Distributed under Creative Commons Cc-by 4.0 Multi-gene Analysis of Symbiodinium Dinoflagellates: a Perspective on Rarity, Symbiosis, and Evolution

Symbiodinium, a large group of dinoflagellates, live in symbiosis with marine pro-tists, invertebrate metazoans, and free-living in the environment. Symbiodinium are functionally variable and play critical energetic roles in symbiosis. Our knowledge of Symbiodinium has been historically constrained by the limited number of molecular markers available to study evolution in the genus. Here we compare six functional genes, representing three cellular compartments, in the nine known Symbiodinium lineages. Despite striking similarities among the single gene phylogenies from distinct organelles, none were evolutionarily identical. A fully concatenated reconstruction, however, yielded a well-resolved topology identical to the current benchmark nr28S gene. Evolutionary rates differed among cellular compartments and clades, a pattern largely driven by higher rates of evolution in the chloroplast genes of Symbiodinium clades D2 and I. The rapid rates of evolution observed amongst these relatively uncommon Symbiodinium lineages in the functionally critical chloroplast may translate into potential innovation for the symbiosis. The multi-gene analysis highlights the potential power of assessing genome-wide evolutionary patterns using recent advances in sequencing technology and emphasizes the importance of integrating ecological data with more comprehensive sampling of free-living and symbiotic Sym-biodinium in assessing the evolutionary adaptation of this enigmatic dinoflagellate.


INTRODUCTION
Dinoflagellates in the genus Symbiodinium are essential components of coral reef ecosystems in their role as photosynthetic endosymbionts of a myriad of marine organisms belonging to at least five distinct phyla: Foraminifera, Porifera, Cnidaria, Mollusca, and Platyhelminthes (Trench, 1993). Although highly predominant within benthic hosts, symbiotic associations have also been reported in the pelagic medusa Cotylorhiza tuberculata (Astorga, Ruiz & Prieto, 2012). Perhaps best known for their relationship with scleractinian corals, Symbiodinium spp. underpin the productivity and calcification that creates coral skeletons and the structures known as coral reefs that serve as habitat for the immense biodiversity these coastal ecosystems support.
Historically, the pioneering work of Rowan & Powers (1992) divided the genus Symbiodinium into three phylogenetic groups referred to as clades A-C using nuclear small subunit ribosomal (nr18S) sequences. Despite the conserved nature of this marker, sequence variation between clades is comparable to other dinoflagellate taxa placed in different orders (Rowan & Powers, 1992). Later, the use of more variable nuclear large subunit ribosomal (nr28S) sequences was applied across broader host taxa and geographic scales (Santos et al., 2002;Pawlowski et al., 2001;reviewed in Baker, 2003), and ultimately led to the molecular classification of Symbiodinium into nine lineages , clades A through I (Table 1). Clades D, F, and G have been further divided into sub-clades D1-D2, F2-F5, and G1-G2, respectively, using nr28S and the chloroplast large subunit ribosomal DNA (cp23S) domain V (Hill et al., 2011;Pochon, LaJeunesse & Pawlowski, 2004;. Comparative phylogenetic reconstructions have yielded similar evolutionary relationships among Symbiodinium clades using nr28S and cp23S genes (Santos et al., 2002;, as well as when using the coding region of the plastid-encoded photosystem II protein D1 (psbA; Takishita et al., 2003), the mitochondrial cytochrome oxidase I (coI; Takabayashi, Santos & Cook, 2004), and mitochondrial cytochrome b (cob; Zhang, Bhattacharya & Lin, 2005;Sampayo, Dove & LaJeunesse, 2009). However, compared to other markers the nr28S gene typically yields better-resolved phylogenies and is therefore considered as a 'benchmark gene' for clade-level analysis of Symbiodinium (Pochon et al., 2012).
To improve our understanding of Symbiodinium clade evolution, in this study, we present a 'clade-level' multi-gene analysis incorporating samples from all known Symbiodinium clades and sub-clades (Table 2). We selected two genes from three distinct organelles (nucleus: nr28S & elf2; chloroplast: cp23S & psbA; and mitochondria: coI & cob) to test the following hypotheses: (1) single gene phylogenies will yield statistically distinct clades relationships; (2) A six-gene concatenated tree will be statistically different from benchmark nr28S; and (3) Pair-wise relative substitution rate analyses will reveal compartment-specific differences in evolutionary rates among Symbiodinium clades and organelles. Our results are integrated within the current state of knowledge of free-living and endosymbiotic Symbiodinium lineages (Table 1) and may serve as a basis for future studies investigating evolutionary implications of rarity and symbiotic/free-living lifestyles among Symbiodinium dinoflagellates.

Genes selection, DNA extraction and sequencing
Six genes from three organelles were chosen for phylogenetic analyses. These include two nuclear genes (1) (Pochon et al., 2012), nine DNA samples were extracted and partially analyzed in other studies (Pochon et al., 2007; and further sequenced here to cover all genes using the primers and PCR cycling conditions described in Pochon et al. (2012), and two DNA samples were extracted from sponge tissues of the genus Cliona (courtesy of C. Schoenberg) and sequenced for all genes following Pochon et al. (2012) (see Table 2). The psbA gene was not reported in Pochon et al. (2012) and was PCR amplified in this study using the forward primer psbA 1.0 (5 ′ -CWGTAGATATTGATGGWATAAGAGA-3 ′ ) located at the 5 ′ end of the coding region and the reverse primer psbA 3.0 (5 ′ -TTGAAAGCCATTGTYCTTACTCC-3 ′ ) located approximately 700 bp downstream from the 5 ′ end and using standard thermocycling conditions with an annealing temperature of 52 • C. All sequences were obtained by direct sequencing, except for nr28S and cp23S sequences, which were cloned prior to sequencing in Pochon et al. (2012), and a single sequence per sample included in the present study. In all cases, the variability between cloned sequences of any given sample was minimal (e.g., see Figure S1 of Pochon et al., 2012), ranging between 0 and 4 bp difference (data not shown). However, sequences showing the shortest branch length in each sample were selected (data not shown). In cases where several sequences showed the same short branch length, one sequence was randomly chosen among them and included in the analysis.

Phylogenetic analyses
DNA sequences were inspected and assembled using Sequencher v4.7 (Gene Codes Corporation, Ann Arbor, MI, USA) and manually aligned with BioEdit v5.0.9 sequence alignment software (Hall, 1999). Thirteen distinct DNA alignments were generated: six alignments corresponding to individual gene alignments, one fully concatenated alignment of all six genes (ALL Concat), and six partially concatenated alignments including all possibilities of five genes each (i.e., each alignment excluded one of the six gene candidates). Concatenated alignments were created using the 'join sequence files' option in TREEFINDER v12.2.0 (Jobb, von Haeseler & Strimmer, 2004). elf2 was included in these analyses despite two missing samples (see samples #27 and #30; Table 2), which were coded as missing data in all concatenated alignments. GenBank accession numbers for all investigated sequences are shown in Table 2.
Each DNA alignment was analyzed independently under both Maximum-likelihood (ML) and Bayesian environments. Best-fit models of evolution were estimated for each alignment (see Table S1) using Modeltest v3.7 (Posada & Crandall, 1998). ML analyses were carried out using PhyML v3.0 (Guindon et al., 2009), and the reliability of internal branches was assessed using 100 bootstraps with subtree pruning-regrafting branch swapping. Bayesian tree reconstructions with posterior probabilities were inferred using MrBayes v3.2 (Ronquist et al., 2012), using the same model of DNA evolution as for the ML analyses. Four simultaneous Markov chains were run for 1,000,000 generations with trees sampled every 10 generations, with 50,000 initial trees discarded as "burn-in", based on visual inspections. Concatenated alignments were run under ML and Bayesian environments as described above, with the alignments partitioned so that the specific model of evolution corresponded to each gene fragment.

Topological tests, rate calculations, and statistical analyses
To compare the topology of the various trees, approximately unbiased (AU) topological congruency tests (Shimodaira, 2002) were performed using site likelihood calculation in RaxML v7.2.5 (Stamatakis, 2006), followed by AU tests using CONSEL (Shimodaira & Hasegawa, 2001) with default scaling and replicate values. elf2 was excluded from the single gene analyses due to missing data (samples #27 and #30; Table 2), but was included in the concatenated analyses (see above).
In order to determine evolutionary rates among Symbiodinium lineages for each of the six investigated genes, relative-rate tests (RRT) were performed using the program RRTREE v1.1 (Robinson-Rechavi & Huchon, 2000). Clades and sub-clades were compared in a pair-wise fashion with G. simplex as the outgroup. Relative rates of evolution (K-scores from RRTREE analysis above) were compared among clades and among cellular organelles using a two way ANOVA, followed by post hoc analysis with Tukey's Honestly Significant Difference (THSD) test.

RESULTS
DNA alignments for the six investigated genes ranged between 473 (elf2) and 1,057 bp (coI). Individual phylogenies were generated (Fig. 1), and each was compared to the topology obtained with the nr28S gene, which is the current molecular taxonomic benchmark for the clade-level classification of Symbiodinium (Hill et al., 2011;Pochon et al., 2012). Overall, the cladal relationships were remarkably similar among the genes investigated, particularly the basal positions of clades A, D, E and G, and the derived positions of clades B, C, F, H, and I. Symbiodinium clades were relatively well resolved in the nuclear and chloroplastic genes, but not the mitochondrial genes, which placed clades C, F, and H in completely unresolved monophyletic groups (see Figs. 1E and 1F). However, with the exception of nr28S, the relationships amongst clades were weakly supported for all markers, especially in the higher parts of the trees, and this was  Table 2. Numbers at nodes represent the ML bootstrap pseudoreplicate (BP) values (underlined numbers; 100 BP performed) and Bayesian posterior probabilities (BiPP). Black dots represent nodes with <95% BP and BiPP of 1.0. Nodes without numbers correspond to BP and BiPP lower than 70% and 0.8, respectively. Nodes displaying BP lower than 50% were manually collapsed. The phylograms were rooted using the dinoflagellates Gymnodinium simplex, Pelagodinium beii, and/or Polarella glacialis. GenBank accession numbers are given in Table 2. Note: All clades are represented, except for clade E in the elf2 phylogeny. particularly evident for psbA where relationships between clades B, C, D, F, G, H, and I were completely unresolved (Fig. 1D). Furthermore, the relationships between sub-clades within clades D, F, and G showed contrasting results. Well-supported monophyly of all sub-clades was only observed in the nr28S gene (Fig.1A). Notably however, clade G sub-clades (G1 and G2) formed a monophyletic group across all genes. In contrast, the monophyly of clade F and clade D sub-clades was only resolved with nr28S (Fig. 1A) and nr28S and cob (Figs. 1A and 1F), respectively. All Symbiodinium strains belonging to the same sub-clade grouped together across all genes, with two noteworthy exceptions. First, the four samples of sub-clade F5 (#14-16) separated into two groups in cob (Fig. 1F). Second, sample #24 (Table 2) of sub-clade D2 diverged significantly to the root of the tree in cp23S (Fig. 1C).
In order to increase the phylogenetic signal and assess which of the individual markers best reflects the most well resolved evolutionary history of Symbiodinium, a series of gene concatenation analyses were conducted. In total, seven distinct concatenated alignments were analyzed, including one fully concatenated alignment of all six genes (ALL Concat) consisting of a total length of 4,703 bp, and six partially concatenated alignments ranging in length from 3,646 bp (ALL except coI) and 4,230 bp (ALL except elf2), and including all possibilities of five genes each (see Methods). Phylogenetic analysis of the fully concatenated dataset (ALL Concat, Fig. 2) resulted in a highly resolved Symbiodinium tree with identical topology to nr28S gene, but with much stronger phylogenetic signal as evidenced by a significant increase in statistical support at all nodes (Fig. 2). Other concatenated alignments yielded weaker nodes support and unstable cladal relationships globally (data not shown).
Approximately unbiased (AU) topological congruency tests (Shimodaira, 2002) were used to verify whether any of the distinct phylogenies resulted in statistically identical topologies. First, pair-wise comparisons of single gene phylogenies (Fig. 1) resulted in significant p-values (p < 0.05) in all cases, indicating that the different genes have not followed identical evolutionary trajectories (see Table S2A). Second, concatenated topologies tested against single gene topologies, also resulted in significant p-values in all instances (data not shown). Third, pair-wise comparisons of single gene phylogenies to the concatenated topologies, revealed that the two longest genes, coI and nr28S, resulted in 5 and 6 significant topological comparisons, respectively (see Table S2B). Despite the relatively smaller size of nr28S (920 bp) compared to coI (1057 bp), nr28S was the only marker yielding a statistically identical topology to the fully concatenated topology (ALL Concat). The nr28S topology, however, was not identical to the best topology of the concatenated alignment excluding the nr28S gene fragment (see ALL except nr28S in Table S2B). Similarly, pair-wise comparisons of concatenated topologies revealed that significant p-values (p < 0.05) were only observed against the 'ALL except nr28S' topology (Table S2B).
The variable branch lengths observed in the six phylograms ( Fig. 1) are directly proportional to the amount of character change; hence the longest branches are indicative of increased evolutionary rates of any given Symbiodinium strain. In most cases, increased rates of Symbiodinium clades/sub-clades appeared to be gene-specific rather than a Figure 2 Best topology of Symbiodinium based on six concatenated genes. Maximum likelihood (ML) topology for Symbiodinium clades and sub-clades A to I based on fully concatenated DNA alignment (ALL Concat; 4,703 bp) of all six genes investigated in this study. The Symbiodinium strains within each clade/sub-clade are referred using the specific numbers and corresponding ITS2 names in brackets (Table 2, Fig. 1). Numbers at nodes represent the ML bootstrap pseudoreplicate (BP) values (underlined numbers; 100 BP performed) and Bayesian posterior probabilities (BiPP). Black dots represent nodes with 100% BP and BiPP of 1.0. The phylograms were rooted using the dinoflagellates Gymnodinium simplex, Pelagodinium beii, and Polarella glacialis. character state maintained across all markers. K-scores from relative rate tests were coupled with ANOVA to compare the relative rates of evolution among the clades and organelles (Fig. 3) examining all clades across the three makers. There was no significant interaction of clade and organelle (F 16,175 = 1.57, p = 0.081), indicating that the pattern of changes in rates of evolution among clades were similar across organelles. Overall the general pattern of slower relative rates of evolution for some of the basal clades (A, E) and faster rates in more derived clades (C, F, H, and I) is held across organelles. However, organelles differed in their relative rates of evolution (F 2,175 = 248.9, p = 0.0001), driven by rapid rates in the chloroplastic and nuclear compartments in comparison to the mitochondrial compartment (Fig. 3A), with the most rapid rates found in the chloroplastic markers due the high evolutionary rates of clade I and sub-clade D2 (see Figs. 1C and 1D). Additionally, there was a significant difference between clades (F 8,175 = 3.87, p = 0.0003) driven by the slow rates of clade A, and the rapid rates of clade I (Fig. 3B).

Multi-gene analysis supports nr28S as a benchmark lineage marker
Our knowledge of Symbiodinium evolution has historically been constrained by the limited number of phylogenetic markers that have been applied to this group. To date, less than 15 DNA loci have been used to examine Symbiodinium diversity in a phylogenetic context (LaJeunesse & Thornhill, 2011;Pochon et al., 2012;Rowan & Powers, 1992;Sampayo, Dove & LaJeunesse, 2009;Takabayashi, Santos & Cook, 2004;Takishita et al., 2003;van Oppen et al., 2001), and evolutionary relationships among all existing Symbiodinium lineages have never been inferred using more than two concatenated genes . This study is the first to perform a multi-gene analysis using six markers representing three cellular organelles and integrating biological samples from all known clades and selected sub-clades that encompass the genus Symbiodinium. In spite of the overall similarity among the trees for each nuclear, chloroplastic and mitochondrial gene (Fig. 1), their topologies were statistically different (Table S2). This reflects within and among clade differences inherent to the individual markers. Most notably being the unstable positions of clades D, E, F5 and H, as well as weak support for among clade relationships observed in most markers investigated. Long-branch attraction artifacts (Felsenstein, 1985) most likely accounted for the placement of sub-clade D2 (sample #24) at the root of the tree in the chloroplast 23S topology, and for the monophyly of samples #7, 8, 13, and 14 in the cob topology. While the markers investigated here are conserved genes that have a priori limited utility for finer scale (i.e., within clade) analysis, each contains a unique set of characteristics, including variable cladal resolution and/or evolutionary rates (e.g., see samples #2 and #3 in coI or samples #7, 8, 13, 14 in cob), hence each marker has the potential to address different questions. These differences thus support our previous conclusion that no one gene fits all of the taxonomic questions being asked in the genus Symbiodinium (Pochon et al., 2012).
Our fully concatenated analysis, incorporating all investigated genes and totaling 4,703 bp, resulted in a highly resolved phylogeny that was statistically identical to the nr28S gene, a gene used as the benchmark for assigning Symbiodinium lineages ( Fig. 2; Table S2). The fact that the concatenated nuclear, chloroplastic, and mitochondrial genes display overall similar evolutionary histories, suggests that the molecular taxonomy of the currently recognized Symbiodinium clades using nr28S is robust , and that the points of clade differentiation are ancient, allowing for a concerted evolution of these conserved genes across genomes. These new results support a sequential evolution of Symbiodinium clades A/E/G1-G2/D1-D2/I/B/F2-F5/H/C, from most ancestral to most derived, respectively. It appears that there is a level of constraint in the system, with recombination likely being a rare event (Santos & Coffroth, 2003;but see Chi, Parrow & Dunthorn, 2014), a feature that maintains separation among lineages.

Compartment specific evolution and link to environmental preference/prevalence
Dinoflagellates are characterized by several genetic distinguishing features, including large genome size, and complex architecture and gene regulation (Barbrook et al., 2010;Hackett et al., 2004;Howe, Nisbet & Barbrook, 2008). One prominent feature is the large number of genes that have relocated from the ancestral organellar genome to the nucleus, resulting in a significant reduction in plastid and mitochondrondrial genomes. For example, the few genes that remain in the plastid of peridinin-containing dinoflagellates are primarily the core subunits of the photosystem (including cp23S), and the cytochrome b6f and ATP synthase complex (about 16 genes including psbA) (Hackett et al., 2004). Similarly, the mitochondrial genome of dinoflagellates has been reduced to three protein-coding genes (coI, coIII, and cob), but also contains a large number of non-functional fragments separated by repetitive non-coding DNA (Barbrook et al., 2010;Waller & Jackson, 2009). Despite the fact that the six Symbiodinium genes investigated here are only a very small subset of the Symbiodinium genome, they are physically separated in three cellular compartments, each with distinct evolutionary constraints and potential. For example, our comparisons of evolutionary rates between markers revealed that the differences among cellular compartments was primarily driven by the dissimilarity in the rates of evolution in cp23S and psbA in Symbiodinium lineages D2 and I (Figs. 1 and 2).
A possible explanation is that the increased evolutionary rates reflect rarity and adaptation to marginal habitats. It has been posited that rare taxa are important in driving evolutionary trajectories and innovations (Holt, 1997). Rarity in terms of small population size and isolation can drive high rates of adaptation and speciation (e.g., peripheral speciation;Mayr, 1963), as mutations in rare species are more likely to accumulate in the periphery of the founding population's habitat where rare species may be subjected to persistent directional selection in the absence of gene flow, as they colonize new areas (Garcia-Ramos & Kirkpatrick, 1997). Such a scenario is supported by the fact that lineages D2 and I have only been documented on few occasions (Carlos et al., 1999;Pochon et al., 2007;, despite numerous Symbiodinium surveys conducted over the last 20 years in both the Western Atlantic and Indo-Pacific Oceans targeting a diversity of host taxa, as well as free-living communities, and crossing a variety of spatial and temporal scales (reviewed in Coffroth & Santos, 2005;Stat, Carter & Hoegh-Guldberg, 2006). In addition, Symbiodinium D2 and I have only been detected in the Hawaiian Archipelago and Micronesia (Guam and Palau), some of the most isolated island groups in the world and areas known for harboring high levels of endemism in marine biodiversity (Hughes, Bellwood & Connolly, 2002;Pauley, 2003). Both lineages have been suspected to either be free-living because of the manner in which the sample was isolated (Carlos et al., 1999), or recently ingested free-living strains due to their apparent rarity in nature .
The high rates of evolution in chloroplastic genes in Symbiodinium sub-clade D2 and clade I might also reflect a relatively recent transition from free-living to symbiotic lifestyles. These habitats are extremely different in nature and composition, with free-living environments exhibiting high levels of environmental variability and unpredictability, while symbiotic habitats are relatively more predictable being spatially constrained and influenced by the biology of the host. These environmental differences undoubtedly drive the very different morphologies of Symbiodinium found in these two habitats, with free-living Symbiodinium flagellated and motile, and symbiotic Symbiodinium encysted and immotile. In terms of evolutionary trajectories, such differences in environment must exert a profound influence. Symbiodinium strains evolving predominantly in symbiosis must have adapted particular biochemical and chloroplastic functions in an environment that bears little or no resemblance to a free-living setting. Previous studies on the transition between symbiotic and free-living habitat show that changes in evolutionary rate occur in bacteria that have transitioned from free-living to a symbiotic lifestyle and mutualism (Lutzoni & Pagel, 1997;Moran, 1996). In addition, in some ectomycorrhizal assemblages, changes in evolutionary rate correspond to reversing from symbiotic to free-living lifestyle (Hibbett, Gilbert & Donoghue, 2000). Further, rapid and extreme environmental changes may favor the survival of rare and transitioning species, as their existing phenotypic diversity may contain traits pre-adapted to a changing environment (Holt, 1997).
Our examination of evolutionary rates for multiple markers and organelles provides an opportunity to begin addressing the implications for gene and genome evolution due to symbiotic lifestyle and dissimilarities in organellar genome constraints. Here we see the significantly slower evolutionary rates of Symbiodinium clade A compared to other clades as well as overall slower relative rates of the mitochondrial compartment across all clades (Fig. 3). As recently highlighted by Decelle (2013), the predominance of a particular lifestyle in marine microalgae (i.e., symbiotic versus free-living) can have important implications in genome evolution. Symbiodinium clade A is a basal lineage known to date back to at least 50 MYA and which has possibly survived through the climatic vicissitudes of the K-T boundary (Tchernov et al., 2004;. This clade easily overgrows other Symbiodinium clades in culture (Carlos et al., 1999) and shows attributes of parasitism in scleractinian corals (Stat, Morris & Gates, 2008). Additionally, clade A contains a high number of unique strains that may never establish symbiotic relationships (e.g., Coffroth et al., 2006;Hirose et al., 2008;Yamashita & Koike, 2013), and evolves at a similar rate to its close pelagic dinoflagellate relatives, contrasting with all other Symbiodinium clades which on average evolve six times faster based on nr18S sequence analyses (Shaked & de Vargas, 2006). As discussed by Decelle (2013) these differential traits and pressures of clade A, such as prevalence in the free-living environment with an occasional symbiotic lifestage (i.e., planktonic symbionts) provide a situation where the genomes are primarily influenced by external environmental pressures rather than host controlled traits. The resulting pressures are more likely to establish sexual exchanges within larger free-living populations, minimizing genomic impacts with often comparatively slower rates of evolution. In contrast, lineages that spend most of their lifecycle in hospite, which is arguably the case for most Symbiodinium clades (Table 1), tend to develop a certain dependence on the host which can lead to comparatively higher rates of change due to genome reduction and higher genetic drift associated with the absence of purifying selection through sexual recombination (Lynch, Koskella & Schaack, 2006).
Our analysis of relative-rates of evolution also indicated that mitochondrial genes evolved approximately twice slower than nuclear and chloroplastic genes. This result appears to contrast markedly with the recent study of Roy-Smith & Keeling (2012), which showed that silent site divergence of the mitochondrial genome in other protists with secondary red-algal derived plastids evolve 5-30 times faster than the divergence of their plastid genomes. These contrasting results may in part be due to the differences in DNA bases of a few selected genes in our study in comparison to the silent site divergence of complete mitochondrial and plastid genomes in Roy-Smith & Keeling (2012). Nevertheless, as there is evidence that our results from a subset of genes matches those of land plants and green algae, with more rapid rates of divergence in the plastid organelle, additional work is needed to further explore the implications of transitions between the free-living and symbiotic state for Symbiodinium, with a goal of gaining a more comprehensive understanding of the dynamics and mechanisms behind the different evolutionary trajectories observed in this study. Additionally, the increasing use of next-generation sequencing for characterizing entire Symbiodinium genomes (e.g., Barbrook, Voolstra & Howe, 2014) is an exciting avenue that provides unprecedented opportunities for the investigation of novel markers and paves the way for much more comprehensive phylogenomics studies to come.

CONCLUSIONS
Our study examines the performance of six genetic markers from three organelles in samples representing all currently documented lineages of Symbiodinium. As such it represents a comprehensive phylogenetic reconstruction of Symbiodinium, and highlights differences in the taxonomic resolution of each marker and their relative value in addressing a variety of evolutionary questions. Our series of phylogenetic analyses were conducted to address three working hypotheses. Despite striking similarities among the single gene phylogenies from distinct cellular compartments, none were evolutionarily identical confirming our first hypothesis. This result reflected within and among clade differences inherent to the individual markers. Our second hypothesis, however, was rejected and showed that a supermatrix tree incorporating all investigated genes (4,703 bp alignment) resulted in a highly resolved phylogeny that was statistically identical to the nr28S gene. This result provides additional support for the use of nr28S as a 'clade-level' benchmark gene for Symbiodinium. Finally, compartment-specific differences in evolutionary rates among Symbiodinium clade and gene organelle were revealed confirming our third hypothesis. Highest evolutionary rates were observed within the chloroplastic compartment, a pattern that was largely driven by fast evolving Symbiodinium clades D2 and I, two lineages that are rare in nature and which may be transitioning between free-living and symbiotic states. As such, rarity appears to associate with evolutionary innovation in a key functional compartment in Symbiodinium. The identification of different evolutionary trajectories in chloroplast genes that link with habitat and prevalence suggests that this organellar compartment is evolutionarily plastic and responsive. This finding may have important implications for our understanding of evolutionary processes that underpin a symbiotic lifestyle in this essential dinoflagellate group. Our analysis further revealed that investigated mitochondrial genes evolved approximately twice slower than nuclear and chloroplastic genes, an observation that contrasts with comparatively fast mitochondrial rates previously documented in non-symbiotic protists with secondary red-algal derived plastids. Together these results further highlight the need for deeper genome sequencing for a variety of Symbiodinium taxa with rapidly advancing next-generation sequencing approaches to understand the evolution of these enigmatic yet critical symbionts.