Depth specialization in mesophotic corals (Leptoseris spp.) and associated algal symbionts in Hawai'i

Corals at the lower limits of mesophotic habitats are likely to have unique photosynthetic adaptations that allow them to persist and dominate in these extreme low light ecosystems. We examined the host–symbiont relationships from the dominant coral genus Leptoseris in mesophotic environments from Hawai'i collected by submersibles across a depth gradient of 65–125 m. Coral and Symbiodinium genotypes were compared with three distinct molecular markers including coral (COX1–1-rRNA intron) and Symbiodinium (COI) mitochondrial markers and nuclear ITS2. The phylogenetic reconstruction clearly resolved five Leptoseris species, including one species (Leptoseris hawaiiensis) exclusively found in deeper habitats (115–125 m). The Symbiodiniummitochondrial marker resolved three unambiguous haplotypes in clade C, which were found at significantly different frequencies between host species and depths, with one haplotype exclusively found at the lower mesophotic extremes (95–125 m). These patterns of host–symbiont depth specialization indicate that there are limits to connectivity between upper and lower mesophotic zones, suggesting that niche specialization plays a critical role in host–symbiont evolution at mesophotic extremes.

2. Introduction

Light attenuation is a primary physical parameter that limits the distribution of coral reefs across depths and habitats [1]. In the tropics, photosynthetic corals are found at depths that range from ca 0 to 150 m in clear waters [1]. The stark differences in irradiance that occur over this depth gradient on spatial scales of only tens of metres have major implications for the distribution of coral species, and the genetic structure of populations [2,3]. The shallow and deep communities differ in species composition, reflecting physiological specialization and capacity tuned to specific corals, with depth being a proxy for the suite of parameters that change moving from shallow to deep communities. Disruptive selection along depth gradients has been proposed to lead to genetic divergence and possibly speciation despite the lack of obvious spatial barriers to gene flow [4–6]. In the case of scleractinian corals, coevolution of the host and symbiont is an important consideration for niche specialization and habitat partitioning, as there are trade-offs between different types of Symbiodiniumdinoflagellates and host–symbiont specificities [7,8]. Symbiodinium spp. are likely to play a significant role in habitat partitioning and the ecological diversification of scleractinian corals along depth and habitat gradients. Striking patterns of depth-specific symbiont types have been reported in various coral species [9–12] and have been linked to differences in photo-physiological responses of different Symbiodinium types from shallow water (less than 14 m depth) dominant reef corals [13], or other depth-related environmental conditions acting synergistically such as temperature, salinity, pH, turbidity and nutrient availability [5,11].

Compared to shallow coral reef studies, mesophotic coral ecosystems have received very little attention because of logistical constraints and are just beginning to be explored [14–16]. The upper mesophotic (less than 60 m) is generally similar in community structure to shallow water ecosystems, whereas the lower mesophotic consists of a more distinct assemblage that is highly specialized to exceptionally low light conditions [14,16–19]. Vertical connectivity between shallow water and upper mesophotic zones is therefore of particular interest to understanding the resilience of shallow ecosystems to disturbance (i.e. the deep water refugia hypothesis; [2,5,20,21]). Deep reef ‘refugia’ areas are protected or dampened from disturbances that affect shallow reef areas and can provide a viable reproductive source for shallow reef areas following disturbance (reviewed in [5]). Mid-to-lower mesophotic zones, on the other hand, are ecologically very distinct, suggesting that connectivity would be limited across these zones, and that persistence in the lower mesophotic zone may require unique adaptations.

The genus Symbiodinium is phylogenetically diverse, consisting of nine divergent clades (A-I; [22]) and hundreds of different sub-clade types based on the internal transcribed spacer region 2 (ITS2) of nuclear ribosomal DNA [23,24], many of which arguably represent different species [25–27], but see 28. Despite numerous studies reporting striking patterns of host–symbiont specificity [29,30], biogeographic partitioning [31,32] and ecological zonation [2,11,13] of Symbiodinium ITS2 types, the high variation among the copies of this gene found in individual genomes complicates interpretation and makes taxonomic assignment problematic [28,33–35]. Recent advances in genomic research [36–39] provide novel opportunities for the identification and characterization of alternative Symbiodinium markers, including a variety of nuclear, chloroplast and mitochondrial genes [40–42]. Previous studies characterizing Symbiodinium spp. diversity in mesophotic corals have all relied on the use of a single marker, ITS2 [2,6,11,12,43–45]. Additional work is required to confront a wider range of available alternative markers and provide a more comprehensive understanding of the diversity and molecular taxonomy of Symbiodinium across contrasting environments.

The geographically isolated Hawaiian Archipelago is an excellent natural laboratory for studying speciation and adaptive radiation on land [46–48], and recent studies have shown similar patterns are present in the marine realm [49,50]. The genus Leptoseris is broadly distributed across depths within the Hawaiian Archipelago, presenting a unique experimental system to examine the potential for host–symbiont coevolution, speciation across a habitat gradient, and potential adaptive radiation across the Archipelago. Initial work on Leptoseris in Hawai'i reported the widespread presence of generalist Symbiodinium clades, and cryptic host diversity [43]; however, this general survey of host–symbiont diversity had limited sampling, and no attempt was made to taxonomically identify small fragments collected by submersible. More recent work has clarified the taxonomy of this genus by integrating molecular data with discrete microscopic features found in type specimens, showing close agreement between the coral genetic clades and skeletal microfeatures [51]. In addition, Luck et al. [51] found polyphyly between Leptoseris and Pavona and a putative new coral species, indicating that this group is in need of taxonomic revision. Luck et al. [51] also found trends suggesting possible depth zonation across the coral genetic clades; however, symbiont diversity was not examined. Here we sampled across the lower mesophotic depth gradient (between 65 and 125 m from the ‘Au‘au Channel; figure 1) in order to examine the genetic diversity of the coral genus Leptoseris and their associated symbiotic dinoflagellates using nuclear and mitochondrial markers.

3. Material and methods

3.1 Sample collection

Mesophotic corals (n=74) were collected across multiple depth gradients (65–125 m) using the Hawai'i Undersea Research Laboratory’s (HURL) manned submersibles, Pisces IV and V, during two cruises (January 2010 and February 2011) to the ‘Au‘au Channel (figure 1; see the electronic supplementary material, appendix A for sites coordinates) between the islands of Maui and Lāna'i aboard the R/V Ka'imikai-o-Kanaloa. In this study, we defined three mesophotic depth ranges: upper (65–75 m), mid (85–100 m) and lower (115–125 m). At each site along these depth ranges, representative corals approximately 20–30 cm in diameter were haphazardly selected from the middle of a Leptoseris reef, with each sample separated by at least 10 m in distance. A small, triangular piece of coral spanning from the middle to the outer edge of the coral head was removed using a Schilling Titan 4 manipulator arm, and placed in an individual sample container in the sampling basket. Collected samples were kept in a darkened container with ambient seawater and in situ temperatures, and processed in a darkened laboratory within 4 h of ascent to the surface. Each sample was photographed, sampled for DNA and then immediately frozen at −80°C.

3.2 DNA extraction, PCR and sequencing

Small biopsies of coral tissue (approx. 2 mm) were individually stored for a week in 600 μl of guanidium DNA extraction buffer [52]. All coral biopsies (n=74) were taken from the upper coenosarc region of coral fragments, and two additional biopsies (taken from the calyx and/or coenosarc region) were also taken from a subset of coral samples (n=12) haphazardly selected to cross the depth range of 75–125 m (electronic supplementary material, appendix A). These samples were used to determine whether different Symbiodinium mitochondrial cytochrome c oxidase I (COI mtDNA) genotypes would be found in different areas of the coral colony. Genomic DNAs from both the Leptoseris species and endosymbiotic Symbiodinium were co-extracted following [35].

Approximately 800 base pairs (bp) of a rapidly evolving intergenic spacer of Leptoseris spp. mitochondrial DNA (cox1–1-rRNA intron) was PCR-amplified using primers and thermocycling conditions described in [51]. PCR products were purified using the QIAquick PCR Purification Kit (Qiagen), and sequenced directly in both directions using the ABI Prism Big Dye Terminator Cycle Sequencing Ready Reaction Kit and an ABI 3100 Genetic Analyzer (Perkin-Elmer Applied Biosystems). All sequences were submitted to BLASTn search as well as compared to Luck et al. [51] sequence dataset for species-level identification.

A 1057 bp fragment of Symbiodinium spp. COI mtDNA was PCR-amplified using primers COX1_FOR2 and COX1_REV1, and the thermocycling conditions described in [41]. PCR products were purified and sequenced directly in both directions as described above. The Symbiodinium ITS2 of nuclear ribosomal DNA (rDNA) marker was PCR amplified using protocols described in [22], and the primers ITS-DINO and ITS2REV2. The gene products were ligated into the pGEM-T Easy vector (Promega) and transformed into α-Select Gold Efficiency competent cells (Bioline). A minimum of 10 colonies were screened for inserts using plasmid-specific primers, and the positive screens were treated with exonuclease I and shrimp alkaline phosphatase and sequenced in both directions, as described above.

3.3 Phylogenetic analyses

DNA sequence chromatograms were inspected and bi-directional sequences were assembled using Sequencher v. 4.7 (Gene Codes Corporation, Ann Arbor, MI, USA), aligned with Clustal W implemented in BioEdit v. 5.0.9 [53] and manually refined. Three main DNA sequence alignments were generated (Leptoseris spp. Cox1–1-rRNA intron, Symbiodinium COI mtDNA and Symbiodinium ITS2 rDNA). Additionally, a fourth comparative sequence alignment of cox1–1-rRNA intron was created, including all Luck et al. [51] sequences and one representative sequence from each clade reported in this study. The Leptoseris spp. Cox1–1-rRNA phylogenies were rooted using Siderastrea radians from whole mitochondrial genomes available in GenBank (DQ643838). The Symbiodinium COI phylogeny was rooted using SymbiodiniumF₁ described in [41] (GenBank JN558066). Both genes were analysed independently using maximum-likelihood (ML) and Bayesian methods. Best-fit models of evolution and ML inferences with global tree searching procedure (10 starting trees) were estimated using Treefinder v. 12.2.0 [54]. Robustness of phylogenetic inferences was estimated using the bootstrap method [55] with 1000 pseudoreplicates in all analyses. Bayesian analyses were performed using the parallel version of MrBayes v. 3.1.2 [56,57], starting from a random tree of four chains with two runs of Metropolis-coupled Markov chain Monte Carlo, and including 1 000 000 generations with sampling every 10 generations. The average standard deviation of split frequencies was used to assess the convergence of the two runs. In all cases, the chains converged within 0.25 generations. Therefore, the first 25 000 trees were discarded as burn-in and a 50% majority-rule consensus tree was calculated from the remaining 75 000 trees. Nodal support was reported as Bayesian posterior probabilities.

Symbiodinium ITS2 cloned sequences were identified by local BLASTn search against the clade C alignment available in the GeoSymbio database [24], as well as BLASTn search against NCBI. To avoid overestimating Symbiodinium diversity owing to the high intragenomic variability of the ITS2 gene [34,35], sequences included in the downstream analyses followed the same conservative criteria as used in our previous studies [8,43,58]. Statistical parsimony haplotype networks of Symbiodinium ITS2 rDNA sequences and Symbiodinium COI sequences were constructed using the software TCS v. 1.21 [59] with a 95% connection limit and gaps were treated as a fifth state.

3.4 Statistical analyses

Patterns of host–symbiont association across collection sites and depth gradients were tested statistically using the square-root of the relative frequency of Symbiodinium COI sequence genotypes present in each Leptoseris spp. sample using the Bray–Curtis coefficient of similarity (S) in the software package Primer v. 6 [60]. To test for the partitioning of Symbiodinium genotypes by host (i.e. Symbiodinium versus Leptoseris mtDNA genotypes), collection site (i.e. between the 31 collection sites), and collection depth (i.e. between depth ranges 65, 75, 85, 95, 100, 115 and 125 m), a permutational MANOVA [61–63] was performed with ‘host’, ‘site’ and ‘depth’ as fixed factors. The test was performed using Type 1 sums of squares and unrestricted permutation of raw data. Because the Symbiodinium ITS2 sequences were obtained from a relatively limited subset of coral samples (n=14 out of the 77 samples investigated), an independent permutational MANOVA analysis was performed to test for the partitioning of Symbiodinium ITS2 sequences by symbiont and host mtDNA genotypes and by depth only.

4. Results

4.1 Phylogenetic analyses

High-quality sequences of COX1–1-rRNA mtDNA were obtained for all investigated Leptoseris spp. samples (n=74). The sequence alignment was 818 bp in length. The model of evolution calculated in Treefinder v. 12.2.0 corresponded to the GTR+G+I model [64]. All Bayesian analyses yielded similar ‘burn-in’ curves. Standard deviation of split frequencies were well below 0.01 after ca 15 000 generations, and the Potential Scale Reduction Factor reached the value of 1 for all parameters. Phylogenetic reconstructions recovered five divergent and highly supported clades, each corresponding to known Leptoserisspecies previously described in [51] (figure 2). Additional phylogenetic analysis, including all sequences from Luck et al. [51] and a representative sequence from each clade reported here, indicated unambiguous correspondence for Leptoseris sp. 1 (clade Ia), Leptoseris tubulifera (here referred to as clade Ia’), Leptoseris hawaiiensis (clade Ib) and Leptoseris scabra (clade VII) (electronic supplementary material, appendix B). The remaining clade (clade II) was most similar by genetic distance measures to Leptoseris papyracea but sequences differed by up to 21 bp. Leptoseris scabra was the most divergent with respect to other Leptoseris species, and consistent with Luck et al.’s [51] finding that L. scabra was polyphyletic with Pavona and Agaricia; this species may require future generic reassignment.

Leptoseris scabra (clade VII) was exclusively represented by samples collected at upper and mid mesophotic zones, with approximately the same number of samples collected from 65 to 75 m (n=8) and from 85 to 100 m (n=6) depth ranges, respectively (figure 2). Among the more closely related Leptoserisspecies, L. tubulifera (clades Ia’) and Leptoseris sp. 1 (clade Ia) were found from upper and mid mesophotic similarly to L. scabra. Leptoseris sp. 1 was also detected once (sample no. L39) from the lower (115–125 m) depth range. Leptoseris papyracea (clade II) and L. hawaiiensis (clade Ib) were exclusively found at mid and deep water (115–125 m) depth ranges, respectively (figure 2). In Luck et al. [51], the water-depth ranges for these five species were 70–80 m (Leptoseris sp. 1), 20–85 m (L. tubulifera), 80–130 m (L. hawaiiensis), 40–70 m (L. papyracea) and 70–130 m (L. scabra).

High-quality sequences of Symbiodinium COI mtDNA sequences were obtained for all investigated Leptoseris spp. samples (n=74). Sequence alignment was 1057 bp in length. All COI sequences belonged to Symbiodinium clade C and were different from the previously published COI sequences produced in [41] for Symbiodinium C1 (4–6 bp differences), C15 (3–7 bp), C90 (13–14 bp) and C91 (14–17 bp) (data not shown). The model of evolution calculated in Treefinder v. 12.2.0 corresponded to the HKY model [65]. Phylogenetic reconstructions yielded three distinct and well-supported COI sequence haplotypes, with haplotypes COI-1 (n=22) and COI-3 (n=32) more closely related to one another than COI-2 (n=20) (electronic supplementary material, appendix C). The relationship and number of bp differences between the three Symbiodinium COIhaplotypes can be visualized in the statistical parsimony network of figure 3a. The COI haplotypes differed by between 3 and 7 bp. Identical COI Symbiodiniumhaplotypes were recovered from all 12 Leptoseris spp. samples that were subjected to additional COI genotyping from calyx and/or coenosarc coral biopsies (see the electronic supplementary material, appendix A).

A subset (n=14) of samples representing all three Symbiodinium COI genotypes and the five Leptoseris species (figure 2; electronic supplementary material, appendix A) was selected for cloning and sequencing of the Symbiodinium spp. ITS2 gene. A total of 140 ITS2 sequences were obtained, including between 8 and 12 cloned sequences per sample (average of 10 sequences per sample; see the electronic supplementary material, appendix A). Ten Symbiodinium spp. ITS2genotypes were recovered, including three previously published types (C1, C1c/C45 and C1v1b), and seven novel sequence variants (C1v1c, C1v1d, C1v1e, C1v3, C1v6, C1v8 and C1v18) that differed from Symbiodinium type C1 by 1–18 bp (figure 3b). These novel sequences were named ‘C1v’ followed by an alphanumeric descriptor following the naming system of Chan et al. [43]. Between two and six co-occurring ITS2 sequence types were recovered from individual coral samples, with type C1 common in all samples (electronic supplementary material, appendix A). The four most common SymbiodiniumITS2 sequence types were C1 (n=66), C1v8 (n=15), C1v18 (n=14) and C1c/C45 (n=12).