2.1 Experimental design and sampling procedures
We performed in situ experiments at a depth of 3 meters in Marine National Park of the Currais Islands (-25.7368°, -48.3664°) in southern Brazil between February 14 and 15, 2021. Through a hierarchical experimental design (Fig. 1), experiments were conducted with L. punicea hosts under high natural colonization of O. mirabilis (treatment), which are colonies visibly and factually more colonized by the brittle stars compared to the overall population, hosts naturally without brittle star individuals (host control), and without hosts or brittle star individuals (sampling artifact control).
The experiments were performed using in situ incubation chambers (Fig. 2), adapted from Ribes et al. (2000). The system can alternate between open circulation, exchange between the seawater outside and the chamber's interior, and closed circulation, only with the seawater of its own chamber. The innovation of our system is the possibility of subaquatic sampling of seawater at specific times through a set of valves that traps seawater in polyethylene bottles. The water that circulates in each 2,590 ml acrylic chamber comes from the same sampling point during open circulation to avoid composition stochasticity in the planktonic community. A set of tubes allows for the individual water flow in each chamber, acting as replicates. A 120 l.h-1 pump allows the turbulent flow of seawater and suspended particle maintenance within each chamber and in the bottles. A 6V 12Ah battery powers the nine circulation pumps.
Colonies of L. punicea with similar biomass were selected at a depth of approximately 6 meters. We collected colonies under high natural colonization of O. mirabilis and colonies naturally without individuals through SCUBA diving. In the study area, living brittle stars present coloration patterns that include yellow-orange, burgundy and yellow, and black with yellow details. Hybridization within different color patterns of individuals is also observed. L. punicea colony treatments that were dominated by yellow-orange O. mirabilis specimens were selected as treatments. The hosts were carefully collected with the substrate's surface layer and transplanted to 144 cm2 (12 × 12 cm) polyethylene supports. Then, we acclimatize under natural conditions on rocky shores at a depth of 6 meters for 24 hours to reduce the disturbance caused by removal from their original substrate. Healthy feeding colonies with open polyps were exposed to the experiment, conducted between 10:00 AM and 01:00 PM. Treatments, host controls and sampling artifact controls were randomly assigned to the chambers. To acclimatize the organisms after handling and under experimental conditions, we kept the system working for 15 minutes in open circulation with the organisms before starting the experiment.
The experiment lasted three hours, and 125 ml of seawater was sampled at the beginning (t0) and every hour (t1, t2, and t3) of the experiment to monitor the swept clearing of seston by octocoral suspension feeding. Three hours is considered more than enough to measure changes in the chambers' seston composition by passive suspension feeding (Ribes et al. 2000). Moreover, the drop in dissolved oxygen during these three hours does not significantly affect the respiration rates and organisms' behavior (Ribes et al. 2000). Seawater samples were taken on board, fixed with 8 ml of glutaraldehyde Grade I (25% aqueous solution), and then stored in thermal coolers. After landing, the seawater samples were stored in a refrigerator (2–4 °C) for a maximum period of 10 days until laboratory analysis. The organisms were also frozen until laboratory analysis.
2.2 Laboratory procedures for estimating variation in feeding rates
Numbers of O. mirabilis were counted within each octocoral colony using Zeiss Stemi 2000, ZEISS ©. The disk diameter of O. mirabilis was measured for 50 individuals within each of the three octocoral colonies using a stereomicroscope SteREO Discovery. V12, ZEISS © to assess the population structure. The degree of body regeneration of O. mirabilis was assessed for 50 individuals within each of the three octocoral colonies into three categories: intact specimens with six arms of similar size; split discs or broken arms; and regenerating discs or arms. The colonies of L. punicea were photographed using a Canon Rebel T6i camera, and the images were analyzed using photoQuad © software (Trygonis and Sini 2012) to measure the colonies' height, width (in cm) and the total surface area (in cm2). Octocoral polyp densities, expressed by the number of polyps per 1 cm2 of area, were obtained by counting three sampling areas of the basal, central, and distal segments of the stem of the three octocoral colonies using a stereomicroscope SteREO Discovery. V12, ZEISS ©. O. mirabilis individuals and L. punicea colonies were dried at 60 °C for 90 hours and weighed separately to determine the relationship between the number of individuals and the dry weight (DW, in grams) of O. mirabilis and the relationship between the area (cm2) and the DW of the host L. punicea.
The effect on feeding rates is the variable that can most directly assess the impacts of O. mirabilis on host feeding performance; therefore, we calculated the predation of particles between 3 and 120 µm in diameter. The water samples were previously filtered through 120 and 65 µm meshes. The number of particles was quantified with a Z Coulter Counter (Beckman Coulter © – USA) within the ranges of 3–8.99 μm, 9–19.99 μm, 20–59.99 μm and 60–120 μm. Cell biovolumes were estimated from the particle diameter, generalizing it as spherical, and then the carbon content was estimated. Using the conversion factors from the literature, we attributed particles < 19.99 µm to nanoeukaryotes, pg C cell-1 = 0.433 × (µm3)0.863 (Verity et al. 1992), and particles > 20.00 µm to phytoplankton, pg C cell-1 = 0.109 × (µm3)0.991 (Montagnes et al. 1994).
Feeding rates of the seston and heterotrophic carbon inputs by the octocoral L. punicea were calculated taking into account the exponential growth of the plankton during the experiment (Frost 1972; Saiz 1993; Ribes et al. 1998), an approach that is widely used in feeding studies of benthic suspension feeders (Ribes et al. 2003; Picciano and Ferrier-Pagès 2007; Coppari et al. 2016). The prey growth rate k (h-1) was calculated as:
where Cb and Ca are the prey and carbon concentrations in the chambers (particles mL-1 and estimated carbon content mL-1) at previous time ta to the consecutive time tb. The clearance rate CR (particles swept clear DW-1 h-1 and estimated carbon content swept clear DW-1 h-1) was calculated as:
where V is the volume of the chamber (in mL), b is the biomass of the octocoral colony (DW, in grams) and g is the grazing coefficient (h-1), calculated as:
where kc is the prey growth rate in the sampling artifact control, and kg is the apparent growth in the chambers with animals. In cases where the values of kc were higher than kg, we arbitrarily disregarded the values in the statistical analysis. Then, we calculated the feeding rate and heterotrophic carbon input by:
where C is the mean prey concentration of all chambers at the initial time. Finally, the feeding rate was expressed in terms of the number of particles consumed per unit of dry weight biomass per unit of time (i.e. particles g DW-1 h-1), and the heterotrophic carbon inputs were expressed in terms of the mass of organic carbon consumed per unit of dry weight biomass per unit of time (i.e. µg C g DW-1 h-1).
2.3 Hard-bottom system effects
Data from Derviche et al 2021 on the distribution and abundance pattern of O. mirabilis and L. punicea in the National Marine Park of Currais Islands were used to estimate the grazing rate of the octocoral population and the putative effects of the brittle star. In the study area, the density of L. punicea was 8.33 ± 10.41 colonies m-2 (mean ± standard deviation), and the density of colonization by O. mirabilis on the octocoral species was 3.64 ± 2.97 inds. cm-2. The morphological characteristics of the octocorals used in the present experiment were established as models for the overall population. The population grazing rate was then estimated taking into account this population structure and heterotrophic carbon inputs. The grazing rate was expressed in terms of the mass of organic carbon consumed per square meter per unit of time (i.e. mg C m-2 d-1).
2.4 Data analysis
We investigated the relationship between octocoral feeding performance and high-density colonization of brittle stars (Dalgaard 2008). Assumptions of normality and homogeneity of variance were checked previously using Shapiro-Wilk’s test and Levene’s test, respectively. The heterotrophic carbon input did not assume normality and homogeneity of variance assumptions. So, we applied the transformation 1/(x) to achieve it. Although Shapiro-Wilk’s test still indicated that the data were close to the non-normality distribution (p value = 0.05), we chose to proceed with this transformation to not use non-parametric analysis. Other transformations (e.g., sqrt(x); log10(x)) did not assume normality or homogeneity of variance assumptions. In the case of swept clearing of the seston, that is, the difference between the particle concentration of the sample (particle. mL-1) at consecutive times, we assigned together the treatments and the host controls to compare the differences to the chambers without animals, i.e., the sampling artifact control. The swept clearing of the seston also did not assume normality and homogeneity of variance. Therefore, we applied the transformation sqrt(max(x+1) - x) to achieve it. Significant differences in the dependent variables feeding rates, heterotrophic carbon inputs and swept clearing of the seston were separately tested for groups (two levels, factor) and time (three levels, factor) using two-way ANOVA. Differences were further identified with Tukey’s post hoc test. In all cases, we compared the models using the Akaike information criterion (AIC). All statistical analyses and graphs were produced in the computational language R (R Core Team 2020).