CO2-induced pH reduction increases physiological toxicity of nano-TiO2 in the mussel Mytilus coruscus

The increasing usage of nanoparticles has caused their considerable release into the aquatic environment. Meanwhile, anthropogenic CO2 emissions have caused a reduction of seawater pH. However, their combined effects on marine species have not been experimentally evaluated. This study estimated the physiological toxicity of nano-TiO2 in the mussel Mytilus coruscus under high pCO2 (2500–2600 μatm). We found that respiration rate (RR), food absorption efficiency (AE), clearance rate (CR), scope for growth (SFG) and O:N ratio were significantly reduced by nano-TiO2, whereas faecal organic weight rate and ammonia excretion rate (ER) were increased under nano-TiO2 conditions. High pCO2 exerted lower effects on CR, RR, ER and O:N ratio than nano-TiO2. Despite this, significant interactions of CO2-induced pH change and nano-TiO2 were found in RR, ER and O:N ratio. PCA showed close relationships among most test parameters, i.e., RR, CR, AE, SFG and O:N ratio. The normal physiological responses were strongly correlated to a positive SFG with normal pH and no/low nano-TiO2 conditions. Our results indicate that physiological functions of M. coruscus are more severely impaired by the combination of nano-TiO2 and high pCO2.

Seawater chemistry. Salinity of the water consistently was around 25‰, and dissolved oxygen in the exposure tanks consistently was maintained at about 7.0 mg l −1 during the experiment. Total alkalinity ranged from 2136 to 2276 μ mol kg −1 , and CO 2 levels were maintained at ca. 350 μ atm and ca. 2600 μ atm in the normal (pH 8.1) and low pH (pH 7.3) treatments, respectively. The nano-TiO 2 concentrations in seawater determined by atomic absorption spectrophotometer in different pH conditions, were in average 1.66 ± 0.18 mg l −1 and 7.88 ± 0.55 mg l −1 (pH 8.1) and 1.52 ± 0.14 mg l −1 and 7.37 ± 0.75 mg l −1 (pH 7.3) for the nominal 2.5 (low) and 10 mg l −1 (high) exposure concentration, respectively, and no significant effects of pH on nano-TiO 2 concentrations were found. The detailed seawater chemistry in each treatment was summarized in Supplementary Table 1 Physiological parameters. Clearance rates were significantly decreased with the increased nano-TiO 2 concentration with the lowest value observed at 10 mg l −1 nano-TiO 2 treatment during the whole experiment at each pH level (Supplementary Table 2; Fig. 1A); pH significantly affected clearance rates at day 1 and 3 under all nano-TiO 2 treatments, with lower values at pH 7.3 than pH 8.1; however, low pH only negatively affected clearance rates at day 7 under nano-TiO 2 10 mg l −1 , and at day 14 under nano-TiO 2 2.5 mg l −1 (Supplementary Table 2; Fig. 1A).
Absorption efficiency significantly decreased with the increased nano-TiO 2 concentration throughout the experiment at each pH level. However, pH showed no significant effect on AE during the whole experiment (Supplementary Table 2; Fig. 1B). Organic weight ratio in feces significantly increased with nano-TiO 2 increment during the whole experiment at each pH level, but no pH effect was found on this parameter (Supplementary Table 2; Fig. 1C).
Respiration rates were significantly influenced by nano-TiO 2 and pH during the entire experiment, and by their interactions at day 7 (Supplementary Table 2; Fig. 2A). Respiration rates decreased with nano-TiO 2 Figure 1. Clearance rate (CR, A), Absorption efficiency (AE, B) and faecal organic weight rate (E, C) of M. coruscus exposed to six treatments for 14 days. The values with different superscripts at each pH are significantly different among three nano-TiO 2 treatments (P < 0.05). The values denoted by an asterisk between two pH groups at each n-TiO 2 concentration are significantly different (P < 0.05).
Scientific RepoRts | 7:40015 | DOI: 10.1038/srep40015 increment during the experiment, with the lowest value observed under 10 mg l −1 nano-TiO 2 condition. In general, low pH reduced the respiration rates, although sometimes there were no significant effects of low pH. At day 1, low pH only reduced respiration rates when nano-TiO 2 was absent; at day 3 and day 14, low pH reduced respiration rates under nano-TiO 2 0 and 2.5 mg l −1 treatments; at day 7, low pH reduced respiration rates under all nano-TiO 2 conditions (Supplementary Table 2; Fig. 2A).
Ammonia excretion rates were significantly influenced by nano-TiO 2 and pH during the entire experiment, and by their interactions at day 1 and 3 (Supplementary Table 2; Fig. 2B). Excretion rates generally increased with nano-TiO 2 increment, with the highest values observed at nano-TiO 2 10 mg l −1 . For the pH effect, at day 1 and 7, excretion rate at pH 7.3 was significantly lower than pH 8.1 when mussels were exposed to nano-TiO 2 10 mg l −1 ; at day 3 and 14, pH 7.3 increased excretion rates of mussels under two nano-TiO 2 treatments, respectively. O:N ratios were significantly influenced by nano-TiO 2 and pH during the experiment, and by their interactions at day 3, 7 and 14 (Supplementary Table 2  . Ammonia excretion rate (ER, A), Respiration rate (RR, B) and O:N ratio (C) of M. coruscus exposed to six treatments for 14 days. The values with different superscripts at pH are significantly different among three nano-TiO 2 treatments (P < 0.05). The values denoted by an asterisk between two pH groups at each nano-TiO 2 concentration are significantly different (P < 0.05).
significantly lower than pH 8.1 when nano-TiO 2 was absent. At day 3, O:N ratio of pH 7.3 was significantly lower than pH 8.1 under two nano-TiO 2 conditions. At day 7, pH 7.3 reduced O:N ratio when nano-TiO 2 was present. At day 14, O:N ratio of pH 7.3 was significantly lower than pH 8.1 when nano-TiO 2 was 10 mg l −1 (Supplementary Table 2; Fig. 2C).
The scope for growth (SFG) values were affected by pH at day 1 and day 3, and by nano-TiO 2 during the entire experiment, but there was no interactive effect of them (Supplementary Table 2; Fig. 3). SFG decreased with nano-TiO 2 increment, and even negative values were observed under the highest nano-TiO 2 concentration. However, pH only affected SFG at day 1 and 3. At day 1, when nano-TiO 2 was absent, SFG of pH 7.3 was lower than pH 8.1. At day 3, SFG of pH 7.3 was significantly lower than pH 8.1 when nano-TiO 2 was 2.5 mg l −1 .
PCA showed that 85.64% of total variance was explained by PC1 (72.41%) and PC2 (13.23%) (Fig. 4). PC1 indicated a clear separation between non-nano-TiO 2 and nano-TiO 2 exposure treatments, showing a high convergence of most physiological parameters, especially high levels of SFG associated with CR and AE under non-nano-TiO 2 treatments. PC2 separated two main experimental periods, i.e., day 1-7 and day 7-14. Generally, the reduced SFG with increased nano-TiO 2 in this study was explained by inhibited physiological activities.   Doyle et al. 7 found measurable concentrations of nano-TiO 2 in the gills of mussel and oyster following exposure 7 . Adsorption of NPs on gill surfaces results in a number of sublethal effects, e.g., gill pathology (such as hyperplasia and edema), respiratory toxicity, oxidative stress and dietary stress 48,49 , which subsequently impair filtration and ingestion of mussels. In the present study, nano-TiO 2 exposure reduced CR, hence the food ingestion rate. The food ingestion rate in the zebra mussels Dreissena polymorpha decreased greatly in nano-TiO 2 exposure media 50 . In our experiment, the mussel valve opening, distinguishing whether M. coruscus was filtering was lower if nano-TiO 2 was present, which was similar to M. edulis exposed to nano-polystyrene 51 . This indicates that M. coruscus was able to recognize nano-TiO 2 and hence reduced its filtration. A decreased filtering activity might cause severe consequences if the mussels were exposed to such concentrations in natural environments. Continuous limitation of CR would impair food intake thereby growth if the mussels were subjected to long term exposure. Some bivalve species have showed different CR responses to CO 2 -induced low pH. Fernández-Reiriz et al. 30 reported a decreased CR in the clam Ruditapes decussates under high pCO 2 conditions 30 . Liu and He (2012) found reduced CRs in both the mussel Perna viridis and the clam Chlamys nobilis under high pCO 2 conditions 52 . In our study, CO 2 -induced low pH reduced the CR of M. coruscus during the first few days, afterwards CR was only negatively affected by low pH under nano-TiO 2 conditions sometimes, probably because the mussels showed some adaptation to low pH along with time, but could not overcome the additional stressor, nano-TiO 2 . Similarly, the CR of M. galloprovincialis was unaffected by pH 7.5 37 . Sanders et al. 53 also found that there was no significant effect of CO 2 -induced low pH on the filtering activity of king scallop Pecten maximus 53 . Thus, some bivalves can adapt reduced pH conditions, but additional stressors may change this situation based on the present study. It is known that some bivalves are sensitive to reduced pH under food limiting conditions 53,54 . Thus, in our study, nano-TiO 2 first influenced the feeding and absorption of the mussels, which would further reduce the tolerance to high pCO 2 because of the reduced energy uptake. Although the purpose of our study was not to address food limiting effects, nano-TiO 2 posed a sub-optimal dietary state for the mussels during the exposure.

Discussion
High pCO 2 did not show significant effect on AE, implying the normal performance of the digestive systems of the mussels exposed to CO 2 -induced low pH conditions. This was in agreement with the gastropod N. conoidalis in which AE was also insensitive to low pH 55 . In such situations, the digestive enzymes may be relatively stable under the low pH conditions in this study. In the present study, reduced AE probably is caused by the decreased CR under nano-TiO 2 exposure. NPs can attach to algal cells and form clusters, which can settle readily, resulting in a large decline of the algal concentration [56][57][58] . Hereby, the intake of algal cells was subsequently impaired. Moreover, exposure to nano-TiO 2 induced oxidative stress and lysosomal membrane alteration in the digestive gland of mussels 9,10 . In addition, some other molecular and functional parameters of digestive gland can be affected by nano-TiO 2 in mussels 12,59 . These results further verify the hypothesis that NPs can enter the digestive system which is a typical target for NP toxic effect in mussels. AE impairment, representing a significant stress response suggests that nano-TiO 2 exposure may cause a serious harm to mussel health.
Some authors reported metabolic decline under high pCO 2 and suggested that altered extracellular pH could cause these reductions 26,34 . Similar metabolic depressions have been observed in the clam R. decussatus 30,60 , the scallop Chlamys nobilis 52 , the mussel M. chilensis 61 , and the scavenging gastropod Nassarius conoidalis 55 exposed to high pCO 2 conditions. The above reports are consistent with the present study, where the respiration of mussel was significantly lowered at high pCO 2 levels. Because carbon dioxide interacts with intra-and extracellular fluids, internally elevated CO 2 levels may result in a respiratory acidosis 62 . In our study, lower respiration rates under nano-TiO 2 exposure indicated that the metabolic activity was weakened by nano-TiO 2 . More particularly, the oxygen consumption of M. coruscus showed a strong sensitivity to nano-TiO 2 during the entire experiment, indicating a limited capacity to adapt to NP exposure. Reduction in RR can be indirectly due to the disruption of ventilation by nano-TiO 2 . In mussels, CR and RR are linked as both occur by filtering water over the gills when their valves are open. Mussels are known to spend less time opening and filtering when there are contaminants present in the water as a behavioural response to avoid uptake of the contaminant. Hence the reduction in CR and RR could also be due to this behavioural response to nano-TiO 2 rather than direct gill toxicity. Limited respiratory efficiency results in a growing mismatch between basic oxygen demand and oxygen supply and finally causes hypoxia and anaerobic metabolism 63 , which may be harmful to the mussels.
ER is considered as an index of high pCO 2 stress in marine mussels 30,37 . An increase in ER may imply a drastic increment in the consumption of amino acids. Higher ER values under high pCO 2 conditions were observed when nano-TiO 2 was present, indicating nano-TiO 2 affects the CO 2 effect on ER. Under nano-TiO 2 treatments, the inverse correlation between ER and O:N ratios indicates the enhanced protein catabolism and subsequently inhibited growth.
The O:N ratio is considered as an index of the nutritional state of the mussels by showing the metabolism of substrates. According to Widdows 36 and Fernández-Reiriz et al. 30 , in mussels, O:N ratios of more than 30 usually indicate a catabolism of carbohydrates and lipids, whereas less than 30 suggest a protein catabolism mainly. Hereby, low O:N ratios are generally a sign of a stressed condition 64 . In this experiment, high pCO 2 and nano-TiO 2 treatments showed lower O:N ratios owing to the decreased respiration and enhanced protein metabolism as a consequence of energy demands.
In our study, the growth of M. coruscus was not negatively affected by high pCO 2 at the end in terms of SFG. Similar to our study, SFG was stable or even increased under moderate pCO 2 level in M. galloprovincialis 30 . The SFG values became negative when M. coruscus was exposed high nano-TiO 2 , probably as a result of a significant reduction in CR. Given that most of the other parameters measured would be affected by reduced filtration activity (valve opening time) of the mussels and hence time spent for feeding or respiration, nano-TiO 2 obviously would impair the growth of mussels. The SFG results suggest that M. coruscus is not able to grow when nano-TiO 2 is more than 10 mg l −1 in seawater, and nano-TiO 2 (above 10 mg l −1 ) is thus stressful to the mussels.
PCA distinguished non-nano-TiO 2 treatments from exposed treatments since non-nano-TiO 2 treatments were grouped together at positive side whereas exposed treatments were grouped at negative side by PC1, reflecting higher values of AE, O:N, RR, CR and SFG were achieved under non-nano-TiO 2 conditions whereas nano-TiO 2 induced high ER and E. PC2 reflected the time change of most physiological parameters, as higher values of E, ER, CR and SFG were positive, corresponding to the experimental time, namely the later period, day 7 and day 14. By integrating ANOVA and PCA results, the characteristics of physiological responses to nano-TiO 2 exposure were lower CR, AE, RR, O:N ratio and SFG associated with higher E and ER. CR, RR and AE can be reflected by filtration activity. In the present study, the filtration activity was impaired by nano-TiO 2 , hence the CR, RR and AE were reduced. SFG is mostly dependent on absorption rate, which is mostly determined by ingestion rate and AE. Thus, if CR was reduced, SFG was also reduced (Fig. 4). In contrast, higher E and ER under high nano-TiO 2 indicate the low absorption rate and subsequently high protein catabolism, which can also impair the growth of mussels.
Nanoparticles exert toxicity via mainly oxidative stress mechanisms 65 . M. edulis increased oxyradical production and antioxidant enzyme activities when they were exposed to nano-TiO 2 11,15 . In our previous study, the hemocyte functions of M. coruscus exposed to nano-TiO 2 and high pCO 2 were impaired associated with an increase of ROS production, indicating an oxidative stress and fitness impairment 66 . Hence, in the present study, some physiological functions, such as clearance, respiration and absorption were all impaired when mussels were exposed to nano-TiO 2 , especially the combined treatment. The CO 2 -induced low pH in seawater could change the physiochemical properties of nano-TiO 2 and lead to a slightly greater aggregation of nano-TiO 2 66 . In the present study, although the exposed nano-TiO 2 concentration under low pH was lower than under normal pH, no significant difference was found in each nano-TiO 2 concentration (Supplementary Table 1). It is reported that mussels are able to capture and ingest aggregates of nanoparticles more efficiently compared with those freely suspended 6 . Thereby, under CO 2 -induced low pH conditions, M. coruscus may accumulate more nano-TiO 2 than pH 8.1, accordingly causing more severe toxic effects. Mussels usually lower their metabolic rate after exposure to acidified water 26 . In the present study, mussels closed their valves when nano-TiO 2 was present, which further decreased their filtration and respiration. Thus, both high pCO 2 and nano-TiO 2 exerted negative effects on the physiological functions of mussels with more severe effects than the single stressor.

Conclusions
High pCO 2 showed minor effect on the physiology of M. coruscus as most physiological parameters were almost unaffected by CO 2 -induced low pH. In contrast, nano-TiO 2 showed significant negative effects on the feeding and physiology of mussels, suggesting that nano-TiO 2 may present a health threat to the mussels, as depressions in feeding and digestion may reduce the ability of mussels to grow and defense predators. Above all, most physiological functions of M. coruscus were more severely impaired by the combination of nano-TiO 2 and high pCO 2 , suggesting a synergistic effect on mussels.

Methods
Experimental animals. Experimental mussels (30.0 ± 2.0 mm shell length, 70.0 ± 5.0 mg dry tissue weight) were collected from the Shengsi island of Zhejiang Province, China (water temperature: 25.0 °C; salinity: 25.0‰; and pH: 8.1). The experimental mussels were wild, and the handling of them was conducted in accordance with the guidelines set by the Institutional Animal Care and Use Committee (IACUC) of Shanghai Ocean University, Shanghai, China. After transportation to aquarium of the university, the mussels were held in fibre-glass tanks (500 l) with a filtering apparatus in laboratory. The laboratory conditions were maintained according to sampling area in September: temperature, salinity, dissolved oxygen and pH were kept constant at 25 ± 0.5 °C, 26 ± 1‰, 7 ± 0.5 mg O 2 l −1 and 8.10 ± 0.02. Mussels were fed with the microalgae Isochrysis galbana daily (25,000 cells ml −1 , ca. 3% of the tissue dry weight). Two weeks were allowed for mussels to acclimatize to the above laboratory conditions.
TiO 2 characterization by X-ray diffraction and electron microscope. X-ray powder diffractometer (Siemens D500, Karlsruhe, Germany) was applied to exam the crystal structures of nano-TiO 2 . Scanning electron microscope (FEI/ Philips XL30 Esem-FEG, Netherlands) and transmission electron microscope (FEI/Philips Tecnai 12 BioTWIN, the Netherlands) were used to observe the surface morphology, primary size and shape of nano-TiO 2 44 . The diameter of 1000 NPs was measured using an image analysis program (Image J, v1.44; National Institute of Health, USA). Experimental design and system. The experimental mussels were randomly divided to six treatments with three nano-TiO 2 concentrations (0 (control), 2.5 and 10.0 mg l −1 ) under two pH values (7.3, 8.1(control)). The pH 7.3 was selected as an extreme pH value expected by the year 2300 and relevant for some coastal waters 25,47,67 . The low pH treatment was achieved by aerating pure carbon dioxide. The experimental system included a computer, pH regulator (DAQ-M; Loligo ® Systems Inc., Tjele, Denmark), pH meter, solenoid valve, CO 2 and air supplies, water pump, filtering system, temperature regulator, head tank and exposure tank. The CO 2 flow from the pure CO 2 cylinder to the header tank was controlled by the pCO 2 /pH regulator which can open or close the solenoid valve when the seawater pH values deviated from the preset values by ±0.1 pH units. Water temperature was kept at 25 °C by temperature regulators. The tanks were covered with acrylic sheets to prevent Scientific RepoRts | 7:40015 | DOI: 10.1038/srep40015 external interference. To achieve exposed concentrations of 2.5 and 10 mg l −1 nano-TiO 2 in the experimental tanks, each test solution was prepared in a water tank (500 l) by adding relevant volume of 10.0 g l −1 sonicated stock solution to the seawater, respectively. The seawater in the exposure tank was renewed and re-dosed everyday with new working solutions of nano-TiO 2 to keep the exposure at relatively consistent nano-TiO 2 concentrations. In each treatment, there were three flow-through tanks (30 l) as three replicates (30 mussels per replicate), and mussels were fed with microalgae Isochrysis galbana (2.5 × 10 4 cells ml −1 ) at 1 st , 3 rd , 5 th , 7 th , 9 th ,11 th and 14 th day during the whole experiment. The mussels were exposed to six treatments for 14 days, and examinations on all physiological parameters of M. coruscus were carried out on the 1 st , 3 rd , 7 th and 14 th day.
Monitoring of carbonate chemistry of seawater. Gran titration was applied to measure total alkalinity (TA) with a total alkalinity titrator system 68 . The pH and TA were used to calculate other carbonate system parameters, i.e., partial pressure of CO 2 (pCO 2 ) and dissolved inorganic carbon (DIC), and the saturation state of omega calcite and aragonite using CO 2 SYS software as described previously 37 . Nanoparticle concentrations in water were measured using the standard test method for determination by atomic absorption spectroscopy of titanium dioxide content 69 .
Physiological measurements. Clearance rate. Clearance rate (CR) was measured using all individuals as a whole in each replicate tank. Prior to CR measurement, the mussels were fasting for at least twelve hours to evacuate their digestive tracts. The initial microalgae concentration was 2.5 × 10 4 cells ml −1 , and no mussels produced pseudo-faeces at this concentration. Three identical tanks without experimental mussels were used as the control. An initial 20 ml water was sampled from the tank center with a syringe at the beginning, and then 20 ml aliquots were collected at 30 min intervals for 120 min, without being influenced by large decrease of algal concentration (< 30%). The algal concentration in each sample was counted by a particle analyzer (Multisizer 3 Coulter Counter, Beckman, Irvine, USA). Cell concentrations in control tanks did not show any significant variation during the measurements. CR was calculated according to the formula of Coughlan 70 : where CR represents the clearance rate (l h −1 ), V represents the water volume in the tank (l), C 0 represents the initial microalgae concentration (cells ml −1 ), C t represents the microalgae concentration at time t (cells ml −1 ), N represents the number of experimental mussels in the tank, t represents the sampling time (h). CR and the rest physiological parameters were standardized to unit dry weight (see below scope for growth).
Ingestion rate (IR) was computed by multiplying POM (particulate organic matter, mg l −1 ) by CR 71 , i.e., the food intake per hour. The POM concentration was transformed to joules using a conversion factor of 23.5 J mg −1 for Isochrysis galbana 72 .
Absorption efficiency. Absorption efficiency (AE) was determined according to Conover 73 . AE was determined by collecting the feces after CR measurements in each replicate tank. The organic content in microalgae was measured by filtering 5 L seawater with 2.5 × 10 4 cell ml −1 algae using glass fiber filters (Whatman ® GF/C).
Ammonium formate solution (0.5 M) was used to rinse these fulter papers which were then dried at 110 °C for 24 hours and weighed, and then ashed in a muffle furnace (450 °C for 6 hours) and reweighed. Blank GF/C filters were also treated using the same procedure to correct weight change caused by daily humidity variations. Faeces were carefully collected using a pipette from the tanks 12 hours after the CR measurements, and the organic content of faeces was determined as above. AE was calculated according to Conover 73 : where AE represents the absorption efficiency (%), F represents the ratio of ash-free dry weight:dry weight in the microalgae, and E represents the ratio of ash-free dry weight:dry weight in the faeces.
Respiration rate. Fifteen mussels were randomly sampled from each tank and divided averagely into three replicates. The respiration rate (RR) was determined for five mussels in a closed glass respirometer (1000 ml) filled with air-saturated seawater from the corresponding experimental tank for one hour. To ensure the mussels started to respire in the chamber, the measurement began 20 minutes later when the mussels opened their valves. DO concentrations in three chambers without mussels were also recorded as the control. The oxygen consumption rate within the chamber was measured by oxygen meters (model YSI 58). The oxygen concentrations at the beginning and the end in each chamber were recorded. The RR was calculated according to the following formula: where RR represents the respiration rate (mg O 2 h −1 ), C t0 and C t1 represent the initial and final DO concentrations in the chamber (mg O 2 l −1 ), V (l) represent the volume of the seawater in the chamber, N represents the number of mussels in the chamber, t (h) represents the time elapsed. RR values were converted into J/h using a conversion coefficient of 13.98 J mg O 2 −1 74 .
Ammonia excretion rate. When RR measurements were finished, ammonia excretion rates (ER) of the same mussels were examined. Water was sampled from each chamber and stored at − 20 °C until analysis. The concentration of ammonia produced by mussels was measured by the phenol-hypochlorite method 75 . ER was obtained from the difference of ammonia concentrations between the chamber containing mussels and the control chamber according to the equation: Scientific RepoRts | 7:40015 | DOI: 10.1038/srep40015 where U represents the rate of ammonia excretion (mg NH 4 -N h −1 ), C s represents the ammonia concentration (mg l −1 ) in the experimental sample, C c represents the ammonia concentration in the control sample, V represents the volume (l) of seawater in the chamber, N represents the number of mussels and t represents the time elapsed (h). Values of excretion rate were converted into J h −1 using a conversion coefficient of 25 J mg NH 4 -N −1 76 . The ratio of oxygen consumption to ammonia excretion expressed as atomic equivalents (O:N) was calculated to assess the utilization of different biochemical compositions for energy metabolism 36 .
Scope for growth. At the end of the experiment, soft tissues of the mussels were collected and dried at 90 °C for 24 h to calculate their tissue dry weight. CR (l h −1 ), RR (mg O 2 h −1 ) and ER (mg NH 4 -N h −1 ) were converted to mass specific rates as a 'standard mussel' of 1 g dry weight using the formula: Ys = (Ws/We) b × Ye, where Ys is the physiological value for a mussel of standard weight, Ws is the standard weight (1 g), We is the measured weight of the mussel (g), Ye is the uncorrected (measured) physiological value, and b is the weight exponent for the physiological value (b = 0.67). Each physiological parameter was transformed to energy equivalents (J h −1 g −1 ) to calculate the scope for growth (SFG), showing the difference between the energy acquired from the food and the energy depleted by respiration and excretion. SFG was calculated according to Smaal and Widdows 74,76,77 : where SFG represents scope for growth (J h −1 g −1 ), Ab represents absorption rate which derives from IR × AE (J h −1 g −1 ), R represents the energy loss in respiration (J h −1 g −1 ), and U represents the energy loss in ammonia excretion (J h −1 g −1 ).

Statistical analyses.
Normality of the data was checked by Shapiro-Wilk's W test and homogeneity of variances was evaluated by Levene's test using SPSS 16.0. To evaluate the interactive effects of pH and nano-TiO 2 on physiological parameters, two-way repeated measures analysis of variance (ANOVA) were used. When there was a significant interaction, one-way ANOVA followed by Tukey's HSD test or student t test was conducted for each factor separately in each level of the other factor. Principal component analysis (PCA) was conducted for multivariate analysis using XLSTAT ® 2014. A biplot was graphed with both the measured variables and the observations. All data were expressed as the mean ± standard deviations, and the results were statistically significant at p < 0.05.