Samples of the scleractinian coral G. fascicularis were collected at depths between 2 and 4 m from the coast of Hainan Island in the South China Sea in September 2010 (Figure 1). At each location, six to seven colonies separated by at least 5 m were collected and fragments of approximately 4 cm² were picked and preserved in 95% ethyl alcohol at field temperature and stored at -20°C until DNA extraction took place.

Total DNA was extracted as described previously (Zhou et al., 2012). The quality and quantity of the DNA were determined with a NanoDrop spectrophotometer (Thermo Fisher Scientific, United States). Purified DNA samples were stored at -20°C for future use.

All samples were PCR amplified using a pair of barcoded Symbiodinium-specific primers: ITSintfor2 (5^′-GAATTGCAGAACTCCGTG-3^′) and ITS2-reverse (5^′-GGGATCCATATGCTTAAGTTCAGCGGGT-3^′) (Lajeunesse and Trench, 2000) targeting the ITS2 region of the ribosomal RNA gene for Symbiodinum. PCR amplification was carried out on a thermocycle controller (Bio-Rad, United States) with the following program: initial denaturing at 94°C for 5 min; 35 cycles at 94°C for 30 s, 51°C for 30 s, and 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 (Qiangen, China) and quantified with the NanoDrop spectrophotometer. All amplification products were mixed in equal amount followed by sequencing on an Illumina Miseq platform using the 2 × 300 bp mode at Novogene (Beijing, China). The raw data were submitted to the NCBI Sequence Read Archive under accession number SRP066283.

Overlapping paired-end reads were merged to obtain fragments using PEAR (Zhang et al., 2014). After de-multiplexing and quality control, a custom BLAST Symbiodinium-specific database of ITS2 types was downloaded (Arif et al., 2014), containing 408 ITS2 sequences. For each sample, datasets were randomly subsampled to 10,411 sequences (the lowest read number) which were subsequently searched against the database using BLASTn. Sequences were assigned to the ITS2 types that gave the highest identity in the BLASTn hits (Tong et al., 2017). The resulting counts of Symbiodinium ITS2 types were merged for downstream statistical analysis.

Aqua-MODIS sea surface temperature (SST) and chlorophyll a concentration (Chl a) with a spatial resolution of 4 km at each sampling location from January 2006 to December 2010 were obtained from NASA¹.

The Shannon–Wiener (H^′) diversity index was calculated to assess the level of alpha-diversity across samples from different locations. One-way analysis of variance and post hoc Tukey’s HDS comparisons were conducted to test the significance of differences in diversity between sampling locations. The similarity of Symbiodinium assemblages was also characterized by non-metric multidimensional scaling (nMDS) using the Bray–Curtis distance metric after data transformation. Analysis of variance (ADONIS) was performed to test the significance of differences in Symbiodinium communities among different sampling locations. The significant relationship between environmental variables (SST and Chl a) and Symbiodinium community composition was assessed using Monte Carlo permutation methods. All statistical analysis were conducted using the vegan package (Oksanen et al., 2015) in the R software environment (R 3.1.2).

In total, 997,760 qualified sequences were obtained from 32 samples (10,411–51,626 sequences per sample). A total of 119 Symbiodinium ITS2 subclades were assigned based on alignment with the ITS2 database at the 97% similarity level, covering clades B, C, D, and F. Overall, clade C comprised the highest proportion of sequences (averagely 85.6%), followed by clade D (averagely 13.6%) and then rare clades B and F (Figure 2A). C2r and D17 were the most dominant ITS2 subclades representing > 99% of the sequences for all samples. All individual colonies contained multiple Symbiodinium subclades belonging to different clades. Despite a high number of distinct Symbiodinium types, most of them had abundances lower than 0.1% (Figure 2B), indicating that rare subclades are present in heterogeneous Symbiodinium assemblages.

It is believed that low abundances of cryptic Symbiodinium have largely been overlooked by conventional screening techniques (Mieog et al., 2009). An increasing body of evidence shows that highly diverse rare taxa with important ecological roles are prevalent elsewhere (Lynch and Neufeld, 2015) and are being increasingly explored in reef-building corals (Silverstein et al., 2012; Green et al., 2014; Quigley et al., 2014; Kennedy et al., 2015; Boulotte et al., 2016). The Symbiodinium types in clades B and F associated with G. fascicularis are unusual, which have rarely been reported from the South China Sea (Huang et al., 2011; Zhou and Huang, 2011; Zhou et al., 2012; Tong et al., 2017) or other regions (Lajeunesse et al., 2004, 2010a). The results presented here suggest that G. fascicularis exhibits a high cryptic diversity and flexibility in symbiotic associations. It has been shown that rare Symbiodinium types (e.g., type D1) have the potential to enable the coral host to resist heat stress through symbiont shuffling or switching (Lajeunesse et al., 2009; Silverstein et al., 2012, 2015; Bay et al., 2016; Boulotte et al., 2016). For example, Acropora can change its thermally tolerant symbiont abundance from rare to dominant in a response to heat stress (Berkelmans and Van Oppen, 2006). Community diversity and functional redundancy may contribute to the stability of community resistance and resilience (Oliver et al., 2015), which has been characterized in coral holobionts (Silverstein et al., 2012, 2015). Highly diverse and flexible Symbiodinium may facilitate the ability of G. fascicularis to survive successfully in various habitats they experience throughout the Indo-Pacific area. However, the real contribution of rare symbionts to the host coral and their ecological significance is still unclear and needs to be addressed in future (Lee et al., 2016).

The G. fascicularis holobiont can be viewed as a highly complex symbiotic system with the flexibility to associate with a wide range of Symbiodinium (Blackall et al., 2015). It has been suggested that the mode of transmission of symbionts can affect the flexibility of coral-algal symbiosis (Baker, 2003; Fabina et al., 2012). Therefore, horizontal transmission of endosymbionts in each generation may provide greater opportunities for G. fascicularis to obtain multiple symbionts from the external environment. However, emerging evidence shows that many corals can host multiple Symbiodinium subclades without correlation with the mode of transmission (van Oppen, 2004). In addition, other factors such as environmental variability, host recognition and maintenance mechanisms can also influence the flexibility of coral-algal associations (Baker, 2003; Rodriguez-Lanetty et al., 2006; Dunn and Weis, 2009). Moreover, the development of coral especially in early life stages has additional effects on symbiont acquisition and selection (Abrego et al., 2009; McIlroy and Coffroth, 2017; Zhou et al., 2017). For example, Abrego et al. (2009) demonstrated that the symbiont associations in juvenile Acropora are more flexible than those in adults.

No significant differences in the Shannon diversity index were detected among sampling locations (one-way ANOVA; F = 1.618, p = 0.198). However, there were significant differences in the Symbiodinium assemblages between sampling locations (Figure 3; ADONIS, p = 0.01), demonstrating that coral-algal symbiosis is highly flexible around Hainan Island. Importantly, these differences were attributed to changes in the relative abundance of existing Symbiodinium types in individuals. Of six colonies of G. fascicularis at Basuo, four were dominated by clade D with sparse clade C, whereas the others contained abundant clade C with rare clade D. At subclade level, D17 dominated at Basuo, whereas C2r dominated in other locations (Figure 4). The abundance of each of the D1a, D2, and D6 subclades was higher at Basuo than at other locations, indicating that these holobionts are likely to be locally adapted through shifts in symbiont community composition.

Sea surface temperatures and Chl a concentrations from 2006 to 2010 at each location were employed to investigate the relationships between environmental conditions and Symbiodinium communities (Figure 5). Monthly average SSTs at Mulan Bay, Long Bay and Dazhou Island decreased sharply in July, possibly due to the Qiongdong Upwelling in summer (Jing et al., 2015). Of the sample locations, Leigong Island experienced the largest annual fluctuation in monthly average SST (∼11°C). Annual average SSTs was highest at Basuo (26.4°C), followed by Dazhou Island (25.5°C), Long Bay (25.3°C), Leigong Island (24.7°C), and Mulan Bay (24.5°C). The monthly average Chl a concentrations of all the locations showed little variation throughout the year, but Long Bay and Dazhou Island had lower yearly average Chl a concentrations. Symbiodinium communities at Basuo was significantly correlated with spring average SSTs (Monte Carlo permutation test; p < 0.05), but there were no significant differences between Symbiodinium communities and Chl a concentrations (Monte Carlo permutation test; p > 0.05). G. fascicularis had a high specificity for Symbiodinium clade D at Basuo where annual average SSTs were the highest (Figure 5). In contrast, Chl a values at all locations did not show any patterns or trends consistent with observed Symbiodinium distribution patterns.

It is well-known that Symbiodinium clade D occurs more frequently in areas with high SST and high turbidity (e.g., Lajeunesse et al., 2010a; Keshavmurthy et al., 2012). Furthermore, it is evident that heat tolerant Symbiodinium can confer thermal tolerance to its host coral but at a cost of reduced growth rate, a decline in reproduction and increased susceptibility to disease (e.g., Little et al., 2004; Berkelmans and Van Oppen, 2006; Silverstein et al., 2015). It has been suggested that symbiont compositions may be regulated to maintain optimal benefit to the host in a given environment (Cunning et al., 2015a,b). Such trade-off mechanisms may depend on both biotic (e.g., host species, host ontogeny, and symbiont competition) and abiotic (e.g., temperature, light, and nutrients) factors, and may also involve stochastic processes and selective pressures. In the present study, relative high abundances of Symbiodinium clade D at Basuo where the annual average SST is the highest are possibly mediated by a cost-benefit trade-off by the coral G. fascicularis. These observations agree with our previous investigations (Huang et al., 2011; Tong et al., 2017), which reported that clade D in G. fascicularis was more prevalent in tropical locations than in subtropical locations from the South China Sea, which might be attributed to the latitudinal temperature gradients. It is known that temperature has a profound influence on the ecological structure of coral communities (Brown et al., 2004). Previous studies also found that temperature is the main determinant to the geographic distribution of Symbiodinium in both conspecific corals (Lajeunesse et al., 2010a) and local adaptation (Howells et al., 2012). We suggest that temperature is the main environmental factor influencing the spatial variability of Symbiodinium assemblages in G. fascicularis around Hainan Island. However, other undetermined environmental factors, such as light intensity, and nutrient levels may also contribute to the biogeographical patterns of host-Symbiodinium associations (Baker, 2003), which can be investigated thoroughly in the future.

A better understanding of the spatial patterns will allow us to predict how corals will respond to environmental change over time (Dunne et al., 2004). In the present study, it may also reflect the capacity of the coral G. fascicularis to respond to environmental disturbances (e.g., thermal bleaching) by shuffling its internal symbionts. Some coral-algal associations remain remarkably stable over time (Thornhill et al., 2009; Williams et al., 2015) or revert to their original status after thermal bleaching (Lajeunesse et al., 2010b), which can be explained by the trade-off mechanism (Cunning et al., 2015b). More recently, it has been shown that symbiont shuffling in reef-building corals is attributed to the magnitude of the disturbance and the recovery conditions (Cunning et al., 2015a; Silverstein et al., 2015). The combinations of long-term, in situ field observations and elaborate laboratory experiments may provide more supporting evidence for symbiont shuffling in a better understanding of how coral will adapt to future climate changes. With the advances in ‘omic’ technologies, it is becoming feasible to elucidate the molecular mechanism of local acclimation and symbiont shuffling by analyzing biochemical complementarity of the symbiotic partners (e.g., Lin et al., 2015).