Reef Endemism, Host Specificity and Temporal Stability in Populations of Symbiotic Dinoflagellates from Two Ecologically Dominant Caribbean Corals

Background The dinoflagellate genus Symbiodinium forms symbioses with numerous protistan and invertebrate metazoan hosts. However, few data on symbiont genetic structure are available, hindering predictions of how these populations and their host associations will fair in the face of global climate change. Methodology/Principal Findings Here, Symbiodinium population structure from two of the Caribbean's ecologically dominant scleractinian corals, Montastraea faveolata and M. annularis, was examined. Tagged colonies on Florida Keys and Bahamian (i.e., Exuma Cays) reefs were sampled from 2003–2005 and their Symbiodinium diversity assessed via internal transcribed spacer 2 (ITS2) rDNA and three Symbiodinium Clade B-specific microsatellite loci. Generally, the majority of host individuals at a site harbored an identical Symbiodinium ITS2 “type” B1 microsatellite genotype. Notably, symbiont genotypes were largely reef endemic, suggesting a near absence of dispersal between populations. Relative to the Bahamas, sympatric M. faveolata and M. annularis in the Florida Keys harbored unique Symbiodinium populations, implying regional host specificity in these relationships. Furthermore, within-colony Symbiodinium population structure remained stable through time and environmental perturbation, including a prolonged bleaching event in 2005. Conclusions/Significance Taken together, the population-level endemism, specificity and stability exhibited by Symbiodinium raises concerns about the long-term adaptive capacity and persistence of these symbioses in an uncertain future of climate change.


Introduction
Determining the population dynamics of marine organisms is critical toward understanding their evolutionary ecology as well as successfully designing marine protected areas (MPAs) [1][2]. For coral reef ecosystems, genetic studies have revealed the degree of gene flow and connectivity in many species, such as fish [e.g., [3][4][5] and numerous other reef dwelling metazoans [e.g., [6][7], including scleractinian corals [e.g., [8][9][10]. However, surprisingly little is known about the population genetics of Symbiodinium, the unicellular symbiotic dinoflagellates that are among coral reefs' most important constituents [11]. This is particularly relevant since rising sea-surface temperatures are increasingly threatening coral reef ecosystems and populations that reside within them [reviewed by 12]. Since populations represent the fundamental units of evolution [reviewed by 13], understanding patterns and processes at this level is paramount toward furthering our knowledge on the basic biology of Symbiodinium as well as how anthropogenically driven global climate change may impact these symbionts and their host associations in the future.
Symbiodinium forms symbioses with a variety of marine protistan and invertebrate metazoan hosts. Currently, the genus is viewed as ecologically, physiologically, biochemically, and genetically diverse [e.g., [14][15][16][17][18][19][20]. Although previous genetic studies have extensively examined symbiont diversity at the levels of sub-generic ''Clades'', chloroplast and mitochondrial ''haplotypes'', or internal transcribed spacer (ITS) rDNA ''types'' [reviewed by 21], only five studies to date have examined Symbiodinium at the population level. In this context, Santos et al. [22] reported that ITS2 ''type'' B1 [ITS2 ''type'' nomenclature sensu 18] Symbiodinium populations associated with the Caribbean gorgonian Pseudopterogorgia elisabethae exhibited strong genetic structure over distances of tens of kilometers in the Bahamas. Similarly, Kirk et al. [23][24] found significant differentiation between Symbiodinium ''type'' B1 populations of the sea fan Gorgonia ventalina on reefs separated by 5 to 200 km in the Florida Keys. In the Pacific Ocean, Howells et al. [25] reported that Symbiodinium Clade C populations associated with an alcyonacean coral, Sinularia flexibilis, were genetically differentiated at a range of spatial scales, from 16 to 1,360 km. On the other hand, Magalon et al. [26] observed no population structure in Symbiodinium Clade C from the scleractinian Pocillopora meandrina over distances of up to 200 km in the central and western Pacific Ocean. These disparate results may be a consequence of differences in host affiliation, Symbiodinium lineage, mode of symbiont acquisition, geographic/hydrological differences (such as currents), and/or other factors. Thus, additional studies are necessary to better elucidate patterns of Symbiodinium structure and gene flow at the population level. Here, we contribute to this area by examining the Symbiodinium Clade B populations associated with the scleractinian corals Montastraea faveolata (Ellis and Solander, 1786) and Montastraea annularis (Ellis and Solander, 1786).
Montastraea faveolata and M. annularis are two of the ecologically dominant reef-building corals in the Caribbean and western Atlantic Ocean, with both species acquiring their Symbiodinium from the external environment at each generation [27]. These coral species are among the most flexible hosts known; symbioses with various ITS2 ''types'' in Symbiodinium Clades A, B, C, and D [19,[28][29][30][31][32], along with intracolonial zonation of Symbiodinium Clades and ''types'' [28][29]32], have been well-documented across their biogeographic range. Thus, these coral species provide an ideal system for studying population-level dynamics in Symbiodinium. Previously, Thornhill et al. [31,33] tracked symbiotic associations in tagged colonies of nine coral species, including M. faveolata and M. annularis. At the ITS2-rDNA level, these hosts were found to engage in relatively stable temporal associations with particular Symbiodinium, such as ''type'' B1. This observation, along with earlier population genetic and host association studies of this symbiont ''type'' in the western Atlantic Ocean [22][23][24]34], led us to hypothesize that the Symbiodinium ''type'' B1 populations of M. faveolata and M. annularis would exhibit endemism to particular reefs, host specificity and within-colony stability through time. To test these predications, microsatellite allelic variation at loci specific to Symbiodinium Clade B was 1) examined from symbiont populations of these scleractinian species on Florida Keys and Bahamian reefs, and; 2) tracked over an approximately three-year period.

Study sites and collection of Montastraea faveolata and M. annularis
Three reefs in the Upper Florida Keys, U.S. and two reefs in the Exuma Cays, Bahamas were selected for monitoring of symbiotic associations (Fig. 1) For further information on the history of coral colonies at these reefs, see Fitt et al. [35] and Thornhill et al. [31]. Nocturnal seasurface temperature data for the study sites were obtained from the National Oceanographic and Atmospheric Administration's (NOAA) satellite-based Coral Reef Watch virtual and experimental coral bleaching stations (http://coralreefwatch.noaa.gov/ satellite/index.html).
To ensure that subsequent sample collections were from the same individual, six colonies of Montastraea faveolata and M. annularis, each separated by 4 to 50 m from the nearest conspecific individual, were tagged on each reef (a total of 30 for M. faveolata and 30 for M. annularis). Note that in the genetic analyses presented below, all six M. annularis colonies from AG and four colonies of M. faveolata from SP were detected as harboring only Clade C Symbiodinium. Because the emphasis of this study was on population structure of Symbiodinium in Clade B, these ten colonies were excluded from further analyses.
All coral colonies were sampled one to two times per year for genetic analyses between March 2003 and December 2005 by S.C.U.B.A. or snorkel. Additional samples were collected from these tagged colonies for cell count measurements between 2002 and 2006 (up to four total samplings per year). Approximately 10 cm 2 fragments were removed via hammer and chisel, with care taken to ensure that the same relative position (i.e., unshaded colony tops) was sampled each time. To test for potential sampling effects due to within-colony variation of Symbiodinium, tissue was also collected from the sides of colonies in January (NP, Exuma Cays) and March (ADM, Upper Florida Keys) of 2004. Upon collection, fragments were placed in seawater-filled plastic bags and immediately transported to the laboratory in an insulated cooler for processing.

Symbiodinium density and nucleic acids extraction
Coral fragments were processed as follows. Tissue was removed using a recirculating Waterpik (Water Pik Inc.) with 0.45 mm filtered seawater and the ''blastate'' pulsed for 1-4 s with a Brinkmann Instruments Polytron Kinematica Tissue Homogenizer TM to disperse mucopolysacharides. Symbiodinium cell densities were determined following Fitt et al. [35]. Symbiont cells for molecular analyses were isolated from the salt-water ''blastate'' via centrifugation in 50 mL tubes at 2000-3000 g for 5 min, preserved in DMSO Buffer (20% dimethyl sulfoxide and 0.25 M ethylenediaminetetraacetic acid [EDTA] in NaCl-saturated water [36]) and nucleic acids subsequently extracted following Thornhill et al. [31].

ITS2-PCR and Denaturing Gradient Gel Electrophoresis (DGGE) of Symbiodinium
Polymerase chain reaction and denaturing gradient gel electrophoresis (PCR-DGGE) of the ITS2 rDNA was used to discriminate Symbiodinium ''types'' [18][19] harbored by the tagged colonies. Following PCR, ITS2 sequences were electrophoresed according to the protocol of Sampayo et al. [37], with the exceptions noted below. Amplified ITS2 fragments were loaded on 20 cm long, 0.75 mm wide, 8% polyacrylamide (37.5:1 acrylamide/bisacrylamide ratio) gels containing a 45-80% denaturing gradient (100% denaturant contains 7 mol L 21 urea and 40% deionized formamide) and electrophoresed at 150 V for 10 h at 60uC on a C.B.S. Scientific TM DGGE-2001 model apparatus. To diagnose symbiont identity, the DGGE fingerprint for each sample was compared to ITS2 standards of known nucleotide sequences. To further validate identifications, all bands in the DGGE-profile were excised, reamplified, and bidirectionally sequenced in 25 of 213 samples using the Genome Lab TM Quick Start Mix (Beckman Coulter) on a Beckman CEQ 8000 Genetic Analysis System (Beckman Coulter) according to the protocol of LaJeunesse [19]. Criteria for band excision and sequencing were as follows: 1) profiles that differed in any way from the ITS2 standards were excised and sequenced to insure they were properly diagnosed, 2) samples were regularly selected at random to insure that no misidentifications due to fragment co-migration had occurred. Chromatograms were checked by comparison to the complement DNA strand in SEQUENCHER v4.6 (Gene Codes) and finished sequences aligned manually using MACCLADE v4.06 (Sinauer Associates). Sequences were deposited in GenBank under Accession Nos. GQ268415-GQ268420.
Microsatellite amplifications were performed in 10 mL volumes containing 10 mM Tris-HCl (pH 8. A multilocus genotype was constructed for the Symbiodinium Clade B population associated with each Montastraea spp. colony using the allele sizes recovered from each locus [22]. In cases were multiple alleles were detected at a single locus, these were interpreted as Symbiodinium populations comprised of more than a single Clade B genotype [following 22]. Preliminary analyses revealed that the three loci possessed allelic variation capable of identifying unique Symbiodinium haploid genotypes. Along with this, we calculated the probability of identity (PI), or probability that two unrelated host individuals would (by chance alone) have the same multilocus Symbiodinium genotype, for both the M. faveolata and M. annularis datasets in GENALEX v6.2 [43]. Symbiodinium genotypes from each host species were each treated as belonging to a single population. The PIs for M. faveolata and M. annularis were 2.92610 23 and 4.96610 23 , respectively. These values suggest that, at maximum, 2.92 out of 1,000 M. faveolata samples and 4.96 out of 1,000 M. annularis samples would be mis-identified using these three microsatellites. It should also be noted that these PI values are overestimates because with repetitive sampling of tagged coral colonies, the same Symbiodinium genotype is likely to be sampled multiple times in the data set (whereas they are treated as different individuals in the above calculations). Therefore, these three microsatellites have sufficient resolution to differentiate Symbiodinium populations from different coral colonies, host species, reefs, and geographic regions with statistical confidence (see Results and Discussion).
To quantify the distribution and relative importance of genetic variation within the examined Symbiodinium populations from each host species, analyses of molecular variance (AMOVAs) were performed with the ''codom-genotypic'' option for haploid data in GENALEX v6.2 [43]. Total genetic diversity in the Symbiodinium populations from each host species was partitioned among groups of populations, among populations within groups, and within populations. Two different models were utilized in the AMOVAs: one based on regional structure and the other on temporal structure. The regional model partitioned variation by geographic region (i.e., Upper Keys vs. Exuma Cays) and among reefs within these regions (i.e., LG, ADM, and AG within the Upper Keys and NP and SP within the Exuma Cays) while the temporal model partitioned variation first by sampling time points and then by reefs within a sampling. It should be noted that the within-colony samples from colony sides collected in January (NP, Exuma Cays) and March (ADM, Upper Florida Keys) of 2004 were not included in the AMOVAs since they represent pseudoreplicates of a single sampling time. Pairwise tests for Symbiodinium population differentiation were also conducted by randomizing genotypes between pairs of populations 1) between reefs; and, 2) on a reef through time, using GENALEX v6.2. Significance of the AMOVAs and pairwise tests were assessed by 10,000 permutations.
For the Symbiodinium Clade B populations of both M. faveolata and M. annularis, the AMOVAs based on a regional structure model identified significant levels of genetic variation as being partitioned among geographic regions (40-53%) and among reefs within regions (32-47%), with a smaller (but significant) amount of variation (14-15%) among colonies on a reef ( Table 2). Under the temporal structure model, none (0.0%) of the genetic variation in the Symbiodinium populations was attributable to among sampling time points from either host species (Table 2). Instead, all of the genetic variation was partitioned among reefs (78-83%) or among colonies (17-23%) within sampling time points (Table 2). Specific regional and temporal patterns of genetic variation within and among the Symbiodinium Clade B populations of M. faveolata and M. annularis are detailed in the following subsections.
A single multilocus genotype dominated the Symbiodinium Clade B populations on both the tops and sides of Montastraea spp. colonies from the Bahamas (Table 3). Likewise, the majority (n = 7 of 9; Table 3) of colonies examined from the Upper Florida Keys also harbored populations along their tops and sides comprised of a single Symbiodinium genotype. However, five colonies (i.e., four M. faveolata and one M. annularis) from Admiral Reef in the Florida Keys possessed differing, or two, alleles at one locus (i.e., B7Sym34) on their tops versus sides (  Complete ITS2 and multilocus genotype data available in Tables S1-S2. doi:10.1371/journal.pone.0006262.t001 (Table 4). This is evident by the AMOVAs, where a relatively low proportion of genetic variation was partitioned among colonies on a reef (14.74% for M. faveolata, 13.81% for M. annularis; Table 2). For M. faveolata, between one (e.g., South Perry Reef, Exuma Cays) and six (e.g., Little Grecian Reef, Upper Florida Keys) Symbiodinium multilocus genotypes were identified per reef (Table 4). A similar range (i.e., 2-4) of symbiont genotypes was also recovered from M. annularis (Table 4). In most cases, these Symbiodinium genotypes were not uniformly distributed across colonies and/or through time. Instead, individual colonies of both species predominantly associated with only one (or more rarely, two to three in the case of M. faveolata) of these Symbiodinium Clade B genotypes during all sampling time points (Fig. 2, Tables S1-S2). When a differing genotype was recovered from a colony, this was typically at a single time point during the approximately three-year study period (Fig. 2, Tables S1-S2).
In the Upper Florida Keys, the Symbiodinium Clade B populations harbored by M. faveolata and M. annularis were distinct, with no genotypes being common between the two species (Table 4). Given this, all interspecies pairwise tests of Symbiodinium population differentiation between the Montastraea spp. on these reefs were significant (P,0.005). In contrast, two multilocus genotypes were shared by M. faveolata and M. annularis from the Exuma Cays (Table 4). Here, pairwise tests of population differentiation were significant between M. faveolata and M. annularis on North Norman's Patch Reef (P,0.005); however, similar comparisons between South Perry M. faveolata and M. annularis were not significant (P.0.005).

Symbiodinium population variation between reefs and regions
The Symbiodinium Clade B populations of Montastraea spp. in the Upper Florida Keys exhibited strong endemism to reefs (Tables 2,4). Specifically, 12 of 13 genotypes harbored by M. faveolata were site-specific. The single outlier to this pattern was a genotype recovered on both Admiral (9 of 30 samples) and Alligator (1 of 23 samples) Reefs. In the case of M. annularis, all genotypes were site-specific. Thus, all pairwise tests of betweenreef Symbiodinium population differentiation by species were significant (P,0.005) in the Upper Florida Keys.
In contrast, the Symbiodinium Clade B populations of Montastraea spp. in the Exuma Cays were less structured. Although two and three Symbiodinium genotypes associated with M. annularis were sitespecific to North Norman's Patch and South Perry Reefs, respectively, one genotype was common to colonies of this species at both sites (Table 4). This particular Symbiodinium genotype was also the most frequently recovered from M. faveolata at North Norman's Patch and South Perry Reefs (Table 4). This led to pairwise tests of between-reef population differentiation being significant (P,0.005) for Symbiodinium Clade B harbored by M. annularis, but not for those of M. faveolata.
Regionally, no Symbiodinium Clade B genotypes were common between the Upper Florida Keys and the Exuma Cays (Table 4). Thus, all regional pairwise tests for population differentiation were significant (P,0.005).

Symbiodinium population stability through time and a coral bleaching episode
This study complements the multi-year temporal tracking of Symbiodinium ITS2 ''types'' associated with Montastraea spp. [31] by extending it to the population level for symbionts from Clade B. As previously reported, colonies generally harbored the same ITS2 ''type'' over time (Tables S1-S2), with all M. faveolata and most M. annularis associated with Symbiodinium ITS2 ''type'' B1 (the exception being M. annularis from Little Grecian Reef, which harbored ITS2 ''type'' B10). Likewise, the same Symbiodinium Clade B multilocus genotype was maintained by the majority of colonies throughout the study duration ((M. faveolata [57.7%, 15 of Table 2. Effects of regional and temporal structure on population genetic variation in Symbiodinium Clade B populations of Montastraea spp. as determined by analysis of molecular variance (AMOVA) of three microsatellite loci.  Fig. 2, Tables S1-S2). In the remaining colonies with two or three symbiont genotypes (M. faveolata with two [n = 7] or three genotypes [n = 4]; M. annularis with two genotypes [n = 6]), these tended to temporally alternate and/or were maintained as a mixed population (Fig. 2). As a consequence of this stability within a colony and population variation between colonies and reefs, AMOVAs revealed that none of the genetic variation was attributable to among sampling time points (Table 2). Along with this, pairwise tests found no statistically significant change in the Symbiodinium Clade B populations from any of the sampled Montastraea spp. colonies occurred over time (P.0.05). The temporal sampling scheme of this study also provided an opportunity to observe how Symbiodinium Clade B populations respond to a temperature-induced bleaching episode. During the summer of 2005, Montastraea spp. in the Exuma Cays were exposed to temperatures .30uC for an extended period of time, with conditions exceeding those in preceding years (Fig. 3). As a result, extensive bleaching was recorded from many scleractinian species at both North Norman's Patch and South Perry Reefs (Fig. 4, [33]). Visible paling of the colonies sampled for this study was accompanied by a concomitant decrease in the density of Symbiodinium to levels at, or below, 5610 5 cells ? cm 22 of host tissue (Fig. 3). Essentially, this bleaching event reduced Symbiodinium densities well below those observed during the normal seasonal fluctuations of symbionts in Montastraea spp. [35].
Within this context, little variation was observed in the ITS2type of Symbiodinium associated with either M. faveolata or M. annularis over the course of the 2005 bleaching episode, with colonies continuing to maintain symbioses with either Symbiodinium ITS2 ''types'' B1 or B10 (Tables S1-S2; note the detection of ITS2 ''type'' D1a in one colony of M. faveolata) [31]. Similarly, this event had no detectable effect on the Symbiodinium Clade B populations of these Montastraea spp. colonies (Table 2, Fig. 2). In most cases, a given host harbored the same Symbiodinium multilocus genotype before, during, and following the event. Furthermore, while four colonies (i.e., one M. faveolata and three M. annularis) did exhibit compositional changes in their populations through the event, the recovered genotypes were consistent with those encountered at previous time points from within the same (or other Exuma Cays') individuals (Table 2, Fig. 2). Thus, no statistically significant change in the Symbiodinium Clade B populations from any of the colonies was observed as a result of the bleaching episode (P.0.05).

Discussion
This study is the first fine-scale comparative and temporal analysis of genetic diversity and population structure for Symbiodinium associated with two ecologically dominant reefbuilding scleractinian corals in the Caribbean, Montastraea faveolata and M. annularis. Here, microsatellite markers revealed Montastraea spp. harboring Symbiodinium Clade B populations that were generally endemic to a reef, specific to a given host species, and temporally stable through both seasonal change and environmental perturbation in the form of a high temperature-induced bleaching event. The significance of these results is discussed below.
Patterns and processes of genetic structure between Symbiodinium populations Strong population structure ranging from between-species comparisons on the same reef to within-species surveys at geographic sites separated by tens to hundreds of kilometers characterizes the Symbiodinium ''type'' B1 populations of Montastraea spp. in the Upper Florida Keys and Bahamas. While an identical Symbiodinium multilocus genotype inhabits the majority of colonies of a host species on a given reef, the sharing of genotypes between sites separated by ,12.5 km or more is nearly non-existent. These findings are consistent with reports of strong differentiation between Symbiodinium ''type'' B1 populations from other Caribbean host species, such as the gorgonians Pseudopterogorgia elisabethae [22] and Gorgonia ventalina [23][24] in the Bahamas and Florida Keys, respectively. Similar patterns have also been documented in another Symbiodinium Clade and oceanic basin. Specifically, the Symbiodinium Clade C populations of the alcyonacean coral Sinularia flexibilis are significantly structured along the Great Barrier Reef in the Pacific Ocean at scales from ,16-1,360 km [25]. Given this, a potentially ubiquitous feature of Symbiodinium populations is high levels of structure over spatial distances as small as 10's of km.
In contrast to the above, Symbiodinium Clade C populations hosted by the scleractinian Pocillopora meandrina were reported to have no structure across ,200 km in the central and western Pacific Ocean [26]. One explanation for this could relate to Pacific LG current systems facilitating dispersal and connectivity between populations. Alternatively, differences in a host's mode of Symbiodinium acquisition are a potential driver for these patterns [44]. Specifically, although P. elisabethae, G. ventalina, S. flexibilis, M. faveolata, and M. annularis acquire their symbionts horizontally from the environment [22,[24][25]27,34], P. meandrina transmits Symbiodinium vertically from maternal parent to offspring [26]. Thus, since vertically transmitted Symbiodinium disperse in concert with their host's progeny, this may convey greater dispersal capability and higher levels of connectivity relative to symbionts horizontally transferred through the environment.
One interpretation for the low diversity of the examined Montastraea spp. Symbiodinium Clade B populations on a reef is that the coral hosts are propagating primarily via asexual reproduction (i.e., fragmentation) and that the same genetic individual (i.e., genet), harboring an identical Symbiodinium genotype, was independently sampled multiple times. Such a scenario would lead to the perception that a single Symbiodinium genotype dominates that particular host species on a reef. However, this hypothesis seems unlikely for two reasons. Firstly, sexual reproduction is the dominant reproductive mode in Montastraea spp. [27] and previous studies demonstrated only low levels of clonality in M. faveolata and M. On the y-axes, each unique Symbiodinium genotype is assigned a specific vertical position. Arrows and numbers indicate the specific Symbiodinium multilocus microsatellite genotype; numbers correspond to genotype numbers presented in Table 4. The x-axes depict each of the tagged colonies (n = 50) tracked during this study. annularis [9,45]. Secondly, when they occur, multiple colony genets are generally in close proximity to one another (i.e., 0.15 to 6.94 m apart) in, for example, M. annularis [45]. Given that our sampling scheme utilized spatially separated colonies (4 to 50 m apart) and assuming levels of clonality consistent with previous reports [9,45], host clonality should have a minimal impact on the dataset since less than two clonal colonies per host species per reef would have been tagged and sampled during the course of this study.
Overall, the available data suggest dispersal of Symbiodinium as environmental pools is generally limited. This is particularly relevant since a majority of hosts acquire their symbiont(s) anew from such pools at each generation. For example, approximately 81 genera in the broadcast spawning scleractinian corals alone depend on this mode of Symbiodinium infection [46]. Thus, although it has been established that these symbionts are present in the environment [47][48][49][50][51][52][53][54] and aposymbiotic hosts quickly acquire them when exposed to natural seawater [e.g., [55][56], analyses of within-host populations imply that they do not disperse widely since most are endemic to a single reef (see Results and references above). Like many dinoflagellates, Symbiodinium is capable of independent dispersal through the water column [57][58][59] as they alternate between a vegetative and flagellated/motile state [60][61]. However, Symbiodinium motility is governed by light conditions and cells swim just 3-10 m per vegetative and dividing period [61]. Such behaviors have the potential to retain these symbionts within localized populations (i.e., a given reef) and contribute to the geographic isolation reported here and elsewhere.
An alternative hypothesis for the strong structure observed in Symbiodinium populations is that while genotypes are fully capable of dispersing between reefs, migrants are rarely successful at establishing a new symbiosis upon arrival due to competition with the resident/endemic genotype. Invertebrates harboring Symbiodinium constantly expel symbiont cells as a means of density regulation [58,[62][63][64][65]; this would result in the resident/endemic genotype for that particular reef and/or host species occurring at higher densities than that of immigrants in the environmental infection pool. Under such a scenario, an aposymbiotic host would have a higher probability of acquiring the numerically dominant Symbiodinium genotype present in that environment, which in this case is from the resident/endemic population. Thus, the densitydependent feedback loop created by these cycles of expulsion and reacquisition could severely limit the opportunity for novel Symbiodinium genotypes to penetrate and homogenize the established population on a reef.

Symbiodinium specificity within and among host species
At the population level, M. faveolata and M. annularis in the Upper Florida Keys harbored distinct Symbiodinium ''type'' B1 genotypes (Table 4). This is particularly noteworthy since both 1) acquire their Symbiodinium from the environment at each generation, and; 2) occur immediately adjacent to one another on the same reef. Similar patterns of Symbiodinium population partitioning in sympatric and congeneric host species have been reported for Pseudopterogorgia spp., G. ventalina [34] and Pocillopora spp. [26], with this phenomenon attributed to fine-scale specificity between partners. Thus, while Montastraea spp. are among the most flexible cnidarians known when it comes to their symbiotic associations (see Introduction), specific host-symbiont pairings are clearly evident in the Symbiodinium populations of these scleractinian corals from the Upper Florida Keys.
In contrast, two Symbiodinium genotypes were common between the Montastraea spp. on Bahamian reefs (Table 4). Although it is possible that additional polymorphic microsatellite markers might differentiate these Symbiodinium populations, one hypothesis for this regional variability in pairings is that host-Symbiodinium specificity is influenced by the degree of hybridization between these coral species. Based on molecular genetic and morphological evidence, Fukami et al. [66] reported higher frequencies of introgression between M. faveolata and M. annularis in the Bahamas relative to Panama, suggesting the existence of a geographic gradient in hybridization between closely related Montastraea across the Caribbean [see also 67]. Given this, it is conceivable that M. faveolata and M. annularis hybridize less frequently in the Upper Florida Keys compared to the Exuma Cays and anecdotal observations of Montastraea spp. coralite morphology support this conjecture (D.J. Thornhill and W.K. Fitt, personal observations). Thus, Montastraea spp.-Symbiodinium specificity at the population level appears to be tracking host genetics and morphology, with introgression in the Bahamas breaking down the mechanism(s) leading to specificity and potential co-evolution between the symbiotic partners. Future studies might test this hypothesis by comparing the fine-scale molecular genetics of these hosts in relation to their Symbiodinium populations.

Temporal and stress stability of Symbiodinium populations
Most of the Montastraea spp. colonies examined here maintained the same Symbiodinium Clade B genotype(s) over a three-year duration. In cases where change was observed, this was typically a brief (i.e., between two sampling time points) transition between genotypes previously recovered from that colony and/or endemic to that particular reef (Table 4, Fig. 2). While the sampling design of this study cannot identify the specific mechanism(s) behind this variability, potential reasons include spatial heterogeneity of genotypes within the Symbiodinium population of a colony (Table 3), shuffling in the relative densities of genotypes within a population [e.g., [68][69], and/or switching through the acquisition of genotypes from the external environment [e.g., 70]. Notably, no more than two Symbiodinium Clade B genotypes were simultaneously recovered from any Montastraea spp. colony over the study period (Table 3, Tables S1-S2), which is consistent with previous population-level analyses or microsatellite screening from Caribbean cnidarians harboring this Clade [22][23][24]39,71]. Such a pattern suggests within-host competition may be occurring between closely related Symbiodinium [72], with a stable maximum of only two Clade B genotypes co-occupying a single host at a given time. This contrasts sharply with comparable analyses of Symbiodinium Clade C populations, where individual hosts appear capable of harboring .2 (and up to 11) genotypes simultaneously [25][26]. Further work is needed to determine whether these variable levels of within-host population heterogeneity are confined to the examined cnidarian species or are general and specific characteristics of Clades B and C, respectively, as well as if the other Symbiodinium Clades follow these precedents.
In addition to temporal stability, the Symbiodinium Clade B populations of Montastraea spp. exhibited resilience in the face of physiological and/or environmental perturbation (Figs. 2-3). Specifically, the 2005 bleaching event in the Bahamas did not induce any statistically significant change in the genotypic composition of populations harbored by the examined colonies and only limited change at the ITS2-rDNA level (Fig. 2, Tables S1-S2). Thus, while bleaching has been postulated to drive acclimatization through changes in host-Symbiodinium associations [e.g., [73][74], no support for this hypothesis is evident in these data from M. faveolata, and M. annularis. Instead, severe bleaching events often result in substantial coral death through differential mortality [e.g., 75], with concomitant reductions in the biodiversity of an affected reef [12 and references within]. Because populations represent the fundamental unit of evolution and due to the population-level endemism and fine-scale specificity of Symbiodinium seen here and in other studies [22][23][24][25][26]34], the potential loss of host and symbiont diversity from coral reefs via bleaching events may ultimately compromise the adaptive capacity and long-term persistence of many symbiotic invertebrates.