Chl a was extracted from ground samples in 100% acetone and the absorbance measured at 663 and 630 nm wavelengths using a Spectronic 20 Genesis visible spectrophotometer. The Chl a concentration was calculated using equations from Jeffrey and Humphrey (1975) corrected for turbidity with the absorbance value measured at 750 nm. Algal endosymbionts were separated from the animal tissue by centrifugation and three subsamples were counted with a hemocytometer and light microscope. The total number of algal cells and the number of mitotically dividing algal cells were recorded.

Total lipids were extracted in a 2:1 chloroform/methanol solution, washed in 0.88% KCl, and then dried to a constant weight (method modified from Grottoli et al., 2004). Soluble animal protein was extracted using the bicinchoninic acid method (Smith et al., 1985) with bovine serum albumin as a standard (Pierce BCA Protein Assay Kit). Soluble animal carbohydrate was extracted using the phenol-sulfuric acid method (Dubois et al., 1956) with glucose as a standard. All analyses were made from whole ground coral samples, converted into Joules (Gnaiger and Bitterlich, 1984), and standardized to both grams of ash free dry weight (gdw) and to surface area as measured using the aluminum foil technique (Marsh, 1970). However, normalization to ash free dry weight is more robust when comparing species with different tissue depths, morphology, and polyp structure than the same data normalized to surface area (Edmunds and Gates, 2002). Therefore, only results of data standardized to ash free dry weight are reported here (However, note that all statistics were also performed on the area normalized data and yielded almost identical patterns). Total energy reserves were calculated as the sum of total lipids, proteins, and carbohydrates. Total dry biomass was standardized to surface area. These analytical methods have been successfully used in corals before (e.g., Rodrigues and Grottoli, 2007).

Coral tissue was completely removed from the skeleton using a Water-pik (Johannes and Wiebe, 1970). The tissue slurry was separated into animal host and algal endosymbiont fractions by centrifugation using established methods (Rodrigues and Grottoli, 2006). Briefly, coral slurry was homogenized to break open the animal cells, then centrifuged at 4,000 rpm for 5 min to separate the animal host and endosymbiotic algal fraction. The supernatant (i.e., animal host) was decanted and the algal pellet resuspended in deionized water and centrifuged again to remove any remaining animal host particles. The supernatants from both centrifugations were combined. Each organic fraction was isolated onto a pre-burned glass fiber filter under vacuum, and individually combusted in a Costech Elemental Analyzer. The resulting CO₂ gas was analyzed for δ¹³C [δ¹³C = permil deviation of ¹³C:¹²C relative to Vienna Peedee Belemnite Limestone standard (VPDB)] with a Thermo Delta Advantage IV via a Thermo ConFlow III open split in the Grottoli lab. At least 10% of all samples were run in duplicate. The standard deviation of repeated measurements of the USGS24 standard for δ¹³C was ±0.08 ‰. The host tissue isotope values were reported as δ¹³C_h while the algal endosymbiont isotope data were reported as δ¹³C_e. In general, as the incorporation of isotopically depleted zooplankton and other heterotrophic carbon sources (i.e., dissolved and particulate organic carbon) into coral tissues increases, the δ¹³C_h decreases (Rodrigues and Grottoli, 2006; Levas et al., 2013; Schoepf et al., 2015). As the rate of photosynthesis increases, isotopic fractionation decreases, and endosymbiotic algae incorporate relatively less ¹²C than ¹³C into their tissues, resulting in increased δ¹³C_e (Muscatine et al., 1989; Rodrigues and Grottoli, 2006). The difference between the δ¹³C_h and δ¹³C_e were also calculated to assess the proportionate contribution of heterotrophy and photoautotrophy in the carbon budget (sensu Muscatine et al., 1989). Specifically, decreases in δ¹³C_h–δ¹³C_e are interpreted to mean that the proportionate contribution of heterotrophic carbon to coral tissues has increased.

The skeleton was prepared for δ¹³C (δ¹³C_s) and δ¹⁸O (δ¹⁸O_s)[δ¹⁸O = permil deviation of ¹⁸O:¹⁶O relative to VPDB] analysis by shaving the top 100–200 μm of skeletal material off the tip of each coral fragment using a Dremel tool with a diamond-tipped drill bit. Approximately ~80 μg of skeletal material was acidified with 100% ortho-phosphoric acid in an automated Kiel III carbonate device, and the δ¹³C and δ¹⁸O values of the resulting CO₂ were measured using a Finnigan MAT 252 mass spectrophotometer at Stanford University. At least 10% of all samples were run in duplicate. The standard deviation of repeated measurements of an internal standard was ±0.03 ‰ for δ¹³C and ± 0.05 ‰ for δ¹⁸O. Skeletal values were reported as δ¹³C_s and δ¹⁸O_s. In general, as the rate of photosynthesis increases or as the incorporation of heterotrophic carbon into coral tissues decreases, the pool of inorganic C available for calcification through animal respiration becomes enriched resulting in increased δ¹³C_s (Grottoli and Wellington, 1999; Grottoli, 2002). In addition, variability in δ¹⁸O_s results from temperature-induced kinetic fractionation and seawater δ¹⁸O (Epstein et al., 1953). As temperature increases, δ¹⁸O_s decreases (Weber and Woodhead, 1972; Kim and O'Neil, 1997). Since seawater δ¹⁸O_sis constant in this study because the source seawater was the same in all tanks, it did not influence the δ¹⁸O_s values.

Univariate two-way analysis of variance (ANOVA) was used to test the effects of species and temperature on Chl a, algal endosymbiont, mitotic index, lipid, protein, carbohydrate, total energy reserves, and tissue biomass concentrations as well as δ¹³C_s, δ¹⁸O_s, δ¹³C_h, δ¹³C_e, and δ¹³C_h–δ¹³C_e. Prior to statistical analysis, all data was tested for normality using a Shapiro-Wilk's test and homogeneity of variance was assessed with plots of expected vs. residual values. Any data failing to meet this assumption was log transformed and then met the assumptions of normal distribution. A posteriori slice tests (i.e., tests of simple effects, Winer, 1971) determined if bleached and control averages significantly differed within species. Bonferroni corrections were not used due to increased likelihood of false negatives (Quinn and Keough, 2002). With only one tank for treatment and one tank for controls, we were unable to assess tank effects independent of treatment effects. However, every aspect of both tanks that were measured were the same (i.e., seawater source, flow, light, position relative to the sun's arc) except for temperature. In addition, one bleached P. damicornis and two control F. favus fragments were lost during shipping or storage, yielding an unbalanced number of samples in the treatment and controls for these two species. Statistical analyses were generated using SAS software, Version 8.02 of the SAS System for Windows. Values of p < 0.05 were considered significant.

To get an overall assessment of the effect of treatment across all variables, non-parametric multi-variate statistics were employed. The non-parametric analyses were conducted on three sub-sets of the data: (1) all of the physiology data, (2) all of the isotopic data, and (3) all of the physiology and isotopic data combined. Since two samples did not have isotopic data, the multi-variate analyses that included the isotopes were comprised of 29 corals whereas the analyses of the physiological variables alone included 31 corals. Euclidean-distance based resemblance matrix was constructed using normalized data of all of the measured variables. Non-parametric multidimensional scaling (NMDS) plots were generated to graphically represent relationships between each coral sample across all treatments and species. Analysis of similarities (ANOSIM) was then used to test for the effect of species and treatment. ANOSIM simultaneously uses all variables measured, and tests whether all data (i.e., all physiological and/or isotopic) from one coral group differs from another coral group. The ANOSIM pairwise test statistic R ranges from 0 (no difference and complete overlap between groups) to 1 (maximum difference and no overlap between groups) and is a strong indicator of separation between groups, especially at low sample sizes such as in this study, as long as the overall ANOSIM test statistic R is significant (Clarke and Gorley, 2006). Multi-variate analyses were conducted using the software package Primer V6, and for the ANOSIMs a p < 0.05 was considered significant. Finally, since fragments pairs of each colony were included in both treatments, differences between treatment and control for any variable were due to treatment alone, independent of genotype.

Average endosymbiotic algal concentrations dropped to 5, 49, and 19% of control concentrations in S. pistillata, P. damicornis, and F. favus, respectively (Table 1, Figure 1A). At the same time, Chl a concentrations in bleached S. pistillata and P. damicornis decreased to 37 and 33% of control values, respectively (Table 1, Figure 1B), and mitotic divisions were 26-fold lower in bleached S. pistillata and 9-fold lower in bleached F. favus corals relative to their controls (Table 1, Figure 1C). While endolithic algae were not present in noticeable quantities in either P. damicornis or S. pistillata, F. favus showed evidence of harboring endolithic algae. Because all samples of F. favus showed approximately the same quantity of endolithic algae irrespective of treatment, it was assumed that their contribution to Chl a was relatively constant and that any differences in Chl a between the bleached and control fragments was due to the bleaching effect alone.

While lipids and protein were unaffected by bleaching overall, a posteriori slice tests revealed that lipid concentrations in bleached S. pistillata and protein concentrations in bleached F. favus were 49 and 48% lower, respectively, than in their controls (Table 2, Figures 2A,B). Carbohydrate concentrations dropped to 40 and 41% of control values in bleached P. damicornis and F. favus, respectively (Table 2, Figure 2C). Despite losses of at least one component of energy reserves in each species, only bleached F. favus had significantly lower total energy reserves (43% decrease) and experienced a 27% decrease in biomass compared to controls at the species level (Table 2, Figures 2D,E).

Average δ¹³C_s values were significantly higher in bleached than in control corals overall (Table 3), but this difference was only significant at the species level for S. pistillata where bleached δ¹³C_s was 1.7 ‰ higher than control corals (Table 3, Figure 3A). Average δ¹⁸O_s was 0.9 ‰ higher in the bleached than in control S. pistillata (Table 3, Figure 3B). At the same time, average δ¹³C_h values were 2.2 ‰ and 1.8 ‰ higher in bleached than in control S. pistillata and F. favus (Table 3, Figure 3C), respectively, while average δ¹³C_e values were 1.0 ‰ higher in bleached than in control F. favus, and at least 2.4 ‰ higher in F. favus overall compared to the other two species (Table 3, Figure 3D). Finally, average δ¹³C_h–δ¹³C_e values were 0.8 ‰ higher in bleached than in control corals overall (Table 3), but this difference was only significant at the species level for S. pistillata where the average δ¹³C_h–δ¹³C_e was 1.5 ‰ higher in bleached than in control corals (Figure 4).

Overall, both bleached S. pistillata and F. favus corals significantly differed from their controls, whereas bleached and control P. damicornis corals did not differ from each other (Table 4, Figure 5). This was true for all combinations of the data (i.e., physiological variables alone, isotopic variables alone, of both physiological and isotopic data combined). The degree of separation between the bleached and control corals of both S. pistillata and F. favus was similar with R values produced from the physiological and isotopic data combined of 0.66 and 0.64, respectively. In addition, all species significantly differed from each other, though F. favus differed the most from the other two species as indicated by R > 0.7 when compared to either of the other two species when both the physiological and isotopic data were considered (Table 5, Figure 5).

S. pistillata bleached the most severely of all three species studied with the greatest percentage drop in endosymbiotic algal density (Figure 1A). Large drops in endosymbiotic algal density are associated with dramatic decreases in photosynthesis, resulting in corals that are unable to meet their total metabolic demand (Grottoli et al., 2006, 2014; Tremblay et al., 2012). Here, S. pistillata catabolized its lipid reserves (Figure 2A), presumably to compensate for losses in photosynthetically fixed carbon. The decline in lipids could have been further exacerbated by the absence of supplemental feeding, since heterotrophic carbon is critical to lipid synthesis in bleached corals (Baumann et al., 2014). Increases in δ¹³C_h (Figure 3C) suggest that the lipids catabolized during bleaching (Figure 2A) were isotopically depleted and/or that in the absence of isotopically depleted zooplankton in the diet, coral tissues became progressively enriched. Given that bleached Hawaiian corals were found to catabolize isotopically enriched lipids (Grottoli and Rodrigues, 2011), it is most likely that the enriched δ¹³C_h observed here in bleached corals is a result of a lack of zooplankton. In addition, increases in δ¹³C_h−e (Figure 4) suggests that S. pistillata incorporated less heterotrophic C into its tissues when bleached [i.e., increases (decreases) in δ¹³C_h−e are indicative of dramatic decreases (increases) in the incorporation of heterotrophic carbon in tissues (Rodrigues and Grottoli, 2006)]. This is consistent with the fact that corals in this study were not fed zooplankton during the experiment. However, even if they had been fed, this species is known to decrease feeding rates on zooplankton, pico-, and nano-plankton, as well as suffer increases in total organic carbon losses when bleached (Ferrier-Pages et al., 2010; Tremblay et al., 2012), and that it has limited heterotrophic plasticity capacity in response to a changing environment (Alamaru et al., 2009). Despite the loss of lipids and isotopic evidence of reduced heterotrophic carbon input, bleached S. pistillata only showed a trend of declining total energy reserves (p < 0.13) and maintained its biomass (Figures 2D,E). This may be because this species is able to rapidly curb apoptosis and acclimate to sustained thermal stress when exposed to elevated temperatures (Kvitt et al., 2014, 2016).

The increase in δ¹³C_s and δ¹⁸O_s when bleached (Figures 3A,B) is consistent with decreases in calcification during bleaching (Rodrigues and Grottoli, 2006) due to greater equilibration with seawater DIC when calcification rates drop (McConnaughey, 1989). Increases in δ¹³C_s could also indicate an increase in photosynthesis rates (Grottoli and Wellington, 1999; Grottoli, 2002). However, this is not likely since this species typically experiences significant declines in photochemical efficiency (Fv/Fm) and photosynthesis rates at elevated temperatures (Tremblay et al., 2012; Fine et al., 2013), and because it is not physiologically possible to increase photosynthesis rates when bleached. In addition, the increase in δ¹⁸O_s deceptively suggests that seawaters were cooler, not hotter, in the bleached coral tank (Figure 3B). Increases in δ¹⁸O_s in bleached corals are caused by isotopic equilibration with seawater as calcification dramatically slows (Rodrigues and Grottoli, 2006). Thus, the skeletal isotopic signature here indicates a decline in calcification in bleached S. pistillata. Decreasing calcification rates is one mechanism by which S. pistillata may reduce its energetic demand and maintain total energy reserves and biomass. In principle, maintenance of energy reserves and biomass should confer an advantage to S. pistillata and facilitate recovery from bleaching, as has been found for the branching Caribbean coral Porites divaricata (Grottoli et al., 2014).

Of all three species studied here, P. damicornis was the least affected by bleaching with the smallest decline in endosymbiont cell density and the only species to maintain mitotic cell division (Figure 1). Fine et al. (2013) showed that photochemical efficiency of PSII in this species is among the least affected by 34°C temperatures compared to other corals from this site, despite hosting the typically thermally sensitive clade C Symbiodinium (Karako-Lampert et al., 2004). At the same time, P. damicornis preserved its energy reserves and biomass (Figure 2), showed no metabolic shifts in its isotopic signatures (Figures 3, 4), and overall was unchanged when bleached (Figure 5). The maintenance of its skeletal isotopic values when bleached (Figures 3A,B) suggests that calcification was maintained when bleached (Rodrigues and Grottoli, 2006). This is not an unreasonable interpretation as some populations of Pacific P. damicornis corals have optimal calcification rates at 31°C (Clausen and Roth, 1975)—only 1° cooler than the experimental temperatures. The lack of any shifts in δ¹³C_h, δ¹³C_e, and δ¹³C_h−e indicate that the proportionate contribution of photoautotrophic and heterotrophic C to tissues did not change with bleaching. Given that these corals were not fed zooplankton during the experiment, these results further suggest that P. damicornis may not be very dependent on heterotrophic C for meeting metabolic demand or may not be able to change its heterotrophic intake under bleaching stress. Work by Ziegler et al. (2014) shows that Pocillopora are not heterotrophically plastic in response to changes in light, which may indicate that they would not be heterotrophically plastic in response to temperature changes either.

Schoepf et al. (2013) also found no changes in energy reserves and biomass in P. damicornis under control and elevated temperature conditions (26.5 vs. 29°C). This is surprising since corals of this genus are often among the most sensitive to thermal stress in other regions (Hueerkamp et al., 2001; Obura, 2001; McClanahan et al., 2004; Guest et al., 2012; Foster et al., 2014; Pisapia et al., 2016) and even suffer lipid losses when bleached (Rodriguez-Troncoso et al., 2016). But there are also cases where Pocillopora corals do not readily bleach compared to other species (Sebastian et al., 2009; Guest et al., 2012). Some evidence suggests that Pocillopora corals are beginning to acclimatize or adapt to thermal stress in Australia (Ulstrup et al., 2006; Maynard et al., 2008) and may have high acclimatization potential in northern Red Sea populations (Sawall et al., 2015), though not through heterotrophic plasticity mechanisms (Ziegler et al., 2014). These findings suggest that P. damicornis in Eilat is a more resilient population to temperature-induced bleaching than those in other locations because (1) they can maintain their energy reserves, biomass, and calcification, and (2) appear to not be dependent on heterotrophy for meeting metabolic demand when bleached.

Bleached F. favus suffered declines in more physiological variables than either of the other two species with significant decreases in endosymbiotic algal density, mitotic divisions, protein, carbohydrates, total energy reserves, and biomass (Figures 1, 2), yet still maintained Chl a concentrations (Figure 1B). Even though photochemical efficiency (F_v/F_m) tends to decline at elevated temperatures for F. favus from Eilat (Winters et al., 2006; Kuguru et al., 2007; Fine et al., 2013), the observed increase in δ¹³C_e (Figure 3D) and maintained Chl a levels suggests that some level of photosynthesis persisted while bleached. At the same time, F. favus catabolized its protein and carbohydrate reserves, which underlie the significant declines in total energy reserves and biomass (Figures 2B–E), presumably to compensate for losses in photosynthetically fixed C. Increases in δ¹³C_h (Figure 3C) suggest that the protein and carbohydrates catabolized during bleaching (Figures 2B,C) were isotopically depleted, or that in the absence of isotopically depleted zooplankton in the diet, coral tissues became progressively enriched. Further research is needed to evaluate the isotopic composition of protein and carbohydrates. However, since these corals were not supplied zooplankton during the experiment, it is likely that the enrichment in δ¹³C_h observed here is at least in part a result of a lack of zooplankton availability. Although the δ¹³C_h−e did not change with bleaching in F. favus (Figure 4), it was lower overall than in S. pistillata and P. damicornis indicating a higher contribution of heterotrophic C to its tissues in general, irrespective of bleaching, than in the other two species (Figure 4). This is consistent with field-based findings showing that F. favus allocates a higher proportion of heterotrophic C to its lipids than does S. pistillata (Alamaru et al., 2009). In addition, some species are able to meet up to 36% of metabolic demand when bleached through the heterotrophic acquisition of dissolved and particulate organic carbon (Levas et al., 2013, 2016). The lower overall δ¹³C_h−e of F. favus suggests that when bleached it was able to take up dissolved and/or particulate organic carbon during the tank experiment to meet at least part of its metabolic demand heterotrophically. But any uptake in heterotrophic C was not enough to prevent large losses in protein and biomass. Had the corals been fed zooplankton, it is possible that F. favus might have maintained its energy reserves and biomass.

Finally, the skeletal isotopic signature in F. favus was unaffected by bleaching (Figures 3A,B), suggesting that calcification was maintained or at least did not decrease dramatically (Rodrigues and Grottoli, 2006). This is also consistent with the interpretation that some level of photosynthesis was maintained, since calcification is tightly linked to photosynthesis (see review by Allemand et al., 2011). Furthermore, in healthy corals, δ¹⁸O_s decreases by ~0.2 ‰ per 1°C seawater warming (Weber and Woodhead, 1972; Wellington et al., 1996). Here a non-significant trend of a 0.36 ‰ decline in δ¹⁸O_s (P < 0.085) suggests that F. favus skeletal isotopic signature recorded the seawater warming event, though not the ~1.2 ‰ decline that would have be expected if it had accurately recorded the full 6°C increase in temperature. Nevertheless, the fact that a warming trend was recorded is further support that F. favus calcified throughout the experiment.

Overall, F. favus suffered significant declines in energy reserves and biomass, but maintained Chl a, appears to have maintained calcification, and appears to have high baseline rates of heterotrophic C contributions to its tissues. The maintenance of Chl a and the apparent high baseline heterotrophic capacity of this coral could explain why this genus experiences little to no mortality following bleaching events (McClanahan, 2004; Sutthacheep et al., 2013). In Hawaii, the mounding coral Porites lobata can also bleach severely, but recovers quickly due to a high baseline contribution of heterotrophic carbon to the diet (Palardy et al., 2008; Levas et al., 2013). Thus, the high baseline heterotrophic capacity of the mounding coral F. favus may facilitate the rapid recovery and low mortality of this species following bleaching stress, and allow it to persist into the future, despite its initial dramatic energy reserve and biomass losses when bleached.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer MN and handling Editor declared their shared affiliation, and the handling Editor states that the process nevertheless met the standards of a fair and objective review.