Introduction

In the Caribbean, the massive coral Orbicella annularis (formerly Montastraea, Budd et al. 2012) is ecologically important in terms of its community dominance and wide geographic range, in addition to playing a key role in creating habitats associated with biomass and diversity of reef organisms (Knowlton et al. 1992; Mumby et al. 2008). Therefore, understanding the resilience of the O. annularis holobiont in a rapidly changing climate is vital when evaluating the future of Caribbean reefs (Ortiz et al. 2013). When O. annularis colonies are healthy, photosynthetic dinoflagellates residing in their gastrodermal cells occur at densities of ~2×106 cells cm−2 (Fagoonee et al. 1999), contributing to host metabolism and skeletogenesis. The symbiotic relationship between corals and their dinoflagellate endosymbionts, Symbiodinium, ultimately sustains the coral reef ecosystem, but is sensitive to environmental change and can be disrupted, causing corals to bleach. During bleaching, endosymbionts may lose pigmentation, become degraded in situ or be expelled from the host (Brown 1997). If not reinstated within a specified time period (dependent on host species, severity of bleaching and other environmental factors), corals will experience partial and occasionally full colony mortality. As global climate change drives sea surface temperatures (SSTs) further above regional norms, the frequency of Caribbean thermal stress events (>2 °C) is predicted to increase by 0.4–0.7 yr−1 (Frieler et al. 2013). These events can trigger mass coral bleaching, such as in 2005, where 80 % of Caribbean corals bleached and 40 % experienced mortality (Eakin et al. 2010). However, experimental and observational work has shown that a variety of physiological properties can be attributed to different endosymbiont haplotypes (Rowan et al. 1997; Stat et al. 2008). In particular, some Symbiodinium display differential susceptibilities to thermal stress, which can affect the bleaching response of their coral hosts (Warner et al. 2006).

Molecular techniques focused on endosymbiont haplotypes have revealed a suite of internal transcribed spacer 2 (ITS2) gene region variants, classified into nine clades named A–I (Pochon and Gates 2010; Pochon et al. 2014). Clades A–D generally associate with scleractinian corals, with B and C being dominant in the Caribbean. Members of Symbiodinium clade D are often found on reefs that experience unusually high SSTs (Baker et al. 2004; Fabricius et al. 2004; Oliver and Palumbi 2011), and in colonies recently impacted by bleaching events (Jones et al. 2008; LaJeunesse et al. 2009), suggesting that temperature stress can, at least temporarily, favour this symbiont (Little et al. 2004; Stat and Gates 2011). Some Symbiodinium D types may also be associated with other ‘stressful’ environmental conditions, including high sedimentation levels (Garren et al. 2006), turbidity (LaJeunesse et al. 2010a), and cool water bleaching events (McGinley et al. 2012). A shift in dominance towards Symbiodinium D in stressed Acropora millepora has been associated with an acquired tolerance of 1–1.5 °C (Berkelmans and van Oppen 2006). This has led to clade D being described by some as a ‘safety parachute’ in the face of rapid climate change (Berkelmans and van Oppen 2006; Stat and Gates 2011). However, the benefits of harbouring D are unlikely to outweigh the costs of associated reduced skeletal growth that are also associated with hosting D types (Ortiz et al. 2013).

Symbiodinium D1 (nomina nuda Symbiodinium glynni, LaJeunesse et al. 2010b) and D1–4 (previously known as D1a and also designated Symbiodinium trenchii, LaJeunesse et al. 2014) are two relatively rare members of clade D with a global distribution and are the only members of clade D that occur in Caribbean corals (Correa et al. 2009a). These symbionts (henceforth referred to as Symbiodinium D1, sensu Pochon et al. 2014) have also shown evidence of thermal tolerance (LaJeunesse et al. 2009; Wang et al. 2012). O. annularis is known to host a variety of symbiont taxa, commonly clades B and C, but also A and D (Rowan and Knowlton 1995; Toller et al. 2001a; Garren et al. 2006). This association with multiple symbiont taxa is partly due to the fact that O. annularis acquires symbionts from its environment (i.e., by horizontal transmission; Szmant 1991). While Symbiodinium D1 has been detected in a few O. annularis populations, many more studies report no Symbiodinium D1 types at all: of 291 colonies screened in previous studies, just 32 hosted clade D (ten studies: see Table 1 for details).

Table 1 Occurrence of Symbiodinium clade D in O. annularis, as documented by other studies that examine symbiont communities using traditional molecular electrophoresis-based techniques (e.g., RFLP and DGGE)
Full size table

The apparent low prevalence of Symbiodinium D1 in O. annularis may be confounded by snapshot sampling. In corals such as O. annularis that are associated with multiple symbiont haplotypes, symbiont shuffling—a change in the relative abundance of resident algal populations within a colony—provides a mechanism by which a holobiont may adjust its capacity to respond to environmental change (Baker et al. 2004; LaJeunesse et al. 2009). A study that monitored Symbiodinium communities in Barbadian O. annularis pre-, during, and post-bleaching demonstrated that corals that harboured ‘cryptic’ (i.e., low level, <10 % of the total symbiont population) sub-clade D1 endosymbionts later became dominated by them and that these colonies remained unbleached (LaJeunesse et al. 2009). However, most investigations suggest that symbiont communities hosted by O. annularis show a high degree of temporal stability (Toller et al. 2001b; Thornhill et al. 2009), with limited evidence for symbiont ‘switching’ (i.e., changes brought about by uptake of new symbionts from the environment, Baker 2003). Typically, only severe bleaching events are capable of disrupting the symbiont community, and post-disruption, most O. annularis colonies regain their original community balance within a matter of months to years (Toller et al. 2001b; Thornhill et al. 2006b; LaJeunesse et al. 2009).

Another explanation for the discrepancy in detecting clade D among populations may be the resolution of the screening techniques used. Commonly used molecular techniques for detecting and cataloguing coral endosymbionts (namely PCR-DGGE, SSCP, and RFLP analysis, reviewed by Sampayo et al. 2009) have been successfully employed to detect clades A to D in O. annularis, but are not always capable of consistently detecting symbionts at abundances below 5–10 % of the total population (Thornhill et al. 2006b; Mieog et al. 2007; LaJeunesse et al. 2008). Furthermore, in mixed communities, the relatively low copy number of D has meant it may be harder to detect than other taxa, e.g., clade C (Smith 2008). These limitations mean that symbiont diversity may be underestimated in many species (Loram et al. 2007; Mieog et al. 2007; McGinley et al. 2012; Silverstein et al. 2012). Studies employing real-time PCR (RT-PCR) are now beginning to identify low-abundance ‘cryptic’ endosymbionts in a number of host species, demonstrating additional complexity of symbiont communities (Silverstein et al. 2012). RT-PCR is 1,000-fold more sensitive than denaturing gradient gel electrophoresis (DGGE) and single strand conformation polymorphism (SSCP; Mieog et al. 2007), and subsequently has the potential to improve estimates of the diversity of coral-associated Symbiodinium (Correa et al. 2009a). For example, D has been shown to be prevalent in a low-abundance, background capacity on the Great Barrier Reef (GBR), where it was detected in 71 % of colonies of four coral species tested (Mieog et al. 2007). Meanwhile, in the eastern Pacific, clade D was present at almost imperceptible levels in 40 % of the Pocillopora screened (McGinley et al. 2012), and in the Caribbean, cryptic levels of clade D were found in five of six coral genera tested, many of which had never been observed previously to host D (Correa et al. 2009a). Other benefits of RT-PCR include its ability to quantify abundance of types and the ease and speed of the technique compared to traditional methods (Granados-Cifuentes and Rodriguez-Lanetty 2011).

This study aims to reveal hidden cryptic Symbiodinium D1 diversity in the Caribbean’s most important reef building coral, O. annularis, from across the wider Caribbean (Fig. 1). Firstly, the high-resolution technique RT-PCR was used to specifically screen for presence of Symbiodinium D1 (i.e., sub-clades D1 and/or D1–4, sensu Pochon et al. 2014) in >500 O. annularis colonies. Secondly, conventional DGGE techniques were used to screen the same samples, with a combination of sequencing and gel comparisons used to confirm presence of Symbiodinium D1. Detection of Symbiodinium D1 in any given sample by RT-PCR, but not DGGE, indicates background or ‘cryptic’ levels of D1. Finally, a statistical approach called SADIE (Spatial Analysis of Distance Indices) was used to explore patterns in the distribution of cryptic Symbiodinium D1.

Fig. 1
figure 1

Proportion of Orbicella annularis population hosting Symbiodinium D1. Thirty-three sites (each identified by a letter) tested. Pie chart size reflects sample size (min = 5 colonies, max = 23), dark shading reflects proportion of samples hosting Symbiodinium D1 detectable by DGGE, pale shading corresponds to low-abundance ‘cryptic’ Symbiodinium D1, detected only by high-resolution technique RT-PCR

Full size image

Quantifying the prevalence of background Symbiodinium D1 in coral communities may facilitate refinement of models that predict ecological responses of Caribbean reefs to climate change (Ortiz et al. 2014). Evidence of naturally occurring symbiont shuffling in O. annularis has previously been documented at just one Caribbean site (LaJeunesse et al. 2009); by screening samples from across the whole region, our understanding of the ability of O. annularis to survive bleaching events can be improved.

Materials and methods

Sample collection and DNA extraction

Fragments of O. annularis tissue were collected from 33 sites spanning a 5.4 million km2 area across the wider Caribbean between 2003 and 2007 (Foster 2007; Fig. 1; Table 2). Approximately 30 small (1 cm3) fragments were chiselled from the edge of spatially independent ramets at each site and stored in 90 % ethanol at 4 °C, sensu Foster et al. (2012). As bathymetry and irradiance are known to influence the O. annularis symbiont community, collections were limited to 2–6 m and only the tops of colonies were sampled. A mix of coral and symbiont DNA was extracted using the DNeasy tissue kit (Qiagen) and then stored at −20 °C.

Table 2 Summary of RT-PCR and DGGE outputs by site, grouped by marine eco-region
Full size table

Screening for Symbiodinium D1 using RT-PCR

A 312-base pair target region specific to Symbiodinium clade D (including both Caribbean types D1 and D1–4, referred to as ‘Symbiodinium D1’ throughout this document sensu Pochon et al. 2014), located in domain 2 of the large subunit (LSU) ribosomal RNA gene, was amplified using published RT-PCR primers (Table 3; Correa et al. 2009a). A 10 µl reaction mix containing 1 mM of both forward and reverse primers, 1 µl DNA template, and 2X Absolute qPCR SYBR Green Fluorescein Mix (Thermo Scientific) was amplified in RT-PCR (CFX96 RT-PCR detection system, Bio-Rad Laboratories, Inc.) using the FAM filter. Reaction conditions were an initial denaturing step of 95 °C for 10 min, followed by 50 PCR cycles of 95, 61, and 72 °C for 30 s each (Correa et al. 2009a). A final high-resolution melting (HRM) step entailed a 55–95 °C temperature ramp of 0.2 °C every 2 s. Fluorescence data were collected during each PCR annealing step and each temperature step of the HRM melt cycle. Each DNA sample was run in duplicate for the Symbiodinium D1 primer set, and positive (standard) and negative controls were included on every plate.

Table 3 Primer pairs used for RT-PCR and PCR-DGGE analyses
Full size table

Selection of a fixed fluorescent threshold (see Fig. 2a) in the exponential phase of the reaction allowed comparable C T values to be calculated. Attempts to quantify copy number of Symbiodinium D1 in each sample were unsuccessful; however, C T values can also provide a useful quality control: duplicate values that differed by >1 were discarded. To further assess DNA quality, every sample that was screened for Symbiodinium D1 was run through RT-PCR for a second time (again in duplicate) using the ITS2 primer set ‘ITSintfor2’ and ‘ITSrev’, Table 3 (cycling conditions: 98 °C for 2 min, followed by 45 cycles of 98 °C (5 s), 55 °C (5 s) and 72 °C (5 s); Granados-Cifuentes and Rodriguez-Lanetty 2011). Any samples that failed to generate positive ITS2 amplifications or produced poor quality melt curves were removed from the dataset to avoid false negative D1.

Fig. 2
figure 2

Comparing outputs of the two Symbiodinium clade D screening methods. a RT-PCR output. Amplification curves (top panel from RT-PCR) and melt curves (bottom panel from HRM analysis) were generated for every reaction. Positive amplification (top panel) of duplicates and generation of a melting peak around 84.4 °C (bottom panel) indicates that Symbiodinium D1 is present in the sample, while template-free controls (NTC) and samples without D1 did not amplify. b DGGE gel. Each lane represents the symbiont community hosted by an individual O. annularis colony (from Cuban samples CC, see Fig. 1). Paired bands in the D1 and D1a position on the gel indicate presence of type D1-4

Full size image

Screening for Symbiodinium D1 using DGGE

DGGE was used to identify Symbiodinium ITS2 types within each individual O. annularis fragment following LaJeunesse (2002). DNA was amplified in a PCR (95 °C for 5 min; followed by 30 cycles of 94 °C (45 s), 57 °C (45 s), and 72 °C (60 s); with a final annealing step of 59 °C for 20 min) using Symbiodinium-specific rDNA primers ‘ITS2 Clamp’ and ‘ITSintfor2’ (Table 3). PCR products, mixed with 5 µl bromophenol blue loading buffer (15 % Ficoll, 0.25 % xylene cyanol FF, 0.25 % bromophenol blue), were electrophoresed at 114 V on a polyacrylamide denaturing gradient gel (40 to 60 % denaturant) at 60 °C (Ingeny System). An ITS2 standard (with B1, C1, and C3, provided by the Coral Reef Ecosystems lab, University of Queensland, Australia) was run in the first lane of each gel. After 14 h, the gel was stained with SybrGreen I (Invitrogen) nucleic acid gel stain at room temperature for 20 min, before imaging in a UV transilluminator. Imaged gels were examined by eye and scored for types, in comparison to a database of other gels used to help identify haplotypes. Dominant bands from each DGGE gel were excised, cleaned, and sent for sequencing (Macrogen) to resolve ITS2 type.

Data analysis

Examination of HRM melt curves from the final step of the RT-PCR process allowed purity of the reaction products to be assessed. Amplified product generated a melt peak at the correct temperature range (e.g., 84.4 °C ± 0.5 (mean ± SD) for D1 fragments targeted by RT-PCR primers), whereas melting at lower temperatures was indicative of primer dimer. Where both RT-PCR duplicates produced a melt peak around 84.4 °C, the sample was scored as containing Symbiodinium D1 (Fig. 2a). Where bands were observed on the DGGE gel in either the D1 or D1a position (or both), the sample was also scored as hosting Symbiodinium D1 (Fig. 2b). Samples that scored positively for D1 in RT-PCR, but negatively in DGGE, were said to contain ‘cryptic’ or background levels of Symbiodinium D1.

SADIE, a statistical approach designed for assessing the patterning of count data from spatially referenced locations, was used to quantify spatial patterns in the data (Perry 1995). SADIE measures the spatial pattern at each sampled unit using an index of clustering based on geographic distance—assigning each site either a positive patch cluster (v i ) or a negative gap cluster (v j ) value. These can then be mapped with filled and empty circles, representing local quantification of spatial patterning. Interpolation between the data points (using a universal kriging method) produces a red-blue contour plot (e.g., Fig. 3), indicating clustering of spatial data (Perry et al. 1999).

Fig. 3
figure 3

Red-blue plot generated by SADIE analysis, indicating spatial clustering in the Symbiodinium D1 distribution among sampled Orbicalla annularis populations. Each circle represents one site where O. annularis was sampled (n = 33 sites). Red circles indicate sites with a higher-than-expected prevalence of Symbiodinium D1, blue circles indicate sites with significant scarcity of Symbiodinium D1, and white circles represent populations hosting the same amount of D1 as might be expected by chance when compared to >5,000 random permutations of the spatially referenced data. The size of each circle represents the statistical significance of the clustering index value, i.e., largest filled circles represent a unit with clustering index that exceeds the 95th percentile for patches (red) or gaps (blue), from the mean of the randomised distributions; medium-sized circles denote a unit that exceeds the 90th percentile; small circles denote unit with clustering below expectation (<1 or >−1). The 33 data points are superimposed onto a contour plot, with coloured areas denoting clustering beyond expectation, e.g., red shaded areas are geographic regions of higher-than-expected prevalence of Symbiodinium D1 in O. annularis and blue areas regions of lower-than-expected (or ‘gappy’ prevalence), while the darkest areas indicate highly significant clustering/gappiness. The large amount of white space on the map is indicative of a lack of significant spatial patterning in the occurrences of Symbiodinium D1 in O. annularis across the majority of the Caribbean region (exceptions being the Mesoamerican Barrier Reef and southern Caribbean, particularly around Curaçao and Venezuela)

Full size image

Results

Detection of Symbiodinium D1 in O. annularis

Five hundred and fifty-two coral colonies were successfully screened for Symbiodinium D1 using both techniques. RT-PCR produced positive amplifications (indicating presence of Symbiodinium D1) in 170 colonies (31 % of all corals), and Symbiodinium D1 was scored as being present at 97 % (32 of 33) of sampling sites (Fig. 1). Melt curves revealed almost all (>98 %) samples with positive amplifications had a characteristic dissociation signature with melting point at 84.4 °C: the melting temperature of the target fragment (Fig. 2a). Those that did not were removed from the dataset.

Any C T values over 45 were also discarded, as at a level of C T of >45 rare cells could be contaminants. This improved reproducibility estimates by >10 %. Although signals were often low (again, common in mixed assemblages), template-free controls were consistently negative.

The proportion of coral samples hosting Symbiodinium D1 detected by RT-PCR ranged from 0 to 100 %, with an average of 30 % hosting cryptic Symbiodinium D1 per site (SE = 4.67). Only Drunkenman’s Cay in Jamaica (site JA) was found to have no cryptic Symbiodinium D1 (n = 14), despite these samples having a good quantity of starting DNA template (ITS2 = 12.8 ng µl−1).

In comparison, DGGE detected Symbiodinium D1 in a total of 12.5 % of corals (69 colonies) at just 12 of the 33 sites. At six of these 12 sites (DM, CC, CA, EN, BVI, T, and R; Table 2), Symbiodinium D1 was manifested on DGGE gels as paired bright bands in the D1 (upper band) and D1a position (lower band), see Fig. 2b. This banding profile represents a D1 ITS2 variant known as D1–4 (GenBank Accession AF499802), which was confirmed by sequencing. At the remaining six sites (A–E and BA; Table 2), clade D was not the dominant symbiont type. Here, D1 bands were pale and were rarely accompanied by a second band in the D1a position. It is possible that these symbionts may be a closely related sequence variant D1 (AF334660), which differs from D1–4 by a few bp (Pochon et al. 2014), as RT-PCR primers cannot distinguish these types, and melt curves were unable to discriminate without a pure reference.

In addition to Symbiodinium D1, DGGE also distinguished numerous other ITS2 Symbiodinium types (Electronic Supplementary Materials, ESM, Table S1), nested within clades A–D. All Symbiodinium D1 appeared more commonly in mixed assemblages with B and C types (44 and four colonies, respectively) than alone (21 colonies). ITS2-B1 (AF333511) was the most frequently occurring sub-clade, found in 71 % of colonies (see ESM Table S1 for other types). Thirty-one percent of samples harboured clade C (13 % exclusively), 65 % exclusively harboured clade B, and a further 17 % hosted a mix of B and C, compared with the 4 % of colonies that hosted Symbiodinium D1 alone.

Spatial distribution of Symbiodinium D1

DGGE-detectable Symbiodinium D1 showed a lack of spatial structuring across the Caribbean (SADIE Index of Aggregation, I a  = 1.57, P a  = 0.05), indicating an apparently random distribution of this symbiont among coral populations. A second spatial analysis of the overall pattern of distribution of low-abundance (RT-PCR-detectable) Symbiodinium D1 was also close to random (I a  = 1.42, P a  = 0.09), again indicating absence of regional-scale structuring in its arrangement. However, a high SADIE patch cluster index (v i ) of 2.22 (p = 0.01), indicative of localised clustering at the site level, revealed higher-than-expected occurrences of cryptic Symbiodinium D1 in some local areas (i.e., sites A–C, G, H, K, and L). The location of these sites created a patch cluster around the Mesoamerican Barrier Reef System, MBRS (red area, Fig. 3). Several more sites (CB, T, SB, VB, and BV) hosted less Symbiodinium D1 than might be expected by chance, producing a significant gap cluster around the southern Caribbean (Fig. 3, blue area), although there were no significant gaps in the overall distribution (v i  = −1.25, p = 0.192).

Discussion

RT-PCR revealed that at all but one site, ~30 % of O. annularis harboured Symbiodinium D1 at low-abundance ‘cryptic’ levels. In comparison, out of the 552 individuals screened, DGGE identified only 69 instances of Symbiodinium D1 at just 12/33 sites, indicating that this technique did not detect D1 at low levels in a further 101 coral colonies and missed its presence at 20 sites.

The discrepancy in the resolution of the two techniques is unsurprising: the strength of DGGE is in its ability to detect a breadth of types, rather than its sensitivity. ITS2-DGGE detection levels of just 5–10 % have been reported in the Caribbean (tested on mixed cultures of B1 and C2; Thornhill et al. 2006b), while in Pacific populations, Symbiodinium D1 were only identified at 10–30 % of the total symbiont population (LaJeunesse et al. 2008). Limited detection of Symbiodinium D1 by DGGE may be due to a swamping effect of other symbiont types during the PCR step or preferential amplification of alternative clades. The genomes of clade D Symbiodinium are known to have a low ribosomal copy number (e.g., 3–5 times lower than clade C), making them harder to distinguish in mixed samples (Smith 2008). For example, one ITS-DGGE study detected Symbiodinium C3-e at proportions as low as 1 %, but only detected D when it existed in proportions >20 % (LaJeunesse et al. 2009). DGGE banding showed 48/69 colonies hosting Symbiodinium D1 contained mixed symbiont assemblages, which would exacerbate this issue. The distribution and abundances of other observed types corroborate well with the findings of numerous smaller-scale studies into O. annularis symbiont community composition (e.g., Toller et al. 2001a; LaJeunesse et al. 2009; Thornhill et al. 2009). Symbiodinium B1 dominated >70 % of sampled corals, supporting the literature that shows B1 to be the most common Caribbean ITS2 type in a variety of cnidarian genera (LaJeunesse 2002), including O. annularis (Thornhill et al. 2009).

Precision of RT-PCR can also be lower in samples hosting multiple clades (Loram et al. 2007). Mixed communities were common in our dataset, with 27 % of colonies hosting >3 sub-clades. However, the primers used work well in mixed assemblages (Correa et al. 2009a), and reproducibility (among replicates) of RT-PCR detection of Symbiodinium D1 was 63 %. Another way in which the reported RT-PCR results may have misrepresented Symbiodinium D1 diversity is through the sampling design. Collection of coral fragments was limited to unshaded colony tops at <6 m, in order to avoid sampling intra-colony irradiance-driven zonation. One consequence of this standardised sampling protocol is that Symbiodinium D1 diversity on colony sides or shaded parts may have been missed. Since shallow-water O. annularis colonies have been shown to bleach more in shaded areas (Rowan et al. 1997), and clade D is often associated with bleaching, the results presented may be an underestimate of the true prevalence of Symbiodinium D1 in O. annularis.

DGGE detection of Symbiodinium D1 in O. annularis

DGGE detected Symbiodinium D1 in just 12 % of the total O. annularis colonies screened (69 colonies, representing 40 % of the ‘true’ Symbiodinium D1 prevalence). Studies employing similar molecular techniques have reported comparable abundances of D in the Caribbean, which generally represent <10 % of the endosymbiotic community in the host population (Thornhill et al. 2006b). At 21 of 33 sites, DGGE did not detect Symbiodinium D1 at all (Fig. 1), a result directly comparable to several smaller-scale studies (e.g., Belize: Garren et al. 2006; Warner et al. 2006; Panama: Rowan and Knowlton 1995, Rowan et al. 1997; and in the Bahamas: Thornhill et al. 2009; see Table 1) that reported absences of clade D in O. annularis when employing similar molecular techniques (DGGE and RFLPs). In cases where clade D has been detected using DGGE and RFLPs (e.g., at Lee Stocking Island, Bahamas by LaJeunesse 2002; in Panama by Garren et al. 2006; and Barbados by LaJeunesse et al. 2009), it often presents at similar frequencies to the 12 % of colonies in this study (12.5, 11, and 3 % (by DGGE), respectively). Higher reported abundances are often associated with corals stressed by temperature (e.g., >60 % O. annularis in Barbados and Florida during bleaching; LaJeunesse et al. 2009; Finney et al. 2010) and/or disease (e.g., 100 % O. annularis in Panama; Toller et al. 2001b). In this study, sporadic response to stress events could explain the high dominance of Symbiodinium D1 (90–100 %) detected by DGGE at sites EN, R, and CC.

DGGE banding profiles (Fig. 2b) and sequencing suggest that the majority of the clade D detected in this study is ITS2 sequence variant D1-4 (S. trenchii; LaJeunesse et al. 2010a), although in some cases, distinguishing D1–4 from the closely related sub-type D1 was problematic. Further structural, biochemical, and physiological investigations would be required to determine the ecological significance of the molecular differences between D1 and D1–4 to the host (Stat et al. 2012, but also LaJeunesse et al. 2014). Examination of these types with fine-scale genetic markers may reveal further diversity, just as higher-resolution markers have demonstrated that Symbiodinium B1 consists of several specialised lineages (Finney et al. 2010). An improved understanding of how Symbiodinium genetic variation among sub-types is manifested in terms of quality as a mutualistic partner will be key to understanding the coral-algal symbiosis, and the ecological and evolutionary drivers that maintain such a variety of genetic types (Heath and Stinchcombe 2013).

RT-PCR detection of Symbiodinium D1 in O. annularis

RT-PCR revealed a widespread prevalence of Symbiodinium D1, both between (at 32/33 sites) and within sites (in more than half of all populations, 11–40 % of colonies harboured cryptic Symbiodinium D1). Prevalence of cryptic Symbiodinium D1 in an average of 29.8 % of colonies per site supports evidence from Barbados that reported 28 % of healthy O. annularis colonies contained D1 that was detectable only by RT-PCR (LaJeunesse et al. 2009). In Barbados, D1 prevalence later increased to 60 % of colonies in response to a bleaching event. Colonies harbouring D1 remained unbleached, demonstrating the ability of O. annularis to respond to temperature anomalies through symbiont shuffling. Correa et al. (2009a) also used high-resolution techniques to record low levels of Symbiodinium D1 in 21 % of sampled Caribbean corals (but not O. annularis) across a geographic range comparable to this study. Our findings demonstrate the breadth of the association of O. annularis with Symbiodinium D1 shown in Barbados, across a Caribbean scale comparable to Correa et al. (2009a).

The spatial arrangement of RT-PCR-detectable Symbiodinium D1 across the Caribbean was fairly uniform (Figs. 1, 3), supporting the consensus that clade D exhibits random geographic distribution (Stat and Gates 2011). Its presence might mirror environmental availability of Symbiodinium D1 at the time of symbiont uptake during coral settlement. This would suggest an innate region-wide low-level presence of Symbiodinium D1, which could imply a natural ability to respond to future warming events (if shuffling allows D1 to increase in abundance and be maintained in response to warming). Site-level spatial analyses revealed a high number of colonies hosting Symbiodinium D1 at MBRS sites (Fig. 3, red areas) and, to a lesser extent, a relative scarcity of Symbiodinium D1 in corals in the southern Caribbean (Fig. 3, blue areas). Bleaching history has been proposed as an explanation for the occurrence of clade D (Baker et al. 2004), and observed site-level variability may be explained by thermal stress history at different Caribbean locations. Higher frequencies of Symbiodinium D1 at MBRS sites may reflect the severity of the 1998 mass bleaching event in the region, where an extended +1.5 °C warming event was experienced in the western Caribbean in September 1998. Here, 76 % of O. annularis colonies experienced bleaching (Kramer et al. 2000), which may have facilitated the establishment of Symbiodinium D1 in these communities. The 1998 bleaching event was not as severe in the southern Caribbean, possibly explaining fewer Symbiodinium D1 at these sites. Timespans ranging from months to years are required for original symbiont communities to re-establish dominance following bleaching-associated increases in clade D in O. annularis (Thornhill et al. 2006b; LaJeunesse et al. 2009), making it conceivable that our sampling period for the MBRS (October 2004) could coincide with recovering symbiont communities. Thus, temporal sampling would be needed to resolve the stability of cryptic Symbiodinium D1 at each site and to indicate whether shuffling or local uptake best explains the observed spatial distribution.

Wider implications of high Symbiodinium D1 prevalence

This study demonstrates that the prevalence of background Symbiodinium D1 reported in Barbados in the sentinel study of LaJeunesse et al. (2009) can be extrapolated to O. annularis across the entire wider Caribbean region. This suggests that the demonstrated ability of Caribbean colonies to recover from severe bleaching events may be more geographically extensive than first supposed. However, there are several caveats. Firstly, uncertainty exists around the ecological significance of background levels of Symbiodinium to the host. Despite 30 % of colonies hosting cryptic Symbiodinium D1, the majority of symbiont communities (65 %) in this study were dominated by Symbiodinium B1. The relationship between ITS2-B1 and Caribbean corals has existed since the Pleistocene and has been used to demonstrate stable symbioses and a lack of support for symbiont shuffling in the past (Baker et al. 2013). Evidence for the evolution of these long-term stable, mutual cooperations between coral and dominant Symbiodinium types presents a paradox in the face of an extremely large amount of variation in potential symbiont partners. An improved understanding of the traits imparted to hosts by different symbiont types might go some way to elucidating the significance of maintaining background ‘cryptic’ abundances of these types and help explain persistence of variation in Symbiodinium (Heath and Stinchcombe 2013). The flexibility of O. annularis in hosting varied symbionts certainly does not appear to confer a greater diversity of potential responses to environmental threats in this important reef builder; it remains ranked among the most bleaching susceptible of Caribbean species (Manzello et al. 2007; van Hooidonk et al. 2012).

Secondly, not all species are as flexible in their associations as Orbicella spp.; many are ‘specifist’, exhibiting low flexibility and hosting taxonomically narrow assemblages (Putnam et al. 2012). Brooding corals (including common Caribbean species Agaricia sp., Siderastrea sp., and Porites sp.) inherit symbionts from parents, usually resulting in more specific and more stable symbioses, implying fewer opportunities for shuffling (Thornhill et al. 2006a). Other D1 hosts do not show comparable ‘shuffling’ in response to stress events (e.g., LaJeunesse et al. 2009; McGinley et al. 2012). Host-related factors, such as abundance of UV-absorbing compounds and the ability of coral to mop up hydrogen peroxide that triggers the response to expel, are also important in determining differential bleaching responses (Baird et al. 2009; Wooldridge 2014); this undermines the importance of the symbiont community in dictating future reef responses to thermal stress.

Thirdly, the benefits incurred by hosting D1 in terms of resilience to thermal stress may be countered by severe costs in terms of trade-offs, including a reduction in energy to juvenile corals, reduced growth rate and reproductive output, and increased disease susceptibility (e.g., Little et al. 2004; Jones and Berkelmans 2010; Littman et al. 2010). Not all clades are equally valuable as symbiotic partners, e.g., clade A is understood to be functionally less beneficial to Acropora in Hawaii than clade C (Stat et al. 2008). The ability of scleractinian corals to form calcium carbonate skeletal structures is linked to symbiosis; A. millepora colonies harbouring clade D grew 38 % slower than those hosting C (Little et al. 2004; Jones and Berkelmans 2010). Trade-offs such as this may help to explain the stable coexistence of multiple symbiont types when evolution should favour the development of long-term cooperation in coral-algal mutualisms (Heath and Stinchcombe 2013). Symbionts such as D1 that translocate fewer resources to their host may be maintained in low abundances as long as greater numbers of higher-quality mutualistic partners exist that can provide for the fitness of the host. However, Symbiodinium D1 may use stressful conditions as an opportunity to exploit the host as a habitat rather than engage in an interactive and mutually beneficial partnership, which may affect fitness of coral hosts and accelerate reef decline (Heath and Stinchcombe 2013; Ortiz et al. 2013). Ecosystem models have been used to explore the likelihood of the evolution of hypothetical ‘super symbionts’ that possess a combination of traits (and trade-offs) capable of maintaining Caribbean coral cover in the face of rising SSTs (Ortiz et al. 2014). In these models, the time required for establishment of the ‘super-symbiont’ into coral populations was a critical factor in their success in maintaining future Caribbean coral cover. The empirical findings of this study—that a symbiont with traits comparable to those required in the ‘super-symbiont’ is already established in at least 30 % of the Caribbean population of the reef builder O. annularis—are encouraging in this context and also may help refine the super-symbiont model. Unfortunately, the model concludes unless the new symbiont could provide 33 % more benefit than existing D1–4 in terms of reducing mortality during bleaching, at 20 % of the cost (of reduced growth rates), any increase in dominance is unlikely to ensure the persistence of Caribbean reefs. In the Caribbean, where maintenance of coral growth rates is important in competitive interactions with macroalgae (Roff and Mumby 2012) and is synonymous to healthy ecosystem functioning (Kennedy et al. 2013), increasing the dominance of cheating Symbiodinium D1 may cause more harm to reefs than good (Ortiz et al. 2013, 2014).