Abstract
Many broadly-dispersing corals acquire their algal symbionts (Symbiodiniaceae) from their environment upon recruitment. This horizontal transmission strategy has the potential to promote coral fitness across diverse environments provided that corals can associate with diverse algae across their range and that these symbionts exhibit reduced dispersal potential. Here we quantified genetic structure of algal symbionts in symbiosis with two hosts (Acropora hyacinthus, Acropora digitifera) across two spatial scales (across islands, within islands) in Micronesia using microsatellites. We contrast these symbiont genetic structures to previously published coral host structures across the same spatial gradient. We find that both hosts associated with two genetically distinct Cladocopium lineages (C3, C40), confirming that Acropora coral hosts associate with diverse algae across this range. In addition, each Cladocopium lineage exhibited signatures of moderate host specialization. Across locations, algal populations were significantly more genetically structured compared to their hosts: they demonstrated higher FST among islands and often exhibited significant divergence among reefs on the same island, which was never observed in coral hosts. While Cladocopium C3, like their hosts, broadly followed an isolation-by-distance pattern across islands, there was one notable case of no genetic structure across more than 2,000 km (between Yap and Pohnpei), demonstrating that Cladocopium genetic structure can be more complex than their hosts. Overall, our results support the view that horizontal transmission – establishing novel symbioses with local symbionts - has the potential to facilitate local fitness for broadly dispersing coral species, and highlight the complexity of factors affecting the population biology of Cladocopium.
Introduction
Many well-known symbioses involve the passing of symbionts from parents to offspring (vertical transmission), fully aligning the evolutionary trajectories of symbiotic partners and typically leading to their deep integration at biochemical and genomic levels (i.e. Buchnera in aphids (Nakabachi, Ishida, Hongoh, Ohkuma, & Miyagishima, 2014; Shigenobu & Wilson, 2011)). The result of such symbiosis is essentially a novel composite organism, often called the ‘holobiont’, upon which selection can act (Bordenstein & Theis, 2015). In other types of symbioses, the association between partners must be established anew each generation (horizontal transmission), which offers the host’s offspring the opportunity to sample a variety of symbiont lineages and select partners that potentially confer some sort of local advantage (Hilario et al., 2011; Schwarz, Krupp, & Weis, 1999; Usher, Bergman, & Raven, 2007). In theory, this kind of relationship should generate novel ecological opportunities for both symbiotic partners through their mixing and matching across environments. For example, association with ecologically specialized algal photobionts can lead to distinct ecological guilds of lichens (Peksa & Skaloud, 2011) or allow a fungal partner to expand its geographic range across a more broad climatic envelope (Fernandez-Mendoza et al., 2011). Similarly, in aphids, association with various horizontally transmitted bacterial symbionts allows these insects to colonize novel host plants across climatic zones (Henry et al., 2013).
Reef building corals are no exception, and associations with algal symbionts in the family Symbiodiniaceae is obligatory for the majority of tropical corals since they rely on photosynthetic byproducts from the algae for energy in oligotrophic waters and, in turn, the algae benefit from a protected and light-exposed residence as well as inorganic nutrients and CO2 concentration mechanisms provided by the host (Barott, Venn, Perez, Tambutte, & Tresguerres, 2015; Muscatine, 1990; Muscatine & Cernichiari, 1969; Trench & Blank, 1987). Given the obligatory nature of this symbiosis for the host, it is somewhat surprising that in the majority of coral species (~85%), algal symbionts are not transmitted vertically, but rather must be acquired by the juvenile coral from its local environment post settlement (Baird, Guest, & Willis, 2009; Fadlallah, 1983; Harrison & Wallace, 1990; Hartmann, Baird, Knowlton, & Huang, 2017). One possible benefit to this horizontal transmission strategy is that aposymbiotic coral larvae are highly dispersive (S. W. Davies, Treml, Kenkel, & Matz, 2015; Foster et al., 2012; Rippe et al., 2017; van Oppen, Peplow, Kininmonth, & Berkelmans, 2011), but the within-reef environmental variation that corals experience is highly variable (Gorospe & Karl, 2011) and conditions within the coral’s new reef environment are likely to be ecologically distinct from conditions on their natal reef (Baird, Cumbo, Leggat, & Rodriguez-Lanetty, 2007; LaJeunesse et al., 2004). In this way, corals can potentially improve their fitness by associating with locally available, and putatively ecologically specialized, algal strains (Byler, Carmi-Veal, Fine, & Goulet, 2013; E.J. Howells et al., 2012; Rowan & Knowlton, 1995).
Indeed, the diversity of algal symbionts in the family Symbiodiniaceae is rich (LaJeunesse et al., 2018) and specific coral-algae associations have been suggested to play pivotal roles in holobiont adaptation to climate change (Berkelmans & van Oppen, 2006; E.J. Howells et al., 2012). The genus Cladocopium (formerly clade C Symbiodinium; (LaJeunesse et al., 2018)) originated and diversified most recently among Symbiodiniaceae, and has achieved the highest diversity of all Symbiodiniaceae lineages (Lesser, Stat, & Gates, 2013; Pochon & Gates, 2010; Pochon, Montoya-Burgos, Stadelmann, & Pawlowski, 2006; Thornhill, Howells, Wham, Steury, & Santos, 2017; Thornhill, Lewis, Wham, & LaJeunesse, 2014). This diversity has been associated with functional variation in symbiont thermal performance across reefs (Sarah W. Davies, Ries, Marchetti, & Castillo, 2018; E.J. Howells et al., 2012) as well as with functional differences in gene expression between reef zones (Sarah W. Davies et al., 2018), lending support for the potential for reef-specific symbiont communities.
However, little is known about the population biology of Cladocopium spp. algal symbionts, including how their populations are structured relative to their coral hosts. Comparing the dispersive capacities of each of these symbiotic partners across diverse environments is an essential component to better understanding this symbiosis. Here, using microsatellites, we examined multilocus genotypes (MLG) of Cladocopium spp. algal symbionts hosted by two species of Acropora coral hosts – A. hyacinthus and A. digitifera – collected from the same reef locations across the Micronesian Pacific (Fig 1A, B). Our previous work demonstrated that both host species exhibited extensive genetic connectivity and their genetic structure was well explained by the patterns of regional surface currents and geographic distance in a classic isolation by distance pattern (S. W. Davies et al., 2015). Here, by analyzing two coral species that co-occur across the same islands as well as local reef environments, we aimed to investigate the relative roles of geographic distance and host species in driving the genetic structure of algal symbiont populations in the genus Cladocopium and to compare the structure of the algal symbiont populations to their coral hosts.
Materials and Methods
Sampling
This study comprised a subset of samples previously analyzed for coral host genetics in Davies et al. (2015) (Table S1, Fig 1A). Twenty-five individuals of each coral host species (Acropora hyacinthus and Acropora digitifera) were examined at two reef sites (Fig 1B) within each of seven islands, with the exception of Ngulu (where only A. hyacinthus was collected and only one site was visited) and Guam (where there were no identified A. hyacinthus), for a total of 13 reef sites.
Laboratory Procedures
DNA was isolated following Davies et al. (2013). Microsatellite primers consisted of five previously described Cladocopium-specific loci (previously described as clade C Symbiodinium) (Bay, Howells, & van Oppen, 2009; Wham, Carmichael, & LaJeunesse, 2014) and one novel locus mined using MsatCommander (Faircloth, 2008) from nucleotide EST data for lineage C3 (Leggat, Hoegh-Guldberg, Dove, & Yellowlees, 2007), for a total of six loci (Table S2). Loci were multiplexed in 20 µl polymerase chain reaction (PCR) mixtures containing 10 ng of template DNA, 0.1 µM of each forward and reverse primers, 0.2 mM dNTP, 1X ExTaq buffer, 0.025 U ExTaq Polymerase (Takara) and 0.0125 U Pfu Polymerase (Promega). Amplifications began at 94°C for 5 min, followed by 35 cycles of 94°C for 40 s, annealing temperature for 120 s, and 72°C for 60 s and a 10 minute 72°C extension period. Molecular weights were analyzed using the ABI 3130XL capillary sequencer. Data were binned by repeat size and individuals failing to amplify at ≥3 loci were excluded from downstream analyses.
Data Analysis
Although Symbiodiniaceae in hospite are assumed to be haploid (Santos & Coffroth, 2003), the genus Cladocopium are generally observed to have two copies of every allele (Thornhill et al., 2014; Wham et al., 2014; Wham & LaJeunesse, 2016). This apparent genome duplication may or may not correspond to a change in chromosome number, or the actual diploid state (Wham & LaJeunesse, 2016), and it has been suggested that these lineages should be scored as if they were effectively diploid (i.e. with the expectation of two alleles per locus) to appropriately construct multilocus genotypes (MLGs) from samples (LaJeunesse et al., 2014; Pettay, Wham, Smith, Iglesias-Prieto, & LaJeunesse, 2015; Thornhill et al., 2014; Wham et al., 2014; Wham & LaJeunesse, 2016). However, given that the ploidy of the Cladocopium samples in our study is unknown and because a single coral host could potentially contain several genetically distinct Cladocopium clones, data were analyzed in two ways. In the first analysis, MLGs were only included if they contained two or fewer alleles across all loci, which would suggest that all alleles originated from a diploid Cladocopium clone or a haploid clone with a duplicated genome (Wham et al., 2014). This MLG dataset consisted of 190 of 277 (69%) A. digitifera samples and 178 of 278 (64%) A. hyacinthus samples. A second analysis made no ploidy assumptions and alleles were simply designated a binary presence/absence values for each individual, thereby retaining all alleles and all individuals in the analysis. This analysis is described in detail below under the ‘additional unconstrained community analyses’ section.
For the first analysis, structure v2.3.3 (Pritchard, Stephens, & Donnelly, 2000) was used to assign similar MLGs to clusters (106 iterations, burn in = 300,000) across ten replicate runs for each K (1-10) using an admixture model with no location prior. Following Evanno et al. (Evanno, Regnaut, & Goudet, 2005), ΔK was calculated in structure Harvester (Earl & Vonholdt, 2012) and CLUMPP (Jakobsson & Rosenberg, 2007) and distruct (Rosenberg, 2004) produced all graphics. In this initial analysis, all samples exhibited strong assignments into two highly supported clusters. Following Wham and LaJeunesse (2016), we investigated the relationship between these cluster assignments and their allele identities and found strong relationships between assignments and allele identity, particularly at two loci (Sgr_21 and Spl_33; Fig S1), and therefore concluded that two distinct Cladocopium lineages were present. Additional analysis of psbAncr locus, described below, indicated that these two lineages were C3 sensu (LaJeunesse et al., 2003) and C40 sensu (LaJeunesse et al., 2004).
MLG data were then split into initial structure lineages using the BAYES ALLELE script from Wham and LaJeunesse (2016) and additional structure runs were completed on these lineage subsets. All C40 MLG data and C3 MLG data are available in the supplementary information (Supplemental files 1-2). Partitioned lineage data were investigated for genetic divergences among host species, islands and sites and hosts within islands using FST (AMOVA, 9999 permutations in GENALEX v6.5, (Peakall & Smouse, 2006)).
Pairwise differentiation (FST) and pairwise island distances (km) were correlated to test for isolation by distance (IBD) in Cladocopium C3 (data were insufficient for C40) using Mantel’s test implemented in the vegan package in R (Oksanen et al., 2013). Two matrices were tested, the full FST dataset and an FST data subset from which a single non-significant pairwise FST (Yap-Pohnpei) was excluded. These trends were compared to our previously published results from the two host species (S. W. Davies et al., 2015). Mean allelic diversities per island and numbers of private alleles per island were also calculated in genalex v6.5.
To visualize differences within C3 and C40 between host species, between islands, and between sites and host species within each island, assignment of samples to genetic clusters with prior grouping of island/host/site was performed in R (R Development Core Team, 2015) using discriminant analysis of principal components (DAPC) in the ADEGENET package (Jombart, 2008; Jombart, Devillard, & Balloux, 2010). DAPC is a machine-learning algorithm aiming to classify samples into user-defined groups based on their coordinates in the principle components’ space and because DAPC does not rely on a traditional population genetics model, it is free from Hardy-Weinberg equilibrium and linkage disequilibrium assumptions and thus is considered to be useful across organisms regardless of their ploidy and genetic recombination rates (Jombart et al. 2010). Successful reassignment, indicated by a high proportion of samples correctly assigning to their group, indicates that user-defined groups are distinct, which in our case implies genetic divergence. Here, data were converted into principal components (PCs) and then a-scores were computed to determine the optimal number of PCs to retain. A-scores determine the proportion of successful reassignment corrected for the number of PCs retained and ensure that model overfitting is not occurring (Jombart et al., 2010). Assignment rates, PCs and DFs retained, and the proportion of variance explained by the model are included in Tables S4. To further substantiate genetic divergence of our populations, we computed FST for each DAPC clustering analysis.
Sequencing Analysis of Cladocopium psbAncr
To confirm phylogenetic affiliation of the two detected Cladocopium lineages, the non-coding region of the circular plastid (psbAncr) was amplified in four representative samples (two for each lineage) using methods described by LaJeunesse and Thornhill (LaJeunesse & Thornhill, 2011). Amplified products were directly sequenced and aligned with ClustalW2 (http://www.ebi.ac.uk/Tools/msa/clustalw2/) to Cladocopium psbAncr sequences from common Indo-Pacific lineages (Table S5). The phylogeny was reconstructed using the default settings of Phylogeny.fr (Dereeper et al., 2008) and identities of lineages were interpreted based on percent sequence identity with previously identified Cladocopium lineages.
Additional unconstrained community analyses
Since the ploidy of Cladocopium remains debated and only four Cladocopium samples were confirmed via psbAncr sequencing, we elected to employ a second analysis of DAPC to confirm the population structure results we detected using the analyses that assume diploidy. All samples were included regardless of allele number at each locus and data were converted to a binary table of allele presence/absences within each sample (Supplementary File 3). Genetic clusters were then explored using DAPC with group priors as described above. In addition, because DAPC analyses maximize variation between groups, we also visualized data using a principle component analysis of binary allele data using the rda(x, binary=TRUE) function implemented in the vegan package in R (Oksanen et al., 2013).
Results
Two Cladocopium algal symbiont lineages observed in Micronesian acroporids
To enable standard population genetic analyses, data were restricted to corals hosting a single multilocus genotype (MLG), which encompassed 69% of A. digitifera and 64% of A. hyacinthus samples. Across the two coral species in Micronesia (Fig 1A, B), two distinct Cladocopium lineages were observed, which were most clearly discriminated by two of the six microsatellite loci (Fig S1). Sequencing of the psbAncr gene for two representative samples from each lineage identified these lineages as Cladocopium C40 and C3 (LaJeunesse & Thornhill, 2011) (Fig S2). It is important to note that the possible presence of other genera of Symbiodiniaceae in the background would not affect Cladocopium genotyping results since our microsatellite assays are highly genus-specific. Within the single MLG data subset, corals of both Acropora species from Palau and Ngulu were found to exclusively host Cladocopium C40 (Fig 1C, pink bars). C40 was also prevalent in A. digitifera at one reef site on Yap (Goofnuw Channel) and was occasionally found in A. digitifera across the rest of Micronesia (Fig 1C). All other Acropora samples across the spatial gradient explored associated with lineage C3 (Fig 1C, turquoise bars). Both Cladocopium lineages possessed high genetic diversity, with a total of 70 unique alleles across three islands in C40 (Table S3A) and 130 unique alleles across five islands in C3 (Table S3B). Mean numbers of alleles per island for each locus for C40 ranged from 3.00-4.67 with the highest number of private alleles observed in Palau (N=7)(Table S3A). Mean C3 allele numbers per island for each locus ranged from 3.83 to 5.17 and numbers of private alleles ranged from one to four (Table S3B).
Cladocopium C3 are more genetically structured than their coral hosts
Overall Cladocopium genetic divergence between the two host species across the entire dataset was significant for both C40 (FST=0.092) and C3 (FST=0.037), confirming that host species play a role in structuring Cladocopium populations across Micronesia (Fig 2A, B). All pairwise between-island FST’s for C40 were significant (Table 1A) and in fact, the high FST values for Palau and Yap C40 relative to C40 from Ngulu suggest the potential for a third lineage of Cladocopium (Fig 2C), although this pattern is not further explored here. C3 exhibited one surprisingly non-significant FST comparison (Yap-Pohnpei, 0.009) with all other island pairs exhibiting significant FST values, which ranged from 0.058 (Guam-Chuuk) to 0.078 (Yap-Kosrae) (Table 1B). Notably, when all pairwise comparisons were included, C3 differentiation did not show significant isolation-by-distance (IBD), which had been previously observed for both coral hosts over this range (S. W. Davies et al., 2015) (Mantel’s r=0.1161, p=0.1809, Fig 3A). However, when the single non-significant Yap-Pohnpei FST comparison was removed, significant IBD was observed for C3 (Mantel’s r=0.4082, p=0.0277, Fig 3B). IBD could not be computed for C40 due to too few between-island comparisons. With the exception of the Yap-Pohnpei comparison, C3 FST values between islands exceeded those for both host species and the slope of FST ~ distance dependence was parallel to the slope observed in both coral hosts (Fig 3B). Overall, with the exception of C3 collected on Yap, all FST values between reefs located within the same island (regardless of host species) were significant and ranged from 0.025 (C3 Guam) to 0.183 (C3 Kosrae) (Table 2). This result is in stark contrast to both coral host species, where no significant within-island genetic divergence was detected (S. W. Davies et al., 2015).
Algal symbiont genetic differentiation by coral host species and local reef environment
Discriminant analysis of principal components (DAPC) strongly differentiated between host species for both Cladocopium C40 and C3 (Table S4, Fig 2A, B), corroborating FST estimates and suggesting that host specificity drives symbiont genetic diversity. However, both Acroporia hosts were capable of maintaining symbiosis with the same Symbiodiniaceae lineages across Micronesia, demonstrating that these hosts have flexibility across the spatial range investigated here. In addition, DAPC demonstrated differences among islands for each Cladocopium lineage irrespective of host species: generally high per-island assignment rates were obtained for C40 (Fig 2C, 82-100%) and C3 (Fig 2D, 51-98%), consistent with both structure (Fig 1C, D) and FST results (Table 1). Notably, algal symbionts from Yap consistently showed the lowest assignments rates for both C3 and C40. Moreover, nearly all pairwise FST values between unique host-environment pairings on each island were significant for both lineages (Table 3), suggesting that some combination of host specialization, environmental partitioning and/or strong dispersal limitation across space is responsible for algal symbiont genetic differentiation. In accord with these results, of the two top eigenvalues in DAPC analysis within islands, one discriminant function (DF) explained divergence by host species and the other DF explained differences between reef sites (Fig 4). However, there were several examples of non-significant pairwise FST comparisons within the dataset (Table 3). These included A. hyacinthus symbionts across sites within Palau, A. digitigera symbionts within Guam, A. digitifera symbionts across sites within Yap and also A. hyacinthus and A. digitifera symbionts within Goofnuw on Yap (Table 3). However, it is important to note that the sample size of A. digitifera symbionts at Goofnuw was low (N=4) given that many of these hosts associated with C40, while the rest of A. digitifera symbionts on Yap associated with C3 (Fig 1C, D) and this may explain the lack of significance observed. Interestingly the strongest pattern of separation within host-environment pairings was observed for Kosrae (Fig 4E), suggesting that whatever processes are responsible for symbiont genetic divergence are strongest in Kosrae.
Lastly, to confirm that the patterns we observed were not simply an artifact of selecting only individuals with two or fewer alleles per locus, we performed additional unconstrained community analyses on all data in binary format for each island. We conducted both DAPC and Principle Component Analyses (PCA) and the general patterns were upheld: there was evidence of both genetic divergence by host species and reef environment within each island (Fig. S3, S4).
Discussion
Acropora corals establish symbiosis with diverse Cladocopium lineages
Across the Micronesian Pacific (Fig 1A, B), both Acropora coral hosts associated with Cladocopium algal symbionts, which were represented by two distinct lineages – C40 and C3 (Fig 1C; Fig S2), with potentially more diverse cryptic lineages present (i.e. Ngulu Cladocopium; Fig 2C). This observation demonstrates that both hosts show considerable flexibility in their symbiotic associations across their range and within their specific environments (Abrego, van Oppen, & Willis, 2009; Berkelmans & van Oppen, 2006). This specificity at the Cladocopium genus level is consistent with the wealth of previous comprehensive community composition studies suggesting that Indo-Pacific acroporids are generally dominated by algal symbionts in this genus (i.e. LaJeunesse et al., 2004; LaJeunesse et al., 2003). While initial algal symbiont infection is likely determined by the local availability of free-living symbionts and made possible by the flexibility of arriving coral recruits (Abrego et al., 2009; Cumbo, Baird, & van Oppen, 2013; Little, van Oppen, & Willis, 2004), a winnowing process involving selection, competition, immune system response, and differing rates of asexual proliferation likely plays a role in the eventual dominance of a single MLG in the majority of coral hosts in their respective habitat (Rowan, Knowlton, Baker, & Jara, 1997; Thornhill et al., 2017). Strict associations with a single MLG have been observed across a variety of coral hosts and Symbiodiniaceae genera (Baums, Devlin-Durante, & LaJeunesse, 2014; Pettay & Lajeunesse, 2013; Pinzón, Devlin-Durante, Weber, Baums, & LaJeunesse, 2011; Thornhill et al., 2014), however this strict association is not always the case (see E.J. Howells, van Oppen, & Bay, 2009; E. J. Howells, Willis, Bay, & van Oppen, 2013). Here, we observed that the majority of corals contained no more than two Cladocopium alleles per locus, which can be interpreted as hosting a single Cladocopium clone (A. digitifera: 69%, A. hyacinthus: 64%), corroborating previous works on Symbiodiniaceae genetics (reviewed in Thornhill et al., 2017). However, it is also important to note that our study only explicitly explored population genetic patterns in Cladocopium, the symbiont genus most commonly known to associate with Acropora in this region, by targeting Cladocopium-specific loci. Only a few samples (N=4) were specifically tested here for community level algal species identification, however our research group had previously characterized community level Symbiodiniaceae communities on Palau using ITS2 metabarcoding and determined that all hosts associated with one of two Cladocopium algal symbiont haplotypes (Quigley at al., 2014). However, these community level analyses were not performed on the other islands explored here and therefore, we are unable to comment on the genetic patterns of other algal genera known to inhabit corals in background amounts (Silverstein et al. 2012, Kennedy et al. 2015, Lee et al. 2016, Ziegler et al 2018).
Cladocopium C3 exhibits IBD and stronger genetic structure that coral hosts across Micronesia
We expected to observe strong genetic differentiation in Cladocopium, since the prevailing view of their life cycle involves symbiotic existence in sedentary hosts alternating with a short-lived free-living form that largely exists in the benthos where dispersal by ocean currents is limited and their primary role is to invade novel hosts (Fitt, Chang, & Trench, 1981; Fitt & Trench, 1983; Littman, van Oppen, & Willis, 2008; Magalon, Baudry, Husté, Adjeroud, & Veuille, 2006; Yacobovitch, Benayahu, & Weis, 2004; Ali et al. in press). Our data supported this hypothesis: with one notable exception of the Yap-Pohnpei comparison, we observed significant FST between all pairs of sampled islands in both C3 and C40 lineages. Our FST estimates for the more broadly distributed lineage (i.e. generalist), C3, did not exceed 0.078 (Table 1B) which is markedly lower than previously published estimates from other locations (Andras, Kirk, & Harvell, 2011; E.J. Howells et al., 2009). This could be attributable to the use of different markers and analytical approaches, differences in the spatial scales explored, or differences between generalist/specialist algal symbionts. In addition, exceptionally high genetic divergence values might be the result of the presence of additional Symbiodiniaceae lineages within the same genus. Here, as an example, we observed that C40 FST values involving Ngulu were higher than any of those observed in C3 even though C40-populated islands were only separated by 100’s, not 1000’s, of kilometers (Palau-Ngulu = 0.279, Yap-Ngulu = 0.326) (Table 1A). These results may suggest the presence of an additional Cladocopium lineage present in Acropora hyacinthus on Ngulu. Regardless of the magnitude, significant C3 and C40 FST estimates support the prevailing view that Symbiodiniaceae dispersal is limited across the seascape, although there are occasional cases of high genetic similarity over very long distances (C3, Yap-Pohnpei).
Our working hypothesis predicted stronger symbiont FST values compared to coral hosts. It also predicted that Cladocopium algal symbionts would follow isolation by distance (IBD) patterns, which was previously shown for both coral host species (S. W. Davies et al., 2015). Here, we were only able to investigate IBD in Cladocopium C3 given that C40 was only observed in high proportions at three islands (Fig 1B). Indeed, Cladocopium C3 FST values were nearly always greater than host FST values, however, when all pairwise island comparisons were included significant C3 IBD was not detected (Fig 3A). This lack of IBD was driven by one pairwise comparison between Yap and Pohnpei, which was the only insignificant FST across more than 2000 km distance (Table 1). When this outlier was removed, we observed significant IBD for Cladocopium C3 (r2=0.4082, p=0.0277) with higher values than their hosts (Fig 3B), which is consistent with other studies investigating host-symbiont population genetics (i.e. (Baums et al., 2014)). Similar levels of distance-dependent FST in Symbiodinium have also been observed in C1 in Pocillopora meandrina in the South Pacific Islands (Magalon et al., 2006) and B1 in Gorgonia ventallina (Andras et al., 2011). This lack of genetic differentiation between Yap-Pohnpei C3 (Fig 2D and Fig 3C) suggest that there is potential for occasional long-range dispersal in Cladocopium C3 symbionts and that their population genetics, and perhaps population genetics of other generalist symbionts, can exhibit complexities that cannot be predicted by traditional distance-based dispersal limitation (Thornhill et al., 2017; Thornhill et al., 2014).
Cladocopium C3 and C40 exhibit imperfect host specificity
The majority of reef-building coral species associate with a specific Symbiodiniaceae type, which have traditionally been broadly defined based on ribosomal and/or chloroplast markers (Fabina, Putnam, Franklin, Stat, & Gates, 2013; Rodriguez-Lanetty, Krupp, & Weis, 2004; Thornhill et al., 2014; Weis, Reynolds, deBoer, & Krupp, 2001). Previous Symbiodiniaceae multilocus genotyping (MLG) studies revealed that each of these symbiont types can harbor within-type diversity, both at the genetic and functional levels (E.J. Howells et al., 2012; E.J. Howells et al., 2009; Santos, Shearer, Hannes, & Coffroth, 2004). Here we observe significant Cladocopium population structure among two different host species within each site (Fig 4; Table 3). Previous work on octocorals similarly observed significant host differentiation among algal symbionts, however they found that this genetic divergence was driven by different aged cohorts and depth in their system (Andras et al., 2011). However, host habitat depth or age class cannot directly explain the host specificity observed in our study given that specific attention was payed to collecting colonies located at similar depths and of similar size classes. Instead, our data suggest that the local association of hosts and symbionts within the same genotypic cluster is due to host specificity in Cladocopium (Fig 3, 4), which has been previously proposed in symbionts hosted by Pocillopora in the south Pacific (Magalon et al., 2006). Since our study sampled two coral host species, we also detected that this specificity is imperfect: at every location, there were symbionts in one host species that would have been assigned to another coral host based on their MLG (Fig 4). This pattern suggests that Cladocopium’s host specialization is present, however there is genetic exchange between these host-specific symbiont assemblages within a location.
Genetic subdivision of Cladocopium within islands
Within each island and sympatric host species, the majority of Cladocopium pairwise comparisons exhibit significant genetic divergence between closely located reef sites (Fig 4; Table 3) suggesting either that symbiont dispersal is very limited or spatially varying selection among reef sites is at play. Genetic subdivision between reef sites at the same island is the best demonstration that symbionts exhibit stronger genetic structure than their coral hosts, which never showed significant substructure within islands (Davies et al., 2015). Notably, in some cases these within-island FST values are higher than those observed between islands, which are much greater distance apart (Tables 1-3). Given this pattern, it is tempting to speculate that genetic divergence among individual reefs might be due not only to dispersal limitation, but also to spatially varying selection, implying environmental specialization (i.e. local adaptation) in the symbionts. Unfortunately, these islands are remote and understudied and therefore we lack further support for this claim as we did not measure environmental parameters and did not assess symbionts’ fitness across environments. Among factors that might contribute to genetic subdivision among reefs irrespective of distance is high variation in reproductive success among Cladocopium clones on a local scale, which would elevate FST due to spatial discordance of short-term allele frequency fluctuations (Thornhill et al., 2017). Yet, previous work has demonstrated that other Cladocopium symbiont populations have exhibited classic signals of local adaptation (E.J. Howells et al., 2012), and therefore reef sites investigated here offer an excellent study system for investigating the fine-scale local adaptation potential of Cladocopium. If these algal symbionts are indeed locally adapted, this would ensure that horizontally transmitting coral hosts increase their local fitness by associating with local symbionts. To confirm this hypothesis, much future work is required to experimentally demonstrate that these symbionts are achieving their maximum fitness in their local reef environment (Kawecki & Ebert, 2004).
Authors’ Contributions
SWD and MVM conceived of the study, designed the study, coordinated the study and drafted the manuscript. SWD, MRK and MVM collected coral samples. SWD carried out molecular lab work, participated in data analysis, and carried out the statistical analyses; DCW participated in data analysis and interpretation; MRK participated in statistical analyses and carried out sequence alignments; All authors gave final approval for publication.
Funding
This study was supported by the Coral Reef Conservation program of the National Oceanic and Atmospheric Administration (NA05OAR4301108) to M.V.M. DCW was funded in part by Pennsylvania State University and grants from the National Science Foundation (OCE-0928764 and IOS-1258058).
Data Accessibility
All multilocus genotyping (MLG) data and binary presence/absence data for Cladocopium are available in the supplementary files 1-3.
Acknowledgements
Thanks to field assistants Carly Kenkel, Tim Keitt, Irina Yakushenok and David Stump. Nida Khawaja Rahman was integral in molecular work and James Derry assisted with data management. We are grateful to Ulrich Mueller and the many reviewers who have helped ameliorate this manuscript. We acknowledge the Federated States of Micronesia Department of Resources and Development (#11-27-09-01, #11-27-09-02), Republic of Palau Bureau of Marine Resources and Koror State Government (Marine Research permit: #09-201; Marine Resource Export Certification #RE-09-23) for all collection permits and CITES export permits.