for each OTU were assigned to different taxa using the Ribosomal Database Project (RDP) classifier version 2.2 57 against the SILVA108 database 58 with a 50% cut-off threshold. Representatives assigned to eukaryotes, chloroplasts and mitochondria were filtered out. The taxon and abundance were summarized at the phylum, class, order, family, and genus levels. The species diversity, Shannon index, rarefaction curves and rank-abundance curves were determined using the QIIME pipeline.
Stable isotope analyses. Skeleton fragments of A. gemmifera were soaked in 30% hydrogen peroxide to remove coral tissue and then sonicated for 4 min at 20 °C. Skeletons were subsequently washed several times with double-distilled water and dried overnight at 50 °C. The newly grown part was scalped and ground into powder. Coral skeletal δ 13 C and δ 18 O data were obtained using a Finnigan MAT 253 Isotope Ratio Mass Spectrometer coupled to a Kiel Carbonate Device IV at the South China Sea Institute of Oceanology, Chinese Academy of Sciences, China. δ 13 C and δ 18 O were determined by repeated measurements of the international reference standard NBS-18.

Statistical analyses.
To test the effect of pCO 2 treatments on microbial community compositions of coral samples, after normalizing all sequence reads for taxonomic analysis to the lowest sequencing depth (19,098 reads), pairwise dissimilarities among coral samples were calculated based on the Bray-Curtis index for ' Adonis' , which is a non-parametric multivariate analysis of variance. In the Adonis analysis, the distance matrix was the response variable with pCO 2 treatment as independent variable. Non-metric multidimensional scaling (nMDS) was also performed to visualize the dissimilarities. Genera making the greatest contribution to dissimilarity among the pCO 2 treatments were further investigated through similarity percentage (SIMPER) analysis. To compare coral skeletal δ 13 C and δ 18 O among pCO 2 treatments, one-way ANOVA and Tukey's test were employed. All analyses were conducted using the package vegan within the R statistical environment (version 3.1.3, R Development Core Team, 2015).