Reduced heterotrophy in the stony coral Galaxea fascicularis after life-long exposure to elevated carbon dioxide

Ocean acidification imposes many physiological, energetic, structural and ecological challenges to stony corals. While some corals may increase autotrophy under ocean acidification, another potential mechanism to alleviate some of the adverse effects on their physiology is to increase heterotrophy. We compared the feeding rates of Galaxea fascicularis colonies that have lived their entire lives under ocean acidification conditions at natural carbon dioxide (CO2) seeps with colonies living under present-day CO2 conditions. When provided with the same quantity and composition of zooplankton as food, corals acclimatized to high CO2 showed 2.8 to 4.8 times depressed rates of zooplankton feeding. Results were consistent over four experiments, from two expeditions and both in field and chamber measurements. Unless replenished by other sources, reduced zooplankton uptake in G. fascicularis acclimatized to ocean acidification is likely to entail a shortage of vital nutrients, potentially jeopardizing their health and survival in future oceans.

between expeditions ( Table 1). The interaction between method and expedition had a significant influence on the total number of zooplankton consumed, although there was no difference for the main effect variables of method and expedition (Table 1).
Following the observation of reduced feeding rates during the first expedition, in the second expedition we assessed whether the reduced heterotrophy was caused by CO 2 -induced impairment of neurotransmitters. The addition of gabazine during the chamber experiment from expedition 2 had no significant impact on the feeding rates (one-way ANOVA: F (2,22) = 0.51; P = 0.48). Thus, heterotrophy rates under high CO 2 were not restored by the treatment with gabazine, the GABA A receptor antagonist.
Composition of consumed food and selective feeding. Although the total number of zooplankton consumed was different between CO 2 levels, the types of zooplankton consumed by G. fascicularis were not different between CO 2 levels. Taxonomic richness of the zooplankton prey consumed was not different between CO 2 levels (three-way ANOVA: F (1,13) = 2.74; P = 0.10), although it was higher in the chamber experiments compared to the field experiments (F (1,15) = 20.2; P < 0.001), and higher in expedition 2 compared to expedition 1 (F (1,15) = 8.17; P = 0.006). Multivariate community analyses on the prey consumed by corals supported these results and indicated that the zooplankton community consumed was also not different between CO 2 levels (three-way ANOVA: F (1,56)  The types of prey identified in the coelenteron of dissected corals had much lower taxonomic richness than the plankton available in the water column: corals contained only 11-17 zooplankton taxa of the 26-33 taxa present in the water. Corals preferentially ingested some zooplankton taxa, including Pontellidae and Paracalanidae copepods, decapods, amphipods, and chaetognaths, whereas Oithonidae copepods that are abundant in the water column were scarce in the food consumed (Fig. 3).
Results from logistic regressions that examined the effects of elevated CO 2 , expedition, and method, on the probability that each zooplankton taxon may be consumed indicated slight variation in the rates of consumption of the various taxa in response to these factors (Table 2). There was no difference in selectivity between high CO 2 and control corals for the most available and most frequently consumed zooplankton taxa. However, the rare Acartidae copepodites, Harpacticoida, Isopoda, Ostracoda, and Polychaeta appeared preferentially consumed at the control CO 2 level. These taxa all represent a small proportion of the plankton available and consumed (<2%).   Furthermore, consumption rates of several zooplankton taxa differed between expeditions and methods. For example, Tortanidae copepods were rarely consumed during the first expedition, and yet during the second expedition they constituted on average 30.2% of the coral diet in the field experiment and 22.6% in the chamber experiment. Similarly, uptake rates of decapods and chaetognaths were relatively high during the second expedition.
Corallite size and polyp expansion between CO 2 levels. No difference was observed in the size of G. fascicularis corallites between colonies originating at the seep and control sites (1-way ANOVA: F (1,62) = 2.7, P = 0.11). Elevated CO 2 also had no effect on polyp expansion of G. fascicularis at the seep and control sites, neither in the field nor in the chamber experiments. While coral polyps were not expanded more under elevated CO 2 compared to control CO 2 levels (4-way ANOVA: F (1,124) = 1.1; P = 0.29), they were expanded significantly more in the field compared to the chamber experiments (F (1,126) = 22.0; P < 0.001), and in expedition 2 compared to expedition 1 (F (1,125) = 12.2; P < 0.001; see supplementary Table S1). Furthermore, corals expanded their polyps more at the end of each experiment compared to the beginning (F (1,123) = 6.3; P = 0.013).

Discussion
The observed effects of ocean acidification on heterotrophy in the stony coral Galaxea fascicularis contradicted our initial hypothesis. We expected corals to ingest more zooplankton under high CO 2 . Instead, we found that food consumption rates were reduced under elevated CO 2 , both in the field and in chamber experiments, and during two expeditions. Since the colonies in our high and ambient CO 2 treatments had been subjected to life-long exposure to their respective CO 2 environments, this study presents the first investigation of heterotrophy in corals that were fully acclimatized to elevated CO 2 throughout their entire post-settlement lives. The taxonomic composition of the zooplankton consumed by G. fascicularis was different compared to the zooplankton community available to the corals. Such selectivity is known for corals 12 . Selectivity may be indicative of plankton behavior; for example, some zooplankton taxa swim more slowly or clumsily making it easier to capture them, while some taxa have chemical defenses that make them unpalatable to corals 35 . Whether it is from their own choosing or more from the behavior or chemical defenses of the zooplankton, there was strong Figure 3. The percent composition of the top available and consumed zooplankton taxa is shown for both expeditions, methods, and between CO 2 levels. Plots for the 16 most commonly consumed zooplankton taxa compare the percent of each taxon consumed by the coral represented in the coelonteron (blue symbols) to the percent of the community that each zooplankton is available in the water column (red symbols). Each zooplankton taxon has two rows, with the top row (circles) representing the control site and the bottom row (triangles) representing the elevated CO 2 site. Each panel represents a separate experiment (two expeditions and two methods). Asterisks indicate a significant difference between the percent consumed and percent available in the water column (t-tests, p-value < 0.05).
Scientific RepoRts | 6:27019 | DOI: 10.1038/srep27019 selection for certain zooplankton, and this selectivity appeared to be largely unaffected by CO 2 treatments, with the exception of only a few uncommon taxa ( Table 2).
Selectivity results may be slightly biased towards larger zooplankton taxa since smaller groups digest faster than larger zooplankton 36 . However, the feeding time in this study was purposely chosen to be one hour so that complete digestion could be avoided. Complete digestion takes hours to days, and even small nauplii are still recognizable after only 60 minutes in the coelenteron 12,26,36,37 . Furthermore, since the mesh size of the plankton net was 100 μm, the smallest zooplankton types were excluded from the experiment. In fact, most zooplankton consumed were easily identifiable to species level even when partially digested, hence the category 'unidentified consumed zooplankton' represented only 13% of the items retrieved from the coelonteron.
G. fascicularis consumed less zooplankton in the high CO 2 water despite having the same access to food, the same state of polyp expansion, and the same corallite sizes between CO 2 treatments. The reasons for the observed reduction in feeding rates could be many, however our study negated several potential causes. Reduced heterotrophy was not caused by a reduction in corallite size since G. fascicularis corallites were the same size between CO 2 levels, even though exposure to elevated CO 2 reduces corallite sizes in some other coral species 38 . For example, the temperate coral Oculina patagonica showed smaller corallites at elevated CO 2 due to high energetic costs for calcification; however, after one month of acidic conditions the skeleton completely dissolved and polyp sizes increased when calcification ceased and the resulting free energy was channeled into somatic growth 39 . With respect to G. fascicularis, it is possible that net calcification rates may change under ocean acidification conditions despite the morphology of the corallites remaining similar for both CO 2 levels.
Reduced heterotrophy was also not caused by a difference in polyp expansion, which remained unaffected by ocean acidification for G. fascicularis. In contrast, another study observed that polyps from the coral P. lutea extended further under high CO 2 conditions 24 . During the second expedition, however, G. fascicularis polyps were expanded more, which happened to occur during a new moon compared to the first expedition that had a full moon. Corals are known to feed differently with the lunar cycle, coinciding with lunar effects on zooplankton migration patterns 28,40 . Also, polyps were expanded more in the field experiments compared to the chamber experiment, probably because the corals were undisturbed in the field.
A deficiency in the functioning of GABA A neurotransmitter receptors in G. fascicularis was also not a likely cause for the observed reduction in heterotrophy. Gabazine plays a role in Hydra vulgaris feeding response 31 , therefore we expected it to also influence coral feeding behavior of G. fascicularis since both of these cnidarians share similar nervous systems. Despite our predictions, the treatment of G. fascicularis with gabazine yielded no change in coral heterotrophy. The effect of ocean acidification on coral neurotransmitters cannot be completely excluded, however, because different chemicals besides gabazine may bind to the neurotransmitter receptors (e.g. the agonist muscimol and the antagonist bicuculline) 32 . To thoroughly understand the effect of ocean acidification on neurotransmitters of G. fascicularis, the reactions of other receptor antagonists and agonists to elevated CO 2 need to be evaluated. Additional experiments are needed to reveal the underlying mechanisms responsible for the reduced feeding rates in G. fascicularis. Potential causes or contributors that deserve further study include reduced particle retention, changes in cellular homeostasis of the tentacle cells, reduced nematocyst functioning, altered mucus production, physiological stress that makes them less capable to feed, an increase in autotrophy, and potential changes in plankton behavior, as briefly outlined here. G. fascicularis exerted similar effort to capture zooplankton between CO 2 levels by extending their polyps to the same level. That they ingested fewer food particles in ocean acidification conditions may reflect upon the polyps' ability to capture food. Food retention may be reduced if the functionality of their stinging cells (nematocysts) is disrupted 41,42 . Nematocyst performance may be vulnerable to changes in pH since the acid-base balance in cells corresponds to the intracellular concentration of free H + ions. A study on the jellyfish Pelagica noctiluca indicated that the cell homeostasis of nematocysts is profoundly compromised by acidification of the surrounding seawater impairing the cells' discharge capability 43 . Although cellular homeostasis in nematocysts may vary between jellyfish and corals, nematocyst functioning may be impaired for corals under ocean acidification and merits further investigation.
Another possible cause for the observed reduced feeding rates could be that the polyps themselves lose their ability to retain food particles. Food particles may be stung or killed, but the mucosal or tentacular action of the polyps may not trap the particles, resulting in the loss of prey items 26 . Mucus enhances coral heterotrophy 44 , therefore heterotrophy will likely be vulnerable to any changes in mucus production, but nothing is known about how ocean acidification may affect coral mucus.
It is perceivable that G. fascicularis may also have reduced rates of heterotrophy in response to a reduced energy demand. Elevated CO 2 enhances the photosynthetic-derived energy supply in some coral species, and this energy is available to support critical functions like calcification. Coral calcification is generally considered to decline with elevated CO 2 levels 45 , although some studies report parabolic and even positive calcification responses to ocean acidification conditions 46,47 . However, corals are more nutrient limited than carbon limited in oligotrophic and shallow (high-light) environments 2 . Furthermore, feeding rates of corals only reach saturation when food concentrations are high, with heterotrophy generally more efficient in oligotrophic habitats 48 . Considering that G. fascicularis from the CO 2 seep sites live in a nutrient-poor and high-light environment, it is highly unlikely that feeding becomes saturated and their need for essential nutrients not attained from photosynthesis would still be prevalent. Therefore, G. fascicularis would likely continue to feed on zooplankton at the CO 2 seep sites if they were still capable even under an increased carbon supply from photosynthesis.
Regardless of the underlying mechanisms, reduced heterotrophy under elevated CO 2 will have biological impacts on corals. Growth, reproduction, zooxanthellae maintenance 49 , and other metabolic processes depend on nitrogen, phosphorus, and other essential trace elements, which are exclusively attained through heterotrophy 6,50-52 . We are only starting to understand the long-term impacts of ocean acidification on tissue growth, phototrophy, respiration, heterotrophy, and their energetic interdependencies, in selected species of coral. Many but not all coral species increase their rates of photosynthesis at higher pCO 2 levels 53 . Reduced heterotrophy may also impact coral lipid content and fatty acid composition, since they are co-determined by zooplankton consumption 54 . Furthermore, lower feeding rates may slow skeletal and tissue growth considering that growth is positively correlated with rates of heterotrophy for several coral species 6,52 , so lower feeding rates may slow growth. Heterotrophy is certainly beneficial to corals and yet clearly heterotrophy declines for G. fascicularis under elevated CO 2 . Any potential impact to their basic biology warrants further research.
Despite the remaining knowledge gaps, decreased heterotrophy will have important implications for the health and resilience of corals. As ocean conditions increasingly become unfavorable for many coral species, their ability to react to such stress will become imperative to their survival. Some coral species will persist while others will not, and our data show that some G. fascicularis colonies are able to survive under high CO 2 in the field, despite their lifetime exposure to elevated CO 2 conditions and associated reduced zooplankton feeding rates. However, it was beyond the scope of this study to measure their physiology (tissue biomass, lipid content, calcification rates, or other biophysical parameters indicative of their overall health). Such measurements should be conducted to better understand coral long-term survivability under ocean acidification.

Study site. The feeding experiments were conducted at Upa-Upasina Reef, a fringing reef in Milne Bay
Province, Papua New Guinea, where a natural volcanic CO 2 seep provides gradients in seawater pH 55 . A spatial map of the seawater carbonate chemistry, along with a detailed description of the Upa-Upasina high CO 2 and control site can be found in Fabricius et al. 55,56 . G. fascicularis colonies were collected near the seep site where seawater approximates 7.8 pH T (total scale), and from a control site with control CO 2 at ~8.1 pH T . The chamber feeding experiments were conducted aboard the back deck of the ship while moored near Upa-Upasina Reef, with G. fascicularis fragments that were freshly collected from the reef. The field and chamber experiments were conducted during two ship expeditions to the site (12-14 April 2014 and 18-20 November 2014).
Seawater carbonate chemistry. The carbonate chemistry for the field sites varied through time and long-term measurements have been reported in previous literature 56 . Additionally, seawater pH at total scale (pH T ) was recorded at the control and elevated CO 2 sites for several days surrounding the commencement of the feeding experiments using SeaFET pH sensors (Supplementary Information Figure S3). pH T values had similar ranges compared to previous expeditions 55,56 . Water samples were also collected, fixed with saturated mercuric chloride solution (HgCl 2 ), and later analyzed for their dissolved inorganic carbon (DIC: μmol kg −1 ) and total alkalinity (A T : μmol kg −1 ) using the Versatile Instrument for the Determination of Total Inorganic Carbon and Titration Alkalinity (VINDTA 3C).
Carbonate chemistry was also measured for the seawater used for the chamber experiments and water temperature (°C) was recorded on site. Water samples saturated with HgCl 2 were stored and later measured for DIC and A T . The water temperature was 25 °C at the time the samples were analyzed in the laboratory for its carbonate chemistry using the VINDTA 3C. DIC and A T were used to calculate other seawater parameters (Table 3), including pH at total scale (pH T ), partial pressure of carbon dioxide (pCO 2 : μatm), bicarbonate (HCO 3 − : μmol kg −1 ), carbonate (CO 3 2− : μmol kg −1 ), aqueous carbon dioxide (CO 2(aq) : μmol kg −1 ), the saturation state of calcite (Ω CA ), and the saturation state of aragonite (Ω AR ), using the Excel macro CO2SYS 57 under the constraints set by Dickson and Millero (1987) 58 .
Food collection. Zooplankton were freshly collected via plankton net tows from the control site at approximately 9 pm, i.e. 2-3 h after sunset, and shortly before the start of the field and the chamber experiments. Each net tow was very slow to minimize stress to the zooplankton. Live samples were handled with care and only living zooplankton were used as food for corals (i.e. zooplankton still suspended in the water column and actively swimming. No zooplankton that had settled at the bottom of the collection container were used). Three to six zooplankton samples were preserved in 4% formalin and kept as references to determine variation in the number and taxonomic composition of zooplankton between samples.

Field feeding experiment.
Tents of 100 μm plankton mesh and 25 cm base diameter (approximately 8 L volume) were used to contain zooplankton close to corals for the duration of the feeding experiment. Five tents were placed over separate G. fascicularis colonies each at the high CO 2 and the control sites. To prevent corals from consuming zooplankton that are naturally in the water column, the tents were deployed during daylight when zooplankton numbers are low and demersal zooplankton have not emerged into the water column yet. At approximately 9 pm, SCUBA divers injected three 60 ml syringes of freshly collected and concentrated zooplankton into each tent. Polyp expansion (25,50,75 or 100 percent expanded) was recorded at the beginning and end of the feeding period. After approximately one hour, the tents were removed and a fragment of each colony was extracted with a hammer and chisel and preserved in 4% formalin. The field experiments were conducted once during expedition 1 (three replicate coral colonies per CO 2 level), and twice on two consecutive nights during expedition 2 (five replicate coral colonies per CO 2 level for both nights). The plankton fed to the corals during the second expedition had similar composition and concentration in the two consecutive nights, so the results from both nights were pooled together and considered one experiment.
Chamber feeding experiment. G. fascicularis fragments were collected from both the high CO 2 site and the control site. They were placed in flow-through aquaria for four days to recover. The aquaria consisted of two 60 L bins with an outboard pump supplying a constant inflow of fresh seawater. For 12 hours prior to the feeding experiment, 100 μm mesh was placed over the input valve to starve G. fascicularis, allowing any previously consumed food to be digested. Three hours prior to the feeding experiment, each coral fragment was transferred onto a raised grid platform in individual cylindrical incubation chambers (89 mm diameter, 106 mm height, 637 ml volume) without exposing them to air. Corals collected from the seeps were placed in the chambers filled with seawater from the seep site, while those from the control site were placed in chambers filled with seawater from the control site (seawater carbonate chemistry for chamber experiments found in Table 3).
Chambers were 80% immersed in a water bath. Airspace in the chamber and a hole in its upper lid facilitated gas exchange. To generate a current within the chamber, a battery driven pulley system activated magnetic stirrer bars underneath the grid 53 . G. fascicularis were fed at around 9 pm. Taking care to supply only living zooplankton, concentrated zooplankton was injected through a hole in the top lid of the chamber with a volumetric pipette. The zooplankton concentration was lower during the second expedition compared to the first, so a larger volume of plankton solution was inserted into the chamber during the second (30 ml) compared to the first expedition (20 ml). An additional three samples of the food were preserved in 4% formalin and kept as references. The feeding experiment was conducted in the dark, although red light was used for a few minutes at the commencement and cessation of the experiment to assess their state of polyp expansion. G. fascicularis fed for approximately one hour and then each coral piece was removed and immediately stored in 4% formalin. The chamber experiment was conducted once through an initial pilot study during expedition 1 (7 replicate coral colonies per CO 2 level), and repeated during expedition 2 with additional replicates (12 replicate coral colonies per CO 2 level).
To determine if elevated CO 2 interferes with neurotransmitter receptor functioning, six of the coral fragments per CO 2 treatment were exposed to gabazine (SR-95531, Sigma-Aldrich) at a concentration of 4 mg L −1 seawater for 30 min (chamber experiment, second expedition). Coral fragments were gently washed and transferred into  ), aqueous carbon dioxide (CO 2(aq) ), the saturation state of calcite (Ω CA ), and the saturation state of aragonite (Ω AR ).
Scientific RepoRts | 6:27019 | DOI: 10.1038/srep27019 their chambers filled with gabazine-free seawater. The other six colonies per CO 2 treatment were exposed to the same handling procedure, but their 30 min transfer was into a container without gabazine. Experiments were then conducted as outlined above.
Food samples for corals. Food samples given to corals were compared within and between experiments.
Food samples given to each replicate coral fragment were similar in quantity and composition within each experiment, and they were not different between high CO 2 (7.8 pH T ) and control treatments (8.1 pH T ) and replicates. However, food samples varied in quantity and composition between the four field and chamber experiments. Details about the analysis of food samples are in the supplementary information, including Figure S1. Laboratory analysis. Coral consumption was measured through coelenteron content analysis. G. fascicularis fragments were removed from formalin and placed in freshwater. Every polyp coelenteron was probed using a tungsten needle and dissecting forceps. Extracted zooplankton were identified to their major taxonomic groups. Total corallite number and corallites containing food particles were enumerated. Each coral fragment was photographed and the surface area calculated within the image-processing program, ImageJ. Corallite size was calculated by dividing the surface area of each coral fragment by the number of corallites.
Statistics. All statistical analyses were computed in R, version 3.2.2 (R Development Core Team, 2015).
Generalized linear models (GLMs) were used to determine if: (1) the number of zooplankton consumed (standardized by surface area) differed across CO 2 regimes (seep vs. control), expedition (one vs. two), or methods (field vs. chamber), (2) species richness (Shannon-diversity index) of the zooplankton taxa consumed by corals differed between CO 2 regimes, expedition, or methods, (3) gabazine affected coral feeding rates, (4) zooplankton concentration in the food samples was different between each of the experimental runs, (5) corallite sizes were different between corals originating from seep and control sites, and (6) polyp expansion differed across CO 2 levels, methods, expeditions, or from the beginning to the end of the experiment. Appropriate data distributions and link functions were chosen for each GLM. Model assumptions of independence, homogeneity of variance, and normality of error were evaluated through diagnostic tests of leverage, Cook's distance, and dfbetas 59 . Checks for all GLMs indicated that no influential data points or outliers existed in the data and model assumptions were met. ANOVAS (Type II) were used to determine the minimal adequate GLM with the ' Anova' function in the R library 'car' (version 2.1-1) 60 . The effects of the explanatory variables on the response variables were then reported based on these GLMs.
Canonical correspondence analysis (CCA) was used to determine if the zooplankton community composition of the food available to the corals, and the food consumed by the corals, differed in relation to the explanatory variables (CO 2 , expedition, method). To account for many zeros in the data where some zooplankton taxonomic groups were rarely present or rarely consumed, the community data was standardized using the Hellinger (square root) method within the decostand function of the vegan package in R 61 . A Monte-Carlo permutation test was used to determine the optimal CCA model and to assess the significance of the variation in species composition attributable to the explanatory variables (CO 2 , expedition, method).
For each zooplankton taxon, its percent representation in the coral coelenteron content was compared against its percent in the available food using two-tailed t-tests, assuming unequal variances between samples. Logistic regressions were used to model the response of each zooplankton taxon contained in the corals to the explanatory variables of CO 2 , expedition, and method. Logistic regressions use binary data of 'successes' and 'failures' . In this example, 'success' equals the probability of each taxon being consumed (p), and 'failure' equals the probability of not being consumed (1-p). Logistic regressions are within the framework of GLMs and use log-odd-ratios, defined by the logit link function, to estimate the (log) odds of each taxon being consumed under each independent variable. GLMs with a binary data distribution and logit link function were checked for overdispersion. Overdispersion (residual deviance greater than the residual degrees of freedom) existed, so the data distribution was changed to quasibinomial. Anovas with a Chi-square test were applied to the results of each GLM for each zooplankton taxon.