Modeling acute toxicity of metal mixtures to wheat (Triticum aestivum L.) using the biotic ligand model-based toxic units method

The combined toxic effects of copper (Cu) and cobalt (Co) were predicted using the biotic ligand model (BLM) for different concentrations of magnesium (Mg2+) and pH levels, with parameters derived from Cu-only and Co-only toxicity data. The BLM-based toxic unit (TU) approach was used for prediction. Higher activities of Mg2+ linearly increased the EC50 of Cu and Co, supporting the concept of competitive binding of Mg2+ and metal ions in toxic action. The effects of pH on Cu and Co toxicity were related not only to free Cu2+ and Co2+ activity, respectively, but also to inorganic metal complexes. Stability constants for the binding of Cu2+, CuHCO3 +, CuCO3(aq), CuOH+, Mg2+, Co2+, CoHCO3 + and Mg2+ with biotic ligands were logK CuBL 5.87, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{log}\,{K}_{{{\rm{CuHCO}}}_{3}{\rm{BL}}}$$\end{document}logKCuHCO3BL 5.67, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{log}\,{K}_{{{\rm{CuCO}}}_{3}{\rm{BL}}}$$\end{document}logKCuCO3BL 5.44, logK CuOHBL 5.07, logK MgBL 2.93, logK CoBL 4.72, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{log}\,{K}_{{{\rm{CoHCO}}}_{3}{\rm{BL}}}$$\end{document}logKCoHCO3BL 5.81 and logK MgBL 3.84, respectively. The combinations of Cu and Co showed additive effects under different conditions. When compared with the FIAM-based TU model (root mean square error [RMSE = 16.31, R 2 = 0.84]), the BLM-based TU model fitted the observed effects better (RMSE = 6.70, R 2 = 0.97). The present study supports the BLM principles, which indicate that metal speciation and major cations competition need to be accounted for when predicting toxicity of both single metals and mixtures of metals.

With the acceleration of industrialization and the agriculture industry, large amounts of metals have been released into the soil, where they can pose risks to the environment 1 . For example, environmental copper (Cu) concentrations are frequently elevated because of sewage sludge and animal manure, burning of fossil fuels, industrial processes and widespread use of pesticides 2 . Nationwide surveys in China recently reported that 19% of agricultural soils are polluted, mainly with heavy metals and metalloids-among which 2.1% of samples exceeded China's soil environmental quality for Cu 3 . The minerals cobaltite, smaltite and erythrite contain cobalt (Co), and anthropogenic activities such as mining and smelting can lead to Co contamination of soil in some areas 4 . Soil contamination is complicated, and organisms in the soil are often simultaneously exposed to multiple metal elements 5 . However, risk-evaluation is usually based on the effects of individual metals in soils 6 . Therefore, investigation and assessment of the environmental effects on mixtures of metals such as Cu-Co may have practical significance.
The concentrations of individual toxicants in a mixture cannot be assessed based simply on their added concentrations. One of the models most commonly used to assess the toxicity of chemical mixtures is the toxic unit (TU) approach 5 , which is employed to calculate the sum of the concentrations of individual chemicals divided by their median effective concentrations (EC 50 ) 7  where, i is the identity of the metal, n is the number of metals, + M i 2 is the activity of free metal ions such as Cu 2+ and Co 2+ , and EC50 i is the median effective concentration of a single metal. The TU model is commonly used to normalize the toxic impacts and categorise the type of combined effect of metal mixtures. In multi-metal systems, a metal mixture is termed additive when TU is equal to 1 at 50% effect, while it is called synergistic or antagonistic when TU is less than or greater than 1 8 . The TU approach has been successfully used to predict joint toxicity and as a tool to assess the corresponding combined effects. However, conventional studies using the TU approach to examine the combined toxicity of multiple metals do not consider the speciation and phytotoxicity of metals under different environmental conditions.
Recently, the biotic ligand model (BLM) has become popular for evaluating metal bioavailability and toxicity in aquatic and terrestrial systems 9 . The critical assumption of the BLM is that metal toxicity depends on free metal ions (or other reactive metal species), which can react with biological binding sites and form a metal-biotic ligand (BL) complex. This model assumes that the cations (e.g., K + , Na + , Mg 2+ , Ca 2+ ) compete with metals for binding sites to mitigate the toxicity of free metal ions 10 . Generally, BLMs are considered to be useful tools for evaluating the toxic effects of metals on organisms. Most previous studies using the BLM to evaluate the impacts of metals have been applied to individual metals. However, it is also increasingly being applied to the assessment of the effects of mixtures of metals. Jho et al. found that data describing the toxicity of a single metal can be used in the BLM-based TU method to predict the joint toxicity of Cd-Pb mixtures to Vibrio fischeri 1 . Although terrestrial higher plants also appear suitable for these types of studies, they have rarely been used in investigations of alternative methods for joint toxicity of metals. To the best of our knowledge, only three reports have employed the BLM to estimate the combined toxicity of metals on joint toxicity [10][11][12] .
When developing the BLM for individual metals, some studies indicated that the toxicity of metals to plants was partly dependent on inorganic complexes (e.g., carbonate and hydroxide species) when exposed to relatively high pH 13 . However, no published studies of combined toxicity of mixtures of metals have considered the potential effects of these metal complexes 12,14,15 . Furthermore, most BLMs for metal mixtures have considered Ca 2+ competition, while ignoring Mg 2+ competition 1,10 . Earlier studies of BLM for single metals showed that Mg 2+ has a greater impact on the toxicity of some single metals (e.g., Co and Cu) than Ca 2+ 16, 17 . Therefore, the present study was conducted to investigate the joint toxicity of Cu-Co mixtures toward wheat (Triticum aestivum L.) in the presence of different Mg 2+ concentrations and under a wide range of pH values using the BLM-based TU method 1 . In addition, the traditional free ion activity model (FIAM)-based TU approach 18 was employed for combined toxicity prediction and the results were compared with those obtained using the BLM-based TU method.

Results
Toxicity of Cu 2+ and Co 2+ individually. A summary of the median effective concentration (EC 50 ) for wheat root elongation expressed as free Cu 2+ and Co 2+ activity at different Mg 2+ concentrations and pH level is given in Table 1. Dose-response curves for both free Cu 2+ and Co 2+ activities were established in Fig. 1. The increasing Mg 2+ activities resulted in corresponding reductions in free Cu 2+ activity by 2.5-fold and free Co 2+ activity by 10-fold, respectively (Table 1). Furthermore, a linear relationship was found between Mg 2+ activities and EC 50 of Cu 2+ or Co 2+ activities (Cu 2+ : p < 0.01, R 2 = 0.97; Co 2+ : p < 0.01, R 2 = 0.97). These results suggest that Mg 2+ can compete with Cu 2+ and Co 2+ for binding sites on wheat root and that it can alleviate the toxicity of Cu and Co. Previous studies showed that K + and Na + did not significantly influence the toxicity of Cu and Co to wheat and barley, and that the effects of Ca 2+ on the toxicity of Cu 2+ and Co 2+ were only slightly larger than K + and Na + , and smaller than the effects of Mg 2+ 13, 17, 19 . Therefore, the effects of K + , Na + and Ca 2+ were ignored and not incorporated into the BLM in the present study.
In the pH test, dose-response curves overlapped in the low pH range, while they differed at high pH values ( Fig. 1c-f). At pHs of 4.5 to 7.6, the values of EC 50 (Cu 2+ ) varied by 4.5-fold and those of EC 50 (Co 2+ ) varied by a  13 . Thus, we determined the relationship between Cu or Co species and their toxicity using a previously described method 13 . We found that the contribution of CuCO 3 (aq) and CuOH + to toxicity should be considered at high pH values. Similar results were found for Co. Furthermore, we determined stability constants (logK) for binding of Mg 2+ , Cu 2+ and Co 2+ of Cu-only and Co-only systems based on the BLM method for single metals (    Table 1.  Fig. 3 and Table 2.

Model
The root mean square error (RMSE) was calculated for each prediction (Fig. 3). The BLM-based TU approach performed better for predicting root elongation than the FIAM-based TU model based on RMSE and R 2 values. Specifically, the RMSE value between the observed and predicted RNE of the BLM (6.70) was much lower than that of the FIAM (16.31). This was likely due to the inclusion of the competition between Cu or Co and Mg 2+ for binding sites on wheat roots and consideration of the effects of CuHCO 3 + , CuCO 3 (aq), CuOH + and CoHCO 3 + , under various pH levels, on the estimation of the f values. Therefore, the proposed BLM-based TU was a better method to predict the toxicity of Cu-Co mixtures for wheat.

Discussion
The observed EC 50 for free Cu 2+ or Co 2+ activity increased up to 2.5-fold and 10-fold with increasing Mg 2+ concentrations, respectively. These findings indicate that Mg 2+ had a protective effect against the toxicity of Cu or Co, which is similar to previous studies [13,17,19] . For instance, Luo et al. reported that Mg 2+ can alleviate Cu toxicity for wheat in nutrient solution, and calculated that the EC 50 for free Cu 2+ increased by up to 3.7-fold. Wang et al. carried out similar root elongation tests on barley using a growth solution with a series of Mg 2+ (0.05-2 mM) concentrations and found an increase in EC 50 of free Cu 2+ activity by a factor of 3. Lock et al. 19 . reported that the increase of Mg 2+ concentrations could alleviate Co toxicity to barley, and calculated a 15.6-fold increase in EC 50 of free Co 2+ activity in solution. The effect of Mg 2+ on Cu 2+ and Co 2+ toxicity may be due to similar ionic radii of Mg 2+ , Co 2+ and Cu 2+ , or due to competition for transporters 21 . The present study revealed that the EC 50 of Cu or Co decreased by up to 4.1-fold with increasing pH values; however, there was no obvious linear relationship between H + and EC 50 . These findings differ from those for the BLM with H + , which can compete with free metal activity ions at BL. Therefore, it is unjustified to incorporate H + competition into the BLM. The present study indicated that the effects of pH on the toxicity of metals may be due to the contribution of inorganic metal complexes to toxicity. For instance, De Schamphelaere and Janssen suggested that CuOH + contributed to Cu toxicity in Daphnia magna 22 . For terrestrial plants, Wang et al. 13 indicated that when incorporating inorganic species of Cu 2+ into the BLM, the regression coefficient (R 2 ) between the measured and predicted EC 50 {Cu 2+ } values was as high as 0.97. These findings suggested that some species of inorganic metal complexes were toxic and should be considered in BLMs with high pH values. The present study indicates that incorporating inorganic metal complexes when assessing the joint toxicity of Cu-Co improves the predictive capacity of the BLM.
The binding constants derived in the present study for wheat can be compared with those reported for Cu-BLM 13 and Co-BLM 19 . The values of logK CuBL (5.87), K log CuHCO BL 3 (5.67), K log CuCO BL 3 (5.44), logK CuOHBL (5.07) and logK MgBL (2.93) in the present study were closer to the results reported by Wang et al. 13 . In addition, the values of logK CoBL (4.72) were slightly lower than those of (logK CoBL = 5.13) reported by Lock et al. 19 , whereas they were very similar to those (logK CoBL = 4.70) published by Garnham et al. 23 19 . Different exposure duration, endpoint, target tissue or BL, or mechanisms of Cu or Co uptake resulted in differences in binding constants 16 .
The present study showed that CoHCO 3 + had 11-fold higher binding affinity than Co 2+ , similar to the results of Deleebeeck et al. and Wang et al., who reported that the affinities of inorganic complexes of metals for the BL were higher relative to the free metal ion [13,24,25] . However, it is mechanistically very unlikely that BL constants for complexes of metals are very close to or higher than for the free metal ion. Therefore, the binding of inorganic metal complexes with BL needs further investigation and direct evidence.
It is widely believed that most metal combinations act additively. For example, Ownby and Newman et al. reported that inhibition of bioluminescence in Microtox assays was approximately additive in Cd-Cu mixtures 26 . Marra et al. found that interactions of Cu-Co mixtures were additive for rainbow trout and duckweed 27 . The present study indicated that the joint toxicity of Cu and Co was additive for wheat. These results are similar to those reported by Ownby and Newman et al. and Marra et al. 26,27 . However, some researchers pointed out that the joint toxicity of mixtures of metals may exhibit synergistic or antagonistic effects, rather than simply being additive. Versieren et al. 10 showed significant (p < 0.05) antagonistic interactions between Cu-Zn and lettuce at low Ca 2+ concentrations. Ince et al. described Cu-Co interactions that were synergistic at most test levels for Vibrio fisheri but that had additive effects under some individual conditions 28 . Thus, the difference in interactions between metal ions may be due to exposure time and test species 29 . The present study indicated that the BLM-based TU approach, which accounted for metal speciation and the integrated competition effects of the major cations, was more accurate at predicting the combined toxicity of Cu-Co mixtures than the FIAM-based TU model. Future studies should investigate the biological actions of metals in plant cell compartments following exposure to mixtures of metals to provide better insight into the mechanisms.

Conclusion
In summary, a BLM-based TU was developed to predict the combined toxicity of Cu-Co mixtures at different Mg 2+ activities and at various pH levels for wheat in nutrient solutions. Using the estimated constants based on the individual Cu and Co toxicity data, the BLM-based TU approach more accurately predicted the joint Cu-Co toxicity than the FIAM-based TU approach. Further research is required to investigate the toxicity of SCIENTIfIC REPORTS | 7: 9443 | DOI:10.1038/s41598-017-09940-5 metal mixtures in a wide range of natural or field soils before BLMs can be used for risk assessments of metal co-contaminated soil in the field.

Materials and Methods
Experimental design. Wheat 22 . After the radicle emerged (about 1 cm in length), six seeds were transferred to a nylon net fixed on the surface of the plastic culture pots containing 350 mL of the solution. The air temperature of the growth chamber was maintained at 20 ± 1 °C for 48 h in darkness and the culture pots were randomly placed in the growth chamber. The lengths of the longest roots on each seedling were measured after 2 d, and the mean value of the three replications for each test was used for data analysis. The relative net elongation (RNE) was calculated and expressed as the percentage relative to the control (Eqn 2): Prediction of Cu and Co speciation in solutions. WHAM 6.0 (Windermere Humic Aqueous Model) was used to calculate the speciation of Cu and Co 31 . The pH and measured total concentrations of Cu 2+ , Co 2+ , K + , Na + , Ca 2+ , Mg 2+ , Cl − and SO 4 2− were inputted into WHAM. For details, refer to Lofts et al. 31 . The calculated proportions of free Cu 2+ and Co 2+ activity (% of total Cu or Co) accounted for 11.7-78.6% and 61.5-78.6% with pH increasing from 4.5 to 7.6, respectively. The calculated proportions of CuHCO 3 + , CuCO 3 , CuOH + and CoHCO 3 + were 0.25-46.6, 0.00-25.0, 0.05-9.54 and 0.01-9.74%, respectively.
The Data Processing System 9.0 (DPS9.0), developed by Tang and Feng 32 , was used to estimate the parameter values of the BLM-based and FIAM-based TU models.

Mathematical description of the BLM and derivation of parameters. The BLM methodology is
based on the assumption that stability constants remain the same under various physico-chemical conditions 20 . The following is a short mathematical description of the BLM along with the equations required to understand the calculations. Based on the BLM assumption, the fraction (f) of the total number of BL sites occupied by Cu 2+ or Co 2+ is given by the following equation when the competing cations and toxicity of inorganic metal complexes are considered: The wheat root elongation was correlated with TU f or TU M and was assumed to follow a log-logistic doseresponse relationship according to Thakali et al. 33 = + β ( ) where β M is a fitting parameter determining the slope of the dose-response curve; x is the value of the toxicity index, i.e. TU f or TU M ; and x 50 is the value of TU at 50% RNE. Eqn. 5 was applied to predict the mixture toxicity effects and compare the BLM-TU and FLAM-TU approaches.