Environmental Toxicology and Chemistry, Vol. 33, No. 9, pp. 1988–1995, 2014 # 2014 SETAC Printed in the USA

DELINEATING ION-ION INTERACTIONS BY ELECTROSTATIC MODELING FOR PREDICTING RHIZOTOXICITY OF METAL MIXTURES TO LETTUCE LACTUCA SATIVA T. T. YEN LE,*yz PENG WANG,x MARTINA G. VIJVER,k THOMAS B. KINRAIDE,#** A. JAN HENDRIKS,y and WILLIE J.G.M. PEIJNENBURGzk

yDepartment of Environmental Sciences, Institute for Water and Wetland Research, Radboud University Nijmegen, Nijmegen, the Netherlands zLaboratory for Ecological Risk Assessment, National Institute for Public Health and the Environment, Bilthoven, the Netherlands xSchool of Agriculture and Food Sciences, The University of Queensland, St Lucia, Queensland, Australia kDepartment of Conservation Biology, Institute of Environmental Sciences (CML), Leiden University, Leiden, the Netherlands #Agricultural Research Service, United States Department of Agriculture, Beaver, West Virginia (Submitted 15 March 2014; Returned for Revision 12 April 2014; Accepted 21 May 2014) Abstract: Effects of ion–ion interactions on metal toxicity to lettuce Lactuca sativa were studied based on the electrical potential at the

plasma membrane surface (c0). Surface interactions at the proximate outside of the membrane influenced ion activities at the plasma membrane surface ({Mnþ}0). At a given free Cu2þ activity in the bulk medium ({Cu2þ}b), additions of Naþ, Kþ, Ca2þ, and Mg2þ resulted in substantial decreases in {Cu2þ}0. Additions of Zn2þ led to declines in {Cu2þ}0, but Cu2þ and Agþ at the exposure levels tested had negligible effects on the plasma membrane surface activity of each other. Metal toxicity was expressed by the {Mnþ}0–based strength coefficient, indicating a decrease of toxicity in the order: Agþ > Cu2þ > Zn2þ. Adsorbed Naþ, Kþ, Ca2þ, and Mg2þ had significant and dose-dependent effects on Cu2þ toxicity in terms of osmolarity. Internal interactions between Cu2þ and Zn2þ and between Cu2þ and Agþ were modeled by expanding the strength coefficients in concentration addition and response multiplication models. These extended models consistently indicated that Zn2þ significantly alleviated Cu2þ toxicity. According to the extended concentration addition model, Agþ significantly enhanced Cu2þ toxicity whereas Cu2þ reduced Agþ toxicity. By contrast, the response multiplication model predicted insignificant effects of adsorbed Cu2þ and Agþ on the toxicity of each other. These interactions were interpreted using c0, demonstrating its influence on metal toxicity. Environ Toxicol Chem 2014;33:1988–1995. # 2014 SETAC Keywords: Metal mixture

Model

Plant

Toxicity

Membrane surface

Interactions

in these studies, whereas interactions were found to occur at the plasma membrane surface as well as within organisms [11]. Electrostatic modeling has been applied in assessment of metal bioavailability and toxicity, taking into account plant–ion interactions at the cell membrane surface (e.g., Cu-wheat; Cu, Ni, Cd-pea; Ni-barley; Mn-cowpea) [12–15]. The basis of this method is the importance of the electrical potential at the plasma membrane surface (denoted as c0) in the uptake and transport of ions. The potential c0 is induced by the intrinsically negative charge at the plasma membrane surface and affects the interactions between ions and plants [12,13,16–20]. The influence of c0 is usually assessed through its dual roles in metal toxicity: c0 has effects on both the ion distribution between the plasma membrane surface and the bulk medium and the electrical driving force for ion transport through the plasma membrane [13,18]. The present study aimed at integrating c0 into the assessment of ion–ion surface and internal interactions and metal rhizotoxicity to the root growth of lettuce, Lactuca sativa. The interactions were incorporated in estimating Cu2þ toxicity in the presence of prevalent cations, that is, Naþ, Kþ, Ca2þ, and Mg2þ, and in predicting toxicity of Cu2þ–Zn2þ and Cu2þ–Agþ mixtures. Surface interactions occur between ions at the proximate outside of the membrane, affecting the activity of ions at the plasma membrane surface. These interactions are directly responsive to c0, that is, the ionic composition of the bulk medium influences the activity of ions at the plasma membrane surface through changes in c0. Internal interactions occur between free ions adsorbed on the plasma membrane, affecting metal transport through the membrane into cells and subsequent metal toxicity. The role of c0 in these interactions is

INTRODUCTION

Besides chemical speciation in the environment, interactions of ions with organisms are important in controlling metal bioavailability [1,2]. Similar to interactions between ions and organisms at the biological surface, interactions between different ions influence metal bioavailability and toxicity [3,4]. Biological actions of metal ions in mixtures might deviate from their actions singly [5]. For example, exposure to metal mixtures, (e.g., Cu, Pb, and Zn), at concentrations below their individual environmental quality guideline levels resulted in adverse effects because of the interactions among these metals [6]. The ion–ion interactions are therefore of high concern and increasingly integrated in estimating bioavailability and toxicity of metals. Interactions of toxic metal ions with Hþ and macro cations (e.g., Naþ, Ca2þ, Mg2þ, and Kþ) were, for example, integrated in predicting metal bioavailability in the biotic ligand model [7,8]. Initial efforts have been put in taking into account ion-ion interactions in predicting mixture toxicity [9,10]. The assumption of (non)competitive binding was included in predicting metal bioavailability by the biotic ligand model [9]. Thereupon, the influence of ion–ion interactions in determining mixture toxicity was also evaluated [10]. However, discrimination between surface and internal interactions was not included All Supplemental Data may be found in the online version of this article. *Address correspondence to [email protected] **Retired. Published online 26 May 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/etc.2643 1988

Surface and internal ion-ion interactions

expressed by the changes in the surface-to-surface transmembrane potential difference, being the driving force for transport through the membrane. Because the transmembrane potential difference was not measured, the internal interactions were expressed by observed effects on the toxicity, which was related to the free metal ion activity at the plasma membrane surface. MATERIALS AND METHODS

Test species and toxic endpoint

Toxic effects of metals were assessed on lettuce, L. sativa, in terms of the inhibition of the root elongation after 4 d exposure. This plant species was chosen because of its high capacity to accumulate metals [21,22]. In addition, L. sativa was recommended by the Organisation for Economic Co-operation and Development as a suitable test species for toxicity studies [23]. Preparation of the test solutions

Steiner solution was used as the test medium [24]. The ionic composition of the Steiner solution used for chemical speciation calculations is given in Table S1, Supplemental Data. The solution pH was stabilized at 7.0 by MOPS (3-[N-morpholino] propane sulfonic acid) buffering at 0.75 g/L and NaOH [8]. The MOPS buffer was chosen because of its negligible effects on biotic and abiotic factors. Specifically, this chemical does not form complexes with metals [25]. It does not affect the toxicity of effluents and sediment porewaters and, consequently, was recommended by the US Environmental Protection Agency [26]. Moreover, at the concentration of 0.75 g/L, the MOPSbuffering does not influence metal toxicity to Daphnia magna and Pseudokirchneriella subcapitata [27]. Metal ions investigated (Cu2þ, Agþ, and Zn2þ) were added into the Steiner solution as nitrate salts. In Cu and Ag toxicity tests, pH and Cu2þ/Agþ activities in the exposure solutions were daily checked and adjusted by adding HNO3/KOH and Cu(NO3)2/AgNO3 during the exposure period. In Zn toxicity tests, solutions were renewed daily because Zn2þ activities were not measured by electrodes in the present study. Similarly, solutions tested for mixture toxicity were renewed every day during the exposure period to avoid interferences of different metal ions on the measurements of the free metal ion activity by the ion-selective electrodes. Metal measurements and speciation in the hydroponic solution

Free ion activities of Hþ, Cu2þ, and Agþ were measured with ion-selective electrodes (Metrohm). Free Zn2þ activities were computed from the total Zn2þ concentrations in the solution by using the Windermere Humic Aqueous model (WHAM) VI with Steiner solution as the default medium [28]. A disadvantage of the WHAM VI is the exclusion of Agþ in this version. A survey was performed to investigate effects of Agþ at the activity range studied on the activities of other cationic constituents in the Steiner solution by using the Chemical Equilibria in Aquatic Systems model [29]. The results from this survey indicated negligible influence of Agþ on the activities of other cations. Therefore, activities of the cations in solution (except for Cu2þ and Agþ) were specified by the WHAM VI model, although Agþ is excluded in this speciation model.

Environ Toxicol Chem 33, 2014

the surface of a glass beaker with the roots immersed in the medium for 4 d. For each beaker, 4 plants were put in. The root growth (growth) of lettuce exposed to each solution was determined as the average increase in the root length of the 4 plants grown in the solution after 4 d exposure compared with the initial length. In total, 180 toxicity tests were carried out to investigate effects of Naþ, Kþ, Ca2þ, and Mg2þ (namely, prevalent cations in the present study) on Cu2þ toxicity. In the tests, Naþ, Kþ, Ca2þ, and Mg2þ were added to the Steiner solution up to the concentrations of 10, 20, 10, and 20 mM, respectively (Supplemental Data, Table S2). With each combination of these cations, 8 to 10 toxicity tests were carried out at varying free ion activities of Cu2þ. In addition, 238 toxicity tests were performed in the assessment of joint toxicity of Cu2þ, Zn2þ, and Agþ, including 122 tests without additions of Agþ and 116 tests without additions of Zn2þ to the Steiner solution (Supplemental Data, Tables S3 and S4). The free ion activities of Cu2þ ¼ Zn2þ, and Agþ studied varied in the ranges: Cu2þ ¼ 1010 to 106 M; Zn2þ ¼ 106 to 103 M; and Agþ ¼ 108 to 107 M. These ranges were selected from preliminary toxicity tests carried out at different activities of these cations in solution varying from the background level of the Steiner solution to the extremely toxic level. Metal ion activity at the plasma membrane surface

The electrical potential at the plasma membrane surface c0 was calculated from free ion activities of all cations in the solution and the equilibrium constants for binding to the plant plasma membrane by using the model developed by Kinraide and Yermiyahu [30]. The equilibrium constants in the model were determined based on both adsorption studies and zetapotential measurements. In other words, both chemical binding and electrostatic interactions were taken into account by these authors in determining the binding constants of metals for the plasma membrane. The calculated surface potential was then used to estimate free metal ion activities at the plasma membrane surface according to the Nernst Equation (Equation 1):
 F  n  c0 fMnþ g0 ¼ fMnþ gb  exp  RT

ð1Þ

where {Mnþ}0 and {Mnþ}b are free ion activities of metal Mnþ at the plasma membrane surface and in the bulk medium, respectively; n is the charge of metal ion Mnþ; F is the Faraday constant; R is the universal gas constant; and T is the experimental temperature (288 K in the present study). Toxicity of single metals

In the present study, metal toxicity was assessed based on free metal ion activities at the membrane surface, which were estimated by the electrostatic model. In other words, the predictions are based on another estimate that was not evaluated by measurements. Toxicity of single metals was found to follow the Weibull Equation [31]. Accordingly, the response of lettuce exposed to single metals expressed by the root growth (growth; mm) can be related to {Mnþ}b (Equation 2) or {Mnþ}0 (Equation 3) [10]: Growth ¼

Toxicity assays

Seeds of L. sativa were germinated in the Steiner solution for 4 d under a normal light cycle of 16:8 h light:dark at 15 8C, which is in the range of the average temperature in the Netherlands. Germinated plants were fixed in parafilm straps that floated on

1989

Growth ¼

b exp½ðc  fMnþ gb Þd  b exp½ðc  fMnþ g0 Þd 

ð2Þ

ð3Þ

1990

Environ Toxicol Chem 33, 2014

T.T.Y. Le et al.

where coefficient b (mm) is the growth of lettuce roots when the metal ion is not present in the solution (i.e., {Mnþ}b ¼ 0) or at the plasma membrane surface (i.e., {Mnþ}0 ¼ 0). Coefficient c (mM1) reflects the metal-specific strength of toxicity. Its value increases with increasing strength of metal toxicity. Coefficient d (dimensionless) is a shape parameter. When coefficient d is greater than 1, the curves are sigmoidal. Toxicity of Cu2þ in the presence of Naþ, Kþ, Ca2þ, and Mg2þ

Prevalent cations, such as Naþ, Kþ, Ca2þ, and Mg2þ, were assumed to act as osmoticants; that is, intoxication resulted from the reduction in water potential [32]. Consequently, in the present study, their effects on Cu2þ toxicity were evaluated by a common term “osmolarity.” Osmotic effects of these prevalent cations on Cu2þ toxicity were evaluated by incorporating expansion coefficients into the strength coefficient in Equation 3. Osmolarity was calculated based on the osmotic coefficients and the total concentrations of salts in the solution [33]. Furthermore, the relationship between root elongation and osmolarity followed a sigmoidal curve [32]. Therefore, toxicity of Cu2þ expressed by the root growth (growth; mm) can be written as a function of the surface activity of Cu2þ ({Cu2þ}0; mM) and osmolarity (Os; mM) as follows: Growth ¼

b h d i ð4Þ exp c1  ð1 þ c10  Os þ c20  Os2 Þ  fCu2þ g0

where coefficients c10 and c20 represent osmotic effects on Cu2þ toxicity. The effects were considered statistically significant if the 95% confidence interval (CI) of these coefficients estimated by the regression analysis did not encompass zero. Toxicity of Cu2þ, Zn2þ, and Agþ in noninteractive mixtures

If mixture components do not interact with each other, the growth of lettuce roots exposed to the mixture can be written according to the conventional concepts of concentration addition and response multiplication, assuming no interactions between the mixture constituents [34]. Noninteractive concentration addition model. According to the concentration addition model, mixture substances are supposed to act by the same mechanism [35]. Subsequently, the growth of lettuce roots (growth; mm), after exposure to a noninteractive mixture of Cu2þ, Zn2þ, and Agþ, can be determined based on the free ion activity of these metals at the plasma membrane surface according to the following equation: Growth ¼

b h d i exp c1  fCu2þ g0 þ c2  fZn2þ g0 þ c3  fAgþ g0

ð5Þ where c1, c2, and c3 (mM1) are the strength coefficients of toxicity of Cu2þ, Zn2þ, and Agþ in their noninteractive mixtures, respectively; d (dimensionless) is the shape coefficient of the dose–response curve describing toxicity of these metal ions according to the concentration addition model [10,32,34]. Noninteractive response multiplication model. The response multiplication model is based on the assumption that mixture components have different modes of action of toxicity [36]. Therefore, the response of organisms to noninteractive mixtures can be expressed as a multiplicative function of the response of the organisms following exposure to each constituent separately [10,32,34]. Accordingly, the root growth of lettuce (growth;

mm) exposed to a noninteractive mixture of Cu2þ, Zn2þ, and Agþ can be written as Equation 6 according to the concept of response multiplication: Growth ¼

h

exp c1 

fCu2þ g

d1 0

b  d  d i þ c2  fZn2þ g0 2 þ c3  fAgþ g0 3

ð6Þ where c1, c1, and c3 (mM1) are the strength coefficients of toxicity of Cu2þ, Zn2þ, and Agþ in their noninteractive mixtures, respectively; and d1, d2, and d3 (dimensionless) are the shape coefficients of the dose–response curves describing toxicity of Cu2þ, Zn2þ, and Agþ according to the response multiplication model, respectively [10,32,34]. Equations 5 and 6 are based on the conventional concept of concentration addition and response multiplication, that is, the presence of 1 metal at the membrane surface does not affect the toxicity of another. In other words, toxicity of Cu2þ, Zn2þ, and Agþ after exposure to their noninteractive mixture is assumed to be similar to the toxicity of these metals in a single exposure. Subsequently, the strength coefficient and the slope parameter describing toxicity of single metals as in Equation 3 can be modeled by the regression analysis using toxicological data on metal mixtures. This assumption could not be examined in the present study because both Cu and Zn were components of the Steiner solution. Toxicity of interactive mixtures of Cu2þ–Zn2þ and Cu2þ–Agþ

If metals adsorbed after exposure to mixtures interact with each other, that is, metals adsorbed affect the toxicity of another, the interactions can be taken into account in estimating the joint toxicity by expanding the conventional concentration addition and response multiplication models. In particular, expansion coefficients that describe interactive effects can be integrated into the strength coefficients in Equations 5 and 6. For example, the strength coefficient of Cu2þ toxicity in interactive mixtures of Cu2þ and Zn2þ can be expressed as: c1  ¼ c1  ð1 þ c12  fZn2þ g0 Þ

ð7Þ

or c1  ¼

c1 1 þ c12  fZn2þ g0

ð8Þ

where c1 (mM1) is strength coefficient of Cu2þ toxicity in Cu2þ–Zn2þ mixture exposure; c1 (mM1) is the strength coefficient of Cu2þ toxicity in the medium free of Zn2þ; and expansion coefficient c1-2 (mM1) describes effects of Zn2þ on Cu2þ toxicity. The interactive effect was considered statistically significant when 95% CI of the expansion coefficient estimated did not encompass zero. The comparison of the expansion coefficient with zero indicates whether one substance alleviates or enhances the toxicity of the others (Supplemental Data, Section SA). Because of the interactions, the toxicological parameters of substances, that is, strength coefficient and slope, in mixture exposure may be different from those in single exposure. The selection between Equation 7 and Equation 8 was based on regression results in favor of higher statistical significance. If estimates of the expansion coefficient were found not to significantly deviate from zero, the expansion coefficient was left out of the equation. A more detailed description of equation derivation is given in Supplemental Data, Section SA.

Surface and internal ion-ion interactions

Regression analyses

Coefficients in the equations (Equation 1–Equation 8) shown in the present study were determined by multiple nonlinear regression analyses using the SYSTAT software. One toxicological term in the equations was considered significant if its 95% CI statistically deviated from zero, in other words, not encompassing zero. The power of significance increases with an increase in the absolute value of the ratio between the estimate of the parameter and the asymptotic standard error, that is, parameter/asymptotic standard error (ASE) in the regression result. Data generated from all tests on mixture toxicity were used to assess effects of Cu2þ, Agþ, and Zn2þ in noninteractive mixtures, assuming that the presence of 1 metal does not affect the biological actions of the other metals. Toxicological data from 122 tests without additions of Agþ were used to assess toxicity of interactive mixtures of Cu2þ and Zn2þ. Toxicity of the interactive Cu2þ–Agþ mixtures was evaluated using 116 tests without additions of Zn2þ assuming negligible effects of Zn2þ at the background concentration in the default medium. RESULTS

Surface interactions between Cu2þ and Naþ, Kþ, Ca2þ, and Mg2þ

Additions of Naþ, Kþ, Ca2þ, and Mg2þ at the concentrations tested increased c0 from 18 to 4.0 mV (Supplemental Data, Figure S1). This variation in c0 then resulted in a decrease of approximately half of 1 order of magnitude in {Cu2þ}0 at a given exposure level expressed by {Cu2þ}b (Supplemental Data, Figure S1). Moreover, a statistically significant relationship was found between {Cu2þ}b and {Cu2þ}0 (p < 0.0001; F ¼ 2910; n ¼ 180; r2 ¼ 0.94). In other words, {Cu2þ}0 generally increased with increasing {Cu2þ}b regardless of the prevalent cations added. At the lowest exposure level, {Cu2þ}0 (in logarithm scale) did not linearly decrease with increasing c0. Furthermore, in the presence of the prevalent cations at varying concentrations, {Cu2þ}0 (n ¼ 180; r2 ¼ 0.79) was a better indicator of Cu2þ toxicity compared with {Cu2þ}b (n ¼ 180; r2 ¼ 0.65) according to nonlinear regression with the Weibull Equation (Supplemental Data, Table S5). Surface interactions in mixtures of Cu2þ–Zn2þ and Cu2þ–Agþ

At the exposure levels studied, surface interactions occurred between the toxic metals after exposure to their mixtures, influencing their activities at the plasma membrane surface. The effects depended on binding constants and concentrations of metals. Additions of Zn2þ resulted in substantial decreases in {Cu2þ}0 as shown by a difference of approximately half of 1 order of magnitude in {Cu2þ}0 at a given {Cu2þ}b with varying concentrations of Zn2þ added (Supplemental Data, Figure S2B). By contrast, Cu2þ had low effects on {Zn2þ}0 as indicated by negligible variations in {Zn2þ}0 at a given {Zn2þ}b in mixtures with different {Cu2þ}b (Supplemental Data, Figure S2D). This difference was attributed to a stronger decrease in the negativity of c0 induced by the addition of Zn2þ than by the addition of Cu2þ at the exposure levels tested (Supplemental Data, Figure S2A and S2C). The presence of Cu2þ and Agþ in test solutions induced only small variations in the negativity of c0 and therefore limited changes in {Agþ}0 and {Cu2þ}0 with additions of Agþ or Cu2þ, respectively (Supplemental Data, Figure S3A-D). Internal interactions between Cu2þ and Naþ, Kþ, Ca2þ, and Mg2þ

Osmolarity had significant effects on Cu2þ toxicity; that is, expansion coefficients c10 and c20 in Equation 4 did not

Environ Toxicol Chem 33, 2014

1991

encompass zero (Supplemental Data, Table S6). Approximately 72% of the variability in the growth of lettuce roots could be explained by Equation 4 (n ¼ 180; r2 ¼ 0.72; Figure 1; Supplemental Data, Figure S4). Moreover, the opposite signs of the expansion coefficients (c10 < 0 and c20 > 0) indicate the dependence of the effects (alleviation or enhancement) on the magnitude of osmolarity or on the concentrations of Naþ, Kþ, Ca2þ, and Mg2þ in solution. The osmotic effects of adsorbed Naþ, Kþ, Ca2þ, and Mg2þ on Cu2þ toxicity were low, as the strength coefficient of Cu2þ toxicity in the medium free of these cations (i.e., c1 in Equation 4) was 5 and 10 orders of magnitude higher than expansion coefficients c10 and c20, respectively (Supplemental Data, Table S6). This accounts for a lack of improvement in predicting Cu2þ toxicity from incorporating the internal interactions between Cu2þ and the prevalent cations (r2 ¼ 0.72; Supplemental Data, Table S6) compared with the exclusion of these interactions (r2 ¼ 0.79; Supplemental Data, Table S5). Toxicity of Cu2þ, Zn2þ, and Agþ in noninteractive mixtures

The conventional concepts of concentration addition (Equation 5) and response multiplication (Equation 6) models, which assume no interactions in mixtures, performed equally well in estimating toxicity of Cu2þ, Zn2þ, and Agþ in noninteractive mixtures (r2 ¼ 0.83 and 0.86, respectively; Supplemental Data, Tables S7 and S8). Based on the strength coefficients estimated by either the concentration addition or the response multiplication model, Agþ and Cu2þ were far more rhizotoxic than Zn2þ, whereas Agþ was slightly more toxic than Cu2þ (Table 1). The strength coefficients of Cu2þ and Zn2þ toxicity predicted by the concentration addition model were not statistically decidedly different from those predicted by the response multiplication model; that is, the 95% CI of the strength coefficients estimated by these models overlapped (Table 1). The strength coefficient of Agþ toxicity predicted by the response multiplication model was slightly higher than the estimate by the concentration addition model (Table 1). Toxicity of interactive Cu2þ–Zn2þ mixture

Toxicity of interactive mixtures of Cu2þ and Zn2þ expressed by the root growth (growth; mm) could be written as Equations 9 and 10 according to the extended concentration addition and response multiplication models, respectively, as coefficients in these equations statistically deviated from zero (Supplemental Data, Tables S9 and S10):


Growth ¼ exp




Growth ¼ exp

b c1 fCu2þ g0 1þc12 fZn2þ g0

c1 fCu2þ g0 1þc12 fZn2þ g0

b d 1

d 

c fZn2þ g

þ 1þc221 fCu2þ0 g



þ c2 

fZn2þ g

ð9Þ

0

d 2

 ð10Þ

0

where c1 (mM1) is the strength coefficient of Cu2þ toxicity in the solution without Zn2þ; c2 (mM1) is the strength coefficient of Zn2þ toxicity in the solution free of Cu2þ; d (dimensionless) represents the slope of the curve describing toxicity of Cu2þ and Zn2þ in their interactive mixtures according to the extended concentration addition model; d1 and d2 (dimensionless) are slope parameters representing toxicity of Cu2þ and Zn2þ,

1992

Environ Toxicol Chem 33, 2014

T.T.Y. Le et al.

Figure 1. The root growth of lettuce (growth; mm) exposed to Cu2þ in the presence of Naþ, Kþ, Ca2þ, and Mg2þ is plotted as a function of the free Cu2þ activity at the plasma membrane surface ({Cu2þ}0; mM) and the osmolarity (Os; mM). The surface represents the estimations based on the regression analysis, and dotted points represent the experimental data.

respectively, in their interactive mixture on the basis of the extended response multiplication model; c1-2 (mM1) represents effects of Zn2þ on Cu2þ toxicity; and c2-1 (mM1) represents effects of Cu2þ on Zn2þ toxicity. According to the extended concentration addition model (Equation 9), adsorbed Cu2þ and Zn2þ had significant effects on the toxicity of each other; that is, the 95% CIs of the expansion coefficients c1-2 and c2-1 did not encompass zero (Supplemental Data, Table S9). Zinc significantly reduced Cu2þ toxicity (c12þ significantly increased Zn2þ toxicity (c22 > 0), whereas Cu 1 < 0). Based on the extended response multiplication model (Equation 10), Zn2þ significantly alleviated Cu2þ toxicity, that is, c1-2 significantly deviated from zero, whereas Cu2þ did not have significant effects on Zn2þ toxicity (Supplemental Data, Table S10). Moreover, the extended concentration addition and response multiplication models had comparable predictive potential in estimating toxicity of interactive mixtures of Cu2þ and Zn2þ (n ¼ 122; r2 ¼ 0.92; Figure 2A and 2B; Supplemental Data, Figure S5). Toxicity of interactive Cu2þ–Agþ mixture

Figure 2. The root growth of lettuce (growth; mm) as a function of the free ion activity of Cu2þ ({Cu2þ}0; mM) and the free ion activity of Zn2þ ({Zn2þ}0; mM) at the plasma membrane surface according to the extended concentration addition model (A) and the extended response multiplication (B) model. The surfaces represent the estimations based on the regression analysis, and dotted points represent the experimental data.

because the 95% CI of the coefficients in this equation did not encompass zero (Supplemental Data, Table S11): Growth ¼

Based on the extended concentration addition model, the growth of lettuce roots (growth; mm) after exposure to interactive mixtures of Cu2þ and Agþ could be described by Equation 11,


 c fCu2þ g exp 1þc113 fAgþ0 g

0

b þc3 ð1þc31 fCu2þ g0 ÞfAgþ g0

d  ð11Þ

Table 1. Estimates of coefficients (mM-1) describing strength of toxicity of Cu2þ (c1), Zn2þ (c2), and Agþ (c3) individually found in the present study for Lactuca sativa according to the conventional concepts of concentration addition and response multiplication models and in the study of Kopittke et al. [37] for Vigna unguiculata after exposure to single metals.a Strength coefficient (mM1)

Source

Species

Present study

Lactuca sativa

Kopittke et al. [37]

Vigna unguiculata

a

95% confidence intervals (CI) are provided

Cu2þ

Zn2þ

Agþ

Model

(c1, 95% CI)

(c2, 95% CI)

(c3, 95% CI)

Concentration addition Response multiplication Single-metal exposure

0.67 (0.59–0.76) 0.73 (0.63–0.83) 0.154

1.66  103 (1.45  103–1.87  103) 1.75  103 (1.54  103–1.95  103) 3.68  103

1.43 (1.24–1.63) 1.81 (1.68–1.95) 6.71

Surface and internal ion-ion interactions

where c1 (mM1) is the strength coefficient of Cu2þ toxicity in the medium free of Agþ; c3 (mM) is the strength coefficient of Agþ toxicity in the solution without Cu2þ; d (dimensionless) is the slope parameter describing toxicity of Cu2þ and Agþ in their interactive mixtures according to the extended concentration addition model; c1-3 (mM1) represents effects of Agþ on Cu2þ toxicity; and c3-1 (mM1) represents effects of Cu2þ on Agþ toxicity. In the assessment based on the extended concentration addition model (Equation 11), the deviation of expansion coefficients c1-3 and c3-1 from zero as shown in Table S11 indicates significant effects of Cu2þ and Agþ adsorbed on the toxicity of each other. Specifically, Agþ significantly increased Cu2þ toxicity, whereas Cu2þ had significant alleviative effects on Agþ toxicity. Moreover, these effects were considerable, as shown by negligible differences between the estimates of strength coefficients and expansion coefficients. Approximately 80% of the variability in the growth of lettuce roots accumulated with Cu2þ and Agþ could be explained by {Cu2þ}0 and {Agþ}0 (mM) (n ¼ 116; r2 ¼ 0.80; Figure 3; Supplemental Data, Figure S6). By contrast, no expansion coefficients that significantly deviate from zero were found to be integrated into the strength coefficients of Cu2þ and Agþ toxicity in the extended response multiplication model. According to the extended response multiplication model, adsorbed Cu2þ and Agþ did not have significant effects on the toxicity of each other, in other words. DISCUSSION

Toxicity of Cu2þ, Zn2þ, and Agþ

The order of toxicity strength found in the present study for L. sativa, that is, Agþ > Cu2þ > Zn2þ, was consistent with the findings of Kopittke et al. [37] for Vigna unguiculata (Table 1). The order of toxicity to L. sativa observed in the present study in addition agreed with the predictions by the biotic ligand model [9]. These studies show that the lowest amount of Agþ at the membrane surface or bound to biotic ligand was required at the 50% effect level, followed by Cu2þ and Zn2þ [9]. Moreover, the

Figure 3. The root growth of lettuce (growth; mm) as a function of the free ion activity of Cu2þ ({Cu2þ}0; mM) and the free ion activity of Agþ ({Agþ}0; mM) at the plasma membrane surface according to the extended concentration addition model. The surface represents the estimations based on the regression analysis, and dotted points represent the experimental data.

Environ Toxicol Chem 33, 2014

1993

negligible differences in the predictions by the concentration addition and response multiplication models based on the assumption of no interactions between the metals are similar to the observation by Backhaus et al. [38] for organic compounds. Interactions between Cu2þ and Naþ, Kþ, Ca2þ, and Mg2þ

The higher predictive potential of {Mnþ}0 over {Mnþ}b found in the present study is consistent with the results reported in previous works [13,18,39], demonstrating the significance of integrating surface interactions in modeling metal toxicity. The variation in c0 induced by additions of the prevalent cations as calculated in the present study is substantially lower than the level previously observed [13,16]. This is because of high background concentrations of cations in the Steiner solution, which led to an electrical potential of approximately –20 mV at the plasma membrane of lettuce exposed to the default medium. Furthermore, the inconsiderable reduction in the negativity of c0 could not lead to decreases in {Cu2þ}0 when {Cu2þ}b increased. This accounts for the significant relationship between {Cu2þ}b and {Cu2þ}0 while Naþ, Kþ, Ca2þ, and Mg2þ were added to the solution. For essential metals such as Cu, their uptake at low exposure levels can be enhanced by homeostasis to maintain certain internal concentrations. This might contribute to the nonlinear relationship between c0 and {Cu2þ}0 (at the logarithm scale) as presented in the Results section below. At a given {Mnþ}b, the decrease in {Mnþ}0 after the reduction in the negativity of c0 potentially contributes to the decreasing metal toxicity with additions of the prevalent cations [8,39,40,41]. The inconsistent variations in Cu2þ toxicity with varying concentrations of the prevalent cations, which could not be interpreted in terms of competitive binding [8], indicates that the effects of the prevalent cations cannot be completely explained by the impacts on the surface activity only. Furthermore, the observed influence of Naþ, Kþ, Ca2þ, and Mg2þ might be related to the dependence of osmotic effects on the concentration of the prevalent cations. Interactions in metal mixtures

In electrostatic terms, the surface interactions between Cu2þ and Zn2þ as well as between Cu2þ and Agþ observed in the present study could be interpreted based on their binding constants in combination with the exposure levels tested. For example, Zn2þ led to substantial decreases in the negativity of c0 and then in {Cu2þ}0 because exposure concentrations of Zn2þ were higher than those of Cu2þ in the test solutions, whereas these metals have similar affinity for the plasma membrane [20,30]. This impact declines with decreasing differences between their activities in the solution. The addition of Cu2þ at activities some orders of magnitude lower than Zn2þ led to negligible changes in c0. These results indicate strong dependence of the surface interactions between different metals on the exposure level tested. Moreover, these dose-dependent interactions potentially account for different effects of Cu2þ on Zn2þ uptake, which were reported at different exposure levels studied and cannot be explained by competition for binding sites [37,42–45]. The internal interactions predicted might be explained by different mechanisms, although in the present study the equations developed were based purely on statistical analyses. Assuming metals act by the same mechanism (concentration addition model), their interactions may be accounted for by competitive binding. In particular, Zn2þ and Cu2þ might inhibit the binding of the other metal to sites of toxic actions. However, different impacts were induced by the competition. The

1994

Environ Toxicol Chem 33, 2014

replacement of Cu2þ by Zn2þ at the binding inhibits toxic effects because Zn2þ is far less toxic than Cu2þ as presented previously. By contrast, toxic effects increase when Zn2þ is replaced by Cu2þ because Cu2þ has higher toxicity strength. A similar explanation might contribute to the predictions by the interactive concentration addition model for the Cu2þ–Agþ mixtures. Assuming mixture components have different modes of actions (response multiplication model), competitive binding cannot account for the different interactions predicted in mixtures of Cu2þ–Zn2þ and Cu2þ–Agþ. By contrast, the interaction patterns estimated by this model might be related to c0. The negligible changes in c0 with the addition of Cu2þ or Agþ as presented in Supplemental Data, Figures S3A and S3C, lead to limited variations in the transmembrane potential difference. This may account for insignificant interactions between Cu2þ and Agþ adsorbed as predicted by the response multiplication model. Substantial changes in c0 induced by addition of Zn2þ might lead to changes in the transmembrane potential difference, contributing to significant alleviative effects of Zn2þ on Cu2þ toxicity. By contrast, insignificant interactive effects of Cu2þ on Zn2þ toxicity may be related to negligible variations in c0 on the addition of Cu2þ at the exposure levels tested in the present study. These results demonstrate that c0 might provide insight into ion–ion interactions. The capacity of the modeling method based on the surface potential for explaining the interactions observed may be attributed to the consideration of both electrostatic interactions and chemical binding in determining binding constants of metals for the plasma membrane, and then c0. However, significant uncertainties are included in the c0–based method for predicting internal interactions. Model selection was mainly based on statistical assessment, although the predictions might be explained by different mechanisms. Metal toxicity was predicted based on the estimate of free metal ion activities, which were not evaluated by comparison with measurements. Uncertainties in modeling ion–ion interactions would be expected to be mainly associated with predicting internal interactions when the electrostatic model has been demonstrated to work well in estimating surface activity of metal ions [13– 15,37,39]. Another disadvantage of our modeling approach is its limited potential for application to field assessment. A large number of coefficients are required to take into account interactions between different metals present in the environment. Accordingly, large data sets are needed to reduce uncertainties in regression analyses. This limitation can be avoided if the transmembrane potential difference can be modeled and considered as a unifying factor in predicting the internal interactions similar to c0. SUPPLEMENTAL DATA

Tables S1–S11. Section SA. Figures S1–S6. (9 MB DOC). REFERENCES 1. Worms I, Simon DF, Hassler CS, Wilkinson KJ. 2006. Bioavailability of trace metals to aquatic microorganisms: Importance of chemical, biological and physical processes on biouptake. Biochimie 88:1721– 1731. 2. Fairbrother A, Wenstel R, Sappington K, Wood W. 2007. Framework for metals risk assessment. Ecotoxicol Environ Saf 68:145–227. 3. Blanco-Penedo I, Cruz JM, Lopez-Alonso M, Miranda M, Castillo C, Hernandez J, Benedito JL. 2006. Influence of copper status on the accumulation of toxic and essential metals in cattle. Environ Int 32: 901–906.

T.T.Y. Le et al. 4. Mozdzer TJ, Kramarz P, Piskiewicz A, Niklinska M. 2003. Effects of cadmium and zinc on larval growth and survival in the ground beetle, Pterostichus oblongopunctatus. Environ Int 28:737–742. 5. Manzo S, Buono S, Cremisini C. 2010. Cadmium, lead and their mixtures with copper: Patracentrotus lividus embryotoxicity assessment, prediction, and offspring quality evaluation. Ecotoxicology 19:1209–1223. 6. Copper NL, Bidwell JR, Kumar A. 2009. Toxicity of copper, lead, and zinc mixtures to Ceriodaphnia dubia and Daphnia carinata. Ecotoxicol Environ Saf 72:1523–1528. 7. De Schamphelaere KAC, Janssen CR. 2002. A biotic ligand model predicting copper toxicity for Daphnia magna: The effect of calcium, magnesium, potassium and pH. Environ Sci Technol 36:48–54. 8. Le TTY, Peijnenburg WJGM, Hendriks AJ, Vijver MG. 2012. Predicting effects of cations on copper toxicity to lettuce (Lactuca sativa) by the biotic ligand model. Environ Toxicol Chem 31:355–359. 9. Le TTY, Vijver MG, Hendriks AJ, Peijnenburg WJGM. 2013a. Modeling toxicity of binary metal mixtures (Cu2þ–Agþ, Cu2þ–Zn2þ) to lettuce, Lactuca sativa, with the biotic ligand model. Environ Toxicol Chem 32:137–143. 10. Le TTY, Vijver MG, Kinraide TB, Peijnenburg WJGM, Hendirks AJ. 2013b. Modeling metal-metal interactions and metal toxicity to lettuce Lactuca sativa following mixture exposure (Cu2þ - Zn2þ and Cu2þ Agþ). Environ Pollut 176:185–192. 11. Kabata-Pendias A, Pendias H. 1984. Trace Elements in Soils and Plants. CRC Press, Boca Raton, FL, USA. 12. Kinraide TB. 2006. Plasma membrane surface potential (CPM) as a determinant of ion bioavailability: A critical analysis of new and published toxicological studies and a simplified method for the computation of plant CPM. Environ Toxicol Chem 25:3188–3198. 13. Wang P, Kinraide TB, Zhou D-M, Kopittke PM, Peijnenburg WJGM. 2011. Plasma membrane surface potential: Dual effects upon ion uptake and toxicity. Plant Physiol 155:808–820. 14. Wang P, Kopittke PM, De Schmphelaere KAC, Zhao F-J, Zhoud D-M, Lock K, Ma Y-B, Peijnenburg WJGM, McGrath SP. 2011. Evaluation of an electrostatic toxicity model for predicting Ni2þ toxicity to barley root elongation in hydroponic cultures and in soils. New Phytol 192:414– 427. 15. Kopittke PM, Blamey FPC, Wang P, Menzies NW. 2011. Calculated activity of Mn2þ at the outer surface of the root cell plasma membrane governs Mn nutrition of cowpea seedlings. J Exp Bot 62:3393–4001. 16. Kinraide TB, Pedler JF, Parker DR. 2004. Relative effectiveness of calcium and magnesium in the alleviation of rhizotoxicity in wheat induced by copper, zinc, aluminum, sodium, and low pH. Plant Soil 259:201–208. 17. Hassler C, Slaveykova VI, Wilkinson KJ. 2004. Some fundamental (and often overlooked) considerations underlying the free ion activity and biotic ligand models. Environ Toxicol Chem 23:283–291. 18. Kinraide TB. 2001. Ion fluxes considered in terms of membrane-surface electrical potentials. Aust J Plant Physiol 28:605–616. 19. Yermiyahu U, Kinraide TB. 2005. Binding and electrostatic attraction of trace elements to plant root surfaces. In Huan PM, Gobran GR, eds, Biogeochemistry of Trace Elements in the Rhizosphere. Elsevier, Amsterdam, The Netherlands, pp 365–389. 20. Kinraide TB, Wang P. 2010. The surface charge density of plant cell membranes (s): An attempt to resolve conflicting values for intrinsic s. J Exp Bot 61:2507–2518. 21. Garate A, Ramos I, Manzanares M, Lucena JJ. 1993. Cadmium uptake and distribution in three cultivars of Lactuca sp. Bull Environ Contam Toxicol 50:709–716. 22. McKenna IM, Chaney RL, Williams FM. 1993. The effects of cadmium and zinc interactions on the accumulation and tissue distribution of zinc and cadmium in lettuce and spinach. Environ Pollut 79:113–120. 23. Organisation for Economic Co-operation and Development. 2006. Terrestrial plant test—vegetative vigour. OECD Guidelines for testing chemicals 227. Paris, France. 24. Steiner AA. 1961. A universal method for preparing nutrient solution of a certain desired composition. Plant Soil 15:134–154. 25. Kandegedara A, Rorabacher DB. 1999. Noncomplexing tertiary amines as ‘better’ buffers covering the range of pH 3–11. Temperature dependence of their acid dissociation constants. Anal Chem 71:3140– 3144. 26. US Environmental Protection Agency. 1991. Methods for aquatic toxicity identification evaluation. Phase I. Toxicity characterization procedures, 2nd ed. EPA 600/6-91/003. Office of Water, Washington, DC. 27. De Schamphelaere KAC, Heijerick DG, Janssen CR. 2004. Comparison of the effect of different pH buffering techniques on the toxicity of

Surface and internal ion-ion interactions

28. 29. 30. 31. 32. 33. 34.

35. 36.

copper and zinc to Daphnia magna and Pseudokirchneriella subcapitata. Ecotoxicology 13:697–705. Tipping E. 1998. Humic Ion-binding model VI: An improved description of the interactions of protons and metal ions with humic substances. Aquat Geochem 4:3–48. Verweij W. 2004. Chemical Equilibria in Aquatic Systems (CHEAQS Pro). [cited 2011 December 12]. Available from: http://home.tiscali.nl/ cheaqs. Kinraide TB, Yermiyahu U. 2007. A scale of metal ion binding strengths correlating with ionic charge, Pauling electronegativity, toxicity, and other physiological effects. J Inorg Biochem 101:1201–1213. Kinraide TB, Parker DR. 1989. Assessing the phytotoxicity of mononuclear hydroxy aluminium. Plant Cell Environ 12:479–487. Kinraide TB. 1999. Interactions among Ca2þ, Naþ and Kþ in salinity toxicity: Quantitative resolution of multiple toxic and ameliorative effects. J Exp Bot 50:1495–1505. Robinson RA, Stokes RH. 2002. Electrolyte Solutions. Dover Publications, Minerola, NY, USA. Le TTY, Peijnenburg WJGM. 2013. Modeling toxicity of mixtures of perfluorooctanoic acid and triazoles (triadimefon and paclobutrazol) to the benthic cadoceran Chydorus sphaericus. Environ Sci Technol 47:6621–6629. Bliss CI. 1939. The toxicity of poisons applied jointly. Ann Appl Biol 26:585–615. Loewe S, Muischnek H. 1926. Über kombinationswirkungen. 1. Mitteilung: Hilfsmittel der fragestellung. Nanyn-schmiedebergs. Arch Exp Pathol Pharmakol 114:313–326.

Environ Toxicol Chem 33, 2014

1995

37. Kopittke PM, Blamey FPC, McKenna BA, Wang P, Menzies NW. 2011. Toxicity of metals to roots of cowpea in relation to their binding strength. Environ Toxicol Chem 30:1827–1833. 38. Backhaus T, Altenburger R, Boedeker W, Faust M, Scholze M, Grimme LH. 2000. Predictability of the toxicity of a multiple mixture of dissimilarly acting chemicals to Vibrio fischeri. Environ Toxicol Chem 19:2348–2356. 39. Kopittke PM, Kinraide TB, Wang P, Blamey PC, Reichman SM, Menzies NW. 2011. Alleviation of Cu and Pb rhizotoxicities in cowpea (Vigna unguiculata) as related to ion activities at root-cell plasma membrane surface. Environ Sci Technol 45:4966–4973. 40. Wang P, Zhou D, Kinraide TB, Luo X, Li L, Li D, Zhang H. 2008. Cell membrane surface potential (C0) plays a dominant role in the phytotoxicity of copper and arsenate. Plant Physiol 148:2134– 2143. 41. Kinraide TB. 1998. Three mechanisms for calcium alleviation of mineral toxicities. Plant Physiol 118:513–520. 42. Luo Y, Rimmer DL. 1995. Zinc-copper interaction affecting plant growth on a metal-contaminated soil. Environ Pollut 88:79–83. 43. Beckett PHT, Davis RD. 1978. The additivity of the toxic effects of Cu, Ni and Zn in young barley. New Phytol 81:155–173. 44. Sanders JR, McGrath SP, Adam TM. 1987. Zinc, copper and nickel concentration in soil extract and crops grown on four soils treated with metal-loaded sewage sludge. Environ Pollut 44:193–210. 45. Tani FH, Barriongton S. 2005. Zinc and copper uptake by plants under two transpiration rates. Part II. Buckwheat (Fagopyrum esculentum L.). Environ Pollut 138:548–558.