Predicting Water Quality Criteria for Protecting Aquatic Life fromPhysicochemical Properties of Metals or MetalloidsFengchang Wu,*,† Yunsong Mu,† Hong Chang,† Xiaoli Zhao,† John P. Giesy,‡,§,∥,⊥ and K. Benjamin Wu#

†State Key Laboratory of Environmental Criteria and Risk Assessment, Chinese Research Academy of Environmental Sciences,Beijing 100012, China‡Department of Veterinary Biomedical Sciences and Toxicology Centre, University of Saskatchewan, Saskatoon, Saskatchewan,Canada§Zoology Department and Center for Integrative Toxicology, Michigan State University, East Lansing, Michigan 48824, United States∥Department of Biology & Chemistry and State Key Laboratory in Marine Pollution, City University of Hong Kong, Kowloon, HongKong, SAR, China⊥School of Biological Sciences, University of Hong Kong, Hong Kong, SAR, China#HDR-HydroQual, 1200 MacArthur Blvd, Mahwah, New Jersey 07430, United States

*S Supporting Information

ABSTRACT: Metals are widely distributed pollutants in water and canhave detrimental effects on some aquatic life and humans. Over the pastfew decades, the United States Environmental Protection Agency (U.S.EPA) has published a series of criteria guidelines, which contain specificcriteria maximum concentrations (CMCs) for 10 metals. However,CMCs for other metals are still lacking because of financial, practical, orethical restrictions on toxicity testing. Herein, a quantitative structureactivity relationship (QSAR) method was used to develop a set ofpredictive relationships, based on physical and chemical characteristics ofmetals, and predict acute toxicities of each species for five phyla andeight families of organisms for 25 metals or metalloids. In addition,species sensitivity distributions (SSDs) were developed as independentmethods for determining predictive CMCs. The quantitative ioncharacter−activity relationships (QICAR) analysis showed that the softness index (σp), maximum complex stability constants(log −βn), electrochemical potential (ΔE0), and covalent index (Xm

2r) were the minimum set of structure parameters required topredict toxicity of metals to eight families of representative organisms. Predicted CMCs for 10 metals are in reasonableagreement with those recommended previously by U.S. EPA within a difference of 1.5 orders of magnitude. CMCs weresignificantly related to σp (r2 = 0.76, P = 7.02 × 10−9) and log −βn (r2 = 0.73, P = 3.88 × 10−8). The novel QICAR-SSD modelreported here is a rapid, cost-effective, and reasonably accurate method, which can provide a beneficial supplement to existingmethodologies for developing preliminarily screen level toxicities or criteria for metals, for which little or no relevant informationon the toxicity to particular classes of aquatic organisms exists.

■ INTRODUCTION

Metals can become contaminants in aquatic environments. Thewater quality criteria (WQC) for some metals were developedin the middle of the 20th century. Some acute toxicity data foraquatic organisms were used as the scientific basis fordevelopment of WQC and also for use in the environmentalmanagement.1,2 To sustain the reproduction and survival ofaquatic organisms, in 1976, the United States EnvironmentalProtection Agency (U.S. EPA) published the first WQCguideline commonly referred to as the “Red Book”, whichrecommended WQC for 12 metals or metalloids. Currently 167ambient WQC for priority and nonpriority toxicants have beendeveloped, and these WQC have been updated seven times.3−9

However, there are only 12 aquatic life criteria values forpriority metals and 4 criteria values for nonpriority metals in

the latest water quality guideline,9 in which specific criteriamaximum concentrations (CMCs) are recommended for 10metals. The CMCs for other metals are still lacking. Thisdeficiency limits the capacities of assessing water quality anddealing with unexpected environmental incidents, pollutioncontrol and environmental risk management. The presentCMCs were mainly derived from standardized aquatic lifetoxicity tests. Thus, environmental behavior in specific ambientmedia and different end points could lead to differences intoxicity of metals.10 Comprehensive toxicity tests are time-

Received: August 16, 2012Revised: November 29, 2012Accepted: November 30, 2012Published: November 30, 2012

Article

pubs.acs.org/est

© 2012 American Chemical Society 446 dx.doi.org/10.1021/es303309h | Environ. Sci. Technol. 2013, 47, 446−453

consuming and expensive, and testing on some endangeredspecies cannot be conducted due to eco-ethics challenges. Forother species, sometimes it is not possible to conductcontrolled laboratory tests because of size, lack of informationon culturing or maintaining them under laboratory conditions.Therefore, the importance of establishing a system to estimateCMCs that is based on limited toxicity test data is recognized.One of the commonly used models to predict criteria of metalsis the recently developed biotic ligand model (BLM).11−15 TheU.S. EPA adopted the copper BLM for establishing WQC andextended the use of the BLM to WQC of silver and zinc.16,17

The European Union also recently applied the BLM to assesschronic toxicity of nickel (Ni) and developed WQC for Ni.18

The BLM is used to estimate the bioavailable fraction of metalsfor ascertaining the role of water chemistry on toxicity of metalsto aquatic organisms. A model, based only on physicochemicalproperties to predict toxicity, has also been developed.19

However, that model was of limited scope.Quantitative structure activity relationships (QSARs) estab-

lish intrinsic relationships between characteristics of acompound to bioactivity or toxicity of metals by use ofstatistical analysis. Most QSARs have been developed fororganic chemicals, with inorganic chemicals, such as metalsbeing under-represented in the environmental toxicologyliterature.20−22 Because of metal speciation, complexation,interactions in biological systems and formation/degradationof metal−ligand bond, correlation of toxicity with physical orchemical properties of metals is still challenging. It is knownthat most metals exist in biological system as cations andtoxicity of metals depends mainly on cationic activity.23 Atpresent, ion characteristics have been used to predict toxicity orsublethal effects of metal ions, and the quantitative ioncharacter−activity relationships (QICAR) models based onmetal−ligand binding have recently been developed.19,23−28

More than twenty ion characteristics including hydrolysis,ionization, covalent binding and spatial characteristics wereused in a study conducted by Walker et al.,29 to predict bindingof soft ligands. However, there are still challenging issues tosolve. First, existing QICAR models contained different metaltoxicity data, which varied largely in exposure times, organisms,effects and effect levels. Second, most of QICAR models relatecertain ion characteristics with toxicity of metals. Thus, novelcharacteristics of metals or metalloids could also be considered

and predictive relationships could be developed for additionalaquatic species. The purpose of this study was to relatecharacteristics of data-rich metal ions with metal toxicities ofeach species of several representative aquatic organisms thatwere necessary for getting WQC, to predict the toxicities ofadditional data-poor metals and ultimately to obtain the CMCsfor these additional metals.The species sensitivity distribution (SSD) analysis is a

promising method to determine CMCs, based on cumulativeprobability distributions of toxicity values for multiple species.For derivation of CMCs, the concentration of a chemical thatcan be used as a hazard level can be extrapolated from a SSDsuch that 95% of species would not be affected by a particularconcentration of a specific metal.30 The SSD can be used toestimate the concentration at which 5% of species would beaffected. The concentration associated with the fifth percentilehas been referred to as the 5% hazard concentration (HC5). Inthe semiprobabilistic approach developed by Stephan,31 theminimum data set required for derivation of a WQC forfreshwater was at least three phyla and eight differenttaxonomic families.32,33 As a result, the diversity and thesensitivities of a range of aquatic life are represented in thecriteria values in order to estimate a concentration to protectorganisms against small effects. The present study wasconducted to compile the relative toxicity data of 25 metalsor metalloids and to examine the relationships between selectedphysicochemical parameters and corresponding toxicities (LC50or EC50) in eight families of representative organisms. Thisinformation was used to determine CMCs of each metal bySSD analysis and obtain a predictive relationship for CMCs.

■ MATERIALS AND METHODS

Modeling Data Sets. Toxicity data used in the presentstudy were selected based on data collection and requirementdescribed in 1985 U.S. EPA WQC guidelines where minimumeight species (three phyla) were required.31 Those data havealso been used in conjunction with metal criteria derivation andrecent risk assessment.34,35 For better comparison andconsistence, toxicity data were further selected based on thefollowing standards: (1) Toxicities of metals to each specieswere required from the same data source and the same researchteam under the same experimental conditions. (2) Results forsix or more metals were investigated for each species. (3) Data

Table 1. Two-Variable Regression Models Based on Seven Characteristics for Metal Ionsa

species phyla predicting equations r2 RSS RMSE F P

Chironomus tentans Arthropoda log 48 h − EC50 = (28.136 ± 18.459)σp + (−0.150 ± 0.112) log −βn +(0.814 ± 3.625)

0.769 3.749 0.790 9.98 0.012

Crangonyxpseudogracilis

Arthropoda log 96 h − LC50 = (39.716 ± 25.627)σp + (−0.254 ± 0.136) log −βn +(1.678 ± 4.533)

0.791 8.520 1.103 13.26 0.004

Daphnia magna Arthropoda log 48 h − EC50 = (−0.272 ± 18.674)σp + (−0.360 ± 0.136) log −βn +(6.604 ± 4.093)

0.869 1.360 0.583 13.25 0.017

Lymnaea acuminata Mollusca log 96 h − EC50 = (−2.160 ± 0.821)Xm2r + (0.237 ± 0.222) AN/ΔIP +

(5.557 ± 1.399)0.827 1.453 0.696 7.191 0.070

Cyprinus carpio Chordata log 96 h − LC50 = (33.441 ± 6.256)σp + (0.412 ± 0.137)Z/r + (−3.159 ±0.559)

0.960 0.226 0.274 35.55 0.008

Brachionuscalycif lorus

Rotifera log 24 h − LC50 = (−0.297 ± 0.082) log −βn + (−0.111 ± 0.106)|log KOH|+ (6.375 ± 2.058)

0.823 1.478 0.702 6.97 0.070

Bufo melanostictus Chordata log 96 h − LC50 = (6.955 ± 20.353)σp + (−1.569 ± 0.474)Xm2r + (5.014

± 3.156)0.902 1.814 0.673 18.35 0.009

Lemna minor Angiosperms log 96 h − EC50 = (24.984 ± 9.959)σp + (1.494 ± 0.439)ΔE0 + (−2.046 ±1.140)

0.824 0.728 0.427 9.40 0.030

ar2 is the coefficient of determination, RSS is the residual sum of squares, RMSE is root mean square deviation, and p is the statistical significancelevel.

Environmental Science & Technology Article

dx.doi.org/10.1021/es303309h | Environ. Sci. Technol. 2013, 47, 446−453447

considered included the assessment end points of survival andgrowth. (4) Results of acute tests conducted in unusual dilutionwater, for example, dilution water in which total organic carbonor particulate matter exceeded 5 mg/L, should not be used. Alltoxicity data of different species were from the ECOTOXDatabase and literatures.36−43 In the present study, five phylaand eight different taxonomic families (three chordates, twoarthropods, a rotifer, a mollusk and an aquatic plant) wereselected (Table 1). Twenty-five metals or metalloids (e.g.,mono-, di-, trivalent- and hexavalent metals) were chosen.Those included 16 metals recommended by U.S. EPA in thelatest WQC guideline, in which 10 metals had their own CMCsfor protecting aquatic life.Characteristics of Metals and Development of

Predictive Relationships. Characteristics of the metals used

as parameters in predictive relationships models were obtainedfrom a variety of sources. The basic characteristics consideredin developing the relationships were atomic number (AN),atomic radius (AR), Pauling ionic radius (r), ionic charge (Z),and ionization potential (ΔIP).44 In total, 14 parameters wereconsidered to characterize the metals in the QICAR model.These parameters included: softness index (σp);45 maximumcomplex stability constants (log −βn), which was derived frommaximum strength of complexes formed between metals andEDTA, CN−, or SCN−; electrochemical potential (ΔE0);19 firsthydrolysis constants (|log KOH|);

46 electronegativity (Xm);47

electron density (AR/AW);47 relative softness (Z/rx), where xrepresents electronegativity values; atomic ionization potential(AN/ΔIP); covalent index (Xm

2r); polarization force parame-ters (Z/r, Z/r2, and Z2/r); and similar polarization force

Figure 1. Regression models of log −EC50 or log −LC50 and two most predictive characteristics of ions for eight model organisms.

Environmental Science & Technology Article

dx.doi.org/10.1021/es303309h | Environ. Sci. Technol. 2013, 47, 446−453448

parameters (Z/AR and Z/AR2). For each species, measures ofacute toxicity were correlated to every characteristic of ions byuse of linear regression. The magnitude of association of eachcharacteristic with toxicity was tested by F-test statistic, with thelevel of significance at α = 0.05. Because the two-variable modelprovided better fits and contained sufficient information onstructure, the two most predictive characteristics were selectedbased on the rank of adjusted correlation coefficients,. As aresult, multiple linear regressions were performed between thelogarithm of LC50 or EC50 and the two most highly correlatedcharacteristics of ions in Table S1, Supporting Information(SI). The predictive potentials of QICAR models wereevaluated by use of the coefficient of determination (r2),residual sum of squares (RSS), root-mean-square deviation(RMSE), F value using multiple analysis of variance (ANOVA),and the level of Type I error (P).SSD Construction and HC5 Derivation. On the basis of

the QICAR equations developed for each of the representativeorganisms, predicted acute toxicity values were derived for eachmetal. After ranking these data from least to greatest, plottingpositions (proportions) for use in a cumulative probabilitydistribution were calculated (eq 1).

= −proportion (rank 0.5)/number of species (1)

To obtain the logarithm of HC5 values, the SSD was fitted byuse of the sigmoidal-logistic model (eq 2), where a wasrepresented as an amplitude, Xc was a center value, and k was acoefficient. The CMCs were defined as (HC5)/2. Multiplelinear regression and SSD fitting were performed by use of theOriginPro 8 software package, with three fitting parameters (a,Xc, and k) and their standard errors (a-SE, Xc-SE, and k-SE).Significant differences among species were examined by use ofANOVA.

=+ − −y

ae1 k x x( )c (2)

■ RESULTS AND DISCUSSIONQuantitative Ion Character−Activity Relationships

(QICARs) to Predict Toxicity of Metals. Statisticallysignificant, positive or negative relationships between log−EC50 or log −LC50 and the 14 ion characteristics wereobserved. Characteristics with the greatest r2 values for eachorganism were obtained. Seven characteristics, σp, log −βn,Xm

2r, AN/ΔIP, Z/r, |log KOH|, and ΔE0 exhibited statisticallysignificant associations with toxicities to representativeorganisms (Figure 1). The two parameters with the greatestr2 were used to predict the toxicity of each metal for eachrepresentative species. The parameter σp was significantly andpositively correlated with log −EC50 of three arthropods (C.tentans, C. pseudogracilis, and D. magna), with coefficients ofdetermination (r2) of 0.6994 (F = 16.289, P = 0.005), 0.6872 (F= 17.575, P = 0.003), and 0.6403 (F = 8.900, P = 0.03),respectively (Figure 1a, b, and c). These results are consistentwith those reported previously where σp was the characteristicthat exhibited the strongest association with metal−ligandquantifying binding constant of seven metals for D. magna.28

Log −βn was strongly correlated with acute toxicities ofarthropods. Values of log −βn were derived from maximumstrength of complexes formed between each metal and EDTA,CN−, or SCN−, which is a measure of binding affinity of eachwith the O-donor group. The two characteristics that weremost predictive of toxicity of metals to mollusks (L. acuminata)

were Xm2r (r2 = 0.7620, F = 12.809, P = 0.023) and AN/ΔIP (r2

= 0.4293, F = 3.009, P = 0.158). Values of Xm2r were

significantly and negatively correlated with potency of metals.However, the AN/ΔIP ratio was not significantly correlatedwith toxicity of metals. Xm

2r, which qualifies covalentinteractions relative to ionic interactions, is an index of stabilityof metal ions in water.48 Absorption of cadmium ions inmussels has been reported to involve covalent interactions withsulfhydryl groups on proteins 49 and that parameter has beenused to successfully predict bioaccumulation of metals. Thisobservation suggests that Xm

2r was a characteristic that wasuseful for predicting the potency of metals to cause toxicity inmollusks.48,50 For two organisms of the chordate, C. carpio andB. melanostictus, σp was correlated with the potency of metalswith r2 = 0.8373 (F = 20.578, P = 0.011) and 0.6325 (F = 8.605,P = 0.033), respectively. Although the Z/r ratio also providedadequate fits, it was not statistically significant (r2 = 0.5739, F =5.388, P = 0.081). Therefore, σp was the only characteristic thatwas used to predict the potency of metals to cause toxicity tofish. For B. calycif lorus, only log −βn was significantly associatedwith log-LC50 (r2=0.7587, F = 12.580, P = 0.024), while |logKOH| was not significantly correlated (r2 = 0.051, F = 0.213,P = 0.668). This is consistent with the findings of otherresearchers, who found that |log KOH| was not a uniquecharacteristic for predicting the potency of toxicity of metals torotifer.27 The two-variable model that best predicted potency ofmetals was a combination of σp and ΔE0. The predictiverelationships for these two parameters were significant andpositive with r2 = 0.5484 (F = 6.071, P = 0.057) and 0.3173 (F= 2.324, P = 0.188), respectively. In conclusion, fourcharacteristics of ions, σp, log −βn, Xm

2r, and ΔE0, werestatistically significantly associated with potencies of metals tothe reference species studied here. All these characteristics ofions except for log −βn had been previously reported to beuseful in developing QICAR models to predict toxicity ofmetals. The other three characteristics (Z/r, |log KOH|, and AN/ΔIP) could be used as additional factors to improve thepredictability.Eight two-variable linear regression models were developed

(Table 1). Among three arthropods, the model for D. magnahad the best relationships (r2 = 0.869, F = 13.25, P = 0.017),while the model for C. tentans exhibited the poorest coefficientof determination (r2 = 0.769, F = 9.98, P = 0.012). The QICARmodels for L. acuminata, B. calycif lorus, and L. minor hadcoefficients of determination, which were 0.827, 0.823, and0.824, respectively. Chordates, compared to other species in thepresent study, exhibited the best regressions. RSS and MSEwere used to evaluate indicators to assess the robustness of thepredictive relationships. Because of the diversity of training setswith the maximum number of metal ions (n = 10) and valencetypes (+1 to +3), RSS and MSE were greater for C.pseudogracilis, than for other species. On the basis of theresults of the multiple regression analysis of variance thesecharacteristics of ions exhibited a statistically significantrelationship to log −EC50 (P < 0.05). Alternatively, equationsto predict potency of toxicity of metals to both L. acuminataand B. calycif lorus were better (0.05 < P < 0.1), which wereaccepted for statistical significance at the 90% confidence level.The QICAR models presented herein were an improvement onthe contributions reviewed by Wolterbeek and Walker et al.29,47

First, the present study further extended the application to 25metals or metalloids with various valences. Second, more than20 physicochemical parameters reviewed by Walker et al. and

Environmental Science & Technology Article

dx.doi.org/10.1021/es303309h | Environ. Sci. Technol. 2013, 47, 446−453449

their potential to produce toxic effects were examined29andthen acute toxicities of five phyla and eight species werepredicted, including a wide range of aquatic life with differenttrophic levels. Finally, the two-variable QICAR models fordifferent species were established with indication of which ioncharacteristic was the most relevant for each species. Base onthe analyses of QICAR models for eight species, two-variablecorrelations appear to be species-specific and also providemajor and minor ion characteristics for each species. Thepresented results demonstrated that all the QICAR modelspresented herein were capable of predicting potencies oftoxicity of metals (RSS < 8.520, MSE < 1.103) that could beused in development of CMCs for these metals.Species Sensitivity to Metals and HC5 Derivation. The

predicted acute toxicities of each species were calculated basedon eight two-variable linear regression models using seven ioncharacteristics (see Table S1, SI). Potencies of 25 metals ormetalloids to each species, which were derived from QICARmodels, varied among species (see Table S2, SI). Planktonicarthropods were most sensitive to Al, As, Fe, Hg, La, Ni, andPb, with log −EC50 values ranging from −1.252 to 1.487.Arthropods were more sensitive to trivalent ions (Al, As, andFe) in contrast to group IVA, IIB, IIIB and VIII metals. In thepresent study, it was demonstrated that C. pseudogracilis was themost sensitive species to Hg, which agrees with the report ofZhang.51 However, B.calycif lorus was among the most sensitivespecies to the effects of metals in groups IA, IIA, VIIB and IIB,including Ba, Be, Ca, Cd, K, Li, Mg, Mn, Sr, and Zn, with log−EC50 values ranging from −2.225 to 0.751.C. pseudogracilis instead of D. magna was the most sensitive to

the effects of Cd, a result that is in compliance with the findingsthat D. magna was not the most sensitive species to Cd.52 ForZn, C. pseudogracilis and D. magna were equally sensitive withsimilar log −EC50 values of 0.594 and 0.669. This finding alsoagrees with reports that proposed that freshwater crustaceanswere the most sensitive to Zn.53 In conclusion, zooplanktonwere most vulnerable to the toxic effects of metals, which is alsoconsistent with previous conclusions based on empiricalmeasures of toxic potency.54−56 Among the three arthropods,C. tentans, C. pseudogracilis, and D. magna, D. magna was themost sensitive species to major metals. This conclusioncorresponds to the results of Song et al.,57,58 who found thatD. magna was more sensitive to triphenyltin than C. tentans,with 24 h-LC50 values of 13.3 and 287.7 μg/L. The mollusk, L.acuminate, was the most sensitive organism to Ag, Co, Cu, Sb,and Tl, which are group IIIA, VA, VIII, and IB metals. This wasalso in compliance with the results of a previous study.59

Compared to other organisms, the vertebrates, C. carpio and B.melanostictus, were the least sensitive to metals except Na. L.minor was the representative of aquatic plants, since metalsaffected its microstructures and cell growth. L. minor wasamong the most sensitive organisms to Cr(III) and Cr(VI) (seeTable S2, SI), which was consistent with previous studies inwhich exposure to Cr resulted in lesser production of biomassof L. minor.60 In conclusion, five phyla and eight families ofrepresentative organisms were used as minimum set of speciesto establish models of sensitivity to metals. In accordance withrequirements of the U.S. EPA, all of the species were found tobe sensitive to the toxic effects of main group metals (IA−VA)and metals in groups IIB, IIIB, VIIB, and VIII.Species sensitivity distributions for 25 metals or metalloids

were constructed and ultimately used to predict CMCs. Thebasic fitting parameters (a, Xc, k, a-SE, Xc-SE, and k-SE) and

statistical indexes (Adj.r2, RSS, F and P) used to develop thepredictive relationships are shown in Table S3, SI. Coefficientsof determination (r2) of the 25 fitting equations were greaterthan 0.9 (RSS < 0.0437, P < 0.0001), which suggests that allSSD models provided adequate fits to the data. However, thepredicted toxicities for every species in the Al-SSD model weresimilar, which was insufficient to provide a reasonable fit. TheSSD for Al that contains more sensitive species needs to befurther investigated.

Distributions of 25 Metal Criteria and CorrelationAnalyses. On the basis of the SSD curves of 25 metals ormetalloids in Figure 2, the toxicity profiles were classified as

highly toxic, moderately toxic, low toxic and lesser toxic withinthe whole concentration thresholds between −1.89 and 3.11.Cr (VI), Ag, Hg, and Tl were classified as highly toxic metals,with log −HC5 values between −1.89 and −1.0. As (III), Cd,Cu, and Sb were classified as moderately toxic metals, with log−HC5 values ranging from −1.0 to 0. Al, Co, Fe, La, Mn, Ni,Pb, and Zn were classified as low toxic metals with log −HC5values from 0 to 1.0. The rest of the metals caused lesser acutetoxicity at the concentrations tested, with log −HC5 valuesfrom 1.0 and 3.11.Correlations between log −HC5 and seven characteristics of

metal ions were good indicators for prediction of the toxicity offree ions. The coefficient of determination ranked in thedescending order was σp > log −βn>ΔE0 > Xm

2r > |log KOH| >Z/r > AN/ΔIP (See Table S4, SI). In the present study, σp andlog −βn were significantly correlated with log-HC5, withadjusted correlation coefficients of 0.7638 (F = 78.626, P =7.02 × 10−9) and 0.7265 (F = 64.760, P = 3.88 × 10−8),respectively. The relationships of actual and model predictedlog −HC5 with softness index (σp) were shown in Figure 3.The softness index σp separated metal ions into three groupsbased on their solubility constants with sulfur and oxygen-containing anions: (1) hard ions, which preferentially bind tooxygen or nitrogen (e.g., Li, Na, Ca, and Mg); (2) soft ions,which preferentially bind to sulfur (e.g., Cd, Hg, Ag, andAs(III)); and (3) borderline ions, which form complexes withoxygen, nitrogen, and sulfur to varying degrees (e.g., Co, Ni,Cu, and Zn). Most soft ions were classified as having high andintermediate toxic potency (σp < 0.10); borderline ions were

Figure 2. Species sensitivity distributions analysis and derivation of thepredicted log −HC5 based on the QICAR regressions for 25 metals ormetalloids. The predicted toxicities were derived from minimum eightspecies (three phyla), including C. tentans, C. pseudogracilis, D. magna,L. acuminata, C. carpio, B. calycif lorus, B. melanostictus, and L. minor.

Environmental Science & Technology Article

dx.doi.org/10.1021/es303309h | Environ. Sci. Technol. 2013, 47, 446−453450

less toxic (0.10 < σp < 0.163); and hard ions exhibited lesserpotency (0.163<σp < 0.250). These findings support theconclusion that σp is the best predictor of toxic potency ofmetals or metalloids to aquatic organisms. Alternatively, log−βn had the greatest negative correlation with toxicity, whichreflected the binding affinities of metal−ligand complexes.Values of log −βn of more toxic metals, such as Hg, Ag, and Tl,were greater than 18.0, while those for metals of intermediatetoxicity were between 11.0 and 18.0. Both ΔE0 and Xm

2r wereweakly correlated (P < 0.001) and the rest of the descriptiveparameters were not significantly (P > 0.05) correlated withtoxic potency. Thus, the minimum set of characteristics ofmetal ions required to reasonably predict WQC of 25 metalswere σp, log −βn, ΔE0, Xm

2r, |log KOH|, Z/r, and AN/ΔIP, whilenot strongly predictive, can be used as complementaryparameters to discriminate among potencies of metals toaquatic organisms.Validation and Applicability of the QICAR-SSD Model.

Values of log −HC5 for 10 metals derived from CMCsrecommended by U.S. EPA in 2009 are shown in Figure 4. TheQICAR-SSD model herein demonstrated a good performanceto predict log-HC5. The order of standard errors betweenpredicted values and values recommended by U.S. EPA were

Zn < Cr (III)<Cu < Hg < Pb < Ni < Ag < Cd < Al < Cr (VI) <As (III). The predicted CMC for Zn was 120 μg/L and theCMC recommended by the U.S. EPA was 110.9 μg/L. Themodel provided the poorest prediction of the potency of As(III). The predicted CMC was 12.16 μg/L, and the CMC valueby U.S. EPA was 340 μg/L. In summary, the comparison resultsindicate that values of log −CMCs predicted by this study andlog −CMCs recommended by U.S. EPA for Zn, Cr (III), Cu,and Hg were in the same order of magnitude. The differencesof values for Pb, Ni, Ag, and Cd were within 1 order ofmagnitude, and for the rest metals or metalloids, the differenceswere within 1.5 orders of magnitude. Of these metals, theCMCs recommended by U.S. EPA for Hg, Cu, Cr (III), Zn, Ni,and Pb were well reproduced by the models with differences ofless than 0.20-fold. The model provided reasonable predictionsfor Ag and Cd with standard error between 0.20 and 1.00. ForAg, the toxicity prediction equation for the sensitive species L.acuminata contains two sensitive characteristics of metal ions,Xm

2r and AN/ΔIP, which resulted in greater toxicity predictionerror and SSD fitting errors. For Cd, it water hardness caninfluence aquatic toxicity. When acute toxicity of Cd to D.magna was determined at three different harnesses, it was foundthat the toxicity in softer water was 5 times greater than thatharder water.4 However, the QICAR-SSD model only focuseson the influence of physicochemical properties of metals andneglects site-specific geochemical conditions such as hardness,pH, temperature, dissolved oxygen, and dissolved organicmatter. Indeed, site-specific geochemical conditions influencethe degree to which organisms take up metals and exhibitadverse effects. It is critical to consider bioavailability inextended QICAR-SSD model, as geographically distinct eco-regions and sites will show distinctive geochemical character-istics.Trivalent and hexavalent metals (Al (III), As (III), and Cr

(VI)) provided weak prediction because of the averagedeviation of 1.4 orders of magnitude. Of the seven character-istics of metal ions, σp exhibited the greatest statisticalsignificance to predict log-HC5, with the linear regressionslope of 26.408 (F = 78.626, P = 7.02 × 10−9). Therefore, σp ofmore charged metals need to be revised by increasing the valueby a factor of 0.05 (1.387/26.408). However, valences of metalions affected the magnitude of σp, which is the reason that theratio of coordinate bond energy of iodide to coordinate bondenergy of fluoride will increase after highly charged metalscombine covalently to bioligands.47 In QICAR-SSD model, theσp per unit charge was used as a criterion for evaluation withthe cumulative contribution of charge neglected. In furtherinvestigation, the QICAR-SSD model for highly charged metalsshould be developed to raise accuracy of predicted WQCs.To provide reasonable predictions of the toxicity of metals to

various organisms, one needs to consider many factors sinceeach metal has unique physicochemical property that will makedifferences in its environmental behavior and toxic mechanismto sensitive species. How to predict toxicity of metals to allspecies under various conditions from limited information isthe key issue to the WQC study. Although the modelspresented can reasonably predict toxicity of metals from limitedinformation, they do need to be further developed in five areas.These include (1) QICAR models for additional representativespecies in order to expend the protection range of WQC; (2)modeling metal toxicity separately by metal valence mightimprove the prediction accuracy of the model; (3) the need toaccount the effect of site-specific geochemical conditions, such

Figure 3. Model for log −HC5 and softness index (σp) at 95%prediction level.

Figure 4. Relationships between predicted log −HC5 and recom-mended log −HC5 derived from WQC.

Environmental Science & Technology Article

dx.doi.org/10.1021/es303309h | Environ. Sci. Technol. 2013, 47, 446−453451

as hardness, pH, temperature, dissolved oxygen and dissolvedorganic matter; (4) optimization can be further generalizedthroughout the entire predictive framework; and (5)bioavailability of multiple metal speciation due to trans-formation should also be considered. These predictive modelscould be useful when data on metal toxicity are lacking orincomplete. While further development of such models isnecessary to determine their range of applicability, the QICAR-SSD model is a promising screen level tool that can be used torapidly predict aquatic toxicity and criteria of metals.

■ ASSOCIATED CONTENT*S Supporting InformationTables with seven metal ion characteristics used in regressionmodels, predicted log −EC50 values of minimum eight species,SSD fitting parameters and correlations between metal ioncharacteristics, and predicted log −HC5. This material isavailable free of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATIONCorresponding Author*Phone: +86-10-84915312. Fax: +86-10-84931804. E-mail:wufengchang@vip.skleg.cn.

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThe present study was supported by the National BasicResearch Program of China (973 Program) (No.2008CB418200), the National Natural Science Foundation ofChina (No. U0833603 and 41130743), and the National WaterPollution Control and Management Technology MajorProjects of China (2012ZX07503-003). J.P.G. was supportedby the program of 2012 “High Level Foreign Experts” (no.GDW20123200120) funded by the State Administration ofForeign Experts Affairs, the P. R. China. J.P.G. was alsosupported by the Canada Research Chair program, an at largeChair Professorship at the Department of Biology andChemistry and State Key Laboratory in Marine Pollution,City University of Hong Kong, and the Einstein ProfessorProgram of the Chinese Academy of Sciences.

■ REFERENCES(1) Shaw, W. H.; Grushkin, B. The toxicity of metal ions to aquaticorganisms. Arch. Biochem. Biophys. 1957, 67 (2), 447−452.(2) Kassem, M. A.; Kassem, A. A.; Ammar, H. O. On heavy metalions contamination from ampoule glass. Pharm. Acta. Helv. 1969, 44(9), 535−540.(3) U.S. EPA. Quality Criteria for Water; National TechnicalInformation Service: Washington, DC, 1976.(4) U.S. EPA. Quality Criteria for Water; Office of Water andHazardous Materials: Washington, DC, 1986.(5) U.S. EPA. National Recommended Water Quality CriteriaCorrection; Office of Water, Office of Science and Technology:Washington, DC, 1999.(6) U.S. EPA. National Recommended Water Quality Criteria; Office ofWater, Office of Science and Technology: Washington, DC, 2002.(7) U.S. EPA. National Recommended Water Quality Criteria; Office ofWater, Office of Science and Technology: Washington, DC, 2004.(8) U.S. EPA. National Recommended Water Quality Criteria; Office ofWater, Office of Science and Technology: Washington, DC, 2006.(9) U.S. EPA. National Recommended Water Quality Criteria; Office ofWater, Office of Science and Technology: Washington, DC, 2009.

(10) De Schamphelaere, K. A. C.; Janssen, C. R. Bioavailabilitymodels for predicting copper toxicity to freshwater green microalgae asa function of water chemistry. Environ. Sci. Technol. 2006, 40 (14),4514−4522.(11) Pagenkopf, G. K. Gill surface interaction model for trace-metaltoxicity to fishes: role of complexation, pH, and water hardness.Environ. Sci. Technol. 1983, 17 (6), 342−347.(12) Campbell, P. G. C. Interactions between Trace Metals and AquaticOrganisms: A Critique of the Free-Ion Activity Model; John Wiley andSons: New York, 1995.(13) Meyer, J. S. A mechanistic explanation for the ln(LC50) vs.ln(Hardness) adjustment equation for metals. Environ. Sci. Technol.1999, 33 (6), 908−912.(14) Meyer, J. S. The utility of the terms “bioavailability” and“bioavailable fraction” for metals. Mar. Environ. Res. 2002, 53 (4),417−423.(15) Meyer, J. S.; Boese, C. J.; Collyard, S. A. Whole-bodyaccumulation of copper predicts acute toxicity to an aquaticoligochaete (Lumbriculus variegatus) as pH and calcium are varied.Comp. Biochem. Physiol., Part C: Toxicol. Pharmacol. 2002, 133 (1−2),99−109.(16) Di Toro, D. M.; Allen, H. E.; Bergman, H. L.; Meyer, J. S.;Paquin, P. R.; Santore, R. C. Biotic ligand model of the acute toxicityof metals. 1. Technical basis. Environ. Toxicol. Chem. 2001, 20 (10),2383−2396.(17) Santore, R. C.; Mathew, R.; Paquin, P. R.; Di Toro, D. M.Application of the biotic ligand model to predicting zinc toxicity torainbow trout, fathead minnow, and Daphnia magna. Comp. Biochem.Physiol., Part C: Toxicol. Pharmacol. 2002, 133 (1−2), 271−285.(18) Deleebeeck, N. M. E.; De Schamphelaere, K. A. C.; Janssen, C.R. A novel method for predicting chronic nickel bioavailability andtoxicity to Daphnia magna in artificial and natural waters. Environ.Toxicol. Chem. 2008, 27 (10), 2097−2107.(19) Kaiser, K. L. E. Correlation and prediction of metal toxicity toaquatic biota. Can. J. Fish. Aquat. Sci. 1980, 37 (2), 211−218.(20) Hansch, C.; Gao, H. Comparative QSAR: radical reactions ofbenzene derivatives in chemistry and biology. Chem. Rev. 1997, 97 (8),2995−3060.(21) Kubinyi, H. QSAR and 3D QSAR in drug design Part 1:methodology. Drug Discovery Today 1997, 2 (11), 457−467.(22) Gramatica, P.; Papa, E. Screening and ranking of POPs forglobal half-life: QSAR approaches for prioritization based on molecularstructure. Environ. Sci. Technol. 2007, 41 (8), 2833−2839.(23) McCloskey, J. T.; Newman, M. C.; Clark, S. B. Predicting therelative toxicity of metal ions using ion characteristics: Microtoxbioluminescence assay. Environ. Toxicol. Chem. 1996, 15 (10), 1730−1737.(24) Williams, M. W.; Turner, J. E. Comments on softnessparameters and metal ion toxicity. J. Inorg. Nucl. Chem. 1981, 43(7), 1689−1691.(25) Newman, M. C.; McCloskey, J. T. Predicting relative toxicityand interactions of divalent metal ions: Microtox® bioluminescenceassay. Environ. Toxicol. Chem. 1996, 15, 275−281.(26) Newman, M. C.; McCloskey, J. T.; Tatara, C. P. Using metal-ligand binding characteristics to predict metal toxicity: Quantitativeion character−activity relationships (QICARs). Environ. HealthPerspect 1998, 106 (6), 1419−1425.(27) Tatara, C. P.; Newman, M. C.; McCloskey, J. T.; Williams, P. L.Predicting relative metal toxicity with ion characteristics: Caeno-rhabditis elegans LC50. Aquat. Toxicol. 1997, 39 (3−4), 279−290.(28) Zhou, D. M.; Li, L. Z.; Peijnenburg, W. J. G. M.; Ownby, D. R.;Hendriks, A. J.; Wang, P.; Li, D. D. A QICAR approach for quantifyingbinding constants for metal-ligand complexes. Ecotoxicol. Environ. Saf.2011, 74 (4), 1036−1042.(29) Walker, J. D.; Enache, M.; Dearden, J. C. Quantitative cationic-activity relationships for predicting toxicity of metals. Environ. Toxicol.Chem. 2003, 22 (8), 1916−1935.

Environmental Science & Technology Article

dx.doi.org/10.1021/es303309h | Environ. Sci. Technol. 2013, 47, 446−453452

(30) Posthuma, L.; Suter, G. W.; Traas, T. P. Species SensitivityDistributions in Ecotoxicology; Lewis Publishers: Boca Raton, FL,U.S.A., 2002.(31) Stephen, C. E.; Mount, D. I.; Hansen, D. J.; Gentile, J. R.;Chapman, G. A.; Brungs, W. A. Guidelines for Deriving NumericalNational Water Quality Criteria for the Protection of Aquatic Organismsand Their Uses; Office of Research and Development: Washington,DC, 1985.(32) U.S. EPA. Methods/Indicators for Determining When Metals Arethe Cause of Biological Impairments of Rivers and Streams: SpeciesSensitivity Distributions and Chronic Exposure-Response Relationshipsfrom Laboratory Data; Office of Research and Development:Cincinnati, OH, U.S.A., 2005.(33) van Straalen, N. M.; Denneman, C. A. J. Ecotoxicologicalevaluation of soil quality criteria. Ecotoxicol. Environ. Saf. 1989, 18 (3),241−251.(34) Fairbrother, A.; Wenstel, R.; Sappington, K.; Wood, W.Framework for metals risk assessment. Ecotoxicol. Environ. Saf. 2007,68 (2), 145−227.(35) U.S. EPA. Framework for Metals Risk Assessment; Office of theScience Advisor: Washington, DC, 2007.(36) Khangarot, B. S.; Ray, P. K. Sensitivity of midge larvae ofChironomus tentans fabricius (Diptera Chironomidae) to heavy metals.Bull. Environ. Contam. Toxicol. 1989, 42 (3), 325−330.(37) Martin, T. R.; Holdich, D. M. The acute lethal toxicity of heavymetals to peracarid crustaceans (with particular reference to fresh-water asellids and gammarids). Water Res. 1986, 20 (9), 1137−1147.(38) Khangarot, B. S.; Ray, P. K. Investigation of correlation betweenphysicochemical properties of metals and their toxicity to the waterflea Daphnia magna straus. Ecotoxicol. Environ. Saf. 1989, 18 (2), 109−120.(39) Khangarot, B. S.; Mathur, S.; Durve, V. S. Comparative toxicityof heavy metals and interaction of metals on a freshwater pulmonatesnail Lymnaea acuminata (Lamarck). Acta Hydrochim. Hydrobiol. 1982,10 (4), 367−375.(40) Rehwoldt, R.; Menapace, L. W.; Nerrie, B.; Alessandrello, D.The effect of increased temperature upon the acute toxicity of someheavy metal ions. Bull. Environ. Contam. Toxicol. 1972, 8 (2), 91−96.(41) Snell, T. W.; Moffat, B. D.; Janssen, C.; Persoone, G. Acutetoxicity tests using rotifers IV. Effects of cyst age, temperature, andsalinity on the sensitivity of Brachionus calycif lorus. Ecotoxicol. Environ.Saf. 1991, 21 (3), 308−317.(42) Khangarot, B. S.; Ray, P. K. Sensitivity of toad tadpoles, Bufomelanostictus (Schneider), to heavy metals. Bull. Environ. Contam.Toxicol. 1987, 38 (3), 523−527.(43) Wang, W. C. Toxicity tests of aquatic pollutants by usingcommon duckweed. Environ. Pollut. 1986, 11 (1), 1−14.(44) Lide, D. R.; Haynes, W. M. CRC Handbook of Chemistry andPhysics; CRC Press: Boca Raton, FL, U.S.A., 2011.(45) Pearson, R. G.; Mawby, R. J. The nature of metal−halogenbonds. In Halogen Chemistry; Gutmann, V., Ed.; Academic Press: NewYork, U.S.A., 1967; Vol. 3, pp 55−84.(46) Base, C. F.; Mesmer, R. E. The Hydrolysis of Cations; John Wileyand Sons: New York, U.S.A., 1976.(47) Wolterbeek, H. T.; Verburg, T. G. Predicting metal toxicityrevisited: general properties vs. specific effects. Sci. Total Environ. 2001,279 (1−3), 87−115.(48) Nieboer, E.; McBryde, W. A. E. Free-energy relationships incoordination chemistry. III. A comprehensive index to complexstability. Can. J. Chem. 1973, 51, 2512−2524.(49) Wang, W. X.; Fisher, N. S. Effects of calcium and metabolicinhibitors on trace element uptake in two marine bivalves. J. Exp. Mar.Biol. Ecol. 1999, 236 (1), 149−164.(50) Veltman, K.; Huijbregts, M. A.; Van Kolck, M.; Wang, W. X.;Hendriks, A. J. Metal bioaccumulation in aquatic species: quantifica-tion of uptake and elimination rate constants using physicochemicalproperties of metals and physiological characteristics of species.Environ. Sci. Technol. 2008, 42 (3), 852−858.

(51) Zhang, R. Q.; Wu, F. C.; Li, H. X.; Feng, C. L.; Guo, G. H.Deriving aquatic water quality criteria for inorganic mercury in Chinaby species sensitivity distributions. Acta Sci. Circumstantiae 2012, 32(1), 1−10 (In Chinese).(52) Wu, F. C.; Meng, W.; Cao, Y. J.; Li, H. X.; Zhang, R. Q.; Feng,C. L.; Yan, Z. G. Derivation of aquatic life water quality criteria forcadmium in freshwater in China. Res. Environ. Sci. 2011, 24 (2), 172−184 (In Chinese).(53) Wu, F. C.; Feng, C. L.; Cao, Y. J.; Zhang, R. Q.; Li, H. X.; Liao,H. Q.; Zhao, X. L. Toxicity characteristic of zinc to freshwater biotaand its water quality criteria. Asian J. Ecotoxicol 2011, 6 (4), 367−382(In Chinese).(54) Bodar, C. W. M.; Zee, A. V. D.; Voogt, P. A.; Wynne, H.;Zandee, D. I. Toxicity of heavy metals to early life stages of Daphniamagna. Ecotoxicol. Environ. Saf. 1989, 17 (3), 333−338.(55) Forget, J.; Pavillon, J. F.; Menasria, M. R.; Bocquene, G.Mortality and LC50 values for several stages of the marine copepodTigriopus brevicornis (Muller) exposed to the metals arsenic andcadmium and the pesticides atrazine, carbofuran, dichlorvos, andmalathion. Ecotoxicol. Environ. Saf. 1998, 40 (3), 239−244.(56) Khangarot, B. S.; Das, S. Acute toxicity of metals and referencetoxicants to a freshwater ostracod, Cypris subglobosa Sowerby, 1840and correlation to EC(50) values of other test models. J. Hazard.Mater. 2009, 172 (2−3), 641−649.(57) Sun, H. W.; Huang, G. W.; Li, S. X.; Dai, S. G.; Zhang, Y.; Qiao,F. Study on toxic effects of triphenyltin and tributyltin on Daphniamagna. Environ. Chem. 2000, 19 (3), 235−239.(58) Lin, H. B.; Song, Z. H.; Wang, F. F. Toxic effect of triphenyltinon Chironomus plumosus larve,Galba pervia and Tubifex tubifex. J.Qingdao Univer. Sci. Technol. (Nat. Sci. Ed.) 2007, 5, 235−239 (InChinese).(59) Wu, F. C.; Feng, C. L.; Cao, Y. J.; Zhang, R. C.; Li, H. X.; Zhao,X. L. Aquatic life ambient freshwater quality criteria for copper inChina. Asian J. Ecotoxicol. 2011, 6 (6), 617−628 (In Chinese).(60) Boonyapookana, B.; Upatham, E. S.; Kruatrachue, M.;Pokethitiyook, P.; Singhakaew, S. Phytoaccumulation and phytotox-icity of cadmium and chromium in duckweed Wolf f ia globosa. Int. J.Phytorem. 2002, 4 (2), 87−100.

Environmental Science & Technology Article

dx.doi.org/10.1021/es303309h | Environ. Sci. Technol. 2013, 47, 446−453453

Recognizing Environmental Science & Technology’s Contributors:Top Papers 2012

Our authors are pivotal. We ask them to submitmanuscripts that not only meet high scientific standards,

but are carefully drafted and redrafted until the message isdelivered just right. Through rejection and resubmission theypersevere. It is therefore fitting to recognize ES&T’s excep-tional papers from 2012.Beginning in December editors nominate papers and our

editorial advisory board carefully combs through the nominees.Our winners were highlighted in issue 7 this year, and below isthe complete list of winners and runners up. Runner-up papersin each category are worthy of praise. They were chosen fromnearly 1600 manuscripts published in 2012. These articles arethe top 1% of our content.For 2013 we introduce a new reader’s choice category to

ES&T’s awards. When you read exceptional Science,Technology, Policy or Feature articles in ES&T you cantweet your nominations to @EnvSciTech with #ReadersChoicethroughout the year. Emails to est@uiowa.edu will also becatalogued (please title: “readers choice nomination”). We willcollect these and announce the selections early next year.

■ FEATURE

Top Paper. Transformations of Nanomaterials in theEnvironmentGregory V. Lowry, Kelvin B. Gregory, Simon C. Apte, and

Jamie R. LeadEnvironmental Science & Technology 2012, 46 (13), 6893−

6899 (see article)First Runner-Up. Ecological and Evolutionary Functional

GenomicsHow Can It Contribute to the Risk Assessment ofChemicals?Nico M. van Straalen and Martin E. FederEnvironmental Science & Technology 2012, 46 (1), 3−9 (see

article)Second Runner-Up. Sea-Surface Chemistry and Its Impact

on the Marine Boundary LayerD. J. Donaldson and Christian GeorgeEnvironmental Science & Technology 2012, 46 (19), 10385−

10389 (see article)

■ ENVIRONMENTAL POLICY

Top Paper. Using Land To Mitigate Climate Change:Hitting the Target, Recognizing the Trade-offsJohn Reilly, Jerry Melillo, Yongxia Cai, David Kicklighter,

Angelo Gurgel, Sergey Paltsev, Timothy Cronin, AndreiSokolov, and Adam SchlosserEnvironmental Science & Technology 2012, 46 (11), 5672−

5679 (see article)First Runner-Up. Fuel Miles and the Blend Wall: Costs and

Emissions from Ethanol Distribution in the United StatesBret Strogen, Arpad Horvath, and Thomas E. McKoneEnvironmental Science & Technology 2012, 46 (10), 5285−

5293 (see article)

Second Runner-Up. Public Health Impacts of CombustionEmissions in the United KingdomSteve H. L. Yim and Steven R. H. BarrettEnvironmental Science & Technology 2012, 46 (8), 4291−

4296 (see article)

■ ENVIRONMENTAL SCIENCE

Top Paper. Exposure Assessment for Estimation of theGlobal Burden of Disease Attributable to Outdoor AirPollutionMichael Brauer, Markus Amann, Rick T. Burnett, Aaron

Cohen, Frank Dentener, Majid Ezzati, Sarah B. Henderson,Michal Krzyzanowski, Randall V. Martin, Rita Van Dingenen,Aaron van Donkelaar, and George D. ThurstonEnvironmental Science & Technology 2012, 46 (2), 652−660

(see article)First Runner-Up (Tie). Photochemical Formation of

Polybrominated Dibenzo-p-dioxins from EnvironmentallyAbundant Hydroxylated Polybrominated Diphenyl EthersKristina Arnoldsson, Patrik L. Andersson, and Peter HaglundEnvironmental Science & Technology 2012, 46 (14), 7567−

7574 (see article)First Runner-Up (Tie). Photochemical Formation of

Brominated Dioxins and Other Products of Concern fromHydroxylated Polybrominated Diphenyl Ethers (OH-PBDEs)Paul R. Erickson, Matthew Grandbois, William A. Arnold,

and Kristopher McNeillEnvironmental Science & Technology 2012, 46 (15), 8174−

8180 (see article)Second Runner-Up. Predicting Water Quality Criteria for

Protecting Aquatic Life from Physicochemical Properties ofMetals or MetalloidsFengchang Wu, Yunsong Mu, Hong Chang, Xiaoli Zhao,

John P. Giesy, and K. Benjamin WuEnvironmental Science & Technology 2013, 47 (1), 446−453

(see article)Third Runner-Up. Quantitative and Qualitative Analysis of

Naphthenic Acids in Natural Waters Surrounding the CanadianOil Sands IndustryMatthew S. Ross, Alberto dos Santos Pereira, Jon Fennell,

Martin Davies, James Johnson, Lucie Sliva, and Jonathan W.MartinEnvironmental Science & Technology 2012, 46 (23), 12796−

12805 (see article)

■ ENVIRONMENTAL TECHNOLOGY

Top Paper. Detection, Characterization, and Abundance ofEngineered Nanoparticles in Complex Waters by HyperspectralImagery with Enhanced Darkfield MicroscopyAppala Raju Badireddy, Mark R. Wiesner, and Jie Liu

Comment

pubs.acs.org/est

© XXXX American Chemical Society A dx.doi.org/10.1021/es401684v | Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Environmental Science & Technology 2012, 46 (18), 10081−10088 (see article)First Runner-Up. Monitoring of a Simulated CO2 Leakage

in a Shallow Aquifer Using Stable Carbon IsotopesAlexandra Schulz, Carsten Vogt, Hendrik Lamert, Anita

Peter, Ben Heinrich, Andreas Dahmke, and Hans-HermannRichnowEnvironmental Science & Technology 2012, 46 (20), 11243−

11250 (see article)Second Runner-Up. Water Footprint of European Cars:

Potential Impacts of Water Consumption along AutomobileLife CyclesMarkus Berger, Jens Warsen, Stephan Krinke, Vanessa Bach,

and Matthias FinkbeinerEnvironmental Science & Technology 2012, 46 (7), 4091−

4099 (see article)Third Runner-Up. High Performance Thin-Film Compo-

site Forward Osmosis Hollow Fiber Membranes with Macro-void-Free and Highly Porous Structure for Sustainable WaterProductionPanu Sukitpaneenit and Tai-Shung ChungEnvironmental Science & Technology 2012, 46 (13), 7358−

7365 (see article)Matt Hotze,* Managing Editor

■ AUTHOR INFORMATIONCorresponding Author*E-mail: m_hotze@acs.orgNotesViews expressed in this editorial are those of the author and notnecessarily the views of the ACS.The authors declare no competing financial interest.

Environmental Science & Technology Comment

dx.doi.org/10.1021/es401684v | Environ. Sci. Technol. XXXX, XXX, XXX−XXXB

The Authors: Fengchang Wu, Yunsong Mu, Hong Chang, Xiaoli Zhao, John P. Giesy 1

and K. Benjamin Wu 2

3

Manuscript ID es-2012-03309h entitled " Predicting Water Quality Criteria for 4

Protecting Aquatic Life from Physico-chemical Properties of Metals or Metalloids" 5

6

Number SI pages: 4 7

8

Number the tables: 4 9

10

S1

Supporting Information 11

Table S1. Values of seven ion characteristics for all metals or metalloids 12

Metals σp log-βn Xm2r AN/∆IP Z/r |logKOH| ∆E0

Ag 0.074 20.6 4.284 6.209 0.87 12.4 0.80

Al 0.136 14.11 1.4 1.351 5.556 4.3 1.66

As(III) 0.106 19.3 2.756 3.395 5.172 2.2 0.68

Ba 0.183 7.78 1.069 11.69 1.481 13.4 2.9

Be 0.172 9.3 1.109 0.45 4.444 3.7 1.85

Ca 0.181 11 1 3.47 2.02 12.7 2.76

Cd 0.081 18.78 2.713 6.068 2.105 10.1 0.4

Co 0.13 10.2 2.65 2.94 2.685 9.7 0.28

Cr(III) 0.107 11.2 1.708 1.66 4.839 4.0 0.41

Cr(VI) 0.107 11.2 1.212 1.134 13.64 4.0 0.13

Cu 0.104 18.5 2.635 2.309 2.74 8.0 0.16

Fe(III) 0.103 15.77 1.842 1.798 5.455 2.2 0.77

Hg 0.065 21.7 4.08 9.62 1.96 3.4 0.91

K 0.232 1.6 0.93 4.38 0.725 14.5 2.92

La 0.171 15.5 1.27 7.36 2.828 8.5 2.37

Li 0.247 2.79 0.71 0.56 1.316 13.6 3.05

Mg 0.167 8.64 1.24 1.62 2.778 11.6 2.38

Mn 0.125 14.2 1.61 3.045 2.985 10.6 1.185

Na 0.211 1.66 0.88 2.14 0.98 14.2 2.71

Ni 0.126 11.33 2.517 2.662 2.899 9.9 0.23

Pb 0.131 18.3 6.46 10.78 1.681 7.7 0.126

Sb 0.119 10.9 3.194 6.439 3.947 0 0.66

Sr 0.174 8.8 1.02 7.12 1.786 13.2 2.89

Tl 0.097 18.47 3.557 5.133 2.67 2.6 0.502

Zn 0.115 16.4 2.015 3.501 2.703 8.2 0.76

13

S2

Table S2. Acute toxicities of 25 metals or metalloids to representative species from eight taxonomic 14

families (log-EC50) 15

Metals C. tentans C. pseudogracilis D. magna L. acuminata C. carpio B.calyciflorus B. melanostictus L. minor

Ag -0.194 -0.615 -0.832 -2.225 -0.326 -1.12 -1.193 0.998

Al 2.524 3.495 1.487 2.853 3.678 1.707 3.763 3.832

As(III) 0.901 0.986 -0.37 0.409 2.517 0.399 1.427 1.618

Ba 4.796 6.97 3.753 6.019 3.571 2.577 4.61 6.859

Be 4.258 6.147 3.209 3.268 4.424 3.202 4.47 5.015

Ca 4.257 6.073 2.595 4.219 4.888 1.698 4.704 6.6

Cd 0.276 0.125 -0.18 1.135 0.417 -0.32 1.321 0.575

Co 2.942 4.25 2.897 0.53 2.295 2.269 1.76 1.62

Cr(III) 2.145 3.083 2.543 2.261 2.413 2.605 3.078 1.24

Cr(VI) 0.653 0.978 2.557 3.208 4.267 2.605 3.488 -0.503

Cu 2.66 3.804 2.491 0.751 2.249 1.911 1.941 1.446

Fe(III) 1.347 1.763 0.899 2.004 2.533 1.447 2.84 1.678

Hg -0.612 -1.252 -1.226 -0.976 -0.178 -0.447 -0.935 0.938

K 7.102 10.49 5.965 4.586 4.898 4.29 5.168 8.113

La 3.3 4.532 0.978 4.558 3.725 0.828 4.211 5.767

Li 7.345 10.78 5.532 4.156 5.643 4.037 5.618 8.682

Mg 4.217 6.116 3.448 3.263 3.57 2.521 4.23 5.682

Mn 2.201 3.036 1.458 2.801 2.251 0.981 3.357 2.855

Na 6.502 9.636 5.949 4.163 4.301 4.306 5.101 7.274

Ni 0.965 1.11 -0.08 0.413 1.448 -0.01 1.603 0.791

Pb 1.755 2.233 -0.02 1.396 5.663 0.085 1.791 1.421

Sb 2.527 3.636 2.648 0.184 2.447 3.138 0.83 1.913

Sr 4.39 6.353 3.389 5.041 3.396 2.296 4.624 6.619

Tl 0.773 0.839 -0.07 -0.91 1.185 0.601 0.108 1.127

Zn 1.59 2.08 0.669 2.034 1.8 0.594 2.652 1.963

S3

Table S3. SSD fitting parameters and CMCs derivation for 25 metals or metalloids (µg/L), with coefficients, standard error, RSS, F and P values. 16

17 Metals a Xc k a-SE Xc-SE k-SE Adj.r

2 RSS F P log-HC5 AW CMCs WQC

Ag 0.9545 -0.7612 2.573 0.0441 0.0552 0.3132 0.9823 0.0083 531.434 1.50×10-6 -1.8865 107 0.695 3.2

Al 923.498 12.268 0.8280 0.0001 1406.4 0.5155 0.9418 0.0273 160.494 2.91×10-5 0.4035 27 34.18 750

As(III) 0.9758 0.9257 2.0628 0.0659 0.0959 0.3222 0.9734 0.0125 352.93 4.15×10-6 -0.4892 75 12.16 340

Ba 0.9387 4.655 1.04 0.1049 0.3305 0.2747 0.9546 0.0213 206.452 1.57×10-5 1.8879 137 5292 /

Be 0.9935 4.1851 1.6012 0.1234 0.2234 0.409 0.9067 0.0437 99.542 9.40×10-5 2.3505 4 448.3 /

Ca 1.0816 4.6755 0.929 0.1936 0.479 0.2874 0.9437 0.0264 166.21 2.67×10-5 1.4173 40 522.8 /

Cd 0.9146 0.2901 3.5246 0.0409 0.042 0.4779 0.981 0.0089 494.711 1.79×10-6 -0.5186 112 16.97 2

Co 0.9617 2.2409 1.8697 0.0682 0.1124 0.3312 0.9655 0.0162 272.42 7.89×10-6 0.688 59 143.8 /

Cr(III) 0.9147 2.4172 4.7783 0.0557 0.0404 0.9387 0.9699 0.0141 312.564 5.61×10-6 1.1656 52 380.7 570

Cr(VI) 1.539 3.5298 0.6637 0.6118 1.2343 0.1768 0.9596 0.019 231.794 1.18×10-5 -1.5837 52 0.678 16

Cu 1.2362 1.0483 1.9225 0.2293 0.2242 0.3665 0.9776 0.0105 420.478 2.69×10-6 -0.5988 64 8.06 13

Fe 0.908 1.6699 3.477 0.0386 0.0411 0.482 0.9818 0.0085 517.632 1.60×10-6 0.8524 56 199.3 /

Hg 0.9273 -0.7808 3.3526 0.0593 0.0647 0.5625 0.9594 0.019 231.176 1.18×10-5 -1.6352 201 2.328 1.4

K 0.8819 5.5426 1.157 0.0672 0.2544 0.2983 0.9353 0.0303 144.28 3.79×10-5 3.1124 39 25263 /

La 1.2924 4.3943 0.8022 0.4173 0.8948 0.2916 0.9237 0.0358 122.172 5.69×10-5 0.3894 139 170.4 /

Li 0.8987 5.7567 0.9453 0.0828 0.3412 0.2919 0.9135 0.0406 107.513 7.79×10-5 2.7612 7 2019 /

Mg 0.8873 3.7525 1.9811 0.0548 0.1118 0.4185 0.9518 0.0226 194.072 1.82×10-5 2.3300 24 2565 /

Mn 1.7877 3.2607 1.2941 0.9485 0.7922 0.3552 0.9669 0.0155 283.517 7.14×10-6 0.5188 55 90.81 /

Na 0.9391 5.4775 1.0534 0.083 0.2785 0.2288 0.9343 0.0308 141.974 3.94×10-5 2.7452 23 6396 /

Ni 0.9664 2.0934 2.4283 0.0487 0.0616 0.3178 0.9799 0.0094 468.199 2.06×10-6 0.8957 59 232 470

Pb 0.9648 1.5224 1.9592 0.0869 0.14 0.5812 0.9148 0.0399 109.203 7.49×10-5 0.0388 207 113.2 65

Sb 1.367 2.8647 1.1087 0.4263 0.6144 0.3247 0.9537 0.0217 202.104 1.65×10-5 -0.0857 51 20.93 /

Sr 0.9595 4.3693 1.1481 0.0829 0.2228 0.2338 0.9632 0.0173 254.99 9.29×10-6 1.8426 88 3063 /

Tl 3.7912 2.0502 1.334 6.469 1.9196 0.4097 0.9757 0.0114 386.453 3.31×10-6 -1.1845 204 6.669 /

Zn 1.1602 1.9188 2.237 0.2848 0.2535 0.8688 0.9061 0.0465 98.865 9.55×10-5 0.5330 65 110.9 120

S4

Table S4. Correlations between metal ion characteristics and predicted log-HC5, with

intercept, slope, coefficients values, F, P and Ranks.

Metals σp log-βn ∆E0 Xm2r |logKOH| Z/r AN/∆IP

Intercept -3.0211 3.3881 -0.7015 2.1214 -0.7487 1.2523 1.1577

Intercept-SE 0.4358 0.3735 0.3161 0.4219 0.5273 0.4457 0.4841

Slope 26.408 -0.2162 0.9998 -0.6886 0.1694 -0.1914 -0.1216

Slope-SE 2.9782 0.0269 0.1853 0.1663 0.0567 0.1089 0.0918

Adj.R2 0.7638 0.7265 0.5394 0.4022 0.2481 0.0801 0.0306

F 78.626 64.760 29.104 17.147 8.918 3.089 1.756

P 7.02×109 3.88×108 1.76×105 3.96×104 0.0066 0.0921 0.1981

Ranks 1 2 3 4 5 6 7