Article pubs.acs.org/est
Internal versus External Dose for Describing Ternary Metal Mixture (Ni, Cu, Cd) Chronic Toxicity to Lemna minor Yamini Gopalapillai*,† and Beverley A. Hale School of Environmental Sciences, University of Guelph, Guelph, Ontario N1G 2W1, Canada S Supporting Information *
ABSTRACT: Simultaneous determinations of internal dose ([M]tiss) and external doses ([M]tot, {M2+} in solution) were conducted to study ternary mixture (Ni, Cu, Cd) chronic toxicity to Lemna minor in alkaline solution (pH 8.3). Also, concentration addition (CA) based on internal dose was evaluated as a tool for risk assessment of metal mixture. Multiple regression analysis of dose versus root growth inhibition, as well as saturation binding kinetics, provided insight into interactions. Multiple regressions were simpler for [M]tiss than [M]tot and {M2+}, and along with saturation kinetics to the internal biotic ligand(s) in the cytoplasm, they indicated that Ni−Cu−Cd competed for uptake into plant, but once inside, only Cu−Cd shared a binding site. Copper inorganic complexes (hydroxides, carbonates) played a role in metal bioavailability in single metal exposure but not in mixtures. Regardless of interactions, the current regulatory approach of using CA based on [M]tot can sufficiently predict mixture toxicity (∑TU close to 1), but CA based on [M]tiss was closest to unity across a range of doses. Internal dose integrates all metal−metal interactions in solution and during uptake into the organism, thereby providing a more direct metric describing toxicity.
■
INTRODUCTION Risk assessments for metal-elevated soils and waters typically use environmental quality standards derived from studies of single metal toxicity thresholds, despite the fact that most metal-elevated sites have metal mixtures due to geologic cooccurrence in the feedstocks. Studies in both aquatic and terrestrial systems have suggested that concentration addition (CA) based on total metal concentration ([M]tot) in exposure solutions will mostly overestimate mixture toxicity and thus is a conservative assessment of risk.1−5 However, CA based on [M]tot may not accurately predict mixture toxicity, particularly if the metals substantially modify each other’s uptake by the organism.6,7 The mechanism of such deviations include competition among free metal ions for binding to the biotic ligand (BL, likely to result in less than additive response and ∑TUtot > 1). Similarly, metals could compete for binding to dissolved organic carbon (DOC), likely to result in greater than additive response (i.e., ∑TUtot < 1). Bioavailability is now accepted as a crucial consideration in toxicity studies and regulations. Increasingly, models that account for the modifying effect of water chemistry factors on the interaction of the metal with the biological receptor to predict toxicity to biota are incorporated into determining environmental criteria. Consideration of metal mixture toxicity in regulatory guidelines is warranted. Two mechanistically based models for mixture toxicity are WHAM-FTOX8 and multimetal biotic ligand model (mBLM).9 Neither has been well-tested with chronic toxicity data, as only a few studies exist, none of which were of plants.1,10,11 Chronic © XXXX American Chemical Society
exposure could result in substantial interactions inside organisms that are difficult to model.7 However, they are necessary for improving applicability of toxicity models for risk assessment for sites impacted by metal mixtures, which was the motive for the current paper. The first objective of this study was to investigate metal− metal interactions in chronic exposure of Ni, Cu, and Cd to a standard aquatic test species (Lemna minor) at a pH of 8.3 and whether these interactions persist when dose is internal (tissue metal concentration, [M]tiss) versus external ([M]tot, or free ion activity, {M2+}, in solution). Note that we use the term dose as a metric of toxicity for both internal and external concentrations. A central composite experimental design (CCD)5 was chosen to efficiently study the range of dissolved metal chemistry in natural waters impacted by mining effluents, as it covers a range of ratios of metal concentrations. Although [M]tiss seems like an obvious predictor of effect, several factors may impact a straightforward relationship to toxicity. For plants, this includes interactions with phytochelatins that sequester M2+ and make them biologically inactive,12 the presence of intracellular membrane transporters where further metal−metal interaction may occur,13 and complex interactions at the site of toxicity (e.g., metal cofactor sites on enzymes). In plants, internal dose means metal bound to a BL Received: Revised: Accepted: Published: A
December 30, 2016 March 15, 2017 April 6, 2017 April 6, 2017 DOI: 10.1021/acs.est.6b06608 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology
allowing estimation of curvature as a second-degree polynomial.22 Single-metal and mixture toxicity tests were conducted consecutively but not simultaneously. Nonsimultaneous testing of single and mixture studies has been suggested as a confounding factor (time),23 but we were confident in the single-metal toxicity thresholds because they related well with previous L. minor toxicity tests for the three metals in our laboratory. Plant Digestion. The procedure generally followed that of Chen et al.24 After the test end points were recorded, plants in replicate test cups were pooled into a rinse cup containing dilute APHA medium and then rinsed in 0.01 M EDTA solution for 2 min, and rerinsed in dilute APHA. This removes metals from the surface of plants for determination of internal [M]tiss. Rinsed plants were blotted dry with Kimwipes, transferred to aluminum weigh boats, and placed in an oven at 60 °C overnight. The dried plants were weighed into tared 60 mL Teflon digestion vessels25 and the weights were recorded. Trace metal-grade concentrated HNO3 was added to each vessel (1 mL) and mixed well to immerse all plant material. After vessel lids were secured, vessels were placed in an oven (set at 105 °C) inside a fume hood to digest overnight. The samples were transferred to 15 mL polypropylene tubes by use of a funnel with a Whatman paper filter, grade 42 (particle retention size >2.5 μm) (Whatman Inc.), and made up to 10 mL with ultrapure water. Appropriate blanks and a standard reference material (NIST 1570a, trace elements in spinach, U.S. Department of Commerce, National Institute of Standards and Technology, Gaithersburg, MD) were digested with every batch of samples. For metal analysis, a blank and a laboratory control sample spiked with known concentrations of the metals (Ni, Cu, and Cd) were run every 10 samples. If the relative standard deviation (RSD) of absorbance was greater than 10%, then the instrument was recalibrated and the prior 10 samples were rerun. Data Analysis. Toxicity Thresholds. EC50 (effective concentration of toxicant inhibiting root growth by 50%) for single-metal exposures was estimated by use of Comprehensive Environment Toxicity Information System (CETIS)26 version 1.7 (which uses Grubb’s test to check for outliers), using nonlinear regression modeling when possible and linear interpolation otherwise. For nonlinear regression, the dose− response curves were fitted with the 3P log−logistic EV model y = A/[1 + (x/D)C], where A = location parameter, C = shape parameter, D = scale parameter and the median of the distribution. Toxicity thresholds for L. minor exposed to single metals (Ni, Cu, Cd) were calculated with dose expressed as [M]tiss. For single-metal exposures, the % RGI ranged from 0 to 87% (Table S1). Previously reported EC25(Ni,tiss) for L. minor in our laboratory was 360 ± 126 μg·L−1,27 but the [M]tiss in that test included surface-bound metal, as there was no EDTA rinse. The EC25(Ni,tiss) value determined in the present experiment was expectedly lower at 176 (133−210) μg·L−1. For mixture toxicity thresholds, multiple regression analysis was used to fit the observed % RGI to dose expressed as [M]tiss. The full model was as follows:
within the cytoplasm rather than that adsorbed by the large cation-exchange capacity of the root apoplasm (cell wall), which is a preconcentration function in aid of uptake of essential elements.14 Accordingly, the second objective of this study was to gain mechanistic insights into the impact of mixtures on metal−BL binding. Metal accumulation in single metal versus ternary mixture exposures were assessed using saturation binding principles derived from enzyme kinetics. The basis is that a fixed number of BL binding sites exist, which become saturated and can be fitted to a Michaelis−Menten saturation curve.11,15,16 Previous studies have exploited saturation binding to explore metal−BL binding in the invertebrate Hyalella azteca,17 including metal mixture exposures,11,18,19 However, none of these studies consider plants; hence the importance of the present work for ecotoxicological risk assessment of mixtures. We previously proposed that CA of {M2+} (assumed to be bioavailable metal) as dose was a more reliable predictor than CA of [M]tot in solution.5 To understand whether internal dose is a better predictor of metal mixture toxicity to aquatic plants than the commonly used [M]tot, the third objective of the present study was to extend the investigation to the use of CA based on [M]tiss, with the goal of determining whether [M]tiss, {M2+} in solution, or [M]tot is the best predictor of chronic metal mixture toxicity to plants.
■
MATERIALS AND METHODS As the data used in the present study were collected in the study described in Gopalapillai and Hale,5 brief explanations are provided here. Culture and Toxicity Testing Using L. minor. Lemna minor L. no. 8434 (Canadian Phycological Culture Centre, Waterloo, Canada) was cultured and tested following the standardized Environment Canada protocol20 with minimal changes, for a chronic test duration of 7 days. The chosen test medium was modified APHA (American Public Health Association), which excludes ethylenediaminetetraacetic acid (EDTA) to avoid altering metal speciation in the exposure solutions. A standard solution of NiII, CuII, and CdII was prepared on the day of toxicity test solution preparation, using ultrapure nitrate powder (Puratronic 99.999%, Alfa Aesar, Ward Hill, MA). For the single-metal toxicity testing, each metal (Ni, Cu, or Cd) treatment was set up as follows: 0 μg·L−1 (6×), 1.56, 3.13, and 6.25 μg·L−1 (4×), and 12.5, 25, 50, and 100 μg·L−1 (3×). All metal analyses were conducted by graphite furnace atomic absorption spectroscopy (Varian, Mississauga, ON) or flame atomic absorption spectrometry with Zeeman background correction (Varian Spectra AA 220). The corresponding [M]tot, [M]tiss, and % RGI values are presented in Table S1. Percentage root growth inhibition (% RGI) was chosen as it was previously determined to be the most sensitive and precise end point.21 Metal mixture toxicity was determined by use of a CCD, an incomplete factorial design that maintains statistical power by allowing the data to be rotatable and thus requires fewer experimental units than a full factorial design. Test combinations included control plus 20 test cases of Ni, Cu, and Cd in a range of nominal [M]tot from 0 to 100 μg·L−1 (i.e., 0−1.7 μM Ni, 0−1.6 μM Cu, and 0−0.89 μM Cd). The corresponding [M]tot, [M]tiss, and % RGI values are presented in Table S2. The factorial points create a firstorder design, the center points are included as a test for curvature, and the star points add a second-order design, thus
%RGI = β0 + β1[Ni] + β2[Cu] + β3[Cd] + β12[Ni][Cu] + β13[Ni][Cd] + β23[Cu][Cd] + β123[Ni][Cu][Cd] + ε B
DOI: 10.1021/acs.est.6b06608 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology Table 1. Single-Metal Toxicity Thresholds of Ni, Cu, and Cda for Percentage Root Growth Inhibition in L. minor EC50 (μg·g−1) [Ni]tiss [Cu]tiss [Cd]tiss a
EC25 (μg·g−1) b
EC10 (μg·g−1)
value
CI
value
CI
value
CI
235 91.8 885
205−269 82.6−102 788−994
176 63.9 609
133−210 53.3−73.7 495−713
131 44.5 419
n/a−166 27.4−55.1 266−518
Doses are given as tissue metal concentration, [M]tiss. bConfidence interval (5−95%).
■
where β values represent the regression coefficients and ε is the normal error. The model was reduced to its most parsimonious form by stepwise regression of nonsignificant terms (p > 0.15) using SAS 9.4, starting with interactions. The adjusted r2 was used to describe goodness of fit so that this evaluation was independent of the number of parameters in the model. The variance associated with a removed term was statistically redistributed to the remaining terms in the model. Standardized regression coefficients (b) were also calculated; these normalize the variances of the dependent and independent variables to 1 and set the intercept to zero. They are useful for comparison of independent variables that are not biased by differences in their units and scale. Note that any time an effect (main, binary interaction, or ternary interaction) is discussed, the p value is 7) was attributed to the increased Cu activity at plasma membrane (PM), as the higher pH can polarize the PM and increase the negativity of the electrical potential.34 This suggests that electrostatic effects play a large role in Cu bioavailability under alkaline conditions, as does Cu speciation. Copper was 16-fold more toxic than Ni when dose was expressed as {M2+} but only 3-fold more toxic when dose was expressed as [M]tiss. When it is considered that Ni speciation was not complex as for Cu, relative differences in their toxicity based on external versus internal dose indicates that bioavailable Cu may have been underestimated by free ion activity. This supports the hypotheses that (1) bioavailability is better predicted by labile Cu (free + easily dissociable inorganic Cu species) or (2) Cu activity increases at the plasma membrane under alkaline conditions due to electrostatic effects. With [M]tiss as dose, Cd was less toxic than Ni (4-fold), while the two metals were equally toxic when dose was expressed as external concentration ([M]tot and {M2+}).5 Cd was also 10fold less toxic than Cu when dose was expressed as [M]tiss and 23-fold less toxic than Cu when expressed as {M2+}. This suggests that the plant had a stronger detoxification capacity for Cd compared to Ni or Cu, which is supported by the fact that Cd is the strongest inducer of phytochelatin (known to sequester and detoxify Cd) production of all the cations.35 Cadmium is known to easily move from soil to plant, about 100-fold greater than that for Cu and 1000-fold greater than that for Ni.36 In addition, Cd is not essential for plant function,
Ci EC50, i
The concentration of the ith toxicant in the mixture causing 50% RGI is denoted by Ci, and EC50,i is the concentration of the ith toxicant causing 50% RGI in a single metal exposure. If ∑TU = 1, then CA occurred; a sum which is more than unity indicates that the mixture effect is less-than-additive, while a sum that is less than unity indicates a mixture effect that is more-than-additive.30 C
DOI: 10.1021/acs.est.6b06608 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology
less negative (thus, smaller reduction in root growth) than for external dose. This is because competition for uptake into the organism is no longer a factor, and it suggests that further interaction among the metals is occurring once inside the plant. This is likely because there are intracellular membranes with similar transmembrane proteins as in the plasma membrane.13 The possibility of additional intracellular membrane transporters is supported by the binary interactions found in the present study, as they mimic interactions for uptake into the organism. For example, there is evidence of Cu-specific channels within plants such as pea.37 The net effect on the dependent variable of change by one or more of the independent variables is difficult to visualize in regression relationships with interactions. Response surfaces for [M]tiss (Figure 1 e−h) illustrate clearly the effect of Ni in mitigating the interactions between high concentrations of Cd and Cu and in exacerbating the individual effects of Cu and Cd. In contrast, the surfaces for [M]tot illustrate a more complex relationship with Cu and the generally exacerbating effect of Ni (Figure 1 a−d). Accumulation of Metals in the Whole Plant: Saturation Binding. Study of saturation binding assumes steadystate or equilibrium conditions. The chronic exposure used in the present study of growing plants is considered to be a pseudoequilibrium,16 as is often the case for equilibrium-based BLM studies. There is environmental relevance for characterizing saturation using the principles of saturation kinetics in nonlethal toxicity studies, as uptake, depuration, and dilution by growth are integrated into the dose at the BL. Thus, to meet the second objective of investigating the mechanism of mixture toxicity, [Ni]tiss, [Cu]tiss, and [Cd]tiss were determined at the ends of the chronic single-metal toxicity studies (Figure 2A) and of the mixture toxicity study (Figure 2B). Overall, [Cu]tiss and [Cd]tiss were lower in the mixture exposures (Figure 2B) compared to the single-metal exposures (Figure 2A), but [Ni]tiss was similar. Copper accumulation in the plant (maximum 92.3 μg·g−1) was lower compared to those of Ni (maximum 484 μg·g−1) and Cd (maximum 579 μg·g−1) (Table S2). Consistent with its ready accumulation in plants, [Cd]tiss was greatest in both the single-metal and mixture exposures. The [M]tiss data for single-metal and mixture exposures were fitted to a Michaelis−Menten ligand binding curve,5 expressed as either {M2+} or [M]tot in the exposure solution (Table 3). The goodness of fit for [M]tiss of every metal was poorer in the mixture than in the single-metal exposure. The most reduced fit was for Cu (r2 = 0.47), while the least changed was for Cd (r2 = 0.86). This increased variability in Michaelis−Menten parameters for metals in mixtures versus single-metal exposure is likely because of the influence of the other metals on binding and uptake at the BL. The half-saturation constant (Kd) is inversely related to the binding affinity of metal to BL, and Bmax is the binding capacity of BL for metal, both of which were calculated from the Michaelis−Menten curve (Table 3). For single-metal exposures, all values for Bmax and Kd were significant (p < 0.05). Expectedly, cadmium had the largest binding capacity of any of the single metals (Bmax(Cd,tot) = 2954 μg·g−1).36 Because of the poorer fits, Kd for the mixture exposures were not different from 0 for Ni and Cu, but Kd(Cd,tot) was 55.0 μg·L−1 and Kd(Cd2+) was 24.2 μg·L−1. However, the Bmax values were significant for all metals in the mixture (Table 3). For both single-metal and mixture exposures, the Bmax was the same regardless of whether dose was expressed as [M]tot or {M2+}, which is expected as the
whereas Cu is a cofactor of various oxidases, plastocyanins, and ceniloplasmin in plants, and Ni is a cofactor of the enzyme urease.36 Toxicity of Mixtures. In response to the metals in mixture, % RGI ranged from 63% to 94% (see Table S2), much smaller than the range for single-metal exposures (0−87%); the higher upper end to the range for mixtures suggested that the organism is responding to a larger total dose. When % RGI for mixtures exposures was fitted to [M]tiss using multiple regression, only seven of the 10 parameters from the full regression model (which includes binary and ternary interactions) remained in the model after stepwise removal (Table 2). This is in contrast to the same % RGI data expressed Table 2. Multiple Regression Analysis of the Effects of Ni, Cu, and Cd in L. minor on Observed Percentage Root Growth Inhibitiona
intercept Ni Ni × Ni Cu Cu × Cu Cd Cd × Cd Ni × Cu Ni × Cd Cu × Cd Ni × Cu × Cd
β for [M]tissb (mg·g−1)
p-value
b for [M]tiss
b for {M2+}c
b for [M]totc
−4.72 (3.5) 185 (22) nsd 1002 (230) −4724 (1782) 181 (31) −139 (45) −1704 (397) −212 (63) ns ns
0.1834
