Recovery from bleaching is mediated by threshold densities of background thermo-tolerant symbiont types in a reef-building coral

Sensitive molecular analyses show that most corals host a complement of Symbiodinium genotypes that includes thermo-tolerant types in low abundance. While tolerant symbiont types are hypothesized to facilitate tolerance to temperature and recovery from bleaching, empirical data on their distribution and relative abundance in corals under ambient and stress conditions are still rare. We quantified visual bleaching and mortality of coral hosts, along with relative abundance of C- and D-type Symbiodinium cells in 82 Acropora millepora colonies from three locations on the Great Barrier Reef transplanted to a central inshore site over a 13 month period. Our analyses reveal dynamic change in symbiont associations within colonies and among populations over time. Coral bleaching and declines in C- but not D-type symbionts were observed in transplanted corals. Survival and recovery of 25% of corals from one population was associated with either initial D-dominance or an increase in D-type symbionts that could be predicted by a minimum pre-stress D : C ratio of 0.003. One-third of corals from this population became D dominated at the bleached stage despite no initial detection of this symbiont type, but failed to recover and died in mid to late summer. These results provide a predictive threshold minimum density of background D-type symbionts in A. millepora, above which survival following extreme thermal stress is increased.


Introduction
Coral cover is declining on reefs around the world through the effects of global and local stressors (mostly) associated with human activities [1][2][3]. A major source of variation in coral physiology and stress tolerance results from their symbiosis with photosynthetic microalgae (Symbiodinium spp.). Coralassociated Symbiodinium are genetically diverse [4] and clades, types and ecomorphs (within types) have pronounced effects on their hosts (e.g. [5,6]). Differences in function and performance by Symbiodinium [7] can affect the growth [8], disease susceptibility [9,10] and thermal tolerance [5,6,11] of coral hosts. Significant variation exists in associations between coral host and Symbiodinium types [12,13] and may correlate with environmental variation suggesting the potential for local adaptation in coral-symbiont associations. As our understanding of the distribution of Symbiodinium within coral species and among geographical locations increases, questions regarding their relative abundance during ambient and stressful conditions, and the effects on coral hosts' physiology and bleaching susceptibility can now be addressed [3].
The molecular tools used to quantify metazoan-microbial association (including coral-Symbiodinium associations) have undergone a revolution in recent years with quantitative polymerase chain reaction (qPCR) and next-generation sequencing vastly increasing our ability to identify and quantify in hospite microbial communities [14,15]. Deep sequencing of DNA-barcoded bacteria has revealed the presence of abundant and rare bacterial taxa in tropical sponges that are vital for the health of their hosts [16]. Temporal variation in the relative abundance of microbial taxa, including pathogenic strains, has been associated with environmental stress [17] and can be detected before visual signs of coral stress emerge. Likewise, the detection of low-abundance microbes and Symbiodinium in corals has vastly improved with the much higher detection rate and quantitative accuracy offered by qPCR relative to gel fingerprinting [18].
Predicting the future of reef corals under ongoing climate change depends to a large extent on how accurately processes of genetic, epigenetic and physiological adaptation can be modelled (e.g. [19]). Considered particularly important is 'shuffling', a process whereby the relative abundance of sensitive and more tolerant symbiont types changes following acute stress (e.g. [20,21]). Shuffling can increase stress thresholds by 1-1.5°C on time scales that are relevant to the rates of environmental change currently being experienced [5,11,19,22,23]. While a change from sensitive symbiont 'C' to tolerant 'D' has been observed in the field following bleaching [23,24], it is not a ubiquitous response [25][26][27] and this makes projection modelling a challenging exercise [19]. Therefore, defining and measuring predictors of shuffling and survival from bleaching stress in wild populations of corals is a research priority.
We examine here the absolute and relative changes in C and D symbionts in Acopora millepora corals from three populations when exposed to warm summer temperatures that exceeded their upper longterm temperature regimes by 2-4°C for an extended period. This coral species is common on the Great Barrier Reef (GBR) where it is dominated by Symbiodinium ITS-1 types: C1, C2, C2* and D (ITS-2 equivalents: C1, C3, C131 and D). Background types have been documented in colonies from many locations, particularly turbid inshore reefs, but their roles in bleaching susceptibility, shuffling potential and survival are not yet clear [28]. We test the role of low-abundance symbiont types in bleaching intensity, survival and recovery of Acopora millepora. We show that the relative abundances of symbiont types are dynamic and that coral survival and recovery following bleaching is associated with an increase in the relative abundance of D-type symbionts through shuffling. We define a threshold minimum density of background D above which recovery and survival was enhanced and discuss ways in which this new knowledge can be applied in natural resource management.

Coral transplantation
Temporal changes in C and D Symbiodinium cell abundance were examined in adult colonies (20-30 cm in diameter) of Acropora millepora at six sampling times that encompassed a summer season. Thirtytwo colonies were collected from a depth of 3-7 m at North Keppel Island in the southern inshore GBR (   . Partial and whole-colony mortality and visual colour score (1-5 scale on the coral colour chart [29]) were also estimated for all colonies at the sampling times. Three hundred and seventy-seven samples were collected, but 32 samples were accidentally lost and hence not included in molecular analyses.
Long-term temperature regimes for these sites are available from in situ temperature loggers (http://data.aims.gov.au/aimsrtds/datatool.xhtml). A 15-year daily mean temperature climatology was computed for each site using 48-144 measurements per day to indicate the normal thermal range for each population, the daily mean temperatures for the 13 month transplant period at Magnetic Island from April 2005 to April 2006 and the departure of ambient temperatures from historical averages (figure 1). Bleaching and mortality thresholds sensu Berkelmans [30,31] were used to determine the date at which ambient temperatures at the transplant location exceeded population-specific stress thresholds. Briefly, these thresholds were developed with time-temperature curves and observed bleaching and mortality data for 13 reefs following the 1998 mass-bleaching event. They have since been validated with data from subsequent bleaching events on the GBR and can more accurately predict bleaching and mortality at local scales compared with other bleaching metrics [31].

Symbiont genotyping and density
DNA was extracted from 345 crushed coral samples with a MoBio plant kit following the manufacturer's protocols.
Symbiont genotyping was achieved on 330 samples with DGGE of the ITS-2 region [32] using a denaturing gradient of 30-50% (15 samples did not amplify). Direct sequencing of excised DGGE bands distinguished C1, C3 and C131 but not D, D1 or D1a types [33]. Only a single-clade C type was evident from DGGE analysis in all samples (see the electronic supplementary material for further details).
qPCR of the Actin locus [34] was used to quantify coral host, Symbiodinium clade C and Symbiodinium clade D on 332 samples following Mieog et al. [35] (details in the electronic supplementary material, File 1; 13 samples did not amplify). Actin copy number was determined from one coral sample each with ITS-2 symbiont types C1, C3 or C131 (accession numbers KJ612067-KJ612069; details in the electronic supplementary material, File S1). The results were: clade C3 = 1.0 ± 0.5, C1 = 5.1 ± 1.2 and C131 = 4.3 ± 1.0 (mean ± s.d.). The copy number for C1 was within the margin of error of that determined by Mieog et al. [35] for the same symbiont type, coral host species and sampling location (7.0 ± 2.9). For clade D, we used a copy number of 1.0 ± 0.7 and one for host Actin as determined by Mieog et al. [35].

Symbiont : host and symbiont D : C ratios
Symbiont types C or D to host ratios were calculated with the formula S [C or D] : H ratio = D, and this correction was applied on a sample by sample basis based on dominant C-type determined by DGGE and cross-referenced by direct sequencing.

Statistical analysis
All statistical analyses were performed in R v. 2.2.2 [36], and raw data and scripts are deposited in Dryad [37]. The D : C ratios were highly skewed and standard square root and logarithmic transformations failed to normalize the data. We therefore applied a custom transformation that consisted of 20 bins (electronic supplementary material, table S1) and, subsequently, the data met the linear model assumptions. This approach had the advantage that it provided adequate granularity of the data (particularly for low D : C estimates of interest here) and naturally constrained the data between 0 and 1 (i.e. all C or all D; more detail in the electronic supplementary material). No transformations were applied to the visual and mortality data as they were tested against suitable distributions (i.e. binomial and quasi-Poisson).
Trends in visual colour, coral mortality and symbiont densities (D-and C-specific S : H ratios) over time from each source population were modelled via generalized additive mixed effects models (GAMM) using mgcv [38] and generalized linear mixed effects models (GLMM) [39,40] using MASS [41] incorporating samples as random effects and first-order autoregressive correlation structure to account for the dependency structure and temporal autocorrelation. GAMMs included a thin-plate smoothing spline function of time and the degree of smoothing was established via general cross-validation. Time smoothers lacking inferential support for nonlinear trends (effective degrees of freedom not significantly different from 1) were thereafter modelled via GLMM.
To examine the role of initial symbiont density on bleaching level a linear model compared square root-transformed symbiont densities (i.e. S : H (C + D)) at the beginning of the experiment (April 2005) among the three populations and their bleaching level in December 2005 (also square root-transformed). Generalized change-point (piece-wise) [42] regression models were used to model December S : H ratio (i.e. bleaching level) against April and September D : C ratios (see the electronic supplementary material, S1). The value of the change-point was estimated by iterative optimization of the linear model to minimize model deviance and 95% CIs were generated by bootstrapping the optimization process (leaving one observation out each time).

Results
Water temperatures at the transplant site were within 2°C of local long-term averages for corals from the three source populations from April to September 2005 but elevated by up to 4°C for extended periods between September and December 2005 (figure 1). In agreement with visual condition, populationspecific temperature thresholds for bleaching were exceeded on 8 November and 14 November 2005 for Keppels and Davies Reef corals and mortality thresholds (sensu [31]) were exceeded on 6 December for Keppels, but 14 November for Davies Reef corals. Magnetic Island native corals did not exceed their bleaching thresholds during this study and were not observed to pale or bleach (figure 2a). Mortality in December, January and February varied among populations with 8, 40 and 75% in Keppels and 3, 11 and 96% in Davies Reef, respectively. By April 2006, all Davies Reef colonies were dead, whereas 25% of Keppels corals survived. No mortality was observed in Magnetic Island corals (figure 2b).
We detected three dominant ITS-2 DGGE symbiont types (C1, C131 and D) in the experimental corals and a single type was detected in 95% of samples (more details in the electronic supplementary material). Dominant ITS-2 DGGE symbiont type changed substantially in the Keppels corals over the course of the experiment, but not in Davies Reef and Magnetic Island corals

Discussion
The application of sensitive molecular tools to understand the distribution and abundance of microbial symbionts in Metazoa has revealed their central role in species and ecosystem health [43]. The role of the dominant Symbiodinium type in coral host physiology and stress tolerance is well recognized [21]. However, many key questions regarding natural variability in symbiont distribution and abundance (both environmental and in hospite) remain, including the costs and benefits of symbiosis on coral performance and stress susceptibility under current and future environmental scenarios [3]. We add new insights to this emerging literature with a field experiment that demonstrates the dynamic nature of coral-Symbiodinium associations. We identify indicators of susceptibility to and survival from bleaching that may be used to monitor the status and trend in coral health and resilience in the face of ongoing environmental change.
The abundance of symbionts within corals' tissues can be highly variable in space and time, and can influence coral physiology including bleaching sensitivity (reviewed in [3,21]). Cunning & Baker [44] found more intense bleaching (i.e. reduced S : H ratio) in Pocillopora damicornis with abundant Ctype Symbiodinium communities compared to those with less abundant C-or D-type communities. The density of symbionts pre-summer (both April and September 2005) did not predict the level of bleaching detected here. Our analyses did, however, reveal contrasting and dynamic patterns in the abundance of C-and D-type symbionts among colonies from the three source populations in response to seasonal and extreme summer temperatures. An increase in C-type symbionts was observed in corals from the Keppels after five months of transplantation but not in the Davies Reef population where C-type abundance remained stable until early summer (December 2005; figure 3b). This result might be due to symbiontspecific responses to the novel environment since the transplanted inshore and offshore populations were dominated by C3 and C131, respectively. The specificity of our primers, however, did not allow us to differentiate quantitatively among C-types. The abundance of symbiont types can also be dependent on the ambient temperature environment [45] which was on average 2.2°C above historical averages over winter for the Keppels corals where C-types became more abundant compared with 0.92°C cooler than average for the Davies Reef source population where C-type abundance remained constant. Symbiont densities may be enhanced by environmental nutrient levels [46] and are predicted to be higher on inshore reefs, as a function of higher dissolved nutrients often experienced there. We recorded higher S : H ratios in the corals from the two inshore locations (Keppels and Magnetic Island) compared to the offshore location (Davies Reef). Spatial variation in symbiont density has been hypothesized to underpin larger spatial scale patterns in bleaching on the GBR in 1998 and 2002 [47]. Our results emphasize the importance of the genetic identity of coral host and/or symbiont type rather than cell abundance for bleaching susceptibility and recovery.
Our results confirmed that D-dominated corals are more tolerant to elevated temperatures [5,23,24,48] with all D-dominated colonies but one resisting bleaching regardless of their source population and historical thermal regime. Interestingly, 37.5% of Keppels corals increased their abundance of D over time despite the absence of measurable levels of this type at the beginning of the experiment. This pattern was also found in Montastraea cavernosa by Silverstein et al. [48] and LaJeunesse et al. [24]. Unfortunately, we are not able to advance the findings of those studies to reveal whether the source of these D-type symbionts was endogenous or environmental. Symbiodinium D are pandemic [21] and composed of several distinct groups with species-level genetic differentiation [33]. The D-type is rarely the dominant symbiont in corals, except in environments with high or variable temperatures and high sedimentation and/or turbidity [28,49]. On the GBR these environments are common on inshore reefs, such as Magnetic Island, and also occur in lagoons or other shallow, ponding reef systems [46]. It is increasingly recognized that cost-benefit relationships between coral and Symbiodinium associations are not only dependent on the symbiont type [8,33,50] but also the environment. For example, the cost in growth of corals that host D may be lost in warmer environments [51,52]. As human impacts on coral reefs intensify, the environmental conditions that promote the distribution and abundance of D may become more frequent and widespread [33]. Predicting the impacts of shifts towards D-dominated symbiont associations on coral health and resilience will require sensitive molecular tools to distinguish between very low abundance or novel uptake (including environmental abundance of these symbionts), as well as knowledge of host and symbiont genotype by environment interactions that may control the in hospite abundance of these types.
Our data highlight the importance of shuffling as a mechanism for corals that are not already dominated by tolerant D-type symbionts to recover from extreme summer heat stress and bleaching. Shuffling can shift stress thresholds within a short time period that is relevant to the rates of environmental change currently being experienced [5,[22][23][24]48]. It is ecologically significant that 25% of colonies from the Keppels survived temperatures that far exceeded their historical temperature regimes (figure 1), bleaching and mortality thresholds [30,31]. All but one surviving colony either started off with D or became D-dominant during the course of the experiment (figure 5). The fact that half of the survivors shuffled supports this process as a mechanism of rapid acclimatization to rising temperatures [23,24]. Our results suggest that the background abundance was important and we estimated a minimum D : C threshold of 0.003 above which the relationship with S : H was more gradual than below (i.e. bleaching was less severe). Eighteen colonies from the Keppels population increased their D : C ratio over the course of the extreme heat exposure but failed to recover and died before the last sampling point; however, those that started above the threshold tended to live for longer and have higher final S : H ratios. Two Davies Reef colonies had pre-summer D : C ratios above the 0.003 threshold level but did not shuffle nor survive (figure 5). Clearly, there is still much to learn about the role of background symbionts in the physiological adaptation potential of coral populations under warmer and more variable conditions.
It is well established that the identity and diversity of Symbiodinium complements greatly affect fundamental aspects of the ecology of their coral hosts including growth rate, resistance to and recovery from environmental stress [3,33] and that multiple Symbiodinium types may be present in low abundance in many (or possibly the majority) of coral species (reviewed in [3]). Sensitivity modelling by Fabini [53] found higher stability (defined as the duration and number of species that persisted through environmental perturbations) of coral communities with more diverse symbiont complements probably because they provide redundant or complementary functions and increased the potential for local recovery following disturbance. It follows that an objective for natural resource management would be to identify and conserve species, locations or environmental conditions where symbiont diversity is high. Data on both high-and low-abundance Symbiodinium types can, therefore, provide valuable information on the status and trend of the health and resilience of corals and could provide baseline information for integrated coral reef monitoring and adaptive management [54]. The diversity and abundance of symbiont complements can be routinely and inexpensively screened with next-generation sequencing methods [18] and could feasibly be analysed on large spatial, temporal and taxonomic scales. Moreover, cataloguing of symbiont diversity and abundance within databases and visualization tools such as SymBioGBR and eAtlas provide perspectives that are relevant for natural resource management [55,56]. This approach would build integrated knowledge of the status and trend in coral symbiosis and health and (in time) provide a resource to evaluate the cost and benefits of management actions under different environmental scenarios [57,58].