Experimental oysters (56.37 ± 8.20 mm shell length, 29.67 ± 3.91 g wet weight) and kelps (70–80 cm length) were collected on 7 May 2017 from a large-scale commercial aquaculture area (37°03′52.17″N, 122°33′11.54″E) in Sanggou Bay, Yellow Sea, China. All specimens were transported on ice to a small semi-enclosed harbor (37°02′14.71″N, 122°33′02.09″E; area 7900 m²; mean depth 1.5 m). The oysters were cleaned and maintained in flowing seawater in in situ mesocosms at ambient conditions (temperature ~19.0°C, salinity ~32, and seawater pH_NBS ~8.10) for one week before the start of the experiment. One hundred and thirty healthy oysters were selected and individually labeled with a plastic-laminated number on the upper valve.

Experiments were conducted using in situ mesocosms with six treatments: three biotic treatments (namely “oyster alone”, “kelp alone,” and “integrated aquaculture”), crossed with two pCO₂ treatments (ambient, ~500 μatm and elevated, ~900 μatm). Each of the six treatments was replicated three times, for a total of 18 mesocosms. The mesocosms were suspended from a floating platform. Each sealed mesocosm was made of translucent polyethylene and was cylinder-shaped (0.5 m × 1.2 m, diameter × height) (Figure 1). Inlet and outlet valves allow for the control of seawater flow through each mesocosm. The top of each mesocosm was immersed to a depth of ~30 cm with the inlet and outlet pipes (Φ ~3 cm) extending ~50 cm above the water surface. The inlet pipe extended to the bottom of each mesocosm, allowing water mixing and preventing gradients from forming within the mesocosm.

Elevated pCO₂ treatment was controlled by enriching natural seawater with CO₂ (from a CO₂ gas cylinder) in a submerged 1,000 L reservoir tank suspended from the same floating platform as the mesocosms (modified from Rastrick et al., 2014; Jiang et al., 2019; Jiang et al., 2021). The flow of pure CO₂ to the reservoir tank was controlled via a pH controller (Aqua Medic GmbH, Germany) by switching on or off a valve when pH readings in the tank deviated from the predetermined set-points by ±0.05 pH units. The acidified seawater was then pumped to each elevated pCO₂ mesocosm via flow rate controllers (60 L h⁻¹). As the CO₂-adjusted seawater was pumped out of the reservoir tank, it was replaced with natural ambient seawater using an automatic pump triggered by changes in volume within the tank. This gave a seawater residence time in the reservoir tank of just under an hour. Pumps in the reservoir tank prevented the formation of CO₂ or temperature gradients, and the ambient temperature was maintained due to the submersion of the reservoir tank and the mesocosms.

Each “oyster alone” treatment contained 8 oysters hugged along a rope tied at the top of the mesocosm to simulate farm conditions. The kelp alone treatment contained 1 blade with holdfasts (70–80 cm length) that were tied at the top of the mesocosm to keep the orientation vertical as in the farm environment. An integrated aquaculture treatment contained both oysters (n = 8) and kelp (n = 1) hung in the same way. The oyster number and the ratio of oyster and kelp were determined based on our modeling research which was published in Lin et al. (2020). The experiment started on the 14 May 2017 (experiment day 1) and continued until 12 June 2017.

During the 30-day experiment, seawater temperature, salinity, and dissolved oxygen concentration (DO) in each mesocosm were measured three times a day. The pH in the reservoir tank was also measured three times a day, and the pH in the inflow and outflow of each mesocosm was measured only at 12:00 daily. Seawater temperature, salinity, and dissolved oxygen concentration (DO) were measured using a Vernier’s LabQuest^® 2 (Vernier Software & Technology, USA) equipped with the corresponding probes. The pH level was measured using a commercial combination electrode (Ross type, Orion) calibrated daily using a 3-point calibration procedure on the U.S. National Bureau of Standards (NBS) scale with a precision of ± 0.001 pH units. The concentration of chlorophyll-a in the ambient seawater was continuously monitored using the ACLW-RS chlorophyll turbidity temperature sensor (ALEC Electronics Co., Ltd., Japan) with a precision of ±0.1 μg L⁻¹. Total alkalinity (A_T) was analysed weekly via a Metrohm 848 Titrino plus automatic titrator (Metrohm, USA) on 100 ml of GF/F filtered seawater with an accuracy of ±5 mmol L⁻¹. Carbonate system parameters, such as total dissolved inorganic carbon, aqueous partial pressure of CO₂ (pCO₂), the CaCO₃ saturation state for calcite (Ω_cal), and aragonite (Ω_arag), were calculated from pH_NBS and A_T by the CO2SYS Package (Pierrot et al., 2006) using the dissociation constants of Mehrbach et al. (1973) and refit by Dickson and Millero (1987). No mortality of oysters was observed in both ambient and elevated pCO₂ treatments throughout the experiment.

The physiological measurements of oysters under different treatments were performed on 12 June 2017 (experiment day 30) by flow-through system after Jiang et al. (2017). Feeding experiments were performed by transferring 9 labeled individual oysters (3 randomly selected from each of the 3 replicate mesocosms per treatment) to individual flow-through chambers which were continuously supplied with the same water as supplied to the respective treatment mesocosms. The flow rates of each chamber were adjusted to 180–200 ml min⁻¹ and determined by simultaneously measuring the volume of water collected from each outflow. Three identical chambers without oysters served as the control. The oysters were left undisturbed in chambers for 1 h to resume feeding before sampling water from the outlet of the chambers. The particle number in each chamber was measured by PAMAS S4031GO particle counter equipped with a tube of 50 μm aperture and set to count all particles >2 μm diameter in a 0.5 ml sub-sample. Three replicate counts are made on each sample and the mean calculated. This measurement was repeated at 60 min intervals over a period of 3 h. Clearance rate (CR) is then calculated as follows:

where C₁ is the particulate concentration in control chamber (particle number ml⁻¹), and C₀ is the particulate concentration from each experimental chamber (particle number ml⁻¹).

The absorption efficiency (AE) was determined using the following equation (Conover, 1966):

Where F represents the ash-free dry weight: dry weight ratio of food, and E represents the ash-free dry weight: dry weight ratio of the feces. Approximately 4 L of food samples were taken from the seawater being supplied to each treatment, and 4 sub-samples were filtered with pre-ashed and weighted glass fiber filters of 1.2 μm (Whatman GF/C 47mm). These filters were rinsed with isotonic ammonium formate (3%) to remove salts and dried to a constant weight (~48 h) at 65°C and weighed (W₆₀), then ashed at 450°C for 6 h to get the ash-free dry weight W_450. The particulate organic matter concentration (POM, mg L⁻¹) is calculated by POM = (W₆₀ − W₄₅₀)/v. Each oyster was placed in the individual feces collector at the respective treatment conditions for ~24 h after the CR measurements. The feces collector (Φ ~10 cm) was made of the casing pipe equipped with an 80 μm mesh screen (after, Rastrick et al., 2018a). Feces were collected by a pipette and filtered with the organic content of the feces measured as described above.

The respiration rate (MO₂) of individual oysters was measured in ~200 ml stop-flow glass respirometers held in a water bath at the same ambient temperature as the mesocosms and supplied with the same amount of water. Twelve respirometers were run simultaneously, with three chambers without oysters being used as controls. The inflow of each respirometer came from corresponding treatments, and constant flow rates were maintained by a multichannel peristaltic pump. Thirty minutes were allowed for the oysters to recover from handling stress, open and resume pumping, then the flow of water was stopped and MO₂ was measured over the next 90 min. The oxygen concentration in each respirometer was not allowed to drop below 70% saturation throughout the experiment. The rate of decline in oxygen saturation in each respirometer was measured by a calibrated optical oxygen probe connected to a Vernier’s LabQuest^® 2 (Vernier Software & Technology, USA). The initial and final oxygen concentrations in each respirometer were measured. The respiration rate (MO₂) was then calculated using the following equation:

Where t₀ and t₁ represent the initial and finish times (min) of the measurement period, C_t represents the concentration of oxygen in the water (μmol O₂ L⁻¹) at time t; V_r represents the volume of the respirometer minus the animal.

After MO₂ measurement, the ammonia excretion rate (ER) of the same group of oysters was measured. Water samples were taken from each respirometer, filtered by 0.45 μm cellulose acetate membrane filters, and then analysed for ammonia using the phenol-hypochlorite method. ER was calculated using the following equation:

Where C_test and C_control represent the ammonia concentration in the sample and control, respectively, V_r represents the volume of the respirometer minus the animal; t is the exposure time.

After all the physiological measurements, the oyster tissues were dissected from the shell and dried to a constant weight at 65°C. The CR, MO₂, and ER were then corrected to a ‘standard body size’ of 1 g of dry weight mass-specific rates using the following equation: Y_s = (W_s/W_e)^b × Y_e, where Y_s is the physiological rate for an animal of standard weight, W_s is the standard weight (1 g), W_e is the observed weight of the animal (g), Y_e is the uncorrected (measured) physiological rate, and b is the weight exponent for the physiological rate function (b = 0.67). Each physiological rate was then converted into energy equivalents (J h⁻¹) and used to calculate the energy available for scope for growth (SfG). Values for MO₂ were transformed into J h⁻¹ using a conversion factor of 0.456 J μmol⁻¹ O₂ (Gnaiger, 1983), ER was converted to J h⁻¹ using the conversion factor: 1 μmol NH₄-N h⁻¹ = 0.349 J h⁻¹. SfG (J h⁻¹ g⁻¹) was calculated by SfG = A_b − R − U, where A_b is the total absorbed energy (J h⁻¹ g⁻¹), R is the energy lost in respiration (J h⁻¹ g⁻¹), and U is the energy lost in ammonia excretion (J h⁻¹ g⁻¹). A_b = CR (L g⁻¹ h⁻¹) × (mg POM L⁻¹) × AE (%). The POM concentration was converted to joules using a conversion factor of 23 J mg⁻¹ ash-free dry weight (Widdows and Johnson, 1988).

The wet weight-specific growth rate (SGR, % d⁻¹) of each oyster was measured as follows: SGR = (ln(W_t/W₀)/t) × 100%, where W_t and W₀ are the wet weight (g) of the oyster at time t and at time 0, respectively, and t is the sampling time (d).

Statistical analyses were conducted using the software SPSS 22.0 for Windows. The general linear mixed model (GLMM) was used to test the effects of culture treatments (oyster alone and integrated aquaculture) or pCO₂ treatments (ambient, ~500 μatm and elevated, ~900 μatm) on physiological parameters with a mesocosm as a random variable nested within fixed factors (culture or pCO₂ treatments). This design considers that 3 replicate mesocosms in each treatment were supplied with the same seawater, and therefore, mesocosms may not be considered true replicates. Differences between culture or pCO₂ treatments were further tested by F-tests with a significance threshold of α = 0.05. All data are represented as means ± SD.

The daily variations in temperature, salinity, and pH during the experiment are shown in Figure 2. The ambient temperature and salinity were 19.08 ± 1.27°C and 32.86 ± 0.23, respectively. The controlled injection of CO₂ separates the treatments successfully (pH_NBS ~7.83 in elevated pCO₂ treatments and 8.07 ± 0.03 in the ambient treatment) and keeps them relatively stable during the 30 days of exposure. Seawater chemistry variables over the 30-day experimental period for each pCO₂ treatment are summarized in Table 1. There was no significant difference in temperature, salinity, chlorophyll a, DIC, Ω_cal, Ω_arag, and HCO 3 − between treatments (P >0.05). The saturation percentage of dissolved oxygen in the exposure containers consistently exceeded 100% and was similar across treatments during the experiment. There was a significant difference between pH_NBS and pCO₂ between treatments (P <0.05).

The average change in daytime pH_NBS between the inflow and outflow of the mesocosms for the different treatments during the 30 days of exposure is shown in Figure 3. Results showed that the seaweed alone treatment increased the daytime pH_NBS by ~0.08 units, while oyster alone decreased ~0.05 units compared to the inflow under ambient conditions. The magnitude of change in daytime pH_NBS was significantly greater in kelp alone treatment under elevated pCO₂ condition than the ambient pCO₂ condition. The kelp alone treatment increased daytime pH_NBS by ~0.11 units, while the oyster and kelp integrated aquaculture treatment increased the daytime pH_NBS by ~0.06 units.

A significant effect of elevated pCO₂ on the clearance rate of oysters was observed. After 30 days of exposure, the clearance rate of oysters was significantly reduced in the elevated pCO₂ treatments and in the absence of kelp, as compared with ambient pCO₂ treatments (P <0.05) (Figure 4A). The presence of kelp in elevated pCO₂ treatment increased the clearance rate of oysters compared with oysters alone treated at the same elevated pCO₂ level (P <0.01). Additionally, no significant difference in CR was found between oysters incubated with seaweed at elevated pCO₂ and oysters incubated at ambient pCO₂ levels (P >0.05).

Absorption efficiency varied between 41 and 72% and was not affected significantly by elevated pCO₂ during the experiment. The absorption efficiency was lower in the oyster alone treatment at elevated pCO₂ treatment but showed no significant difference from the integrated aquaculture treatment (P >0.05) (Figure 4B).

After 30 days of exposure, the respiration rate of oysters was significantly higher in the absence of kelp at elevated pCO₂ compared with oysters in ambient pCO₂ treatments (P <0.05). The presence of kelp in elevated pCO₂ treatment lowered the respiration rate and ammonia excretion rate of oysters compared to ambient conditions, but no significant difference was found (P >0.05). Compared with all other treatments, the respiration rate and ammonia excretion rate were significantly increased in the oyster alone treatment (P <0.05) (Figure 5).

Elevated pCO₂ exposure significantly reduced the SfG of C. gigas when unintegrated with kelp (Figure 6A). Compared to other treatments, SfG was negative (−9.45 J g⁻¹ h⁻¹) in the elevated pCO₂ treatment in the absence of kelp (P <0.01). However, in the elevated pCO₂ treatment in the presence of kelp, SfG (35.82 J g⁻¹ h⁻¹) was not significantly different to that of oysters in ambient pCO₂ treatments (P >0.05). Moreover, after 30 days of exposure, the wet weight-specific growth rates of C. gigas were also observed to be dramatically reduced in the oyster alone treatment compared with the integrated aquaculture treatment at elevated pCO₂ levels (Figure 6B).