Sharing the slope: depth partitioning of agariciid corals and associated Symbiodinium across shallow and mesophotic habitats (2-60 m) on a Caribbean reef

Background Scleractinian corals and their algal endosymbionts (genus Symbiodinium) exhibit distinct bathymetric distributions on coral reefs. Yet, few studies have assessed the evolutionary context of these ecological distributions by exploring the genetic diversity of closely related coral species and their associated Symbiodinium over large depth ranges. Here we assess the distribution and genetic diversity of five agariciid coral species (Agaricia humilis, A. agaricites, A. lamarcki, A. grahamae, and Helioseris cucullata) and their algal endosymbionts (Symbiodinium) across a large depth gradient (2-60 m) covering shallow to mesophotic depths on a Caribbean reef. Results The five agariciid species exhibited distinct depth distributions, and dominant Symbiodinium associations were found to be species-specific, with each of the agariciid species harbouring a distinct ITS2-DGGE profile (except for a shared profile between A. lamarcki and A. grahamae). Only A. lamarcki harboured different Symbiodinium types across its depth distribution (i.e. exhibited symbiont zonation). Phylogenetic analysis (atp6) of the coral hosts demonstrated a division of the Agaricia genus into two major lineages that correspond to their bathymetric distribution (“shallow”: A. humilis / A. agaricites and “deep”: A. lamarcki / A. grahamae), highlighting the role of depth-related factors in the diversification of these congeneric agariciid species. The divergence between “shallow” and “deep” host species was reflected in the relatedness of the associated Symbiodinium (with A. lamarcki and A. grahamae sharing an identical Symbiodinium profile, and A. humilis and A. agaricites harbouring a related ITS2 sequence in their Symbiodinium profiles), corroborating the notion that brooding corals and their Symbiodinium are engaged in coevolutionary processes. Conclusions Our findings support the hypothesis that the depth-related environmental gradient on reefs has played an important role in the diversification of the genus Agaricia and their associated Symbiodinium, resulting in a genetic segregation between coral host-symbiont communities at shallow and mesophotic depths.


Background
Historically, coral species were thought to be extremely ecologically and morphologically plastic, because of the broad variation observed "within species" [1]. Advances in coral taxonomy and ecology have demonstrated that this variation often reflects sibling species occupying distinct ecological niches [2]. The recent advent of molecular approaches has exposed a further wealth of cryptic diversity in scleractinian corals [3][4][5] often in relation to particular environments, highlighting the role of niche partitioning [6][7][8][9] in addition to ecophenotypic plasticity [10,11]. Likewise, the algal endosymbionts that associate with reef-building corals and allow them to exploit light as an energy source in oligotrophic conditions, were originally thought to consist of a single algal symbiont, Symbiodinium microadriaticum [12]. By now, we know that Symbiodinium diversity in tropical scleractinian corals consists of several different phylogenetic clades [13][14][15], species [16,17] and hundreds of different subclades based on the ITS2 region of the ribosomal DNA [14,16]. These different Symbiodinium genotypes often represent physiologically distinct entities that can be strongly partitioned across host species, geographic regions and particular reef environments [18][19][20][21][22][23]. Although ecological niche partitioning across environmental gradients (e.g. light, temperature, turbidity) appears fundamental to maintaining the diversity of both scleractinian corals and their endosymbiotic dinoflagellates, the evolutionary context of this hostsymbiont niche partitioning remains poorly understood.
Despite the great diversity of Symbiodinium types found in reef invertebrates, many coral host species appear to associate with a specific genetic lineage (or group) of Symbiodinium [16]. Particularly in brooding corals, where the onset of symbiosis occurs through the vertical transmission of Symbiodinium from the maternal colony to the larvae, high levels of host-symbiont specificity have been observed [8,[24][25][26][27]. Molecular studies targeting both symbiotic partners have identified tightly coupled host-symbiont associations in brooding corals at the family [20], genus [9] and species level [8,9,26]. As such, natural selection may be acting on the level of the coral holobiont (i.e., host and symbiont combined) in species with a vertical symbiont acquisition mode, resulting in coevolution of both symbiotic partners. Incongruence between the phylogenies of corals and their associated Symbiodinium indicates that recombination of host-symbiont associations does occur over the course of evolution [13,16,28,29], however the occurrences of such events may be less likely over ecological timescales. It is therefore important to study the evolutionary origin and implications of such host-symbiont specificity, particularly in light of the environmental change predicted over the next decades [30].
The evolutionary context of coral host-symbiont niche partitioning is also important in the context of reef connectivity between distinct environments and more specifically the ability of reef habitats that offer some kind of protection against certain disturbances, such as deep reefs, to act as a "protected" local source of propagules [31,32]. Although the association of a coral species with different Symbiodinium types may facilitate its distribution across distinct reef environments [20,[33][34][35], larval connectivity between these alternative environments may be hampered if symbiotic associations are constrained through vertical symbiont transmission and recruits are not able to adopt symbiont types better-suited to their "new" environment [26]. Along those same lines, host populations may be locally adapted to opposing environments and distinct symbiont populations may in fact reflect genetic differentiation of both symbiotic partners [31]. Such genetic partitioning of host and symbiont populations has been observed across shallow and deep populations of the pocilloporid coral species Madracis pharensis in the Caribbean [9] and Seriatopora hystrix on the Great Barrier Reef [8,26,32]. To gain a better understanding of the ability of mesophotic reefs to act as refugia (for species that occur over large depth ranges) and a subsequent source of reproduction for degraded shallow reefs it is important to determine the overlap in genetic diversity of host and symbiont communities in shallow and deep reef environments. Given the prevalence of a brooding reproductive mode (>65%) and therefore vertical symbiont transmission on Caribbean reefs [31,36], there is a potential for strong genetic differentiation between host-symbiont associations in shallow and mesophotic communities in this region [27]. However, this hypothesis remains largely untested, primarily as molecular data from mesophotic coral communities are limited.
Mesophotic reef communities in the Caribbean are commonly dominated by members of the Agariciidae [37 -39], and in particular by the species Agaricia lamarcki and Agaricia grahamae [31,40]. The plating growth forms, such as those of Agariciidae, are thought to be particularly well suited to maximize light capture in these low-light environments [41]. Nonetheless, species of the genus Agaricia can be abundant throughout the entire depth range of Caribbean reefs and A. humilis, for example, occurs only in very shallow sections of reefs [42]. Species of Agaricia can exhibit high linear growth rates [43] and these brooding species are major contributors to the pool of coral recruits on Caribbean reefs [38,42,44,45]. Data on the phylogenetic affiliation of Symbiodinium associated with Western-Atlantic agariciids is limited, however records so far demonstrate association with Symbiodinium from clade C and occasionally clade D [16,[46][47][48][49][50]. It remains unclear whether the brooding reproductive mode of Western-Atlantic agariciids has resulted in host-symbiont specificity and to what extent the associated Symbiodinium diversity allows this genus to be abundant across large depth gradients.
Here, we assessed Symbiodinium genetic diversity for the five most common agariciid corals in Curaçao over a large depth range (2-60 m) encompassing both shallow and mesophotic depths. We addressed the level of specificity between the different coral species and their dominant Symbiodinium (i.e. the algal type(s) that are most abundantly present in the coral tissue), and evaluated whether these host-symbiont associations are partitioned over depth. Additionally, sequencing of a host mitochondrial region disclosed agariciid evolutionary patterns, and allowed evaluation of further genetic structuring of the coral host in relation to symbiont and depth. Results are discussed in light of extensive previous work on the brooding species of the coral genus Madracis [9,21,34,51] to address common patterns of genetic diversity over depth, and explore potential processes of codiversification in brooding Caribbean corals and their algal symbionts. Overall, the potential role of depth-related processes in the evolution of brooding corals and their associated Symbiodinium is discussed, as are the implications with regards to larval connectivity between shallow and deep coral reef habitats.

Study site and environmental conditions
Collections were performed between 2008 and 2009 at the well-studied site Buoy Zero/One (Bak 1977;Bak et al. 2005;Vermeij et al. 2007;Frade et al. 2008a) in Curaçao, southern Caribbean (12°07'31" N, 68°58'27"W) (Figure 1a). Information obtained from additional samples of previous collections at the same study site in 2004 and 2007 were added to the dataset. The study site is, as are many leeward sites on Curaçao, characterized by a shallow terrace (50-100 m wide) that slopes down to 8-12 m, followed by a steep drop-off with a slope extending down to a deep terrace at 50-60 m depth [52]. Environmental conditions, including seawater temperature and light attenuation were characterized during the beginning of the sampling period and described in Frade et al. [9]. Downwelling light availability (or PAR irradiance) decreases exponentially with a K d s of 0.07 m -1 [9], leading to~87% of surface irradiance at 2 m, 70% at 5 m, 50% at 10 m, 17% at 25 m, 6% at 40 m, 3% at 50 m, and 1.5% at 60 m. Temperature at the study site shows seasonal variation, with mean seawater temperatures decreasing over depth (0.02°C per metre) [9]. The deeper reef slope (40-60 m) experiences cold-water influxes due to thermocline movements, which can cause a sudden decrease in temperature of several degrees [9,39].

Agariciid species
The five most common agariciid species in Curaçao Table 1). Fragments of colonies that were collected but that did not clearly conform to these species descriptions were labeled as "Agaricia sp.".

Species abundances and sampling approach
Species abundances of the different agariciid species were determined in 2009 by counting all individuals in 30-50 quadrats (1 × 1 m) laid out randomly (i.e. using a list of randomly generated numbers) along a 100 m long transect parallel to the shore at each sample-depth (2, 5, 10, 25, 40, 50, and 60 m). A total of 335 agariciid coral colonies were sampled over a depth range from 2 to 60 m from a single reef location (Buoy Zero/One), by removing a small fragment from the colony (~3 cm 2 ). . Tissue from collected coral samples was separated from the skeleton using a modified airgun attached to a SCUBA cylinder or with a sterilized scalpel, and was subsequently stored in 20% DMSO/EDTA preservation buffer or 95% ethanol at -20°C until further processing. DNA was extracted from the tissue using a Qiagen Plant Mini Kit, MoBio Ultra Clean Soil DNA Kit (following the manufacturer's instructions), or a slightly modified method used for black tiger shrimp [32,56].

Symbiodinium ITS2 analyses
The internal transcribed spacer (ITS2) region of Symbiodinium rDNA was amplified for all specimens (n = 335) using Symbiodinium-specific primers [46] as described in Bongaerts et al. [8]. The amplified ITS2 fragments were separated using denaturing gradient gel electrophoresis (DGGE) on a CBScientific System following conditions outlined in Sampayo et al. [20], to identify the dominant Symbiodinium types in each sample (ITS2-DGGE fails to detect symbionts present in low abundances [57,58]). On each DGGE gel, we ran representative samples to allow for profile crosscomparison and identification. Representative, dominant bands of each characteristic profile were excised (from several replicate profiles), eluted overnight in dH 2 O, re-amplified and purified (using ExoSAP-IT) prior to sequencing. The re-amplified PCR products were sequenced in both the forward and reverse directions (ABI BigDye Terminator chemistry, Australian Genome Research Facility). All chromatograms were analyzed using Codoncode Aligner, with sequences being aligned with MUSCLE and blasted on GenBank (http://www.ncbi.nlm.nih.gov/BLAST). Phylogenetic analyses of sequences were performed using maximum parsimony and maximum likelihood in MEGA 4 [59] under the delayed transition setting and calculation of bootstrap support values based on 1000 replicates. Host atp6 sequence analyses The mitochondrial atp6 gene (and parts of the flanking nad6 and nad4) was amplified for a subset of samples (n = 93:~5 samples were randomly picked per species per depth) using the newly designed primers Aga-atp6-F1: GGCTTTATTTGGGGCTGAA and Aga-atp6-R1: CCCACAAAACCAAAGCACTTA. Primers were designed from the mitogenomes of A. humilis (NC008160) and Pavona clavus (NC008160) [60]. . PCR products were purified, sequenced, aligned and analysed as specified for the Symbiodinium ITS2 sequences. Additional sequences were retrieved from GenBank to compare the atp6-based phylogenies with those based on the mitochondrial cytochrome oxidase 1 (COI) and the nuclear 28S region [60][61][62][63]. Phylogenetic analyses of sequences were constructed using maximum likelihood (ML), maximum parsimony (MP) and Bayesian (BAY) methods through the programs MEGA5 (Tamura et al. 2011) and MrBayes (Huelsenbeck and Ronquist 2001) respectively. The best-fit model of molecular evolution was selected by hierarchical Akaike information criterion (AIC) using jModeltest [64] with a GTR model with invariant sites best describing the atp6 data under a log likelihood optimality criterion. For the COI and 28S sequences retrieved from GenBank, a GTR model with gammashaped variation and invariant sites best described the data of both these regions. Maximum likelihood analyses were performed using 1000 non-parametric bootstrap replicates. Bayesian analyses were performed with the Markov Chain Monte Carlo search run with 4 chains for 10 6 generations, with a sample frequency of 100 generations and a "burn-in" of 2500 trees.

Statistics
The contributions of taxonomic species, depth and Symbiodinium type on host genotypic variability (atp6) were assessed for positively identified Agaricia specimens under the AMOVA framework using GenALEx V5 [65], with either depth or Symbiodinium type nested within taxonomic species. In host species for which multiple Symbiodinium profiles were observed, an (nested) analysis of similarity (ANOSIM) was carried out using Bray-Curtis Distance in the software package Primer v6 to test for differences between sampling years (Two-way ANOSIM with depth nested within year) and depths (One-Way ANOSIM comparing individual depth groups).

Species abundances over depth
The species abundances at Buoy Zero/One differed among depths and revealed different distribution ranges for the five agariciid species (Figure 1c). At the shallowest depths (2-5 m), A. humilis is the most dominant species, with A. agaricites starting to occur at 5 m. At 10-15 m, A. agaricites becomes the most dominant species, with colonies of A. humilis and A. lamarcki being present in lower abundances. At 25 m depth A. agaricites and A. lamarcki occur in roughly equal abundances, but A. lamarcki takes over as the dominant species beyond this depth. From 40 m onwards, A. grahamae starts occurring, although in low numbers. H. cucullata was observed from 10-60 m, but always in relatively lower abundances.

Symbiont diversity across host species and depth
Across the five different host species, a total of six distinct ITS2-DGGE profiles were distinguished ( Figure 2; see Additional file 1 for DGGE profiles). A distinct profile was observed for A. humilis (P1, n = 71), A. agaricites (P2, n = 113), and H. cucullata (P5, n = 16). Three different profiles (P3, n = 34; P4, n = 69; P4*, n = 2) were observed for A. lamarcki, of which one (P4) was shared with the species A. grahamae (n = 20) (Figure 2). Given that five of the six profiles contained at least three codominant ITS2 sequences, we decided to assign a number to the DGGE profiles (P1, P2, P3, P4, P4* and P5) rather than referring to individual Symbiodinium types. Although these co-dominant sequences likely represent intra-genomic variants within the rDNA of a single symbiont lineage (little variation in relative band intensity was observed) [66], it cannot be excluded that some of the profiles may represent a mix of distinct Symbiodinium types. Additionally, other Symbiodium types may be present at background levels that cannot be detected due to the limitations of ITS2-DGGE [57,58].
All the ITS2-DGGE fingerprint profiles for Agaricia spp. contained the Symbiodinium C3 sequence, in combination with 2-3 sequences that were all closely related to C3 (Figure 3). All of these sequences were novel (GenBank Accession Numbers KF551185-KF551192), except for C3b in A. humilis, C3d in A. lamarcki, and C11 in A. lamarcki/grahamae. Novel sequences are identified (in this manuscript) by a nomenclature that specifies the sequence to which they are most related, followed by a capital N (indicating novel sequence) and an arbitrary number in italics (e.g. C3bN1). Except for the shared C3 sequence, all profiles contained distinct Symbiodinium sequences (i.e. no overlap between profiles) indicating that each of these profiles contain (a) different Symbiodinium type(s). However, one Symbiodinium ITS2 profile (P4*) was observed in only 2 colonies of A. lamarcki (at 25 m) and was identical to the P4 profile but with an extra band that corresponded with the C1 sequence. The several Agaricia morphs that could not be unambiguously identified (Agaricia sp.) all contained the P4 profile (n = 10). For H. cucullata we could only obtain a single recoverable ITS2 sequence (CN8), which represented a more distant, novel Symbiodinium type (GenBank Accession Number KF551192) (Figure 3).
A. humilis, A. agaricites, A. grahamae and H. cucullata all harboured a single symbiont profile across the entire depth range across which they were sampled. A. lamarcki, the only species observed to harbour multiple symbiont profiles, showed a marked zonation, with colonies at 10 and 25 m depth harbouring either P3 (n = 34) or P4 (n = 36), whereas colonies (>40 m) exclusively harboured P4 (n = 33). Colonies hosting a mix of P3 and P4 profiles were not observed. There were no significant differences (Two-Way ANOSIM; depth nested within sampling year) between sampling years. The pairwise comparisons between the different depths confirmed that the symbiont community associated with A. lamarcki at 40 m was significantly different from 10 and 25 m (One-Way ANOSIM; not taking into account sampling year; respectively R = 0.15; p < 0.01 and R = 0.082; p < 0.05).

Host genetic structure across taxonomic species and depth
Seven different mtDNA haplotypes were identified across the 93 specimens for which the atp6 region (1260 bp including indels) was sequenced (Figure 4; GenBank Accession Numbers KF551193-KF551199). A single haplotype was observed for H. cucullata (n = 10), and indels were only observed when Agaricia sequences were aligned with those of H. cucullata. The remaining haplotypes were observed for Agaricia specimens and the number of substitutions between these sequences ranged from 0 to 16. The Agaricia haplotype network consists of two clades and four main haplotypes (i.e. subclades), with the majority of A. lamarcki and all A. grahamae specimens belonging to clade 2 (respectively 90 and 100%), most A. agaricites belonging to subclade 1a (86%), and A. humilis belonging to subclades 1b and 1c (69%). The Agaricia sp. specimens that were sequenced (n = 9 of 10) were found across the two main clades. Under the AMOVA framework (including only identified Agaricia specimens), 60% of molecular variance (Φ SPP-TOT = 0.599; P < 0.001) was explained by morphotaxonomic species designation. No significant contributions of Symbiodinium type (2%; Φ SYM-SPP = 0.058; P = 0.232) and depth (0%; Φ DEP-SPP = -0.059; P = 0.668) could be detected, when nested within taxonomic species. Phylogenetic analyses (ML, MP and BAY) of the atp6 region of the Agaricia specimens ( Figure 5), using H. cucullata and Pavona clavus as outgroups, supported the monophyly of the genus Agaricia and subdivision of the genus into two main lineages (with bootstrap values across phylogenetic methods higher than 97), with sequences of the two lineages largely corresponding with A. lamarcki / A. grahamae and A. humilis / A. agaricites specimens respectively. The latter lineage is further subdivided into two lineages largely corresponding with A. humilis and A. agaricites (ML/MP/BAY: 94/69/93). Despite the mostly invariant host sequences among conspecifics, a small number of specimens belonging to different species were observed within each lineage. However, in all these cases the symbiont type hosted by these specimens confirmed the taxonomic identity (see Figure 4). Specimens that could not be identified (Agaricia sp.) were found to belong to the three different groups, although all hosting Symbiodinium profile P4, with 5 specimens from 25-40 m in the A. lamarcki/ grahamae group, 2 specimens from 50 m in the A. agaricites group, and 2 specimens from 60 m in the A. humilis group.
Analyses of COI and 28S sequences for the same Agaricia spp. [60][61][62][63], using Pavona spp. and respectively Leptoseris sp. and H. cucullata as outgroups, corroborate the phylogenetic pattern revealed through the atp6 region (Figure 5b, c). However, where COI shows no differentiation between A. humilis and A. agaricites, the 28S region differentiates between A. lamarcki and A. grahamae specimens. As such, the 28S region supports the subdivision of the Agaricia genus into two main lineages, corresponding to shallow (A. humilis and A. agaricites) and deeper species (A. lamarcki and A. grahamae).
Despite this cladal division into shallow and deep Agaricia species, no clear genetic subdivision was observed within species over depth ( Figure 5), or in the case of A. lamarcki in relation to symbiont type ( Figure 4). Nonetheless, several shallow specimens of A. lamarcki were found within the A. humilis (n = 3; 10 m) and A. agaricites groups (n = 1; 25 m). Additionally, several specimens of A. agaricites from 25 (n = 1) and 50 m (n = 2) depth were observed in the A. lamarcki/ grahamae lineage.

Discussion
This study assessed the depth distribution and genetic variation of five agariciid coral species and their associated Symbiodinium across a large depth gradient (2-60 m) on a single reef. Overall, we found a tight coupling between coral species and their dominant Symbiodinium at the study location, and strong evidence for depth-related niche partitioning of these associations. Depth-related parameters appear to have played a major role in the diversification of the coral genus Agaricia and its dominant algal endosymbionts, resulting in a clear genetic partitioning between coral host-symbiont communities across shallow and mesophotic depths.

Symbiodinium diversity and host-symbiont specificity
Associations between the coral host and its dominant algal endosymbionts (Symbiodinium) at the study location were found to be specific in that each of the five agariciid species harboured a distinct Symbiodinium profile (except for A. lamarcki and A. grahamae that had one profile in common) ( Figure 2). The four Symbiodinium profiles found in association with the four Agaricia species all shared the C3 sequence as an intragenomic variant and contained closely related Symbiodinium sequences (Figure 3). This corroborates observations in other parts of the Caribbean (Bahamas, Barbados, Belize, Florida, Mexico, US Virgin Islands), where Agaricia species have been found in association with C3-related Symbiodinium types [16,[46][47][48][49][50].
Nonetheless, except for the C3b-related Symbiodinium type in A. humilis (profile P1), the agariciids studied here hosted distinct Symbiodinium types compared to those previously reported (e.g. B1, C3, C3a, C3q, D1a) [16,[46][47][48][49][50]. A. lamarcki associated with two profiles (P3 and P4) containing respectively a C3d and C11 sequence, which indicate Symbiodinium types related to those found in respectively Montastraea (C3d) and mussid and faviid genera such as Scolymia, Mussa and Mycetophyllia (C11) [16,35,46]. H. cucullata hosted a Symbiodinium type (CN8/profile P5) that is very distinct from the C3 type previously reported in association with this species in other parts of the Caribbean [46,48,50], and does not contain the C3 sequence as an intragenomic variant. Some of these differences in Symbiodinium associated with Agaricia spp. (other studies) can be contributed to geographic separation of populations from the Western and Eastern Caribbean [50], however it also demonstrates that differences in symbiont associations exist even at smaller geographic scales, such as within the Eastern Caribbean (i.e., between Curaçao and Barbados). These observations corroborate the notion that the level of hostsymbiont specificity can vary across geographic locations [22,29,32], and therefore has to be assessed locally in terms of the ecological implications.
The maternal transmission of endosymbionts leads to relative isolation of host-specific Symbiodinium populations, and is likely to be responsible for the local host specificity and the geographic variation in host-symbiont associations, as Symbiodinium types coevolve locally with their coral partners [22]. Similar host-symbiont specificity has been observed in other coral genera with a vertical symbiont acquisition mode, such as in various pocilliporid genera (e.g. Seriatopora, Pocillopora, Stylophora, Madracis) and the genera Montipora and Porites [20,[22][23][24][25]29]. Although the incongruence between host and symbiont phylogenies points towards the importance    of occasional host-symbiont recombination [13,16,28,29], such recombination through "symbiont replacement" is unlikely to occur over ecological timescales [67]. Therefore, despite the variation in host-symbiont associations observed for Agariciids across the broader Caribbean, it remains questionable whether the agariciid species at the study location can respond to rapid environmental change due to the lack of standing variation (i.e. diversity) in Symbiodinium associations. The host-symbiont specificity observed here relates to the dominant Symbiodium types that are hosted by each species, and high-resolution approaches to the detection of Symbiodinium (e.g. using quantitative PCR or nextgeneration sequencing) may demonstrate that these agariciid species also host non-specific background Symbiodinium [57,58]. In addition, we observed the generalist Symbiodinium type C1 in 2 out of 105 sampled specimens of A. lamarcki (in addition to profile P4), indicating that non-specific associations occasionally do occur [29]. Although the potential of "symbiont shuffling" (i.e. shift in dominant Symbiodinium within a coral colony over its lifespan) as a response to environmental change remains debated [67], unusual associations (such as the C1 hosted in A. lamarcki) may become more abundant if these exhibit higher levels of holobiont fitness under the predicted changing environmental conditions.

Depth-partitioning of agariciid host species and associated Symbiodinium
The five agariciid coral species exhibited a clear partitioning across the depth-mediated environmental gradient, progressing from respectively A. humilis, A. agaricites and A. lamarcki as most abundant species from shallow to deep (Figure 1c). A. humilis is a clear "shallow-water specialist" largely restricted to the shallow reef terrace (2-10 m), whereas A. grahamae is a "deep-water specialist" that exclusively occurs in the mesophotic zone (40-60 m). In contrast, both A. agaricites and A. lamarcki are more generalistic in terms of their depth distribution, albeit that A. agaricites predominates at shallower depths (5-25 m) and A. lamarcki at greater depths (25-60 m), with 25 m representing a transitory depth between both species (Figure 1c). These depth-distributions are from a single study location, and may vary across locations with different environmental conditions (e.g. water clarity) [68]. However, similar zonation of Agaricia species has been reported across the Caribbean [37, 38,[69][70][71][72], with A. humilis and A. agaricites most abundant on the shallow end of the depth spectrum, and A. lamarcki and A. grahamae predominating at greater depths. H. cucullata was found across a large depth range (10-60 m), however only in very low abundances, reflecting the remarkable decline of this species on the reefs in Curaçao compared to~30 years ago [44,45].
Four of the agariciid species hosted a single Symbiodinium type over their entire depth range. In contrast, the depth-generalist species A. lamarcki exhibited symbiont depth zonation, changing from hosting P3 (i.e. C3/C3d -related) or P4 (i.e. C3/C11-related) in the shallow to exclusively P4 at mesophotic depths. The "deep-water specialist" A. grahamae harboured this same P4 symbiont profile, indicating that this Symbiodinium type may be particularly well-adapted to deeper water conditions. Similar putative "deep-specialist" Symbiodinium types have been identified in the genera Madracis and Montastraea [21,34,35]. Conversely, the ability of A. lamarcki to associate with another Symbiodinium lineage (represented by the P3 profile) may facilitate its broader depth distribution (i.e., extending into the shallow) compared to A. grahamae (although the P4 profile was found at shallow depths in A. lamarcki as well). Symbiont zonation appears to be a common feature of depth-generalist coral species in the Caribbean [31], however the occurrence of a single Symbiodinium profile in A. agaricites from 5-50 m depth indicates that such zonation, however, is not a prerequisite for broad distribution patterns [27,73].
When considering both the average densities of host species and dominant symbiont associations across depths, it is obvious that host-Symbiodinium assemblages exhibit a clear niche partitioning over depth, for all assemblages apart from H. cucullata / P5 (conceptual diagram shown in Figure 6). Although both A. humilis/ P1 and A. agaricites/P2 predominate in the shallow reef environment, A. humilis/P1 thrives in the shallowest of depths (2-10 m) whereas A. agaricites/P2 is most abundant at intermediate depths (10-25 m). As such, A. humilis/P1 is restricted to depths with high-irradiance conditions (87-50% of surface irradiance) and A. agaricites/P2 occurs in all but the brightest light conditions (1.5-70%). At greater depths (≥ 25 m), however, A. lamarcki/P4 becomes the most dominant host-symbiont assemblage. Given that all sampled colonies of A. lamarcki at 10 m depth (n = 40) were growing in cryptic locations on the reef (Bongaerts and Frade pers. obs.), the shallow end of A. lamarcki's depth distribution is clearly determined by a maximum tolerable irradiance. Additionally, the A. lamarcki/P3 host-symbiont association, which only occurs at 25 m and cryptically at 10 m, occupies a niche that encompasses a very narrow light range. Instead, A. grahamae that shares symbiont profile P4 (C3/C11-related) with A. lamarcki, starts to appear at slightly greater depths than the A. lamarcki/P4 combination, indicating the potential (acclimatization) importance of the host component in mediating the distribution of this host-symbiont partnership [34].
Although A. lamarcki was found to be most abundant at the deeper end of the sampled depth spectrum (60 m), anecdotal observations at greater depths (65-85 m) indicate that A. grahamae may in fact become the more dominant species (Figure 7; Vermeij and Bongaerts, pers. obs.).
Although incident irradiance is likely an important factor to explain the observed depth partitioning of hostsymbiont associations, other environmental variables, such as light spectral composition [74], temperature [75] and nutrient/plankton availability [26] may be important as well. For example, the depth range between 25 and 40 m appears to be a transition zone of strong genetic change, which corresponds to the depth range where influxes of deep oceanic water become noticeable in the temperature regime [21,39]. Given the various abiotic factors that vary over depth it remains impossible to evaluate the individual contributions of these factors in the observed partitioning, however light and temperature are likely to play important roles given their dramatic changes over depth at the study site [21,76].
While representing a massive undertaking, repeating similar symbiont genotyping efforts across a range of different locations would provide us with further insight into the role of different environmental conditions in the depth partitioning of host-symbiont associations.

Host and symbiont (co) evolution
The mtDNA-based phylogenetic analysis showed that none of the Agaricia species studied is monophyletic, however, the presently accepted species designation can explain the majority of the observed molecular variance (~60%). Two major lineages were observed for the Agaricia specimens based on the atp6 region, basically dividing the genus into a shallow (A. humilis/A. agaricites -Clade 1) and a deep (A. lamarcki/A. grahamae -Clade 2) species group. The atp6 phylogeny is consistent with the phylogeny of the ribosomal 28S region (outgroup H. cucullata not included) produced with GenBank sequences, although the 28S region appears more variable ( Figure 5). Although the atp6 region does not separate A. lamarcki and A. grahamae, the phylogeny of the 28S region demonstrates that these are unlikely to represent synonyms. The H. cucullata clade is genetically very distinct and even shows (across markers) a closer relation to the Indo-Pacific genus Pavona than the Atlantic Agaricia genus ( Figure 5). The phylogeny of Agaricia, forming two major clusters representing shallow and deep species, indicates that depth-related factors have played a major role in the original divergence within this genus. Given that also within each of these clusters, each species again exhibit distinct depth distributions on respectively the shallow and deep reef (Figure 1c), highlights the potential role of depth-related factors in the diversification within this genus throughout its evolutionary history.
Within the genus Agaricia all of the four major host atp6 haplotypes are, in fact, shared by more than one Agaricia taxon. This shared genetic polymorphism and repetitive non-monophyly has been shown in the past   Figure 6 Conceptual diagram depicting niche partitioning of agariciid host-symbiont assemblages over depth. Relative abundances are extrapolated from average densities of host species and relative symbiont abundances across depths.
for many other coral taxa [9,[77][78][79][80][81]. Such incongruence between morphospecies and genetically delimited phylogenies have been explained by several authors as being the result of phenomena such as morphological convergence (homoplasy), phenotypic plasticity, recent speciation with incomplete lineage sorting, morphological stasis, and/or interspecific introgressive hybridization [80]. Although introgressive hybridisation has been suggested in other brooding coral species [9,80,82], the fact that we only used a single mitochondrial marker prevents inference of such processes in the current study. The "misplaced specimens" in the phylogeny all hosted the Symbiodinium profile specific to that species (Figure 2), confirming that these are unlikely to represent taxonomic misidentifications. However, the observation that several shallower specimens of A. lamarcki (n = 4) grouped with the "shallow" species group (Clade 1a-c), and several deeper specimens of A. agaricites (n = 3) grouped with the "deep" species group (Clade 2) ( Figure 5), could potentially be indicative of hybridization due to low densities of conspecific sperm. The evolutionary divergence between "shallow" and "deep" Agaricia host species and genetic relatedness among them is reflected in the symbiont associations, with A. lamarcki and A. grahamae sharing an identical Symbiodinium profile (P4), and A. humilis and A. agaricites harbouring a related ITS2 sequence in their Symbiodinium profiles (C3b in P1 and C3N2 in P2) ( Figure 3). This highlights, despite potential hybridisation within the Agaricia genus (which could result in the crossing over of Symbiodinium types), the diversification of both host and symbiont through reciprocal evolutionary changes (at least at a microevolutionary scale). These observations corroborate the growing body of evidence suggesting that corals with a vertical symbiont acquisition (i.e., usually brooding corals) and their associated, dominant Symbiodinium are involved in coevolutionary processes [8,9,26,29,83].

A common pattern for Caribbean brooding corals?
The depth niche partitioning, host phylogenetic structure and local host-symbiont specificity observed for the genus Agaricia has strong similarities to that of the model genus Madracis, which has been extensively studied at the exact same study site [9,21,34,51,76,[83][84][85][86][87]. The Caribbean genus Madracis comprises six common taxonomic species whose distributions are partitioned over the reef slope, including shallow (M. mirabilis and M. decactis) and deep (M. carmabi and M. formosa) specialists and two depth-generalist species (M. pharensis and M. senaria). However, similar to A. lamarcki, the distribution of M. pharensis and M. senaria is restricted to cryptic locations at depths of 10 m and shallower [51]. As such, the genus Madracis can be roughly divided into "shallow" species (M. mirabilis and M. decactis) and "deep" (M. pharensis, M. senaria, M. carmabi and M. formosa) species. This subdivision corresponds to the two main clades observed in the Madracis phylogeny based on the mitochondrial nad5 region [9], similar to the genetic divergence observed between "shallow" and "deep" species in the genus Agaricia.
With regards to the associated Symbiodinium, the genus Madracis harbours three physiologically distinct but phylogenetically closely related symbiont types (B7, B13, B15), which are also partitioned over depth and to a certain extent specific to certain host species [21,34]. The shift in symbionts observed for M. pharensis corresponds to distinct physiological capacities of the symbionts involved [34], but also corresponds with a genetic divergence between shallow and deep populations of the host species M. pharensis [9]. Such intra-specific genetic divergence was not observed for A. lamarcki (the only Agaricia species that exhibited symbiont zonation), however it does corroborate the general pattern of strong host-symbiont coupling in Agaricia (this study).
Overall, the observed niche partitioning and evolutionary split between shallow and deep host lineages observed for the genera Agaricia and Madracis, highlight the role of the depth-related environmental gradient (i.e. differences in light, temperature, etc.) as a driving factor in the evolution of the congeneric species associated with these genera. The molecular phylogenies of both genera are relatively incongruent relative to the taxonomic classification based on morphological traits [9,84], highlighting the potential additional importance of introgressive hybridization. However, the role of introgressive hybridization in Agaricia is hard to infer given the use of a single, maternally-inherited marker region in this study. Nonetheless, the fact that species in both genera remain recognizable and have distinct ecological distributions does point towards the importance of divergent/disruptive selection across a depth gradient in the diversification of these genera. Given the prevalence of a brooding reproductive strategy [36] and vertical symbiont acquisition mode in the Caribbean, ecological adaptation to different ends of the depth spectrum may have played a major role in the diversification of many Caribbean corals.

Implications for vertical connectivity
In recent years, there has been a growing interest in the role of deep reef areas as local refugia and a source of propagules [31], particularly given the observed relative stablity of some mesophotic reef environments [39] and accumulating evidence of localized recruitment on coral reefs [88][89][90]. The extent of genetic specialization within the genus Agaricia to both shallow and deep reef environments ( Figure 6) indicates that, although the mesophotic Agaricia communities have remained relatively stable in the geographic location of our study [39], these communities are unlikely to provide significant amounts of propagules to ensure rapid shallow-reef recovery post-disturbance. The unrelated genus Madracis, shows a similar genetic segregation between its hostsymbiont communities at mesophotic (dominated by M. pharensis / B15 and M. formosa / B15) and shallow depths [9,21], confirming that this pattern may extent across a broad range of coral genera. Nonetheless, the limited overlap between shallow and deep communities indicates that deep reef may still play an important role in ensuring local protection of certain genotypes (e.g. the small number of A. agaricites colonies that occur in the deep) when adjacent shallow-reefs get affected by a major disturbance.

Conclusions
The findings in this study highlight that depth-related parameters have played a major role in the diversification of brooding corals and their associated Symbiodinium, and corroborate the notion that brooding corals and their Symbiodinium are engaged in processes of coevolution. Further studies should investigate whether any adaptive differentiation of these species has occurred at the population level, and whether the high levels of specificity observed here are also prevalent in other brooding coral genera. Overall, we need a better understanding of the actual mechanisms through which the depth-related environmental gradients have led to population/species divergence on coral reefs and how this evolutionary history affects the current day and future interconnectedness of shallow and deep reef systems.

Additional file
Additional file 1: Denaturing gradient gel electrophoresis of Symbiodinium ITS2 types associated with Agaricia species. Sequences used to characterize each symbiont profile are shown adjacent to bands in the gel image. Types in italics represent novel sequences, with the name specifying the sequence to which they are most related, followed by a capital N (indicating novel sequence) and an arbitrary number (e.g. C3bN1).