Temperature shapes coral-algal symbiosis in the South China Sea

With the increase in sea surface temperature (SST), scleractinian corals are exposed to bleaching threats but may possess certain flexibilities in terms of their associations with symbiotic algae. Previous studies have shown a close symbiosis between coral the and Symbiodinium; however, the spatial variation of the symbiosis and the attribution underlying are not well understood. In the present study, we examined coral-algal symbiosis in Galaxea fascicularis and Montipora spp. from three biogeographic regions across ~10° of latitude in the South China Sea. Analysis of similarities (ANOSIM) indicated a highly flexible coral-algal symbiosis in both G. fascicularis and Montipora spp. and canonical correspondence analysis (CCA) showed that temperature explained 83.2% and 60.1% of the explanatory subclade variations in G. fascicularis and Montipora spp., respectively, which suggested that temperature was the main environmental factor contributing to the diversity of Symbiodinium across the three regions. The geographic specificity of the Symbiodinium phylogeny was identified, revealing possible environmental selection across the three regions. These results suggest that scleractinian corals may have the ability to regulate Symbiodinium community structures under different temperatures and thus be able to adapt to gradual climate change.

Province, have a tropical climate with SST ranges from 20 to 30 °C and from 24 to 30 °C, respectively. Coral reefs in Sanya are intermediate and are located near the northern border of the global distribution of tropical coral reefs 18 , whereas coral reefs in Sansha are typically tropical. In general, the average annual SST rises across the three regions with changes in latitude (Fig. 1). The impact of human activities on the five sampling sites is as follows: Lamma Island (LI) > Crescent Bay (CB), Luhuitou (LHT) > Sunny Bay (SB) > Drummond Island (DI). Crescent Bay is located in the northeast of Hong Kong, where there is marginal influence by the Pearl River and the greatest coral cover (30-50%) in Hong Kong has been observed. In contrast, Lamma Island in the southwest of Hong Kong, has a relatively low coral cover (10-30%) due to the influence of the Pearl River 16 . The coral cover in Luhuitou of Sanya decreased from 80-90% in the 1960 s to 11% in 2007 due to frequent human activities; the coral cover around Sunny Bay in Sanya was approximately 35% in the period from 2007-2009 19 . Drummond Island in the Sansha region is one of the Xisha Islands with a coral cover of approximately 50% based on 30-year data from the 1970 s 20 . We selected these five sampling sites with different environmental conditions to examine the change of coral-algal partnerships in response to environmental variations and to identify the environmental factor(s) with the greatest impact.
The diversity of Symbiodnium has mostly been examined by denaturing gradient gel electrophoresis (DGGE) based analysis, which is unable to detect Symbiodinium subclades with a relative abundance of less than 10% and thus potentially lacks the sufficient sensitivity to detect symbiosis change 7 . The development of highly sensitive molecular techniques, such as high-throughput DNA sequencing, would facilitate the identification of low abundance Symbiodinium that benefit coral hosts in terms of tolerant to climate change. Additionally, amplicon sequencing using ITS2 primers has been shown to be reliable for measurements of Symbiodinium communities 10 . In the present study, to explore spatial variations of coral-algal symbiosis, we used ITS2 amplicon sequencing to examine coral-algal symbiosis in G. fascicularis and Montipora spp. from five sampling sites across Among all samples, only two seawater samples from Sansha contained 0.06% and 0.6% A clade, respectively. The total relative abundance of clade A, B and F in samples were all less than 0.01%, thus only Symbiodinium clade C and D were kept for further analysis. As shown in Fig. 2, G. fascicularis in Sanya and Sansha had a high symbiont specificity for clade D, whereas G. fascicularis in Hong Kong showed a high symbiont specificity for clade C. In contrast, in all sampling sites, Montipora spp. contained a very low abundance of clade D but a high abundance of clade C. All of the seawater samples were dominated by clade C, suggesting Symbiodinium selection and enrichment of G. fascicularis in Sanya and Sansha. At the subclade level, 13 dominant/subdominant subclades were found, accounting for more than 85% of the total sequences (Fig. 3). These subclades were picked as their relative abundances were > 10% in at least one sample or their average relative abundances in all samples were at least around 1%. In G. fascicularis, C161, C21a, C2r, D17 were regarded as dominant subclades with abundances > 10%, whereas C163b, C3d and D2 were subdominant with abundances ranging from 1-10%. In Montipora spp., a total of 8 subclades, C116, C15h, C17, C35, C26, C2r, C31 and C3d, could become dominant dependent on the location. Based on an overview of Fig. 3, although Symbiodinium in Montipora spp. did not differ much across regions at the clade level, M. monasteriata exhibited more different Symbiodinium subclade compositions in two sites of Sanya (LHT and SB), whereas Montipora spp. still displayed very similar Symbiodinium subclade compositions in two sites of Hong Kong (CB and LI), supporting the notion that the coral-algal symbiosis is very flexible across host species and geographic distance. The non-parametric multivariate analysis of similarity (ANOSIM) demonstrated that Symbiodinium community structures differed significantly between G. fascicularis and Montipora spp., as well as between coral and seawater samples (R > 0.8, P = 0.01). Moreover, the Symbiodinium community structures among the three regions (Hong Kong, Sanya, and Sansha) were significantly different (R > 0.69, P ≤ 0.03), whereas less significant variations were observed between sampling sites in the same region (CB and LI (Hong Kong), R = 0.257; LHT and SB (Sanya), R = 0.559). These findings indicated a possible environmental adaption across regions and strict host specificity of the Symbiodinium community. Furthermore, most of the subclades identified in corals were also found in the surrounding seawater, suggesting the free-living ability of Symbiodinium and a possible horizontal transmission of Symbiodinium between seawater and corals. Some dominant subclades demonstrated much higher relative abundances in corals compared with those in seawater, suggesting an enrichment of certain Symbiodinium subclades in corals, which is consistent with the findings at the clade level.
The results of non-metric multidimensional scaling (nMDS) revealed a clear clustering pattern, supporting the dissimilarities among all samples (Fig. 4). All of the seawater samples were located between the G. The Symbiodinium communities of Montipora spp. were less different within the same region (CB and LI (Hong Kong), LHT and SB (Sanya)), but exhibited more differences between different regions (Hong Kong and Sanya). G. fascicularis in Hong Kong and all Montipora spp. were dominated by diverse C subclades, whereas G. fascicularis in Sanya and Sansha mainly harbored two D subclades.
In summary, different coral species at the same sampling site appeared to prefer different Symbiodinium subclades. Congeneric corals from different regions also tended to have significantly distinct Symbiodinium community compositions despite fewer changes in Symbiodinium community compositions of the surrounding seawater. These results suggested that, for congeneric corals, coral-algal symbiosis change mainly resulted from environmental changes across the investigated regions.

Correlations between environmental factors and Symbiodinium communities.
According to the canonical correspondence analysis (CCA), temperature, salinity and NH 4 + served as the main factors among all potential impact factors (Fig. S2, Supporting Information) affecting Symbiodinium subclades and community compositions, explaining 81.8% and 52.7% of the total subclade variations in G. fascicularis (Fig. 5A) and Montipora spp. (Fig. 5B), respectively. This finding indicated that these three factors provided the largest contribution to coral-algal symbiosis change in the present study. Remarkably, temperature was the most significant factor in shaping subclade compositions, explaining 83.2% and 60.1% of the explanatory variations in G. fascicularis and Montipora spp., respectively. This finding suggested that coral-algal symbiosis change in studied samples was largely driven by temperature. In G. fascicularis, two D subclades (D2, D17) were positively correlated to temperature, while most C subclades had negative correlation with temperature, indicating that G. fascicularis was more likely to select heat-tolerant Symbiodinium with an increasing temperature because the D clade is believed to be more tolerant to heat 21,22 . In Montipora spp., only three C subclades (C2r, C3d and C163b) were negatively correlated to temperature, whereas the other dominant subclades were all correlated positively to temperature. In both plots, samples from Hong Kong (CB, LI) were clearly separated from those from Sanya and Sansha along axis 1, which was highly correlated to temperature, indicating that temperature is a major factor in shaping Symbiodinium community structures between subtropical (Hong Kong) and tropical (Sanya and Sansha) regions. Thus, temperature provided the greatest environmental contribution to the coral-algal change between subtropical and tropical regions. However, the samples from the tropical region were divided along axis 2, which was mainly correlated to kinds of nutrients, suggesting that nutrients are minor environmental factors affecting Symbiodinium community structures under similar temperature conditions.

Phylogenetic relationships of dominant Symbiodinium subclades from G. fascicularis and
Montipora spp. in the three regions. Symbiodinium subclades dominating congeneric corals tended to be closely distributed in the phylogenetic tree (Fig. 6). For example, C161, C2r, and C163b from G. fascicularis in Hong Kong had closer phylogenetic relationships with D2/D17 from G. fascicularis in Sanya and Sansha compared with the other C subclades that were symbiotic with Montipora spp., suggesting a possible coral-algal coevolution. For the congeneric corals, Symbiodinium associated with corals with a closer geographic distance demonstrated closer phylogenetic relationships, revealing that the phylogenetic relationships of Symbiodinium were linked to geographical distances.

Discussion
Here, we presented the first systematic investigation of Symbiodinium communities in different coral species across tropical and subtropical regions and indicated the highly spatial flexibility of coral-algal symbiosis in G. fascicularis and Montipora spp. in the South China Sea. Although several previous studies have investigated Symbiodinium community structures of certain scleractinian corals in the South China Sea 23,24 , this is the first study to demonstrate that, among all potential environmental factors, temperature plays the major role in shaping Symbiodinium community structures and coral-algal symbiosis. The results of the present study are consistent, in part, with some previous findings. Temperature 9,25,26 , nutrients 27 , and depth 28 all have the potential to affect either the coral host or Symbiodinium community structures, but none of those studies compared the relative effects of these factors or their effects on different coral species. Cooper et al. 11 showed, mud content, carbonate content, SST and depth together explained 51.3% of the total variations in dominant Symbiodinium communities in A. millepora and long-term SST explained 10.8% of the total variations. However, they only examined Symbiodinium C1, C2 and D1, whereas we detected 131 Symbiodinium subclades and 13 dominant Symbiodinium subclades in the present study, suggesting a more in-depth coverage of the Symbiodinium communities.
Temperature shapes the structures of the Symbiodinium community. The concordance of the taxonomic data with phylogenetic and ecological evidence led us to propose that, among all potential environmental factors, temperature acted as the major environmental factor impacting coral-algal symbiosis spatial change in the South China Sea. The important role of temperature and the underlying mechanisms can be partly explained by previous empirical evidence. For example, extremely high temperatures can lead to coral bleaching 29 , but corals that contained more heat-tolerant Symbiodinium were less likely to bleach and bleached corals gained more heat-tolerant Symbiodinium after recovery 22,30 . Coral larvae gained different Symbiodinium when treated with different temperatures 31 , whereas adult corals acquired more heat-tolerant Symbiodinium in response to thermal stress 9 . Additionally, corals from some extremely hot regions have been found to carry some highly heat-tolerant Symbiodinium 15 . These findings support the conclusion that temperature is a significant factor affecting coral-algal symbiosis. Using a network analysis of stressor interactions, Ban et al. 12 suggested that sedimentation, storms and Every symbol represents a group in which the average relative abundance of a certain subclade is > 1%. Solid and non-solid symbols indicate average relative abundances of > 10% and 1-10%, respectively. The phylograms were analyzed with ITS2 sequences using maximum likelihood and Bayesian inference analyses, and nearly identical profiles were obtained. The presented tree is derived from Bayesian inference.
Scientific RepoRts | 7:40118 | DOI: 10.1038/srep40118 temperature were the most influential stressors in the coral reef ecosystem, which also indicated that temperature could drive coral-algal symbiosis change and play an important role in the establishment of coral-algal symbiosis.
Hong Kong environmental data was collected from HKEPD which was monitored less than one week before the sampling time. As Symbiodinium community in a given coral colony can remain relatively stable in a period of time 32 , results shall not be affected much. Sansha sampling was finished two months later than the others, it would cause problems when comparing communities between different sites, while Symbiodinium communities in G. fascicularis from Sansha were still highly similar with those in G. fascicularis from Sanya, suggesting Symbiodinium communities in G. fascicularis from Sansha were relatively stable during the two months. The high dissimilarities between Symbiodinium communities from Hong Kong and Sansha/Sanya due to temperature could result from either spatial or temporal variations, as averaged SST in Hong Kong rose ~2 °C/~10 °C from March to April (Sanya sampling)/June (Sansha sampling), respectively. It's possible that if Hong Kong and Sanya samplings are both finished in June, there would be less differences between Symbiodinium communities in G. fascicularis from Hong Kong and Sansha, and Symbiodinium communities in G. fascicularis from Sanya and Sansha would be more similar. However, spatial variation still played a more important role in the present study: Symbiodinium communities in G. fascicularis from Sanya and Sansha were highly similar even Sansha sampling was two months late; while Symbiodinium communities in G. fascicularis from Hong Kong and Sanya were quite different even sampled at the similar dates. The results of the relationships between Symbiodinium communities and environmental conditions would not be affected by sampling at different time points as long as the environmental data can well reflect the environmental conditions at the sampling points.

Potential roles of Symbiodinium in the coral host in response to bleaching and future climate change. Our results support an argument of coevolution between the coral host and Symbiodinium because
Symbiodinium from congeneric corals demonstrated closer phylogenetic relationships and a geographic specificity was observed in the Symbiodinium phylogeny. While we found three C subclades have closer phylogenetic relationships with D subclades than other C subclades, it might result from classification errors, as perfect matches of C2r and C163b were not found in NCBI database and the perfect match of C161 was not identified as Symbiodinium C clade in NCBI. Symbiodinium clade C is believed to possess a higher photosynthetic efficiency, whereas Symbiodinium clade D is more heat tolerant 33 . Therefore, it can be assumed that the Symbiodinium may contribute to the host adaptation and survival in response to fluctuating environmental conditions. Dixon et al. 34 found that Acropora millepora larvae derived from warmer locations possessed greater thermal tolerance, indicating that thermal tolerance was heritable 34 . Cooper et al. 11 observed that adult A. millepora acquired a greater number of heat-tolerant Symbiodinium in response to higher temperatures, suggesting that a potentially thermal-tolerant coral would prefer thermal-tolerant Symbiodinium subclades during growth to facilitate their future survival during exposure to relatively high temperatures. However, Palumbi et al. 35 found that A. hyacinthus became more heat tolerant without changes in symbiont clade composition as well as gene expression 35 . Thus, different coral species may possess different heat tolerance mechanisms.
Tchernov et al. 36 showed that a rise in temperature damaged the membranes of free-living Symbiodinium and thus disrupted their photosynthetic ability, ultimately leading to death, and that the compositions of membranes determined the thermal stress sensitivity of Symbiodinium 36 . In the present study, we found that as the temperature rose across the regions, the relative abundance of some heat-sensitive Symbiodinium C subclades, such as C116, C17, decreased in seawater. In contrast, some Symbiodinium that are supposed to be heat sensitive according to the clade characteristics, such as C31, were not greatly affected much across the regions and C31 was found to have a high relative abundance in Montipora spp. from Sanya. These findings indicated that some Symbiodinium might possess compensation mechanisms to survive under thermal stress that could be taken advantage by the coral host to adapt to future climate change. However, previous studies have shown that some coral species acquired their Symbiodinium, e.g., C31, through vertical transmission 37,38 . In the present study, we also detected C15h, C21a, C35 and C26 in coral samples but not in ambient seawater samples in certain locations, either due to a potential vertical transmission or to an insufficient sensitivity of the current techniques. We suspect that if a vertically transmitted Symbiodinium fails to develop compensation mechanisms, coral host would become more susceptible to future climate change.
G. fascicularis was found to have greater heat tolerance than Montipora sp. 39 , and similarly in our study, very few members of the Symbiodinium D clade were detected among the Montipora spp., even in Sanya, where temperature is relatively high. This observation raises the possibility that heat-tolerant Symbiodinium facilitates coral adaptation. In response to global warming with a predicted increase in bleaching events, coral species that can acquire more heat-tolerant Symbiodinium will survive better. Coral species that lack the symbiotic flexibility with Symbiodinium subclades have also been found to be more susceptible to bleaching 29 . However, some heat-tolerant Symbiodinium can reduce host calcification and photosynthetic efficiency, which may impair the competitive ability of the coral 26,33 . While Cunning et al. 40 demonstrated that the growth disadvantage of heat-tolerant Symbiodinium could be reduced by elevated temperature, thus enhancing the survival of coral host. More investigations are needed to decipher the mechanisms that how coral species with different Symbiodinium survive during future climate change.
Pettay et al. 33 found that Symbiodinium trenchii invaded the Caribbean and noted that this Symbiodinium might drive coral-algal symbiosis dynamics during climate change, which is inconsistent with our findings. According to the present study, all Symbiodinium (including Symbiodinium trenchii) could potentially be transmitted between corals and surrounding seawater, and certain subclades of Symbiodinium were enriched in the coral host. For certain Symbiodinium subclades such as C161, congeneric corals in different regions surrounded by a similar relative abundance of certain Symbiodinium are internally enriched with different relative abundances of that Symbiodinium. Therefore, the acquisition of Symbiodinium by corals shall not be a random but a selective event. Likely, corals are capable of selecting certain Symbiodinium that are potentially more beneficial to them, Scientific RepoRts | 7:40118 | DOI: 10.1038/srep40118 and the selected Symbiodinium facilitate the adaptation of the host. Thus, the coral host drives the dynamics of the symbiosis. However, we cannot exclude the possibility that, in certain coral species, invading Symbiodinium may drive the dynamics of the symbiosis.
Ecological implications. The adaptation of the coral reef ecosystem to future climate change strongly relies on the adaptation of their internal Symbiodinium community. In the South China Sea, among all potential environmental factors, temperature can serve as a major factor for shaping coral-algal symbiosis. Therefore, a rise in SST is a prominent threat to the South China Sea corals, but still corals have the potential to adapt to future climate change with their flexibility of symbiosis with Symbiodinium. Although more caution should be adhered to when transplanting a new species into a region 33 , the introduction of more heat-tolerant Symbiodinium into the coral reef ecosystem may facilitate their adaptation to the rising SST 41 .
Previous studies have shown that heat-tolerant corals contain a greater number of clade D Symbiodinium 42 . Although both G. fascicularis and Montipora spp. have the potential to change their Symbiodinium community structures in response to environmental changes, G. fascicularis appears to prefer more heat-tolerant Symbiodinium subclades, which may lead to its dominance during future climate change. Indeed, in recent years some dominant coral species have been declining while some subdominant coral species are becoming more dominant 29,43 . During the development of temperature-driven symbiosis, the flexibility of coral-algal symbiosis and preference for certain Symbiodinium of coral species can directly determine coral species' competition during future climate change, which will ultimately affect future coral population structures.
However, human activities such as seawater pollution, overfishing, and tourism etc. have also caused a rapid decline in the coral cover 44,45 . In the present study, nutrients had only a marginal impact on coral-algal symbiosis, considering nutrients contribute the most to coral cover and species richness loss compared with other water quality parameters 46 , suggesting a lack of symbiotic flexibility in response to seawater pollution such as eutrophication. Increases in coastal nutrients are frequently linked to human activities 47,48 , which suggests that human activities can also affect coral-algal symbiosis. This observation indicates that seawater pollution may be another important threat to the coral reef.

Conclusions
The present study demonstrated that temperature drove coral-algal symbiosis spatial change in the South China Sea and corals had the potential to adapt to future climate change with selecting more heat-tolerant Symbiodinium under gradually rising SST. Thus, for the future conservation of corals in the South China Sea, the introduction to coral hosts of heat-tolerant Symbiodinium might facilitate their adaptation to future climate change. Despite the high level of spatial flexibility of coral-algal symbiosis in G. fascicularis and Montipora spp. in the South China Sea, future studies are needed on a larger geographic scale and shall be expanded to other dominant coral species like Acropora spp. to further address coral-algal symbiosis in response to future climate change.  (Table 1). At each sampling site, one small piece (~1 cm × 1 cm) from each healthy coral colony was collected using a hammer/chisel set and wrapped in a tagged bag filled with seawater. On board, the coral pieces were immediately washed with filtered seawater, fixed in 70% ethanol, stored and transported to the laboratory using a cool box containing dry ice. The seawater surrounding the coral colonies was also collected as controls. Freeliving Symbiodinium cells in seawater were collected by filtering seawater through 0.22-μ m polycarbonate membranes and fixed with 50% ethanol. All of the fixed coral and seawater samples were stored at − 30 °C until DNA extraction.

Methods
Environmental data collection. Hong Kong environmental data were provided by the Hong Kong Environmental Protection Department (HKEPD). Data for the sampling month from five selected monitoring stations around each sampling site (Table S1, Table S2, Supporting Information) were collected and averaged, and the averages were used as environmental factor data for a given sampling site. For Sanya and Sansha, the seawater temperature, salinity and depth were measured during the sampling in situ with CTD (Idronaut, Italy), while dissolved oxygen (DO) was assessed in situ using a YSI 6600V2-02 multi-parameter instrument (YSI, USA). The seawater samples were collected and immediately filtered (Whatman GF/F, 47 mm) to analyze dissolved nutrients (nitrite, nitrate, ammonia and phosphate) using a Lachat QC8500 Flow Injection Autoanalyzer (Lachat Instruments, USA).

DNA extraction and amplicon sequencing.
A small fragment (~0.5 cm × 0.5 cm) cut from each fixed piece of coral was first rinsed with 1*PBSE (137 mM NaCl, 2.7 mM KCl, 4.3 mM Na 2 HPO 4 ·7H 2 O, 1.4 mM KH 2 PO 4 and 10 mM EDTA) and then mashed in 1*PBSE using a mortar and a pestle. The mashed coral serous was centrifuged at 12000 g, and the pellets were collected for DNA extraction. DNA was extracted from each pellet using the FastDNA ® Spin Kit for Soil (MP Biomedicals, France) following the protocol provided with the kit. After quality and purity examinations, the extracted DNA samples were applied as PCR templates. The Symbiodinium ITS2 region of the rDNA was amplified by PCR using primers (F: 5′ GAATTGCAGAACTCCGTG-3′ ; R: 5′ GGATCCATATGCTTAAGTTCAGCGGGT-3′ ) designed to produce 330-360 bp ITS2 fragments 49 . At the 5′ terminus of the forward primer, six-nucleotide unique barcodes were attached for multiplexed sequencing. The PCR amplifications were conducted in a 50-μ L reaction volume containing ~50 ng of DNA, 25 μ L of 2× Taq Platinum PCR Master (Tiangen, China), 200 nM of each primer, and ddH 2 O up to the final volume. The reactions were performed under the following conditions: 94 °C for 5 min, followed by 35 cycles of 94 °C for 30 s, 51 °C for 30 s, 72 °C for 30 s, and a final extension at 72 °C for 5 min. Each sample was amplified in three independent reactions to minimize potential PCR bias, and then purified using the PureLink ® PCR Purification Kit (Invitrogen, USA). The purified PCR products were quantified using a Thermo NanoDrop 2000 UV-Vis Spectrophotometer and then mixed based on equal mass for subsequent multiplexed amplicon sequencing. The final DNA samples were sequenced following a paired-end (PE) 300 bp × 2 strategy on an Illumina MiSeq sequencer operated by the Novogene company (Beijing, China). The sequencing datasets were submitted to the NCBI Sequence Read Archive under accession number SRP066283.
Data processing, ITS2 database establishment and data analysis. Strict quality control and sequence filtration were applied to ensure the accuracy of the following analysis. Adaptors, short reads, and low quality reads were first removed by the sequencing company. The PEAR (paired-end read merger) tool was applied to obtain full-length ITS2 rDNA fragments with merging overlapping PE reads to generate ITS2 sequences 50 (http://sco.h-its.org/exelixis/web/software/pear/, see commands in Supporting Information). ITS2 tags were demultiplexed into all samples in the QIIME platform 51 by identifying unique barcodes (http://qiime. org/index.html, see commands in Supporting Information). We found that the previous ITS2 database included some duplicate sequences 52 , therefore we uploaded the previous ITS2 database to CD-HIT Suite website (http:// weizhongli-lab.org/cdhit_suite/cgi-bin/index.cgi ? cmd=cd-hit-est), set sequence identity cut-off as 100%, compared both strands, set all other parameters as default 53 , removed the duplicates, merged the annotations and used the results to establish a non-redundant ITS2 database (ITS2 Database.doc, Supporting Information). Next, all of the sequences were aligned to the ITS2 database by BLASTN 54 based on E-value of 1e-5 (see commands in Supporting Information), the output were then filtered in Microsoft Excel (Microsoft Corp., USA) based on 97% similarity for Symbiodinium subclade identification. For ITS2 rDNA, 97% sequence similarity has been proved to be reliable for Symbiodinium subclade classification and could permit differences of ten base pairs for identification within the same Symbiodinium subclade as a consequence of intra-genomic sequence divergence 10 .
Thirteen (sub)dominated Symbiodinium subclades were picked from 131 total aligned Symbiodinium subclades for following analysis. The relative abundances of picked Symbiodinium subclades were > 10% in at least one sample or their average relative abundances in all samples were at least around 1% and the filtering results can indicate the original communities well (whole_community_data.xls, Supporting Information). To profile Symbiodinium communities in different samples, a heat map and scatter plots were created with OriginPro 2015 (OriginLab, USA). To justify significances between Symbiodinium community structures across the sampling sites, a two-way ANOSIM and a permutational multivariate analysis of variance (PERMANOVA) were performed using PRIMER 7 software 55 . To present the relationships of the Symbiodinium community structures from different sampling sites, a nMDS ordination as well as clustering analysis were developed based on the square root-transformed Bray-Curtis similarity matrix of Symbiodinium profiles (Symbiodinium relative abundance) in PRIMER. To test the relationships among environmental factors, Symbiodinium community, and sampling sites, CCA was conducted in Canoco 5 56 . CCA uses ecological data sets to extract synthetic environmental gradients which are the basis for revealing the differential habitat preferences of microbes and has been proved to be reliable when combining spatial and seasonal sampling together, as long as the sample size is granted and the environment conditions can reflect habitat preference of the communities 57,58 . To identify the phylogenetic relationships among the dominant Symbiodinium subclades found in this study, two phylogenetic trees were constructed based on the Kimura 2-parameter model with uniform rates among sites using Maximum Likelihood in MEGA 6 59 and Bayesian Inference in Mrbayes 60 , respectively.