Understanding trophic relationships among Caribbean sea urchins

Relaciones tróficas entre los erizos de mar Caribe

Study site and data collection: Surveys were conducted in October of 2013 at four reef crests of the Puerto Rico Archipelago. Two sites were located on the Northeastern coast of Puerto Rico, with the other two located off the island of Culebra (Fig. 1). Sites were Cerro Gordo (CGD: 18°16’51.40” N - 65°17’12.21” W), Luquillo (LQY: 18°23’18.46” N - 65°43’5.52” W), Tamarindo Bay (TMD: 18°18’55.29” N - 65°19’5.82” W), and Melones (MLN: 18°18’15.70” N - 65°18’42.27” W). All sites are shallow-water fringing reefs between 1-2m in depth. Culebra Island has no rivers, precipitation average is low, and volcanic and intrusive rocks, with some limestone deposits (U.S.G.S., 1996), characterize the substrate. The sites CGD and LQY are composed of carbonate rocks (Rodríguez-Barreras, Pérez, Williams, Mercado-Molina, &amp Sabat, 2014).

We collected five adults of the species Diadema antillarum, Echinometra viridis, Lytechinus variegatus, and Tripneustes ventricosus from LQY. Additionally, five individuals of Echinometra lucunter were collected from all four localities (Fig. 1). The genus Echinometra and D. antillarum were collected in hardground biotopes (fringing reef), whereas L. variegatus and T. ventricosus were gathered in a backreef lagoon biotope covered by seagrass meadows. Sample sizes were based on previous studies (Rodríguez, 2003; Tomas et al., 2006; Wing et al., 2008; Cabanillas-Terán, 2009). Thus, we collected samples of all seaweed species within a 5m radius from collecting sites associated only with E. lucunter. All organisms, including algae and sea urchins, were preserved in a dry-ice container and transported to the laboratory for later analyses. We identified twenty algae up to the lowest possible taxonomic level (Littler &amp Littler, 2000). The species identified were Amphiroa fragilissima, Acanthophora spicifera, Bryothamnion triquetrum, Caulerpa racemosa, Ceramium nitens, Chaetomorpha crassa, Cladophora vagabunda, Dasya antillarum, Dictyota caribaea, Dictyota menstrualis, Gelidiella acerosa, Dichotomaria obtusata, Galaxaura rugosa, Gracilaria spp., Halimeda scabra, Hypnea valentiae, Jania adhaerens, Padina boergesenii, Palisada perforata, and Tricleocarpa fragilis.

Sample analysis: Carbon and nitrogen were measured from the muscle of the Aristotle’s lantern (echinoids), and from bulk tissues (algae). Once defrosted, calcified algae samples were washed twice with diluted solutions of HCl at 12 % and 10 % respectively, to remove any excess carbonate (Bunn, Loneragan, &amp Kempster, 1995). Samples were rinsed with distilled water and dried in an air circulating Shel Lab Oven at 60 ^oC for 48 hours. After drying, samples were ground to a fine powder using a frequency grinder. Because the acid-wash procedure alters δ¹⁵N results, calcified algae required two samples each: one acid washed for δ¹³C and one without acid-wash for δ¹⁵N. Samples of 0.9-1.3 mg for sea urchin tissue, and 3.9-4.4 mg for algae samples were stored in tin capsules. Measurements of stable isotope ratios were obtained in a continuous flow isotope ratio mass spectrometer at the Laboratory of Stable Isotope Ecology in Tropical Ecosystems, at the University of Miami. Results for δ¹³C and δ¹⁵N were expressed as parts per thousand (%) differences from the standard (δ): Vienna Pee Dee Belemnite (vPDB) for δ¹³C, and atmospheric N₂ for δ¹⁵N. Carbon and nitrogen samples were analyzed in a dual isotope mode. Ratios were then used to complete equation 1, where R_sample and R_standard are the values of ¹³C/¹²C or ¹⁵N/¹⁴N in the sample and the standard, respectively.

For E. lucunter only, we run a Bayesian mixing model and SIBER using SIAR 4.2 (Stable Isotopes Analysis in R) package in the repository Comprehensive R Archive Network (Parnell &amp Jackson, 2013). This package uses organism isotopic data to fit a model to their dietary habits based upon a Gaussian likelihood and a mixture Dirichlet-distributed prior on the mean. Models were run with 100 000 iterations and a burn in of 50 000. This routine takes bivariate isotope data in x and y, and returns the posterior draws for the metric SEA.B (Standard Ellipse Area. Bayesian) for each group. We selected eight algae at every site, except in LQY, where we used seven species, and the percent of contribution of each algae species to the measured isotopic signatures was determined. We run SIBER (Stable Isotope Bayesian Ellipses) to calculate ellipse based metrics of niche width, and determined the existence of any overlap among co-occurring species. SIBER was run twice, once for the hardground biotope (D. antillarum, E. lucunter, and E. viridis), and again for the seagrass biotope (T. ventricosus, and L. variegatus). We considered an isotopic fractionation across trophic levels of 3.4 ‰ for δ¹⁵N and 1.0 ‰ for δ¹³C (vander-Zanden &amp Rasmussen, 2001; Post, 2002; Wing et al., 2008).

Normality and homogeneity of variance were tested by running the KolmogorovSmirnov and Cochran tests, respectively (Zar, 2010). Nitrogen data followed the premises of parametric analysis, but the carbon data required a square-root transformation (Box &amp Cox, 1964). We ran a one-way ANOvA for comparisons of δ¹⁵N and δ¹³C among sea urchins, and a posteriori Tukey-HSD test was used for multiple comparisons of the means. In addition, we ran a One-way ANOvA to analyze spatial differences among isotopic signals of carbon and nitrogen on E. lucunter and the seaweed communities. Statistics were performed with α < 0.05 in R-2.9.13 (R Development Core Team, 2014).

The δ¹⁵N signal for the five studied echinoids of hardground biotopes varied from 5.77 ± 0.25 &amp#091;mean ± SD&amp#093; to 8.22 ± 0.36 ‰; whereas the species collected on seagrass biotopes showed values of nitrogen from 6.09 ± 0.79 to 8.02 ± 0.36 ‰ (Fig. 2A). The species T. ventricosus and D. antillarum showed the highest values of δ¹⁵N. Sea urchins exhibited similar values of δ¹³C varying from -11.6 ± 0.63 to -10.4 ± 0.99 ‰. The echinoid E. lucunter displayed the lowest values of carbon, from -15.40 ± 0.76 ‰ (Fig. 2B). Significant differences among species were found for δ¹⁵N and δ¹³C (Table 1). A posteriori analysis displayed differences in δ¹⁵N between the group formed by E. lucunter, E. viridis and L. variegatus and the upper group formed by D. antillarum and T. ventricosus (p< 0.001, Fig 2A). Whereas differences in terms of δ¹³C were found in E. lucunter with respect to the other four species (p< 0.001, Fig 2B). We found no differences in δ¹⁵N signal of E. lucunter among sites, but we did find spatial differences in δ¹³C (F= 7.410, df= 3, p= 0.001). Those spatial differences were between the sites CGD and TMD (p= 0.001), and between CGD and LQY (p= 0.020).

The overall isotopic signal of δ¹⁵N for seaweed communities associated with hardground biotopes varied from 1.7 ± 0.56 (MLN) to 3.3 ± 1.24 ‰ (CGD), whereas the δ¹³C varied from -15.2 ± 2.51 (LQY) to -13.5 ± 3.58 ‰ (CGD). The lowest nitrogen isotopic signals were obtained in TMD (D. menstrualis: 1.1 ‰) and MLN (G. rugosa: 1.1 ‰); while the highest value was recorded in CGD (B. triquetrum: 4.8 ‰). Both maximum (J. adhaerens: -8.1 ‰) and minimum (D. menstrualis: -8.7 ‰) δ¹³C values were recorded in CGD (Fig. 3). No differences among sites were detected for overall δ¹⁵N (F= 1.300, df= 3, p= 0.301), but we found spatial differences for δ¹³C (F= 7.410, df= 3, p= 0.001). A posteriori analysis displayed differences between CGD with respect to the other three sites (p< 0.05).

The ellipse-based metrics of niche width analysis -restricted to one locality- found that the hardground biotope species (D. antillarum, E. lucunter, and E. viridis) did not overlap each other (Fig 4A). Similar results were obtained for the co-occurring species of the seagrass biotope; however, the distance between these species was closer than that of the hardground biotope species (Fig 4B). The Bayesian mixing models run for E. lucunter at all four localities found differences in food resources contribution. For instance, the algae D. menstrualis, C. crassa and B. triquetrum dominated in CGD; whereas C. nitens, Gracilaria spp., and D. caribaea represented the main algae species to the diet of E. lucunter at LQY (Table 2). In Culebra Island, no dominance of any particular algae was detected in TMD, where six of the eight species exhibited a similar contribution (Table 2). Although MLN displayed a similar low dominance of species as that of TMD, two algae remained with a mean contribution over 15 %: Dichotomaria obtusata and Dasya antillarum.

There is often a high degree of spatial variability in shallow-water ecosystems. Both nitrogen and carbon elemental concentrations may vary due to abiotic and biotic factors, including atmospheric deposition, volatilization, microbial processing, terrestrial runoff, etc. (van-Dover et al., 1992; Nixon et al., 1996). Anthropogenic enrichment may alter marine benthic food webs, increasing the ratios of heavy isotopes in primary producers (McGlathery, 2001; Tewfik, Rasmussen, &amp McCann, 2005). In fact, the presence of elevated values of δ¹⁵N has been shown to be related with local NO₃ enrichment (Clark &amp Wilcock, 2000; Cloern, 2001). In general, our seaweed communities exhibited no spatial differences in nitrogen, which indicates a relative homogeneity in δ¹⁵N among sites. This is probably due to their geographical proximity. In contrast, spatial differences among seaweed communities in terms of δ¹³C might be related to community composition, metabolic species-specific characteristics, or to local geomorphologic features. The CGD site exhibited differences with respect to the other three sites. This difference may be linked to the existence of a mangrove lagoon canal located approximately 200 m away. This is important because mangrove systems provide nutrient inputs, mainly carbon, to nearshore waters (Zieman, Macko, &amp Mills, 1984; Rezende, Lacerda, Ovall, Silva &amp Martinelli, 1990).

Food resources leave tracks within consumers throughout food web (Phillips, 2012), and usually one or two sources may contribute 50 % or more of the diet of a given consumer (Rodríguez, 2003; Tomas et al., 2006). Mixing models indicated spatial differences among sites in the composition of algal food assimilated by the sea urchin E. lucunter. No single algal species exhibited high dominance in the diet of this sea urchin in TMD and MLN. However, some algae reported mean dietary contribution greater than 20 % in CGD and LQY, but never as high as 30 or 40 % for a single species. These results contrast with other studies where some algae species may compromise up to 60 % of urchin diets (Rodríguez, 2003; Prado et al., 2010; Cabanillas-Terán et al., 2016). Although contrasting, trophic plasticity is widespread in nature, and may explain the apparent difference in urchin diets (Pinnegar &amp Polunin, 2000; Wing et al., 2008).

The existence of spatial differences reflects the feeding plasticity of E. lucunter and its ability to use a wide range of food resources. The fact that the Culebra sites showed a lower algal contribution in sea urchin diets might be related to local seaweed assemblages than to feeding preferences. A generalist feeding behavior is very common in sea urchins, and E. lucunter is not an exception (Hendler et al., 1995; Lawrence et al., 2013). The lack of data on seaweed communities did not allow us to quantify algal abundances or establish a relationship between seaweed abundance and its magnitude of contribution to the diet of E. lucunter. We do not suspect that the reason why the sea urchin diet is distributed among so many algae types is due simply to the low species number (8) included in this two-isotope mixing model. Previous studies have stated that those resources, whose isotopic signals most resemble those of the sea urchin, are in general the most important contributors to the diet of a given species (Szepanski, Ben-David, &amp van Ballenberghe, 1999; Ben-David &amp Shell, 2001). There is a possibility that E. lucunter may be selectively feeding on a few algal species, and different combinations could produce the same isotopic signal; nevertheless, further studies will need to focus on isotopic markers to prove this diversity in the diet of E. lucunter.

Co-occurring species coexist due to differences in feeding behavior and internal assimilation mechanisms (vanderklift et al., 2006; Wolf et al., 2009). Differences in δ¹³C detected among the five echinoids of two different biotopes may suggest that E. lucunter uses different sources of carbon with respect to the other two species of the hardground biotope (Fig. 4A), whereas D. antillarum and E. viridis coincided in terms of δ¹³C. However, the lack of overlap between both species suggests the use of different sources of nitrogen and a higher trophic position for D. antillarum, which may be more related to an omnivorous strategy. A similar case of niche partitioning was discussed for three sympatric sea urchins in Australia (vanderklift et al., 2006).

The other two co-occurring species within seagrass bed habitats also exhibited no niche overlap (Fig 4B). However, both echinoids coincided in terms of δ¹³C, which may indicate the use of similar food resources. A higher trophic position of T. ventricosus, however, could indicate the use of other food items, δ¹⁵N-enriched. These two species are common dwellers of seagrass meadows. There, T. ventricosus usually grazes on Thalassia testudinum leaves, but also ingests small invertebrates; whereas L. variegatus feeds on T. testudinum, detritus, and some epiphytes (Keller, 1983; Barrios &amp Reyes, 2009). The isotopic signal of δ¹⁵N for T. ventricosus was also in consonance with its omnivorous behavior, and relatively far from the isotopic signature of T. testudinum leaves (unpublished data). A stomach content analysis on T. ventricosus found bryozoans, ascidians, and hydrozoans (Barrios &amp Reyes, 2009). Thus, δ¹⁵N in T. ventricosus was similar with the isotopic signal of the omnivorous sea urchins Centrostephanus tenuispinus, Phyllacanthus irregularis, Strongylocentrotus pallidus and Arbacia lixula (vanderklift et al., 2006; Wangensteen et al., 2011). Nevertheless, the Mediterranean echinoid Paracentrotus lividus, considered an important grazer of the vascular plant, Posidonia oceanica, reported values of nitrogen between 9-10 %. This particularly high value of δ¹⁵N, which could be considered too high for a grazer, was due to the existence of an epiphyte community growing on Posidonia leaves that covers part of the carbon and nitrogen requirements in the diet of the echinoid P. lividus (Tomas et al., 2006). Therefore, the ingestion of small invertebrates by T. ventricosus may be not an accidental intake, but an important source of nitrogen that gives this sea urchin a higher trophic position in seagrass biotopes.

The high δ¹⁵N values found in D. antillarum contrasts with its feeding preferences (Rodríguez-Barreras et al., 2015). The sea urchin D. antillarum has considered primarily a generalist herbivore that feeds principally upon algae (Hendler et al., 1995). However, this sea urchin also ingests invertebrates (Serafy, 1979). Stomach content analyses have revealed the presence of sponges, hydroids, bryozoans, nematodes, rotifers, gastropods, bivalves, and copepods. These invertebrates may represent up to 32 % of the gut species-richness index (Herrera-López et al, 2004; Hernández et al., 2006). Indeed, some authors proposed Diadema as an omnivore (Rotjand &amp Lewis, 2008), but the idea has met resistance. The presence of invertebrates within the Diadema diet has considered a non-selective behavior, due only to accidental scraping off, but never an important component of the sea urchin diet. The high value of δ¹⁵N displayed for D. antillarum, similar to T. ventricosus, suggests that both sea urchins occupy similar trophic levels in their respective habitats, and support an omnivorous feeding behavior for D. antillarum. However, this tentative statement requires further studies to determine the main food resources, including invertebrates.

To conclude, a Bayesian mixing model indicates feeding plasticity of E. lucunter, using multiple food algal resources depending on local availability. Similarities in δ¹⁵N between D. antillarum and T. ventricosus may hint at a similar trophic level for both species (omnivorous) which is widely accepted for T. ventricosus, but not for D. antillarum, often considered a generalist herbivore species. Thus, the lack of overlap among species in the two biotopes seems to indicate a resource partitioning strategy to avoid niche competition among co-occurring species.

Special thanks to the Center for Applied Tropical Ecology and Conservation (CATEC) for their economic and logistic support. The first author also wants to thank to Donald Phillips, Catalina Dávila, Alain Durán, and Owen Wangensteen for their valuable information and comments to improve this manuscript.