Effects of elevated pCO₂ and feeding on net calcification and energy budget of the Mediterranean cold-water coral Madrepora oculata

Ocean acidification has been recognized as a major threat to many calcifying organisms (Jokiel et al., 2008; Kroeker et al., 2013; Orr et al., 2005; Ries, 2011a,b). The rapid shoaling of the aragonite saturation horizon (ASH) of 1–2 m year⁻¹ (Feely et al., 2012) is a concern for the maintenance of organisms confined to the deep sea and cold waters at high latitudes where cold-water corals (CWCs) typically thrive. More than 70% of the present CWC communities will be subject to under-saturated conditions by the end of the century (Guinotte et al., 2006). Thriving in the deep and dark ocean, they lack the photosynthetic algal endosymbionts (zooxanthellae) of their tropical shallow-water congeners and consequently grow slower (Maier et al., 2009). CWC communities consist of a few frame-building scleractinian species that construct large 3-dimensional structures, sustaining high biodiversity in the deep ocean (Roberts et al., 2006). Because of the dependence on a few species and their relatively slow growth, it had initially been postulated that CWC ecosystems would be even more sensitive to ocean acidification than tropical reefs and that the shoaling of the ASH will thus be a major threat to deep-sea CWCs in the near future (Turley et al., 2007). However, a number of studies have since revealed that CWCs can maintain constant skeletal growth over a large CO₂ partial pressure (pCO₂) range, with levels as high as 1000 µatm (Carreiro-Silva et al., 2014; Hennige et al., 2014; Maier et al., 2013b; Movilla et al., 2014), and still maintain positive skeletal growth at a saturation state for aragonite (Ω_a) slightly below 1 (Form and Riebesell, 2012; Hennige et al., 2015; Jantzen et al., 2013; Maier et al., 2009; Thresher et al., 2011).

In the Mediterranean Sea, the branching Madrepora oculata is the dominant habitat-forming CWC species typically found at water depths below 200 m. The deep Mediterranean Sea is characterized by relatively constant and high seawater temperatures (12.5–14°C) and high salinity (∼38–39) (Freiwald et al., 2009; Tursi et al., 2004). The relatively high total alkalinity (∼2600 µmol kg⁻¹) results in a high buffering capacity, with more atmospheric CO₂ being absorbed than in the open ocean (CIESM, 2008). This has already affected the pH in the Mediterranean Sea and resulted in an estimated decrease of 0.005–0.06 pH units since pre-industrial times (Palmiéri et al., 2015). It is possible that the decline in pH may have already reduced calcification rates of M. oculata in the Mediterranean Sea (Maier et al., 2012), but current research suggests that it will be unaffected by the additional pH decline projected until the end of the century (Maier et al., 2013a, 2012).

To date, it is unknown how CWCs are able to maintain relatively constant rates of calcification under elevated pCO₂. An up-regulation of genes involved in calcification and other key physiological functions as a rapid response to counteract changes in seawater carbonate chemistry have been proposed for tropical corals (Kaniewska et al., 2012; Moya et al., 2012, 2015) and recently for the CWC Desmophyllum dianthus (Carreiro-Silva et al., 2014). The use of lipid reserves might also help sustain calcification during short-term exposure to elevated pCO₂ (Hennige et al., 2014). Fast calcification is supported in both tropical corals and CWCs by the up-regulation of pH at the tissue–skeleton interface, the calicoblast, where the calcareous skeleton is laid down (Al-Horani et al., 2003; McCulloch et al., 2012). In warm-water corals, this active process is fuelled by energy provided by light-dependent photosynthesis of algal symbionts, which results in a pronounced increase in calicoblastic pH during daytime calcification (Al-Horani et al., 2003; Gattuso et al., 1999). The up-regulation in the calicoblastic layer is thought to be an energy-requiring process where active ion transport is mediated via ATP (Allemand et al., 2004; McCulloch et al., 2012; Venn et al., 2013). This suggests that corals require more energy to sustain skeletal growth as seawater pH declines.

The aim of the present study was to determine the energy required to up-regulate the calicoblastic pH and to establish the threshold of Ω_a where calcification cannot be sustained and the skeleton starts to dissolve. Furthermore, we investigated whether increased food availability might help mitigate the negative effects of elevated pCO₂ on key physiological functions (respiration, calcification). The CWC M. oculata was subjected to pCO₂ levels from ambient (400 µatm) to very high pCO₂ (∼1700 µatm), corresponding to Ω_a values ranging between 2.3 and 0.6. The carbon supply to the corals, energy and carbon budgets, rates of respiration and calcification as well as the respiratory quotient (RQ) were determined.

Two small colonies of the CWC Madrepora oculata Linnaeus 1758 were collected at 434 m during the ARCADIA cruise (R/V Urania) on 25 March 2010 with a remotely operated vehicle (ROV) from the Bari Canyon, Adriatic Sea (41°17′16″, 17°16′38″). Corals were kept at 13°C in a plastic container (1040×640×515 mm) equipped with chilling, filtration and recirculation units. Several weeks prior to experiments, colonies were carefully fragmented and distributed into single vials (4.5×15 cm, inner diameter×height). Maintenance conditions with flow through and aeration to single vials were as described previously (Maier et al., 2013a,b). At the beginning of the experiments, the buoyant and the skeletal dry mass (Davies, 1989), polyp number and surface area were determined. Skeletal surface area was measured using digital photographs taken from two angles of each single coral fragment using the area calculation function of the software CPCe v3.4 (Kohler and Gill, 2006). The two-dimensional projected area was multiplied by π to approximate the cylindrical surface area of the coral branches. Coral fragments weighed on average 2.9 g for both feeding regimes with, respectively, 32.4 or 32.7 polyps and a surface area of 17.5 and 16.6 cm² for the high (HF) and low (LF) feeding groups (Table 1). None of the size variables was significantly different between the two feeding regimes (one-way ANOVA, P=0.934 for skeletal mass, P=0.956 for polyp number and P=0.771 for skeletal surface area).

Madrepora oculata were fed with freshly hatched Artemia salina nauplii according to two feeding regimes. In the HF regime (N=13), corals were fed daily from Monday to Friday while corals were fed only twice a week in the LF regime (N=12). To assess the average food uptake by M. oculata throughout the experiment, feeding rates were determined 5 times at 2 week intervals (Table S1, Fig. S1). Corals were left to feed for 21.1±2.2 h (mean±s.d.). Ten sub-samples of Falcon tubes containing A. salina nauplii were used to determine initial concentrations of particulate matter, particulate organic carbon (POC) and nitrogen (PON) from which the concentrations derived at the end of the feeding period were subtracted. Five of the Falcon tubes were used to count the initial number of A. salina nauplii, which were fixed with Lugol and counted under a binocular microscope. At the end of the incubation, uptake rates were determined by filtering A. salina nauplii left over in the vials onto GF/F filters from which the dry mass of particulate matter, and the concentration of POC and PON were measured. The POC and PON concentrations were measured using a CHN analyser (PerkinElmer 2400 Series II CHNS/O Elemental Analyzer) which was calibrated with glycine (32% of carbon and 18.66% of nitrogen).

D_t+1 is the remaining A. salina nauplii POC, taking into account the dilution by overflow for each time increment; i_D is the linear rate of dilution caused by the continuous flow through for each time increment; FT is the continuous flow of seawater in relation to the volume of the vials (here: 32 ml h⁻¹/300 ml); n_t is the number of time steps (here: 12 h⁻¹ for the stepwise approach using 5 min steps); F_t+1 is the prey (food) taken up at each time step, dependent on the concentration of prey left over after food uptake and dilution due to overflow at each time increment; i_F is the rate of food uptake, following an exponential function taking into account the changes in prey concentration at each time step (due to dilution and feeding); P_t is the POC concentration at the start of each time increment with a known amount of prey provided at t₀ (P₀); P_end is the final POC concentration; P_t+1 is the prey concentration after dilution due to flow through and feeding; the hourly uptake was calculated on the basis of average food uptake over a 1 week feeding period; k is the index of time steps (here: 1 … 288 for 5 min increments over 24 h; and F_f is the frequency of feeding per week (HF=5/7, LF=2/7).

The carbonate chemistry was changed in five consecutive steps by adjusting the air–CO₂ mix to T₁=400 ppm (ambient air, control); T₂=1600 ppm; T₃=800 ppm; T₄=2000 ppm and T₅=400 ppm (control). Higher than ambient pCO₂ levels were obtained by mixing ambient air with pure CO₂ using high-precision mass flow controllers (ANALYT MTC GFC17, 0–10 l for air and 0–10 ml for pure CO₂) and an air compressor (Jun-Air OF302-25B), while for ambient pCO₂, ambient air was used only (Maier et al., 2013a,b, 2012). The same colonies were used repeatedly for each pCO₂ treatment, which means each individual coral fragment underwent the consecutive changes in pCO₂ levels with an acclimation period of 2 weeks. This approach allowed us to use a higher replication than a ‘classic’ design with three parallel pCO₂ treatments and two feeding regimes (2×3=6 treatments), which would have reduced the number of replicates to 4 (26 fragments/6 treatments) instead of the 13 used in the consecutive design (26 fragments/2 feeding regimes). To avoid misinterpretation with respect to changes in the measured parameters that might have occurred over time independent of treatment level, a zig-zag pattern with respect to consecutive changes of pCO₂ was applied, with ambient pCO₂ controls at the start and end of the experiments (Figs 1 and 2).

Net calcification (G), respiration (R) and total organic carbon (TOC) dynamics were determined simultaneously from closed-system incubation vials (350 ml). Prior to closed-system incubation, filtered bulk seawater (1 µm micron bags) with respective pCO₂ levels was prepared (T₀) and distributed into four incubation vials with one control (no coral) and three vials with coral samples. The vials were placed in a water bath (13±0.1°C, mean±s.d.) equipped with submersible magnetic stirrers and a magnet, below a grid on which corals were placed. Coral samples and the control were incubated for an average duration of 17.9±0.6 h (mean±s.d., N=45). Respiration rates were determined using optodes (Presens OXY-4 mini) equipped with polymer optical fibres (POF, Presens) and small sensors (5 mm diameter) placed inside the incubation vials. Respiration rates were determined from the depletion of O₂ during the incubation period, where oxygen consumption curves (% O₂) were established at 30 s intervals and respiration rates are given as µmol O₂ g⁻¹ skeletal dry mass h⁻¹. An incubation time of 18 h was used, which was long enough to detect measurable changes in total alkalinity (A_T), necessary to determine calcification rates, but did not generate hypoxic conditions. Oxygen saturation averaged 89.5±9.8% (mean±s.d., N=121) at the end of the incubation.

From the bulk seawater, control and coral incubation, 125 ml samples were taken to determine A_T and total dissolved inorganic carbon (C_T) as described by Maier et al. (2012). Additionally, pH on the total scale (pH_T) and salinity were determined from samples of bulk seawater using a pH meter (Metrohm, 826 pH mobile) equipped with a glass electrode (Metrohm, electrode plus) and a conductivity meter (Seven Easy, S30). Parameters of the carbonate chemistry were calculated using A_T, C_T or A_T and pH_T, an average salinity of 38.21±0.26 (mean±s.d., N=45) and a temperature of 13.6±0.16°C (mean±s.d., N=45) using the R software package seacarb (v2.4, http://CRAN.R-project.org/package=seacarb). The carbonate chemistry of the last coral incubation at ambient pCO₂ was not determined because of missing C_T samples, but it was assumed that the offset of parameters of the carbonate chemistry between the control and coral incubation was similar to that of the first ambient pCO₂ (Fig. 1).

TOC was determined from 20 ml sub-samples taken with a sterile syringe pre-rinsed with 5 ml sample water and transferred to glass vials, cleaned overnight in 10% hydrochloric acid, rinsed 3 times with Milli-Q water and combusted at 450°C for 4 h. Samples were acidified with 20 µl of 85% phosphoric acid and stored at 4°C pending analysis. Analysis was conducted by high catalytic oxidation with a Shimadzu TOC-VCPH analyser [maximum CV <1.5%, i.e. ±1 µmol C l⁻¹, referenced by the consensus reference materials (CRM) of the Hansell Research Laboratory, USA].

E(G) and E(ΔG) (mJ m⁻² h⁻¹) are the energy required for calcification [E(G)] and up-regulation of calicoblastic pH [E(ΔG)], assuming that 1 mol of ATP or 30.5 J are required for the transport of 2 mol of Ca²⁺ to the site of calcification (McConnaughey and Whelan, 1997). For the energy demand of carbon respiration [E(R)], an energy equivalent of 37.56 J mg⁻¹ was used. As RQ values were highly variable (Fig. 4A), a value of 1 was used. This might be a conservative estimation for energy demand with respect to carbon respiration (Fig. 4A). All calculations were based on data normalized to m⁻² h⁻¹ from conversion of mass to surface area as determined for coral fragments used in the two feeding regimes (Table 1).

Physiological parameters were analysed statistically using the software STATISTICA (v7.0, StatSoft Inc. 2004) and illustrated using the software package SigmaPlot^® (v6.0, SPSS Inc. 2000). The experimental design is a repeated measures design with consecutive changes of pCO₂ and two independent groups for feeding; a repeated measures ANOVA was used with pCO₂ level as the repeated variable and HF and LF as the independent factor. Mauchly's test of sphericity was met for calcification (P=0.26), but not for the other variables measured. For post hoc comparisons of calcification, the least-significance difference (LSD) test was used (Table S2). For other physiological parameters (respiration, RQ and TOC), several non-parametric tests were conducted (Mann–Whitney U-test, Friedman test, Kruskal–Wallis ANOVA; Table S3). Additionally, a Wilcoxon matched pairs test with pairs of mean TOC values at T₁–T₅, was used to test the effect of feeding regime on TOC dynamics. A paired t-test was used for the comparison of parameters of food uptake (feeding rates and POC:PON ratio) conducted 5 times throughout experiments between HF and LF. Values are given as mean±s.e.m.

The feeding rate (i_F; Eqn 4) was determined separately for the HF and LF group and was on average 0.020±0.001 and 0.026±0.002, respectively. This difference in i_F was significant (paired t-test, T=3.452, P=0.026, N=5) and resulted in relative capture rates of 38.6±2.7 and 48.0±2.1% day⁻¹ of food provided for the HF and LF group, respectively (paired t-test for the five feeding assessments, T=4.187, P=0.013, N=5). Mean hourly uptake of POC, taking into account dilution by flow through, prey uptake and weekly feeding times, was on average 0.775±0.195 and 0.195±0.044 µmol g⁻¹ skeletal dry mass h⁻¹ for the HF and LF group, respectively (Table 1; Fig. S2). The PON uptake was 0.139±0.047 and 0.038±0.008 µmol g⁻¹ skeletal dry mass h⁻¹, respectively (Table 1). The ratio of POC:PON taken up by M. oculata was higher for the HF compared with the LF group, with values of 6.37±0.67 and 5.09±0.12, respectively. However, this difference was statistically not significant (paired t-test, T=−1.76, P=0.152, N=5).

Subsequent 2-weekly treatment levels (T₁–T₅) for pCO₂ were 438±19, 1638±62, 805±51, 1725±51 and 453±21 µatm. pCO₂ increased during the closed-system incubations to reach 659±30, 2784±189, 1440±95 and 2290±133 µatm, respectively, at the end of the incubations. The carbonate chemistry of the last incubation at ambient pCO₂ (T₅) was not determined as a consequence of missing values for C_T or pH_T, but presumably had a similar offset to the first treatment (T₁) at ambient pCO₂. This resulted in Ω_a levels that were well above 1 (2.0–2.4), around 1 or well below 1 (0.6–0.8) (Fig. 2A).

Mean rates of net calcification (G) varied between −0.039±0.029 and 0.108±0.039 µmol g⁻¹ skeletal mass h⁻¹ depending on pCO₂ level and feeding regime (Fig. 2B). Calcification rates of M. oculata as function of Ω_a followed a logarithmic trend and the threshold between calcification and dissolution was 0.92 for Ω_a (Fig. 3; R=0.763, P_one-tailed=0.01, N=10, independent of feeding regime). For the HF and LF regime, the calculated Ω_a threshold was 0.82 and 1.07, respectively (Fig. 3; R=0.736, P_one-tailed=0.07, N=5 and R=0.866, P_one-tailed=0.03, N=5). There was a significant effect of both feeding regime and pCO₂, but no interaction between the two factors (repeated measures ANOVA, feeding regime: P=0.037, pCO₂ as repeated factor: P<0.001 and pCO₂×feeding regime: P=0.474). Post hoc comparisons revealed that there was only an effect of feeding on calcification for the first incubation at ambient pCO₂ (LSD comparison, P=0.024; Table S2). In general, net calcification was positive and not significantly different between ambient pCO₂ levels (T₁ and T₅) and the pCO₂ level at 800 µatm (T₃), while at very high pCO₂ (1638 and 1725), dissolution set in and calcification was significantly different from that in ambient conditions or 800 µatm (LSD comparison, P<0.05; Table S2). However, post hoc comparison within the LF group revealed a less distinct pattern, with calcification rates being only significantly different between a pCO₂ of 1638 µatm (T₂) and 800 µatm (T₃) and between 1638 µatm (T₂) and the last ambient condition (T₅) (LSD comparison, P<0.05; Table S2).

There was no treatment effect on respiration (Fig. 2C; Table S3). The overall mean of respiration (data pooled) was 0.431±0.056 µmol g⁻¹ skeletal mass h⁻¹ (N=25). The mean respiration of the HF group was higher than that of the LF group, with 0.481±0.080 µmol g⁻¹ skeletal mass h⁻¹ (N=13) and 0.376±0.080 µmol g⁻¹ skeletal mass h⁻¹ (N=12). This difference was statistically not significant (Table S3).

The RQ was on average 1.49±0.14 (N=53; Fig. 4A). There was no statistically significant effect of feeding or pCO₂, or a significant interaction between the two factors on RQ (Table S3). However, the relative carbon demand from CO₂ respired to POC taken up was significantly different between feeding regimes (Mann–Whitney U-test; Table S3), with a range from 60% to 83% for the HF regime and 180% to 304% for the LF regime (Fig. 4B).

There was a high variability in TOC dynamics, ranging from −4.16 to 55.46 µmol TOC g⁻¹ skeletal mass h⁻¹, with an overall mean of 4.47±1.13 µmol TOC g⁻¹ skeletal mass h⁻¹ (N=71). Because of high variability and several very high outliers, the mean TOC values varied between 1.25±1.49 µmol g⁻¹ skeletal mass h⁻¹ and 10.93±6.74 µmol g⁻¹ skeletal mass h⁻¹ (Fig. 2D). The highest mean TOC excretion was observed for the first ambient pCO₂, with values of 7.12±4.33 and 10.93±6.74 µmol g⁻¹ skeletal mass h⁻¹ for the HF and LF group, respectively. TOC excretion of consecutive pCO₂ treatments was well below that of the initial ambient incubation. Also, median TOC values were generally higher for the LF than for the HF regime, with 1.89 and −0.12 µmol g⁻¹ skeletal mass h⁻¹, suggesting that corals of the LF regime excreted more TOC than those of the HF regime. In fact, the slightly negative median value for the HF group indicates that most corals of the HF regime took up TOC. No effect of feeding or pCO₂ was detected using non-parametric tests (Table S3). However, by conducting a Wilcoxon matched pairs test with the means of TOC at T₁–T₅, a significant difference in TOC was revealed between feeding regimes (N=5, T=0; Z=2.02, P=0.043).

The present study confirms earlier findings that net rates of calcification and respiration of M. oculata remain relatively unchanged over a large pCO₂ gradient ranging between ambient (∼ 400 µatm) and pCO₂ levels projected until the end of the century (Maier et al., 2013a, 2012). A similar resistance to ocean acidification has also been reported for other cold-water scleractinian species (Form and Riebesell, 2012; Jantzen et al., 2013; Lunden et al., 2013; Maier et al., 2013a, 2012; McCulloch et al., 2012; Movilla et al., 2014; Thresher et al., 2011; Hennige et al., 2015; Gori et al., 2016 preprint), while only a few reports are available where CWC were impacted by increased pCO₂ (Form and Riebesell, 2012; Hennige et al., 2014; Maier et al., 2009). This is, however, the first experimental study, where a threshold level with respect to pCO₂ or Ω_a was reached and skeletal dissolution occurred. After the 2 week exposure to corrosive pCO₂ levels, calcification rates returned to normal when M. oculata were subjected to pCO₂ levels of either ∼800 µatm or ambient. This indicates a strong recovery potential from dissolution due to ocean acidification as had been shown for zooxanthellate corals that had been completely decalcified (Fine and Tchernov, 2007). Higher food uptake had no apparent mitigating effect with respect to the negative impact on calcification at very high pCO₂. Other key physiological parameters – respiration, TOC excretion (or uptake) and RQ – were not significantly related to changes in pCO₂ or feeding regime, which indicates that these metabolic key functions are less sensitive to ocean acidification even when elevated pCO₂ triggered dissolution.

Mean C uptake (0.78 and 0.32 µmol g⁻¹ skeletal mass h⁻¹ for HF and LF, respectively) was within the lower range of that reported by Purser et al. (2010) for Lophelia pertusa (0.75–2.31 µmol g⁻¹ skeletal mass h⁻¹) and well below that reported by Tsounis et al. (2010) for M. oculata (73 µmol C g⁻¹ skeletal mass h⁻¹). While corals of the LF regime took up relatively more prey than those of the HF regime, both groups took up a higher percentage of prey with increasing prey concentration (Fig. S1) and, contrary to the findings by Purser et al. (2010), no saturation in food uptake was observed. This might indicate that corals, even in the HF regimes, were still below their capture capacity or even starving. However, while the carbon demand for respiration revealed that corals of the LF could not meet the carbon demand by food uptake and were thus slightly starving, corals of the HF regime took up enough prey to meet their respiratory carbon demand (Fig. 4B).

The findings of the present and previous studies hint at a dynamic switch between excretion and uptake of TOC (Houlbrèque et al., 2004; Tremblay et al., 2012). Strikingly, corals of the HF group revealed mostly negative TOC values (55%, N=40), actually indicating net TOC uptake. This is surprising as TOC uptake has so far only been observed as a consequence of 3 weeks of starvation in the CWC D. dianthus (Naumann et al., 2011) and rarely in tropical corals (Houlbrèque et al., 2004; Naumann et al., 2010). Earlier studies reported that high amounts of TOC, dissolved organic carbon (DOC) or POC were excreted by the CWCs M. oculata, L. pertusa (Maier et al., 2011; Naumann et al., 2013b; Wild et al., 2008; Zetsche et al., 2016) and D. dianthus (Naumann et al., 2011). For North Atlantic and Mediterranean M. oculata, mean DOC release was 2.57 and 0.53 µmol g⁻¹ skeletal mass h⁻¹, respectively (Maier et al., 2011; Naumann et al., 2013b), with the latter corresponding to the overall median values of the present study (0.51 µmol g⁻¹ skeletal mass h⁻¹). In the present study, TOC dynamics were measured during closed-system incubation, and no food was provided during that time. As the incubation water had been filtered through 1 µm micron bags, the TOC fraction consisted mainly of DOM, femtoplankton and picoplankton. This is a size fraction that is unlikely to be taken up by active capture, and is more likely to be obtained through passive trapping in the mucus layer.

The composition and function of coral mucus can greatly vary (Brown and Bythell, 2005; Ducklow and Mitchell, 1979), and corals can reverse their ciliate movement, changing the direction of mucus flow either inward or outward, to or from the coelenteron. This is an important mechanism which allows a switch from food ingestion to a removal of undesired components, such as food remnants, metabolites with high N content (Wild et al., 2008), sediment or other particles (Larsson et al., 2013; Larsson and Purser, 2011; Zetsche et al., 2016). Also, mucus has an important role in fuelling bacterial growth in the coelenteron (Herndl and Velimirov, 1986; Weinbauer et al., 2012) and the close vicinity of the coral environment (Wild et al., 2004a,b, 2008, 2009, 2005). In the light of the diverse role of mucus, a frequent switch between TOC excretion and uptake might be related to the direction of mucus flow transporting mucus-trapped TOC into (uptake) or out of the coelenteron (excretion), where for corals of the HF the more frequently observed TOC uptake might be a consequence of the more frequent food uptake and inward-directed ciliate and mucus movement.

During the first incubation at ambient pCO₂, respiration and calcification were concomitantly elevated in the HF compared with the LF regime. This might indicate that the two processes are coupled and both depend on the energy provided as suggested earlier (Anthony et al., 2002; Naumann et al., 2011). However, at consecutive pCO₂ levels, net calcification was similar for the HF and LF regimes, while respiration showed opposing trends with respect to pCO₂ and feeding regime: (1) the consecutive 2 week changes in pCO₂ (or Ω_a) induced an inverse pattern for the two feeding regimes: respiration of the HF group increased or decreased with increasing or decreasing Ω_a, while the opposite was true for the LF regime (Fig. 2A,C); and (2) respiration was about twice as high in the HF as in the LF group at ambient pCO₂ and at a pCO₂ of ∼800 µatm, while it was identical for the two feeding regimes at the very high pCO₂ levels when skeletal dissolution set in (Fig. 2A–C).

So far, three studies have addressed the effect of nutrition in relation to calcification and respiration, which revealed several features. Starvation triggers a continuous decline in respiration rates of L. pertusa and D. dianthus (Larsson et al., 2013; Naumann et al., 2013a). Starvation of D. dianthus caused an initial and significant decrease in calcification after 1 week, while starvation for another 2 weeks did not result in further significant reduction of calcification rates (Naumann et al., 2013b). When L. pertusa was provided with various concentrations of food over 15 weeks, only small and non-significant changes in respiration were measured, and no effect on calcification occurred, despite the fact that food supply varied by a factor of 7.5 (Larsson et al., 2013). Also, for M. oculata (this study) the HF regime only affected the first measurements on calcification, which were taken 2 weeks after the respective feeding regimes had been applied, while during later measurements (4–10 weeks), calcification was independent of feeding regime. This is astonishing as it appears that, while the carbon supply was above the carbon demand in the HF group, the respiratory carbon demand of the LF was not met by the uptake of A. salina (Fig. 4B). These different findings therefore suggest that energy allocation to calcification and energy allocation to respiration are largely independent of each other and that these two processes depend not only on the amount of food provided but also on the time scale, the time of last feeding and the regularity of food provided.

The RQ values reported in the present paper, the first in CWCs, are around 1.5. Earlier measurements of RQ in zooxanthellate corals ranged between 0.7 and 1.0 (Gattuso and Jaubert, 1990), which is also the range of values for aerobic metabolism. The high values found in the present study might consequently be due to anaerobic glycolysis driving calcification as suggested by Wooldridge (2013), via the glyoxylate cycle, which was shown to be present in cnidarians (Kondrashov et al., 2006). As RQ is calculated as the number of moles of CO₂ released per mole of O₂ consumed, the lack of O₂ consumption during anaerobic conditions could lead to RQ values higher than 1. Based on this rationale, the contribution of anaerobic oxidation to calcification was calculated to range between 30% and 85%. A complete oxygen depletion close to the calicoblastic layer had been shown by micro-electrode profiles during the dark in zooxanthellate corals (Kühl et al., 1995), as well as for the CWC D. dianthus (F. Bils, M. A. Peck and C.M., unpublished observation). This, along with the high RQ values of the present study and the molecular evidence for an anaerobic glyoxylate cycle in Cnidaria (Kondrashov et al., 2006) thus support the hypothesis that anaerobic processes might indeed be involved in calcification.

The large range with respect to the contribution of anaerobic oxidation, but also the observed large range in calcification rates could be explained by a sporadic leakage of ambient seawater into the calicoblast as suggested earlier (Adkins et al., 2003). While oxygen-rich ambient seawater close to the calicoblastic layer would favour aerobic metabolism, the lower ambient pH or Ω_a could slow calcification down. The negative impact on calcification during replenishment of the calicoblastic layer with oxygen-rich, ambient seawater might be partly counteracted by the increased contribution of aerobic oxidation, which is energetically more efficient (Wijgerde et al., 2014). A gradual restoration of gradients of carbonate chemistry between the calicoblast and seawater with a continuous O₂ depletion and increase of the pH and Ω_a in the calicoblast would then gradually favour higher rates of calcification accompanied by an increasing contribution of anaerobic glycolysis. This process would thus act like a see-saw, counterbalancing negative effects of a lower energy efficiency with more favourable conditions for calcification. This hypothesis requires further testing.

This is the first experimental study where a threshold with respect to pCO₂ for CWC calcification was reached and skeletal dissolution set in (Fig. 3). So far, experiments have shown that CWCs can maintain calcification constant over a large pCO₂ gradient (Form and Riebesell, 2012; Hennige et al., 2014; Maier et al., 2013b, 2012; Movilla et al., 2014) and maintain positive skeletal growth even at conditions under-saturated with respect to Ω_a (Form and Riebesell, 2012; Jantzen et al., 2013; Maier et al., 2009; Rodolfo-Metalpa et al., 2015; Thresher et al., 2011). Despite the two thresholds determined for the two feeding regimes, the present study revealed no significant difference in calcification with respect to feeding except at the first ambient pCO₂, which forced the two regression functions to higher and lower Ω_a threshold values. Still, the given range in thresholds between an Ω_a of 0.8 and 1.1 reflect those observed in situ with respect to the distributional limitation of CWCs (Jantzen et al., 2013; Lunden et al., 2013; Thresher et al., 2011). This indicates that the projected decline in ocean pH might be less detrimental than initially thought (Guinotte et al., 2006; Turley et al., 2007), although slowing down of net calcification has been reported under various conditions (Form and Riebesell, 2012; Maier et al., 2009, 2012; Movilla et al., 2014). Recent studies revealed that in the calicoblast the pH up-regulation can greatly differ on a micro-scale (Movilla et al., 2014; Raddatz et al., 2014), which may further explain the observed variable results and threshold values for calcification.

There is proxy evidence that the pH of the calicoblast layer is up-regulated (McCulloch et al., 2012), which explains why CWCs are able to calcify relatively fast despite the unfavourable carbonate chemistry of ambient seawater. The up-regulation is an energy-requiring process via active ion transport (Allemand et al., 2004; McCulloch et al., 2012). Hence, when the gradient increases, corals have to invest more energy in calcification to counteract increasingly unfavourable conditions (Cohen and Holcomb, 2009). Therefore, one can expect that more food should favour faster calcification. In the present study, this was only the case after initial changes in food availability (2 weeks), where at the first ambient pCO₂ more food resulted in faster calcification. However, during the subsequent stages (>4 weeks), calcification was independent of feeding. Calculations (Eqns 16–18) revealed that less than 0.5% of energy would be required for calcification in relation to overall respiration in M. oculata (Table 2). Also, for L. pertusa only 3–5% of the available energy is allocated to calcification in comparison to respiration (Larsson et al., 2013), and metabolic carbon contribution to calcification was 1% in relation to that of the tissue pool and respiration for the CWC L. pertusa (Mueller et al., 2014). The relatively low energy requirement for up-regulation of Ω_a in the calcifying fluid and the maintenance of relatively high calcification rates might explain the observed resistance to ocean acidification. An increasing gradient of pCO₂ between the calicoblastic space and ambient seawater might consequently be counteracted by allocating relatively small amounts of energy to sustain calcification or delay dissolution. Despite the apparently low surplus energy required to sustain calcification over a large pCO₂ range, the question arises whether this additional allocation of energy to calcification will hamper other physiological functions, such as tissue regeneration (Horwitz and Fine, 2014), reproduction and recruitment as demonstrated for tropical corals (Albright, 2011; Albright and Langdon, 2011; Albright et al., 2008; Crook et al., 2013; Nakamura et al., 2011).

Dissolution rates were calculated using equations established at 25°C, probably over-estimating the calculated dissolution rates and thus the energy allocated to up-regulate calicoblastic pH, as dissolution slows down with decreasing temperature (McCulloch et al., 2012). The energy demand (Table 2) to maintain calcification constant at 800 µatm was in general only 1% of that of respiratory metabolism; however, this is 2 times that required for calcification at ambient pCO₂. Once the Ω_a drops below 1 and dissolution occurs, the energy required to slow down or counteract dissolution is calculated to be 0.5 to 3 times that at ambient. This may already constitute a limitation with respect to the energy that CWCs can allocate to calcification. However, there is no indication that any surplus energy is allocated to calcification in the HF group to counteract the negative effects at very high pCO₂. Ultimately, this means that there is a limit with respect to the capacity of a coral to up-regulate its calicoblastic pH, possibly due to a coral’s capacity to shelter the calicoblastic space from ambient seawater (Ries, 2011a,b), and surplus energy can consequently not mitigate the detrimental effect on calcification at very high pCO₂ levels.

We would like to thank M. Taviani, captain and crew of R/V Urania, with special thanks to Maria Berzunza-Sanchez, Cecile Rottier and M. Naumann for coral sampling and maintenance prior to experiments.