Combined Toxicity of TiO2 Nanospherical Particles and TiO2 Nanotubes to Two Microalgae with Different Morphology

The joint activity of multiple engineered nanoparticles (ENPs) has attracted much attention in recent years. Many previous studies have focused on the combined toxicity of different ENPs with nanostructures of the same dimension. However, the mixture toxicity of multiple ENPs with different dimensions is much less understood. Herein, we investigated the toxicity of the binary mixture of TiO2 nanospherical particles (NPs) and TiO2 nanotubes (NTs) to two freshwater algae with different morphology, namely, Scenedesmus obliquus and Chlorella pyrenoidosa. The physicochemical properties, dispersion stability, and the generation of reactive oxygen species (ROS) were determined in the single and binary systems. Classical approaches to assessing mixture toxicity were applied to evaluate and predict the toxicity of the binary mixtures. The results show that the combined toxicity of TiO2 NPs and NTs to S. obliquus was between the single toxicity of TiO2 NTs and NPs, while the combined toxicity to C. pyrenoidosa was higher than their single toxicity. Moreover, the toxicity of the binary mixtures to C. pyrenoidosa was higher than that to S. obliquus. A toxic unit assessment showed that the effects of TiO2 NPs and NTs were additive to the algae. The combined toxicity to S. obliquus and C. pyrenoidosa can be effectively predicted by the concentration addition model and the independent action model, respectively. The mechanism of the toxicity caused by the binary mixtures of TiO2 NPs and NTs may be associated with the dispersion stability of the nanoparticles in aquatic media and the ROS-induced oxidative stress effects. Our results may offer a new insight into evaluating and predicting the combined toxicological effects of ENPs with different dimensions and of probing the mechanisms involved in their joint toxicity.


Introduction
In the past decade, research on the health and environmental risk of engineered nanoparticles (ENPs) has increased considerably [1][2][3]. The requests for toxicity data on the effects of ENPs on organisms continue to grow. Most of the toxicity data are derived from a single toxicity [4,5]. In the natural environment, organisms exposed to a mixture of multiple contaminants (n ≥ 2) rather than individual ones is a universal law [6,7]. Many studies have addressed the toxic effects of a mixture of ENPs and other contaminants, e.g., nano-TiO 2 and tetracycline [8], nano-TiO 2 and hexavalent chromium [9], as well as nano-TiO 2 and bisphenol [10]. However, researchers in the field

Algal Growth Assays
The unicellular freshwater algae S. obliquus and C. pyrenoidosa were obtained from the Chinese Academy of Sciences Institute of Hydrobiology (Wuhan, China). Exponentially growing algae cells (with a final density of 3 × 10 5 cells/mL for S. obliquus and 4 × 10 5 cells/mL for C. pyrenoidosa) were added to the control and treated experiments. Internal control experiments were required to eliminate the absorbance effects of the test materials. All flasks containing various treatments were incubated in an artificial growth chamber at a consistent temperature of 24 ± 1 • C for 96 h with a photoperiod of 12-h light (3000-4000 l×) and 12-h dark. The algal cell density was determined using an ultraviolet-visible spectrophotometer (UV1102; Shanghai Tian Mei Scientific Instrument Co., Shanghai, China) after 96 h for S. obliquus and 72 h for C. pyrenoidosa to provide cell numbers and allow the specific growth rate to be calculated. Growth inhibition (%) was calculated by dividing the specific growth rate for a treatment by the mean specific growth rate for the controls. Three replicates were included for each treatment and the data presented are the mean of the runs (n = 3). When intracellular ROS are generated, 2 ,7 -dichlorofluorescein (DCF) can be converted from DCFH, which is obtained by lipase decomposing DCFH-DA in cells [26]. Thus, the fluorescence intensity (FI) of DCF indicates the extent of intracellular ROS generation. FI was measured using a fluorospectrophotometer (F96PRO, Shanghai Kingdak Scientific Instrument Co., Ltd., Zhejiang, China). The excitation and emission wavelength for the optical measurements were based on previous studies [15,27]. Each sample was measured three times. Data were expressed as a percentage (%) of the fluorescence or the absorbance of the control cells according to the equation:

Oxidative Stress Biomarker Assays
where %F is the percentage of fluorescence of algal cells; F c is the mean fluorescence of control cells; and F t is the mean fluorescence of treated cells.

Assessment and Prediction for Mixture Toxicity
The logistic model (Equation (2)) was used to fit concentration-response curves (CRCs) for TiO 2 NPs, TiO 2 NTs, and TiO 2 NPs + NTs. The effect concentrations, such as the 10% effect concentration (EC 10 ) and median effect concentration (EC 50 ) of each treatment group were derived from the CRCs. The combined toxicity tests followed the mixture ratios of the effect concentrations of TiO 2 NPs and NTs at the same effect level. E = 100 where C is the test material's concentration and θ represents the slope parameter.
The EC 50 value was used to calculate a toxic unit (TU), as shown in Equations (3) and (4), Nanomaterials 2020, 10, 2559 4 of 11 TU Total = TU NPs + TU NTs (4) where C i is the concentration of material i and TU Total is the sum of the TiO 2 NP and NT TUs. Therefore, one TU corresponded to the EC 50 . The MOJA of TiO 2 NPs and NTs were evaluated by plotting the observed TU (TU obs ) derived from the toxicity data against the expected TU (TU exp ), as described by Kim et al. [28]. The total concentration of a mixture provoking an x% effect (EC xmix ) was calculated from the CRCs of the individual components using the CA model, as shown in Equation 5.
where P i is the fraction of component i in the mixture and EC xi is the concentration of component i that would, when applied singly, provoke the x% effect.
The general equation shown in Equation (6) was used for the IA model, where E(C mix ) is the effect expected at the total concentration of the mixture (scaled to between 0% and 100%) and E(C i ) is the effect that the ith mixture component would provoke if applied singly at concentration C i .

Statistical Analysis
All data are expressed as means ± standard deviation (SD). Statistically significant differences between the test treatments were determined by a t-test at significance levels of p < 0.05, p < 0.01, and p < 0.001.

Physicochemical Characterizations
The TEM images are shown in Figure 1. The individual TiO 2 NPs and NTs were spherical ( Figure 1A) and tubular ( Figure 1B) particles, respectively. Additionally, the TEM images show that the TiO 2 NPs agglomerated heavily in the test medium, whereas the TiO 2 NTs agglomerated slightly. As can be seen from Figure 1C, the morphology of the two TiO 2 particles with different shapes in the binary mixture systems was not affected by each other. Nanomaterials 2020, 10, x FOR PEER REVIEW 4 of 12 plotting the observed TU (TUobs) derived from the toxicity data against the expected TU (TUexp), as described by Kim et al [28].
The total concentration of a mixture provoking an x% effect (ECxmix) was calculated from the CRCs of the individual components using the CA model, as shown in Equation 5.
where Pi is the fraction of component i in the mixture and ECxi is the concentration of component i that would, when applied singly, provoke the x% effect. The general equation shown in Equation (6) was used for the IA model, where E(Cmix) is the effect expected at the total concentration of the mixture (scaled to between 0% and 100%) and E(Ci) is the effect that the ith mixture component would provoke if applied singly at concentration Ci.

Statistical Analysis
All data are expressed as means ± standard deviation (SD). Statistically significant differences between the test treatments were determined by a t-test at significance levels of p < 0.05, p < 0.01, and p < 0.001.

Physicochemical Characterizations
The TEM images are shown in Figure 1. The individual TiO2 NPs and NTs were spherical ( Figure  1A) and tubular ( Figure 1B) particles, respectively. Additionally, the TEM images show that the TiO2 NPs agglomerated heavily in the test medium, whereas the TiO2 NTs agglomerated slightly. As can be seen from Figure 1C, the morphology of the two TiO2 particles with different shapes in the binary mixture systems was not affected by each other. To characterize the change in the physicochemical properties of TiO2 particles, the ZP and HD values of TiO2 NPs and NTs from single to binary mixtures were measured in the test medium ( Table  1). The ZP value of TiO2 NPs at 96 h were significantly increased compared with the ZP value of TiO2 NPs at 0 h (p < 0.01). It was found that there was no obvious change in the ZP value of TiO2 NTs or TiO2 NPs + NTs between 0 h and 96 h (p > 0.05). The measurement of HD showed that the of the TiO2 particles became smaller in size over a 96-h period. Furthermore, the size of the particles in TiO2 NPs + NTs was significantly decreased during 96 h of exposure (p < 0.01). The reason for the reduction in particle size may be due to the sedimentation of large particles. Table 1. Zeta potential (ZP) and hydrodynamic diameter (HD) ± standard deviation (n = 3) of the test To characterize the change in the physicochemical properties of TiO 2 particles, the ZP and HD values of TiO 2 NPs and NTs from single to binary mixtures were measured in the test medium ( Table 1). The ZP value of TiO 2 NPs at 96 h were significantly increased compared with the ZP value of TiO 2 NPs at 0 h (p < 0.01). It was found that there was no obvious change in the ZP value of TiO 2 NTs or TiO 2 NPs + NTs between 0 h and 96 h (p > 0.05). The measurement of HD showed that the of the TiO 2 particles became smaller in size over a 96-h period. Furthermore, the size of the particles in TiO 2 NPs + NTs was significantly decreased during 96 h of exposure (p < 0.01). The reason for the reduction in particle size may be due to the sedimentation of large particles. To further evaluate the agglomeration and stability of the TiO 2 particles in the test medium, the total potential energy profiles of the TiO 2 particles in the single and binary mixture systems at the end of different intervals were calculated using the DLVO theory ( Figure 2). At 0 h, the peak values of the total potential energy profiles decreased in the order: TiO 2 NPs > TiO 2 NPs + NTs > TiO 2 NTs ( Figure 2A). This means that the TiO 2 NPs showed the highest stability in the test medium compared to the other studied systems. However, over 96 h, the stability of TiO 2 NPs and TiO 2 NPs + NTs obviously decreased, while the stability of TiO 2 NTs enhanced slightly ( Figure 2B). As mentioned above, the particle size of TiO 2 particles in the test medium became smaller with time. Taken together, this suggests that the TiO 2 particles were settling due to particle agglomeration. To further evaluate the agglomeration and stability of the TiO2 particles in the test medium, the total potential energy profiles of the TiO2 particles in the single and binary mixture systems at the end of different intervals were calculated using the DLVO theory ( Figure 2). At 0 h, the peak values of the total potential energy profiles decreased in the order: TiO2 NPs > TiO2 NPs + NTs > TiO2 NTs ( Figure 2A). This means that the TiO2 NPs showed the highest stability in the test medium compared to the other studied systems. However, over 96 h, the stability of TiO2 NPs and TiO2 NPs + NTs obviously decreased, while the stability of TiO2 NTs enhanced slightly ( Figure 2B). As mentioned above, the particle size of TiO2 particles in the test medium became smaller with time. Taken together, this suggests that the TiO2 particles were settling due to particle agglomeration.

Toxicity of Single and Mixtures of TiO2 NPs to Algal
Typical CRCs were observed for the toxic effects of TiO2 NPs, TiO2 NTs, and TiO2 NPs + NTs on the two test species (Figure 3). The effect concentrations determined by the CRCs are listed in Table  2. The CRC analysis indicated a concentration-dependent variation in the individual and combined toxic effects of TiO2 NPs and NTs. For S. obliquus, the CRC for TiO2 NPs was distant from the CRC for TiO2 NTs and started at lower concentrations ( Figure 3A). Moreover, the EC10 and EC50 values of TiO2 NPs were lower than those of TiO2 NTs, suggesting that the single toxicity of TiO2 NPs to the algae was higher than TiO2 NTs. Some previous studies have also indicated that the shape of the ENPs is a significant factor in determining the potency and magnitude of the toxicity effect on organisms [29][30][31]. The CRC for TiO2 NPs + NTs was in between that for TiO2 NPs and that for TiO2 NTs. Similarly, the EC50 value derived from the CRC for TiO2 NPs + NTs was between the EC50 value of each components in the binary mixtures. As mentioned above, the stability of TiO2 NPs + NTs in the test medium at the initial time was also between that of TiO2 NPs and that of TiO2 NTs. The findings for the growth inhibition toxicity to S. obliquus combined with the findings for the stability indicate that the higher the initial stability, the stronger the toxicity. Dispersion of ENPs has received special research attention because the environmental behavior and effects of ENPs are greatly dependent on their dispersion status [32]. Previous studies have also suggested that nano-TiO2 aggregates can reduce the light available to the entrapped algal cells and thus inhibits their growth [33,34]. As mentioned above, the TiO2 particles agglomerated under this study. This also means that

Toxicity of Single and Mixtures of TiO 2 NPs to Algal
Typical CRCs were observed for the toxic effects of TiO 2 NPs, TiO 2 NTs, and TiO 2 NPs + NTs on the two test species (Figure 3). The effect concentrations determined by the CRCs are listed in Table 2. The CRC analysis indicated a concentration-dependent variation in the individual and combined toxic effects of TiO 2 NPs and NTs. For S. obliquus, the CRC for TiO 2 NPs was distant from the CRC for TiO 2 NTs and started at lower concentrations ( Figure 3A). Moreover, the EC 10 and EC 50 values of TiO 2 NPs were lower than those of TiO 2 NTs, suggesting that the single toxicity of TiO 2 NPs to the algae was higher than TiO 2 NTs. Some previous studies have also indicated that the shape of the ENPs is a significant factor in determining the potency and magnitude of the toxicity effect on organisms [29][30][31]. The CRC for TiO 2 NPs + NTs was in between that for TiO 2 NPs and that for TiO 2 NTs. Similarly, the EC 50 value derived from the CRC for TiO 2 NPs + NTs was between the EC 50 value of each components in the binary mixtures. As mentioned above, the stability of TiO 2 NPs + NTs in the test medium at the initial time was also between that of TiO 2 NPs and that of TiO 2 NTs. The findings for the growth inhibition toxicity to S. obliquus combined with the findings for the stability indicate that the higher the initial stability, the stronger the toxicity. Dispersion of ENPs has received special research attention because the environmental behavior and effects of ENPs are greatly dependent on their dispersion status [32]. Previous studies have also suggested that nano-TiO 2 aggregates can reduce the light available to the entrapped algal cells and thus inhibits their growth [33,34]. As mentioned above, the TiO 2 particles agglomerated under this study. This also means that the agglomeration of particles contributed to the overall growth inhibition toxicity to some degree.
implies that the joint toxicity of TiO2 NPs and NTs to C. pyrenoidosa was higher than the single toxicity of each component in the binary mixtures.
It was also found that the EC10 and EC50 values of TiO2 NPs, TiO2 NTs, and TiO2 NPs + NTs to C. pyrenoidosa were lower than those to S. obliquus, which implies that the single and binary mixtures exhibited stronger toxicity to C. pyrenoidosa than to S. obliquus. This finding reveals that C. pyrenoidosa is more sensitive to the TiO2 particles than S. obliquus. S. obliquus is usually composed of four cells that are 12-34 μm wide and 10-21 μm long, with a flat shape. C. pyrenoidosa has a spherical shape with a diameter of 3-5 μm. The cells of C. pyrenoidosa are smaller but have a higher specific surface area than those of S. obliquus. This feature allows for a more effective particle uptake by C. pyrenoidosa. Furthermore, our previous study indicated the cell membrane permeability of C. pyrenoidosa was significantly increased after ENP stimulation, compared with the control and S. obliquus [35]. Consequently, the TiO2 particles might interrupt the cell membrane functions of C. pyrenoidosa to a higher degree, and thus trigger more severe growth inhibition toxicity.

Assessment and Prediction of Joint Toxicity of TiO2 NPs and NTs
Concentration (mg/L)  For C. pyrenoidosa, the CRC for TiO 2 NPs crossed over the CRC for TiO 2 NTs ( Figure 3B). The EC 10 value of TiO 2 NPs was higher than that of TiO 2 NTs. The EC 50 value of TiO 2 NPs was slightly higher than that of TiO 2 NTs. However, as can be seen from the CRCs, the effect concentrations of TiO 2 NPs were lower than those of TiO 2 NTs, as the observed effects gradually increased. This means that the differences in the single toxicity of TiO 2 NPs and NTs depend on the exposure concentration. Moreover, in the higher exposure concentration range, TiO 2 NPs with the higher initial stability showed more toxicity than TiO 2 NTs with the lower initial stability. The CRC for TiO 2 NPs + NTs intercrossed the other two CRCs and decreased slightly with lower concentrations. The EC 50 value derived from the CRC for TiO 2 NPs + NTs was lower than that of TiO 2 NPs and that of TiO 2 NTs. This implies that the joint toxicity of TiO 2 NPs and NTs to C. pyrenoidosa was higher than the single toxicity of each component in the binary mixtures.
It was also found that the EC 10 and EC 50 values of TiO 2 NPs, TiO 2 NTs, and TiO 2 NPs + NTs to C. pyrenoidosa were lower than those to S. obliquus, which implies that the single and binary mixtures exhibited stronger toxicity to C. pyrenoidosa than to S. obliquus. This finding reveals that C. pyrenoidosa is more sensitive to the TiO 2 particles than S. obliquus. S. obliquus is usually composed of four cells that are Nanomaterials 2020, 10, 2559 7 of 11 12-34 µm wide and 10-21 µm long, with a flat shape. C. pyrenoidosa has a spherical shape with a diameter of 3-5 µm. The cells of C. pyrenoidosa are smaller but have a higher specific surface area than those of S. obliquus. This feature allows for a more effective particle uptake by C. pyrenoidosa. Furthermore, our previous study indicated the cell membrane permeability of C. pyrenoidosa was significantly increased after ENP stimulation, compared with the control and S. obliquus [35]. Consequently, the TiO 2 particles might interrupt the cell membrane functions of C. pyrenoidosa to a higher degree, and thus trigger more severe growth inhibition toxicity.

Assessment and Prediction of Joint Toxicity of TiO 2 NPs and NTs
The observed toxicities were converted to TUs and plotted against the expected TU total values of the binary mixtures of TiO 2 NPs and NTs, as calculated from the sums of the individual TiO 2 NPs and NTs. As depicted in Figure 4, the observed TU total,obs are almost equal to the expected TUs, indicating that the TiO 2 NPs + NTs mixture effects were additive. However, for C. pyrenoidosa, the observed TU total,obs of TiO 2 NPs + NTs at the highest concentration under this study was obviously higher than the expected TUs, indicating that the joint toxicity was synergistic at higher concentrations of the mixtures. The observed toxicities were converted to TUs and plotted against the expected TUtotal values of the binary mixtures of TiO2 NPs and NTs, as calculated from the sums of the individual TiO2 NPs and NTs. As depicted in Figure 4, the observed TUtotal,obs are almost equal to the expected TUs, indicating that the TiO2 NPs + NTs mixture effects were additive. However, for C. pyrenoidosa, the observed TUtotal,obs of TiO2 NPs + NTs at the highest concentration under this study was obviously higher than the expected TUs, indicating that the joint toxicity was synergistic at higher concentrations of the mixtures. The classical methods, namely, CA and IA, were used to quantitatively assess and predict the combined effects of ENPs [36,37]. The differences between the experimental and predicted joint toxicities to S. obliquus and C. pyrenoidosa are shown in Figure 5 and Table 2. For S. obliquus, the CRC derived from the CA model slightly deviated from the observed CRC, while the CRC derived from the IA model seriously deviated from the observed CRC ( Figure 5A). For a direct graphical assessment of the whole concentration-response range, we also depicted the 95% confidence band (CB) and prediction band (PB) of the experimental data points. The CA prediction almost overlapped with the CB of the observed concentration-response data, implying that the CA model showed good predictive quality over the widest range of effects. However, except for the lower effect regions (about <40%), the IA prediction was outside the PB range of the observed concentration-response data. This also means that there was a big difference between the observation and the prediction for the IA model. As can be seen from Table 2, the EC50 value predicted by the CA model (99.46 mg/L) approaches the observed EC50 (85.04 mg/L), and the differences in the EC50 value between the observed and the CA predicted is 17%. However, the differences in the EC50 value between the observed and the CA predicted (56.30 mg/L) is 34%. Generally, the CA model performed better than the IA method although the CA slightly underestimated the observed toxicity of the binary mixtures of TiO2 NPs and TiO2 NTs to S. obliquus. This might be because the predictive power of the CA model was strictly restricted by the concentration addition MOJA of TiO2 NPs and TiO2 NTs to S. obliquus.
For C. pyrenoidosa, the CRC derived from the CA and IA models deviated moderately from the observed CRC ( Figure 5B). Moreover, the concentration-response data predicted by the CA and IA models were inside the PB range of the observed data. Further, in the higher effect regions (about >65%), the CA prediction was inside the CB range of observed concentration-response data. Except  The classical methods, namely, CA and IA, were used to quantitatively assess and predict the combined effects of ENPs [36,37]. The differences between the experimental and predicted joint toxicities to S. obliquus and C. pyrenoidosa are shown in Figure 5 and Table 2. For S. obliquus, the CRC derived from the CA model slightly deviated from the observed CRC, while the CRC derived from the IA model seriously deviated from the observed CRC ( Figure 5A). For a direct graphical assessment of the whole concentration-response range, we also depicted the 95% confidence band (CB) and prediction band (PB) of the experimental data points. The CA prediction almost overlapped with the CB of the observed concentration-response data, implying that the CA model showed good predictive quality over the widest range of effects. However, except for the lower effect regions (about <40%), the IA prediction was outside the PB range of the observed concentration-response data. This also means that there was a big difference between the observation and the prediction for the IA model. As can be seen from Table 2, the EC 50 value predicted by the CA model (99.46 mg/L) approaches the observed EC 50 (85.04 mg/L), and the differences in the EC 50 value between the observed and the CA predicted is 17%. However, the differences in the EC 50 value between the observed and the CA predicted (56.30 mg/L) is 34%. Generally, the CA model performed better than the IA method although the CA slightly underestimated the observed toxicity of the binary mixtures of TiO 2 NPs and TiO 2 NTs to S. obliquus. Nanomaterials 2020, 10, 2559 8 of 11 This might be because the predictive power of the CA model was strictly restricted by the concentration addition MOJA of TiO 2 NPs and TiO 2 NTs to S. obliquus. Nanomaterials 2020, 10, x FOR PEER REVIEW 8 of 12 to C. pyrenoidosa is mainly based on response addition. Taken together, the CA and IA methods provided valid predictions of the toxicity of the mixtures.

Cellular Oxidative Stress Effects of Single and Mixtures of TiO2 NPs and NTs on Algal Cells
Cellular oxidative stress caused by the elevation of particle-induced ROS is considered the most likely toxic mechanism of nano-TiO2 [38][39][40]. As shown in Figure 6A for S. obliquus, the FI (%) of the TiO2 NPs is significantly higher (p < 0.05) than the control, which indicates a significant increase in ROS. However, there was no significant difference in the ROS level between the TiO2 NTs and control. This implies that TiO2 NPs, but not TiO2 NTs produce ROS in S. obliquus cells. The binary systems of TiO2 NPs and NTs significantly promoted the generation of intracellular ROS. Note that the ROS levels induced by the binary TiO2 NPs + NTs mixtures at the EC50 ratio were significantly lower than the ROS levels induced by single TiO2 NPs at the EC50 value.
For C. pyrenoidosa, the TiO2 NPs and NTs at their EC50 value, as well as the TiO2 NPs + NTs at the EC50 ratio significantly increased the ROS levels ( Figure 6B). Furthermore, the binary mixtures of TiO2 NPs and NTs induced the generation of intracellular ROS to a higher level than the single TiO2 NPs and NTs, which may intensify the oxidative stress effects on the C. pyrenoidosa cells exposed to the combination of TiO2 NPs + NTs. This also causes more serious apparent toxicity, as we observed in the growth inhibition toxicity testing. In general, TiO2 particle-induced ROS production depended on the particle characteristics, algal cell types, and exposure concentrations. In addition to the mechanisms underlying ROS generation, it also remains unknown as to how the TiO2 NPs interact with the TiO2 NTs and how this interaction regulates the intracellular ROS levels. Nano-TiO2 can cause genotoxicity [41]. Furthermore, the ROS-mediated stress within cells could be the main mechanism for the genotoxicity of nano-TiO2 [42]. Further studies are needed to explore whether the TiO2 NPs and NTs can jointly cause DNA damage due to the production of ROS. For C. pyrenoidosa, the CRC derived from the CA and IA models deviated moderately from the observed CRC ( Figure 5B). Moreover, the concentration-response data predicted by the CA and IA models were inside the PB range of the observed data. Further, in the higher effect regions (about >65%), the CA prediction was inside the CB range of observed concentration-response data. Except for the range of effects from about 50% to 90%, the IA prediction was inside the CB range of observed concentration-response data. As shown in Table 2, the EC 50 value predicted by the CA model was 2.9 times greater than the observed EC 50 value. However, the EC 50 value predicted by the IA model was 1.9 times lower than the observed EC 50 value. Similar to S. obliquus, the CA underestimated the joint toxicity, while the IA model overestimated joint toxicity to C. pyrenoidosa. Generally, the IA model performs better than the CA method. This further implies that the MOJA of TiO 2 NPs and TiO 2 NTs to C. pyrenoidosa is mainly based on response addition. Taken together, the CA and IA methods provided valid predictions of the toxicity of the mixtures.

Cellular Oxidative Stress Effects of Single and Mixtures of TiO 2 NPs and NTs on Algal Cells
Cellular oxidative stress caused by the elevation of particle-induced ROS is considered the most likely toxic mechanism of nano-TiO 2 [38][39][40]. As shown in Figure 6A for S. obliquus, the FI (%) of the TiO 2 NPs is significantly higher (p < 0.05) than the control, which indicates a significant increase in ROS. However, there was no significant difference in the ROS level between the TiO 2 NTs and control. This implies that TiO 2 NPs, but not TiO 2 NTs produce ROS in S. obliquus cells. The binary systems of TiO 2 NPs and NTs significantly promoted the generation of intracellular ROS. Note that the ROS levels induced by the binary TiO 2 NPs + NTs mixtures at the EC 50 ratio were significantly lower than the ROS levels induced by single TiO 2 NPs at the EC 50 value.
For C. pyrenoidosa, the TiO 2 NPs and NTs at their EC 50 value, as well as the TiO 2 NPs + NTs at the EC 50 ratio significantly increased the ROS levels ( Figure 6B). Furthermore, the binary mixtures of TiO 2 NPs and NTs induced the generation of intracellular ROS to a higher level than the single TiO 2 NPs and NTs, which may intensify the oxidative stress effects on the C. pyrenoidosa cells exposed to the combination of TiO 2 NPs + NTs. This also causes more serious apparent toxicity, as we observed in the growth inhibition toxicity testing. In general, TiO 2 particle-induced ROS production depended on the particle characteristics, algal cell types, and exposure concentrations. In addition to the mechanisms underlying ROS generation, it also remains unknown as to how the TiO 2 NPs interact with the TiO 2 NTs and how this interaction regulates the intracellular ROS levels. Nano-TiO 2 can cause genotoxicity [41]. Furthermore, the ROS-mediated stress within cells could be the main mechanism for the genotoxicity of nano-TiO 2 [42]. Further studies are needed to explore whether the TiO 2 NPs and NTs can jointly cause DNA damage due to the production of ROS. Nanomaterials 2020, 10, x FOR PEER REVIEW 9 of 12 Figure 6. Relative levels of reactive oxygen species (ROS) detected using 2′,7′dichlorodihydrofluorescein diacetate (DCFH-DA) staining in Scenedesmus obliquus (A) and Chlorella pyrenoidosa (B) exposed to single TiO2 NPs and TiO2 NTs at each EC10 or EC50 value, as well as TiO2 NPs + NTs at the EC10 or EC50 ratios. Statistical significance versus control group: * p < 0.05, ** p < 0.01, and *** p < 0.001.

Conclusions
To sum up, for the first time we present the toxicity of multiple ENP systems with different dimensions. It was found that the single toxicity varied as a function of the TiO2 dimensions, the test species, and exposure concentrations. The toxicity of the binary mixtures of TiO2 NPs (zerodimension) and NTs (one-dimension) to two freshwater algae was found to be an additive joint activity according to the TUs. The classical toxicological models (CA and IA) for mixtures predicted the joint toxicities and revealed that the TiO2 NPs and NTs acted as a concentration addition and response addition towards S. obliquus and C. pyrenoidosa, respectively. The mechanisms of TiO2 NPs-NTs joint toxicity were related to the aqueous stability of the TiO2 particles and their ROS-induced oxidative stress effects. Our findings highlight the importance of the dimensions of nanoparticles in assessing the combined risks of multiple ENPs.   Figure 6. Relative levels of reactive oxygen species (ROS) detected using 2 ,7 -dichlorodihydrofluorescein diacetate (DCFH-DA) staining in Scenedesmus obliquus (A) and Chlorella pyrenoidosa (B) exposed to single TiO 2 NPs and TiO 2 NTs at each EC 10 or EC 50 value, as well as TiO 2 NPs + NTs at the EC 10 or EC 50 ratios. Statistical significance versus control group: * p < 0.05, ** p < 0.01, and *** p < 0.001.

Conclusions
To sum up, for the first time we present the toxicity of multiple ENP systems with different dimensions. It was found that the single toxicity varied as a function of the TiO 2 dimensions, the test species, and exposure concentrations. The toxicity of the binary mixtures of TiO 2 NPs (zero-dimension) and NTs (one-dimension) to two freshwater algae was found to be an additive joint activity according to the TUs. The classical toxicological models (CA and IA) for mixtures predicted the joint toxicities and revealed that the TiO 2 NPs and NTs acted as a concentration addition and response addition towards S. obliquus and C. pyrenoidosa, respectively. The mechanisms of TiO 2 NPs-NTs joint toxicity were related to the aqueous stability of the TiO 2 particles and their ROS-induced oxidative stress effects. Our findings highlight the importance of the dimensions of nanoparticles in assessing the combined risks of multiple ENPs.