Introduction

The fate of coral reefs has turned into a worldwide concern in the last decades. Yet our current understanding of coral reefs is mainly based on the shallowest quarter of its overall depth range. This overlooks mesophotic coral ecosystems (MCEs), which occur beyond the limits of regular diving in the so-called twilight zone (> 30 m) (Pyle 1998). Ecological conditions vary over the shallow-mesophotic depth range, which define the distribution of coral communities (Slattery and Lesser 2012). Coral symbiotic associations drive how corals acclimatize to a broader depth range under a climate change scenario. Here, we studied a symbiotic component (endolithic algae) in a coral extending well beyond the lower mesophotic zone.

Symbioses support the framework and resilience of coral reefs. The roles of Symbiodinium and bacteria in reef-building species suggest an interdependent holobiont, and still other associations with plantlike forms remain neglected. Endolithic algae, for example, play a dynamic role within the skeleton of scleractinian corals, potentially involved in metabolite translocation (Schlichter et al. 1995), survival during bleaching events (Fine and Loya 2002) and light capture in low-light environments (Koehne et al. 1999; Fine and Loya 2002; Verbruggen and Tribollet 2011). Ostreobium-like algae involve physiological traits, which allow corals to thrive at different depths (Gutner-Hoch and Fine 2011). However, the diversity of these endolithic algae in mesophotic coral ecosystems stays largely unnoticed. Here, we provide the first study in the Caribbean of the endolithic algal community in a reef-building coral, Agaricia undata (Ellis and Solander 1786), down to the lower mesophotic zone (90 m) extending beyond the compensatory point between photosynthesis and respiration (around 77 m in tropical waters), where zooxanthellate corals face a net loss in productivity (Lesser et al. 2009).

Materials and methods

Sample collection

To estimate the prevalence and diversity of endolithic algal associations in A. undata, 103 colony fragments were collected between 2015 and 2016 in three Colombian locations: San Andres island (N12°32′34.112″ W081°42′41.079″) in April and June 2015 and March 2016, Cartagena (N10°15′55.944″ W075°37′32.915″) in July and November 2015 and Natural National Park ‘Corales de Profundidad’ (N10°06′29.3″ W076°0′33.998″) in December 2015, the first location with oceanic influence and the last two, about 500 km apart, with continental siliciclastic influence due to river runoff. The oceanic and siliciclastic conditions have notable differences in light penetration but similar temperature regimes without strong seasonal changes (Velásquez and Sánchez 2015). Depths of sample collection ranged between 17 and 83 m and were obtained using SCUBA or Closed Circuit Rebreather with Trimix below 40 m. Dives in the mesophotic zone (planned to 100 m, one per day) allowed us only 20–30 min of bottom time (Trimix 10/60, oxygen/helium) and about 2.5 h of decompression time. The samples were grouped in three categories: shallow (≤ 30 m), upper (> 30 and < 60 m) and lower (≥ 60 m) zones (see details in Electronic Supplementary Material, ESM Table S1). Although sampling numbers seem low, in comparison with shallow-water projects, here we included the entire depth range of the species at each locality and we restricted sampling to colonies at least 5 m apart, which reduced encountering clones. A coral subsample (~ 2 × 2 cm) was collected for each colony; a small fragment (0.5 × 0.5 cm2) was preserved in individual tubes with DMSO salt buffer (0.5 M EDTA; 20% (v/v) DMSO, saturated with NaCl) and stored at − 80 °C until DNA extraction. A dry voucher is available at the Museo de Historia Natural ANDES (ANDES-IM 4681–4784). Research and collection of specimens were approved by the National Environmental Licensing Authority (ANLA, Spanish acronym): Collection Framework Agreement granted to Universidad de los Andes through resolution 1177 of October 9, 2014-IBD 0359.

Scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDS)

Using light microscopy, we determined the presence by visual estimation of the endolithic algal in the coral dry fragment, filaments connecting the costae (henceforth refer as a ‘pipeline’), and calcified dead filaments were identified. A motor tool was used to cut fragments of around 1 cm2. Each fragment was fixed in 2.5% glutaraldehyde and 0.5% formaldehyde made in 0.1 M phosphate buffer at pH 7.2 for 36 h (Golubic et al. 2005). After fixation, samples were washed with distilled water and dehydrated with an ethanol series of 50, 70, 90 and 96%, 1 h each ethanol concentration. A total of six samples with high concentration of ‘pipelines’ and dead filaments were fixed for the EDS analysis. The samples were dried with critical dry point and coated with gold before SEM, and then, EDS was executed. Coral bridge basements (coral skeleton) were used as controls, to compare coral, dead filaments and ‘pipeline’ composition.

DNA extraction, amplification and analysis

DNA was extracted using the CTAB protocol (Coffroth et al. 1992). For PCR amplification, we used the following primers: rbcLF (′5-ATHGARCCNGTNGTNGGNGAR-3′) and rbcLR (′5-CNGTYTCNGCYTGNGCYTTRTA-3′). The chloroplast gene ribulose 1,5-bisphosphate carboxylase large subunit (rbcL) has been used successfully to determine the genotypic diversity of Ostreobium within scleractinian corals (Gutner-Hoch and Fine 2011). PCR was performed in a 15 μl reaction containing 1× PCR buffer, 3 mM MgCl2, 1 μl of 10× bovine serum albumin (BSA 20 mg/ul), 0.16 mM of each dNTP, 0.52 mM of each primer, 1 U Taq DNA polymerase and 1 μl of DNA (30 ng). PCR conditions were 94 °C for 2 min followed by 35 cycles of 94 °C for 90 s, 52 °C for 90 s and 72 °C for 60 s and a 5-min extension at 72 °C. Resulting PCR products were cleaned with FastAP Thermosensitive Alkaline Phosphatase (Thermo Scientific™) and sequenced using the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems) on the AB13730xl DNA Analyzer (Applied Biosystems). GenBank Accession No. MF135622-MF135643.

Sequence alignment analysis was performed using the AliView software (Larsson 2014). Evolutionary model was assessed using Jmodeltest 2.1.2 (Guindon and Gascuel 2003; Darriba et al. 2012) which resulted in HKY + I + gamma according to the BIC calculations. We reconstructed phylogenetic relationships with two methods: Bayesian inference with BEAST v1.8.2 (Drummond et al. 2012) and MrBayes v3.2.2 (Ronquist et al. 2012) and maximum likelihood trees were estimated using RaxML (Stamatakis 2006). Both methods resulted in the same tree topology; thus, only the Bayesian tree is reported here. Three methods were used to identify phylospecies or clades: generalized mixed Yule coalescent model in our Bayesian ultrametric trees using the R package bGMYC (Reid and Carstens 2012), coalescent Poisson tree process (PTP) method (Zhang et al. 2013) and pairwise genetic distance using MEGA6 (Tamura et al. 2013). bGMYC analysis was based on 100 trees from the post-burn-in generations obtained from the BEAST analysis. For each of the 100 trees, the Markov chain Monte Carlo sampler was run for 50,000 generations, discarding the first 40,000 generations as burn-in and sampling every 100 generations. The upper threshold, defined as number of tips on the tree, was set to 48, and a cutoff value was set as 0.5 to delimit the lineages. For PTP method (Zhang et al. 2013), we used the web interface http://species.h-its.org with Bayesian tree as input, 100,000 MCMC generations, thinning every 100 generations and a burning fraction of 0.1. Endolithic algal clades were named following and continuing the nomenclature used by (Gutner-Hoch and Fine 2011) in which each clade is represented by a letter of the alphabet followed by numbers for the internal grouping. rbcL sequences of clade A-G (JF801728–JF801734) reported by Gutner-Hoch and Fine (2011) in the Red Sea were included for the reconstruction of the phylogenetic relationships. Caulerpa webbiana (AB054017) were included as an outgroup for the Bryopsidales order.

Results and discussion

Endolithic algae were macroscopically observed in 58% (n = 103) of A. undata sampled (Fig. 1a), but of those, only 28 samples amplified the rcbL gene (despite multiple attempts). Microscopic assessment of the endolithic interface demonstrated a close association clearly reaching the coral surface. Thin sections revealed that these algae have a network pattern that grows parallel to the coral surface (Fig. 1b). Some exposed algal filaments emerged as a tubular pipeline bridging the network through coral costae (Fig. 1c–f). Although we did not register colony size, largest colonies had endolithic algae reaching the coral surface in centrally located tissues.

Fig. 1
figure 1

Morphological observations of the filaments of the algae by light microscopy and scanning electron microscope (SEM). a Detail of the endolithic algae showing dense coverage in the base of colony, b view of a longitudinal section of coral cut, c coral—Ostreobium tubular pipeline between costae, d, e higher magnification of the filament pipeline bridging two costae, f clearly shrunk filament pipeline, after the EDS electron beam

Full size image

Energy-dispersive X-ray spectroscopy (EDS) identified the elemental composition of the filaments, which composition differed between the ‘pipeline’ and the coral tissue (Supplementary Fig. S1). Coral skeleton involved carbon (C), calcium (Ca) and oxygen (O), but the filament pipeline, bridging costae, had additional compounds like sulfur (S) and potassium (K). Sulfur could be associated with small cysteine-rich proteins (SCRiPs) in the membrane of calcifying organisms (Sunagawa et al. 2009). Ostreobium spp. could be using the CaCO3 waste of the boring process to make these pipeline-like structures, across A. undata costae, to improve the light uptake in the mesophotic zone. All tubular structures nearly shrunk after the electron beam of the EDS sampled (Fig. 1f).

Phylogenetic analyses of the chloroplast rbcL region (395 bp) from endolithic algae revealed a community consisting of 12 phylospecies/clades (Fig. 2a). Multiple endolithic algae types were found in association with A. undata. Some clades were geographically structured, whereas others distributed worldwide, (H2 and E) previously reported by Gutner-Hoch and Fine (2011), or showed little ecological structure (type I). In Cartagena, we found eight Ostreobium-like algae; five of these (D2, D3, J, K and O) were exclusive to the upper zone and one to shallow water (D1). In San Andrés, two (E and N) out of four OTUs were limited to the lower zone and one to the upper zone (M) (Fig. 2b). Ostreobium DNA was also found in the pipeline-like structure, which corresponded to clades N and H2.

Fig. 2
figure 2

Sampling locations, genetic diversity and relative abundance of endolithic boring algae in Agaricia undata across depths and locations. a Map showing sampled populations in the Caribbean Sea off the coast of Colombia. (1) San Andrés, (2) Cartagena and (3) Natural National Park ‘Corales de Profundidad.’ b Bayesian MCMC phylogenetic tree of Ostreobium-like algae OTUs based on chloroplast gene rbcL and clades previously reported by Gutner-Hoch E, Fine M (2011). Colored bars to the right of the phylogeny represent OTUs grouping based on genetic distance thresholds (3%), bGMYC and bPTP with their corresponding clade nomenclature, where gray bars represent clades of endolithic algae found in other species different to A. undata. Bootstrap values are based on Bayesian analyses, with only probabilities over 80% shown. c Pie charts summarizing diversity and distribution of algae OTUs across depth range per locality, which colors correspond to the clades defined in a

Full size image

The notion that the green band found in the skeleton of corals belongs to one cosmopolitan species (Ostreobium quekettii) is inaccurate. As in Gutner-Hoch and Fine (2011) and Sauvage et al. (2016), we found multiple cryptic species within the monophyletic group of Ostreobium-like algae. Ostreobidineae diversity in A. undata (12 clades) was higher than previously observed, while previous work found seven clades in two species (Gutner-Hoch and Fine 2011). Except for clade I that was found in all localities and along a wide range (17–80 m), Ostreobium spp. was partitioned by the reef setting, i.e., oceanic or continental siliciclastic influence, which is expected in response to low-light, organic matter and/or macroscopic predators (Sauvage et al. 2016), which are different at the two types of Caribbean reefs (Velásquez and Sánchez 2015). In a parallel study using the same samples, we found that both Symbiodinium and Ostreobium had similar and consistent regional variations in A. undata populations, suggesting an analog host–symbiont specificity. Remarkably, there were more types of Ostreobium associated with A. undata than to Symbiodinium (types: 12 vs. 7; n: 28 vs. 77) despite the fact that more samples were screened for zooxanthellae (Gonzalez-Zapata et al. 2018).

Ostreobium-like algae impose a paradox for coral reef science, where the most pervasive microbioeroder of coral skeletons (Tribollet 2008) as it reaches the coral tissue apparently coexists with the coral colony. These algae infest corals at different depths but are particularly shade-adapted (Ralph et al. 2007). The previous research has shown that Ostreobium spp. under extreme low light, as in MCEs, can absorb red and far-red wavelengths and change the chl b:chl a ratio according to depth (Shashar and Stambler 1992; Magnusson et al. 2007). These algae use coral metabolic waste and can translocate fixed carbon to the coral host during stress events or at extreme depths, which indicates commensalistic to mutualistic interactions (Fine and Loya 2002). This should be tested experimentally, but the high diversity and differences found over depth could be a key factor for A. undata to colonize mesophotic ecosystems. Since not all colonies presented this association, further testing is needed to determine whether there is a successional process occurring, as Ostreobium was most often noted in older colonies of A. undata.