Cytotoxicity towards CCO cells of imidazolium ionic liquids with functionalized side chains: Preliminary QSTR modeling using regression and classification based approaches

Ecotoxicology and Environmental Safety 112 (2015) 22–28

Contents lists available at ScienceDirect

Ecotoxicology and Environmental Safety journal homepage: www.elsevier.com/locate/ecoenv

Cytotoxicity towards CCO cells of imidazolium ionic liquids with functionalized side chains: Preliminary QSTR modeling using regression and classification based approaches Marina Cvjetko Bubalo a, Kristina Radošević a, Višnja Gaurina Srček a, Rudra Narayan Das b, Paul Popelier c, Kunal Roy b,c,n a Laboratory for Cell Technology, Application and Biotransformations, Faculty of Food Technology and Biotechnology, University of Zagreb, Pierottijeva 6, HR-10000 Zagreb, Croatia b Drug Theoretics and Cheminformatics Laboratory, Department of Pharmaceutical Technology, Jadavpur University, Kolkata 700032, India c Manchester Institute of Biotechnology, University of Manchester, Manchester M1 7DN, United Kingdom

art ic l e i nf o

a b s t r a c t

Article history: Received 24 July 2014 Received in revised form 13 October 2014 Accepted 20 October 2014

Within this work we evaluated the cytotoxicity towards the Channel Catfish Ovary (CCO) cell line of some imidazolium-based ionic liquids containing different functionalized and unsaturated side chains. The toxic effects were measured by the reduction of the WST-1 dye after 72 h exposure resulting in dose- and structure-dependent toxicities. The obtained data on cytotoxic effects of 14 different imidazolium ionic liquids in CCO cells, expressed as EC50 values, were used in a preliminary quantitative structure–toxicity relationship (QSTR) study employing regression- and classification-based approaches. The toxicity of ILs towards CCO was chiefly related to the shape and hydrophobicity parameters of cations. A significant influence of the quantum topological molecular similarity descriptor ellipticity (ε) of the imine bond was also observed. & Elsevier Inc. All rights reserved.

Keywords: Imidazolium ionic liquids CCO cell line In vitro toxicity Quantitative structure–toxicity relationship Predictive toxicology

1. Introduction Over the past decades, ionic liquids (ILs) have been intensively studied as a green replacement for standard volatile and flammable organic solvents, which is reflected in the more than 30,000 scientific papers related to their preparation, characterization, application and impact on the environment. However, it was demonstrated that ILs were not intrinsically green (Cvjetko Bubalo et al., 2014a; Egorova and Ananikov, 2014) and could become potent water and soil contaminants if not handled properly. Consequently, their persistence in the environment, (bio)degradation, migration, bioaccumulation, and (eco)toxicity should be proactively assessed prior to their large-scale application (Cvjetko Bubalo et al., 2014a; Pham et al., 2010). So far, toxicity studies on ILs have been conducted by performing a series of tests on bacteria (Ranke et al., 2004; Matzke et al., 2007; Ventura et al., 2012), yeast (Zhu et al., 2013), algae (Cho et al., 2008; Latała et al., 2010), nematode (Swatloski et al., 2004), mammals (Yu et al., 2009), plants n Corresponding author at: Jadavpur University, Department of Pharmaceutical Technology, Drug Theoretics and Cheminformatics Laboratory, Raja S C Mullick Road, Kolkata 700032, India. Fax: þ 91 33 28371078. E-mail address: [email protected] (K. Roy).

http://dx.doi.org/10.1016/j.ecoenv.2014.10.029 0147-6513/& Elsevier Inc. All rights reserved.

(Matzke et al., 2007; Zhang et al., 2013; Cvjetko Bubalo et al., 2014b; Wang et al., 2009) as well as in different mammalian (Ranke et al., 2004; Stepnowski et al., 2004, Stolte et al., 2007; Wang et al., 2007) and fish cell lines (Radošević et al., 2013), where depending on the used test-system and chemical structure, ILs showed moderate to high toxicity, in general. Nevertheless, ILs present a group of exceptional chemicals with vast possible variations in their structure, and thus it is crucial to understand the chemical and environmental factors controlling the behavior of ILs in the environment in order to design them as environmentally benign (Cvjetko Bubalo et al., 2014a). The development of predictive chemometric models constitutes an essential part of the toxicity assessment of different chemicals including ILs. It helps in predicting the toxicity of untested compounds and it aids in deriving a rational basis for correlating the toxicity of chemicals with their structural attributes while reducing the number of experiments involving animals. Structure-toxicity relationship studies for ILs have already prompted scientists to develop a variety of predictive models in order to estimate the toxicity of ILs, while avoiding the costs associated with the production and testing of new ILs (Luis et al., 2007, 2010; García-Lorenzo et al., 2008; Torrecilla et al., 2010; Fatemi and Izadiyan, 2011; Alvarez-Guerra and Irabien, 2011; Cho

M. Cvjetko Bubalo et al. / Ecotoxicology and Environmental Safety 112 (2015) 22–28

et al., 2013; Das and Roy, 2013, 2014). The cytotoxic effects of ionic liquids have largely been attributed to lipophilic interactions with cell membranes and cellular proteins, leading to disruption of membrane or protein function (Stolte et al., 2007). The influence of the cationic head group and alkyl side chain on cytotoxicity is also mainly driven by the lipophilicity of the compound. Functionalized side chains (polar ether, hydroxyl and nitrile functional groups within the side chains) lead to lower toxicity. The anions also play a significant role in the cytotoxicity of ILs (Stolte et al., 2006). CruzMonteagudo et al. identified several cytotoxicophores of ILs: cationic linear alkyl side chain of length 45; anions with highly fluorinated alkyl side chains (a fluorocarbonated side chain of length Z2, or two or more trifluoromethyl groups); cationic aromatic N-heterocycles with linear alkyl side chain of length Z4; six-membered aromatic rings with a methyl substituent, which can be either the cation head group or its substituent (CruzMonteagudo et al., 2013). Recently, we reported the toxicity towards a CCO fish cell line of imidazolium ILs containing different anions and alkyl chain lengths as the substituent at the cation ring (Radošević et al., 2013). Because some authors (Samori et al., 2007) reported that the presence of functionalized groups affects the toxicity of ILs, we tested the toxicity toward a CCO cell line of N-alkoxyl- and aminosubstituted imidazolium-based ILs, as well as alkenyl-, alkynyl-, and benzyl-substituted imidazolium-based ILs. These data, together with previously obtained data on toxicity of imidazolium ILs, were used to build preliminary models based on the QSTR approach to elucidate chemical and structural factors controlling the toxicity of ILs. Note that this is the first attempt to build QSTR models of ILs for the endpoint of CCO cell line toxicity.

23

content of the ILs was determined by Karl Fischer titration (Mettler Toledo C20X compact Coulometric) and was typically found to vary between 1–3% (w/w). The water content value of each IL was considered when preparing stock solutions in order to obtain more accurate EC50 values. The CCO cells from the exponential growth phase were seeded at a density of 5  104 cells/well into 96-well plates in 100 μL of media. After overnight cell growth, the culture medium was replaced with a fresh one containing tested ILs within the range of 0.1–10 mmol L  1, and cells were incubated for 72 h. After the incubation period, 10 μL of tetrazolium salt WST-1 {4-[3-(4-iodophenyl)-2-(4-nitrophenyl)-2H-5-tetrazolio]-1,3-benzene disulfonate) was added to each well and the cells were incubated for an additional 4 h. The absorbance was spectrophotometrically quantified at 450 nm on the microplate reader (Tecan, Switzerland). The experiments were performed three times for each IL concentration and the results were expressed as a cell viability, i.e. percentage of treated cells versus control cells (mean7standard deviation). The EC50 values were obtained from the dose–response curves using equations of best-fitted trend-lines. 2.4. Morphological assessment by fluorescent microscopy Briefly, CCO cells were seeded in six-well plates (Corning, USA) at a concentration of 1  105 cells/mL, allowed to attach for 24 h and exposed for 72 h to 1 mmol L  1 of selected ILs. Following exposure, the cells were washed with sterile PBS and stained by a mixture of acridine orange (AO) (100 μg/mL in PBS) and ethidium bromide (EB) (100 μg/mL in PBS) for 10 min. The cells were examined using a fluorescent microscope Olympus BX51 (Olympus, Japan) with integrated camera. 2.5. Quantitative structure–toxicity relationship (QSTR) modeling

2. Materials and methods 2.1. Ionic liquids Seven ILs evaluated for the cytotoxicity towards the CCO cells in this study (1,2-dimethyl,3-pentylimidazolium bis(trifluoromethanesulfonyl)imide [C5mmim][Tf2N], 1-methyl-3-pentoxyimidazolium bis(trifluoromethanesulfonyl)imide [C5omim][Tf2N], 1-heptyl-2,3-dimethylimidazolium bis(trifluoromethanesulfonyl)imide [C7mmim] [Tf2N], 1-allyl-3-methylimidazolium bis(trifluoromethanesulfonyl)imide [C3-2mim][Tf2N], 1-methyl-3-propargylimidazolium bis(trifluoromethanesulfonyl)imide [C3-3mim][Tf2N], 1-benzyl-3-methylimidazolium bis(trifluoromethanesulfonyl)imide [bzmim][Tf2N] and 1-(3N,N-dimethylaminopropyl)-3-methylimidazolium bis(trifluoromethanesulfonyl)imide [apmim][Tf2N] were synthesized as reported previously (Cvjetko et al., 2012). The IUPAC names, abbreviations and molecular structures of the newly synthesized ILs are given in Table 1, together with seven other ILs whose cytotoxicity towards CCO cells was reported previously by Radošević et al. (2013). We have used all 14 ILs for our in silico modeling work in this study. 2.2. Cell culture The CCO fish ovary cell line was purchased from the American Type Culture Collection (ATCC: CRL-2772). Cells were cultured in 25 cm2 T-flasks in Dulbecco's Modified Eagle's Medium (DMEM, Gibco, UK) supplemented with 10% (v/v) fetal bovine serum (FBS, Gibco, UK) and maintained at 30 °C in a humidified atmosphere of 5% CO2. 2.3. Cytotoxicity assay The cytotoxicity of ILs was determined using a WST-1 assay (Roche, Germany). Prior to the cytotoxicity assays, the water

We have attempted to develop classification- and regressionbased QSTR models using the set of 14 imidazolium ILs (out of which seven compounds were previously synthesized and tested; Cvjetko et al., 2012) for their toxicity toward a CCO cell line. The regression-based model should be able to accurately predict the quantitative toxicity values while the classification-based model could initially be used for filtering the ILs out into toxic and nontoxic classes. Considering the limited size of the data set and limited chemical diversity, this attempt may be considered as preliminary one. The toxicity values were converted into the molar negative logarithmic unit (pEC50, M) and a threshold value of pEC50 ¼ 3.0 M, based on the average value of the quantitative toxicity entities, was chosen for the two-group classification analysis such that those bearing a pEC50 value less than 3 were treated as non-toxic while those with more than 3 as toxic. Various twodimensional descriptors (Table S1 in Supplementary material) computed using PaDEL-Descriptor software (version 2.11) (Yap, 2011) and chemometric tools (i.e. linear discriminant analysis (LDA) (Fisher, 1936) and stepwise multiple linear regression (MLR) (Darlington, 1990) were employed to perform classification- and regression-based analyses, respectively. Although we tried to incorporate predictor variables defining chemical features of both the cations and anions, the best model was obtained employing cationic descriptors only. The LDA analysis was performed using a stepping algorithm, F to enter¼ 4.0 and F to remove¼ 3.9, with a tolerance value of 0.01 as a strategy for the selection of essential chemical features. An effort was also made in deriving a regressional QSTR model on the available quantitative toxicity values employing stepwise regression analysis (objective function F to enter¼4.0 and F to remove¼ 3.9) as a descriptor selection tool. During the regression analysis, we have used a pool of quantum topological molecular similarity (QTMS) descriptors (Popelier, 1999) for cations along with the 2D-descriptors as listed

24

M. Cvjetko Bubalo et al. / Ecotoxicology and Environmental Safety 112 (2015) 22–28

Table 1 List of imidazolium-based ILs used for QSTR modeling and their effective concentration (EC50) values obtained in CCO cells.

Systematic name

Abbreviation

Chemical formula

NN

*

and shaded: Seven results published previously by Radošević et al., 2013.

EC50 (mmol L-1)

M. Cvjetko Bubalo et al. / Ecotoxicology and Environmental Safety 112 (2015) 22–28

25

in Table S1 in Supplementary material. The QTMS descriptors focus on bond critical points, which occur when the gradient of the electron density vanishes at a particular point in space, located somewhere between two bonded nuclei. We have previously found that QTMS descriptors can be successfully used alongside with 2D topological descriptors to model toxicity of ionic liquids (Roy et al., 2014).

3. Results and discussion 3.1. Cytotoxicity assessment of ILs In vitro toxicity tests using fish cell lines for the ecotoxicology assessments of different chemicals and samples satisfy a societal desire to avoid the use of whole animals (Schirmer, 2006). CCO cells used in this study have proven to be a good in vitro test-system in cytotoxicity studies of metals (Tan et al., 2008), as well as in our previous reports on cytotoxicity of estrogens (Radošević et al., 2011), environmental samples (Ternjej et al., 2013), and imidazolium ionic liquids (Radošević et al., 2013). In the present work we have evaluated the cytotoxicity towards CCO cells of several imidazolium ILs with functionalized and unsaturated side groups. Fig. 1a and b show the effects of ILs tested on CCO cell viability after 72 h exposure, while corresponding EC50 values are presented in Table 1. All tested ILs showed inhibitory effects on CCO cell viability in a marked dose-dependent manner. Within the group of tested ILs, the toxicity of an imidazolium-based IL with an oxygenated functionalized group, that is, 1-methyl-3-pentoxyimidazolium bis(trifluoromethanesulfonyl)imide [C5omim][Tf2N], was determined (Fig. 1a). No significant difference in toxic effect (EC50 ¼0.2670.04 mmol L  1) was observed compared to non-functionalized analog 1-methyl-3-pentylimidazolium bis(trifluoromethanesulfonyl) imide [C5mim][Tf2N] (EC50 ¼ 0.2670.05 mmol L  1). It was expected that the introduction of an oxygenated side chain in the side chain would greatly decrease toxicity as reported by Frade and Alfonso (2010) as well as Samori et al. (2007). Deng et al. (2011) explained this effect by a decrease in the value of their octanol–water partition coefficient if the alkyl side chains possess oxygen-containing functional groups (hydroxyl, ester and ether group) compared to non-functionalized analogs, meaning lower lipophilicity and thus lower toxicity of this group of ILs. The discrepancy between our results and the results presented by Samori et al. (2007, 2010) could be explained by the position of the oxygen atom within the side chain, causing a different charge distribution thorough the chain and thus its hydrophobicity. For example, in the 1-methoxyethyl-3-methylimidazolium based IL, oxygen is placed at the end of the chain, lowering the hydrophobicity of the alkyl chain and consequently its toxic effect compared to non-alkoxylated analog (Samori et al., 2007). On the other hand, in the 1-methyl-3-pentoxyimidazolium based IL tested in this study, the oxygen atom is bonded to nitrogen at position 1 of the imidazolium ring, leaving a five-carbon-long hydrophobic chain, which retains a similar level of hydrophobicity and consequently a similar mode of toxicity as a non-functionalized analog. From the above, it could be concluded that the position of oxygen in the functionalized group influencing the hydrophobicity/hydrophilicity of the side chains might have an impact on toxicity of oxygenated ILs. The introduction of a methyl group at the C-2 position of [C5mmim][Tf2N] and [C7mmim][Tf2N] resulted in higher or the same level of toxicity than of non-functionalized analogs (EC50 o0.1 mmol L  1), which is in agreement with the results for imidazolium ILs cytotoxicity in IPC-81 leukemia cells from rats (EC50 values for [C6mmim][BF4] and [C6mim][BF4] were 0.09 mmol L  1 and 0.8 mmol L  1, respectively) (Ranke et al., 2007). When comparing the toxicities of ILs with unsaturated [C3-2mim] [Tf2N] and [C3-3mim][Tf2N]) or aromatic [bzmim][Tf2N] side chains

Fig. 1. Effects of (a) different functionalized imidazolium ILs and (b) unsaturated side chains on CCO cell viability using the WST-1 assay. Data are expressed as a percentage of treated to unexposed control cells 7 standard deviation of three replicates for each exposure concentration. The symbol n denotes significant difference from control (p o 0.05).

in CCO cells (Fig. 1b), similar levels of toxicity were obtained (EC50 ¼0.31–0.55 mmol L  1). Relatively speaking, the EC50 values obtained for [C3-2mim][Tf2N] (0.55 mmol L  1) and [C3-3mim][Tf2N] (0.41 mmol L  1) were in the same range as the ones determined in HeLa cells (1.63 mmol L  1 for the allyl chain and 0.33 mmol L  1 for the benzyl chain, respectively), as reported by Wang et al. (2007). However, the toxicities of [C3-2mim] [Tf2N] and [C3-3mim] [Tf2N] were larger than the toxicity of the unsaturated alkyl chain [C4mim] [Tf2N] with EC50 ¼2.8870.43 mmol L  1 while the introduction of the dimethylamino-group in the side chain ([apmim][Tf2N]) contributed to a toxicity higher (EC50 ¼1.1070.21 mmol L  1) compared to the toxicity of [C4mim][Tf2N]. The results obtained for tested ILs toxicities showed that the structure of substituents on the imidazolium cation had a significant effect on their toxicity and support the results previously reported by other authors (Stolte et al., 2007; Ventura et al., 2012). However, according to Merck's UFT database on the biological effects of ILs (http://www.il-eco.uft.uni-bremen.de), where the classification of cytotoxicity was generated by the IPC-81 cell line, most of the tested ILs in our study showed moderate (0.1 mmol L  1 oEC50 o5 mmol L  1) cytotoxicity towards CCO cells. Microscopic observations of CCO cells treated with 1 mmol L  1 of functionalized ILs (Fig. 2b–d) show uptake of the fluorescent

26

M. Cvjetko Bubalo et al. / Ecotoxicology and Environmental Safety 112 (2015) 22–28

Fig. 2. Photomicrographs of CCO cells stained with acridine orange and ethidium bromide. (a) Control cells and cells treated for 72 h with 1 mmol L  1 of: [C5omim][Tf2N], [apmim][Tf2N], [bzmim][Tf2N], [C3-2mim][Tf2N] and [C3-3mim][Tf2N] (b–f). Control cells are shown (a) while apoptotic/necrotic cells were observed in treated samples (b–f). Magnification 400  .

dyes acridine orange and ethidium bromide, which readily cross the damaged plasma membrane, indicating membrane integrity loss and induction of apoptotic and necrotic changes. The enhanced membrane permeability and disruption were structurerelated being consistent with the obtained cytotoxicity results and our previously reported results of effects of 1-alkyl-3-methylimidazolium ILs on CCO cells (Radošević et al., 2013). 3.2. QSTR modeling The LDA analysis on cationic predictor variables yielded Eq. (1), a discriminant function equation consisting of three topological variables of the analyzed cations and characterized by the highly encouraging Wilks' λ value of 0.026 (very close to the ideal value, zero) accompanied with Fischer (126.293) and Chi-square (38.437) parameters (where higher values are better).

DF = − 257.742 + 419.591 × BCUTc − 1l + 242.679 × Petitjean Number +164.839 × Lipoaffinity Index n=14 Wilks‵λ = 0.026, F (df )=126.293(3, 10), p = 0.000, χ 2 = 38.437, 2 DM = 141.448, Rc = 0.987

(1)

The model also shows appreciable values of the squared Mahalanobis distance measure (141.448), Matthew's correlation coefficient (1.0) and the canonical correlation coefficient (0.987) value with a perfect classification performance (accuracy¼1) (Fawcett, 2006). The descriptor pool was subjected to a standardization operation (xstandard =

(xi − xmin) ) (xmax − xmin)

prior to performing LDA

and hence the coefficients in Eq. (1) are standardized ones. From the equation, it can be observed that all three predictor parameters have positive regression coefficients, thus showing a proportional relationship towards the derived discriminant function. The variable ‘Petitjean Number’ defines the shape of the cations, the ‘BCUTc-1l’ index gives a topological measure, while the final

term ‘Lipoaffinity Index’ refers to the lipophilicity contribution of different carbon atom fragments derived from E-state terms. The descriptors have been formally defined in Table S2 in Supplementary material. Considering the compound structures of the present data set, it was observed that the ‘Lipoaffinity Index’ gives higher values in the presence of long alkyl chain substituent on imidazolium nitrogen atoms, e.g. the presence of the decyl substituent on position 1 of an imidazolium ring gives the highest value of 9.633. Hence, the presence of a long alkyl head group can be related to the toxic effect exerted by imidazolium ionic liquids towards CCO cell line inhibition. The value of the shape parameter ‘Petitjean Number’ also increases when substituents are changed from butyl to pentyl and heptyl. The ‘BCUTc-1l’ parameter considers overall topology of the molecule and aids in developing a good correlation with the response. From a contribution plot (a plot of the averaged descriptor values belonging to two groups multiplied by their coefficient in DF equation) of the variables, it can be observed that the predictor variable ‘Petitjean Number’ is the most discriminatory one, followed by ‘BCUTc-1l’ and ‘Lipoaffinity index’ (Fig. 3). Hence, it is observed that the cytotoxicity of ionic liquids towards the CCO cell line can be successfully modeled employing hydrophobicity, molecular shape and other topological features, which chiefly connect the long alkyl chain substituents of cations with toxicity. We additionally attempted to develop a multiple linear regression model using the available quantitative data. Since, only nine compounds with CCO cell line toxicity data are available, we did not divide the data into training test and test sets. Instead we chose to present the following equation as a preliminary QSTR model for the ionic liquids toward the CCO cell line endpoint. We have tested the available data using Shapiro–Wilk statistics (W) (Stephens, 1976) and a normal distribution plot confirming that the response values do not deviate significantly from the normal distribution, which is a prerequisite for regression analysis (Helguera et al., 2008). A stepwise regression analysis on a pool of various two-dimensional topological indices along with QTMS parameters of cations gave a two-descriptor model (Eq. (2))

M. Cvjetko Bubalo et al. / Ecotoxicology and Environmental Safety 112 (2015) 22–28

27

Fig. 3. Contribution plot of descriptors in appearing in Eq. (1).

characterized by excellent values of the squared correlation coefficient (R2 ¼0.961) and leave-one-out cross validation coefficient (Q2 ¼0.926). Note that we have limited the number of descriptors present in the equation to two, in view of the limited number of data points, so as to keep the ratio of number of data points and number of descriptors above 4.

Fig. 4. Scatter plot for Eq. (2).

pEC50(M) = − 15.571 + 15.460 × PetitjeanNumber + 48.244 × ε n=9 R2 = 0.961, Ra2 = 0.948, PRESS = 0.155, s = 0.117, Q2 = 0.926, 2 2 rm (LOO)scaled = 0.898, Δrm(LOO)scaled = 0.04

(2)

Eq. (2) also depicts an encouraging explained variance of 0.948 (Ra2) accompanied with a low value of the summed squared predicted residual (0.155) and standard error of estimate (0.117) metrics. The variable Petitjean Number mainly corresponds to the shape of the imidazolium cations, while the QTMS parameter ε gives a measure of ellipticity of the C ¼N bond in imidazole. The two descriptors used in Eq. (2) have a low inter-correlation (r2 ¼0.222). A plot between the experimental and calculated (Eq. (2)) toxicity values showed uniform distribution of the data points along the diagonal line (Fig. 4). A residual plot for Eq. (2) shows (Fig. 5) that the residuals are randomly distributed around the zero residual line confirming absence of any systemic error. 3.3. Interpretation of the QSTR models Fig. 5. Residual plot for Eq. (2).

In the knowledge that the limited number of data points used for developing the QSTR models may limit the application of the developed models for diverse functional ionic liquids, we attempt here to understand the physicochemical significance of the derived models. The models in general corroborate earlier findings (Stolte et al., 2006, 2007; Cruz-Monteagudo et al., 2013) on the dependence of toxicity of ionic liquids on lipophilicity, shape of the cationic head groups, and length of alkyl substituents, while the observation on the influence of ellipticity of the C¼ N bond in imidazole on the toxicity is novel, which needs to be further explored.

4. Conclusion Imidazolium ILs with functionalized and unsaturated side groups were tested for their biological effects in the fish CCO cell

line. The obtained results showed that the cytotoxic effects of ILs were dose- and structure-dependent. In this respect, a QSTR analysis using discriminant and regression approaches was performed showing preliminary models with encouraging predictive capacity. Toxicity of ILs towards CCO was chiefly related to the shape and hydrophobicity of cations determined by the chain length of their substituent. Furthermore, significant influence of the quantum topological molecular similarity descriptor ellipticity (ε) of the imine bond was also observed. The data on toxic effects of ILs in CCO cells presented in this paper, in addition to the known literature data on the toxicity in other model systems, can be useful information for environmental risk assessment before production and application of ionic liquids at an industrial scale. However, the models developed here are based on a small data set

28

M. Cvjetko Bubalo et al. / Ecotoxicology and Environmental Safety 112 (2015) 22–28

of limited diversity and may not have wide spread applicability. Thus, additional data are needed on the toxicity of ILs towards CCO to develop more robust and predictive models.

Acknowledgments Support from the Ministry of Science, Education and Sports, Republic of Croatia (Grant nos. 0582261-2256058 and 05821842414 is gratefully acknowledged. KR thanks the European Commission's Marie Curie Funding for the IONTOX project (Grant No. 329138). RND acknowledges the financial help received from Council of Scientific and Industrial Research, India (Grant No. 09/ 096(0729)/2012-EMR-I).

Appendix A. Supplementary material Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.ecoenv.2014.10. 029.

References Alvarez-Guerra, M., Irabien, A., 2011. Design of ionic liquids: an ecotoxicity (Vibrio fischeri) discrimination approach. Green Chem. 13, 1507–1516. Cvjetko, M., Vorkapić-Furač, J., Žnidaršič-Plazl, P., 2012. Isoamyl acetate synthesis in imidazolium-based ionic liquids using packed bed enzyme microreactor. Process Biochem. 47, 1344–1350. Cvjetko Bubalo, M., Radošević, K., Radojčić Redovniković, I., Halambek, J., Gaurina Srček, V., 2014a. A brief overview of the potential environmental hazards of ionic liquids. Ecotoxicol. Environ. Saf. 99, 1–12. Cvjetko Bubalo, M., Hanousek, K., Radošević, K., Gaurina Srček, V., Jakovljević, T., Radojčić Redovniković, I., 2014b. Imidiazolium based ionic liquids: effects of different anions and alkyl chains lengths on the barley seedlings. Ecotoxicol. Environ. Saf. 101, 116–123. Cho, C.W., Ranke, J., Arning, J., Thöming, J., Preiss, U., Jungnickel, C., Diedenhofen, M., Krossing, I., Stolte, S., 2013. In silico modelling for predicting the cationic hydrophobicity and cytotoxicity of ionic liquids towards the Leukemia rat cell line, Vibrio fischeri and Scenedesmus vacuolatus based on molecular interaction potentials of ions. SAR QSAR Environ. Res. 24, 863–882. Cho, C.W., Jeon, Y.C., Pham, T.P., Vijayaraghavan, K., Yun, Y.S., 2008. The ecotoxicity of ionic liquids and traditional organic solvents on microalga Selenastrum capricornutum. Ecotoxicol. Environ. Saf. 71, 166–171. Cruz-Monteagudo, M., Ancede-Gallardo, E., Jorge, M., Cordeiro, M.N.D.S., 2013. Chemoinformatics profiling of ionic liquids—automatic and chemically interpretable cytotoxicity profiling, virtual screening, and cytotoxicophore identification. Toxicol. Sci. 136, 548–565. Darlington, R.B., 1990. Regression and Linear Models. McGraw Hill, New York. Das, R.N., Roy, K., 2013. Advances in QSPR/QSTR models of ionic liquids for the design of greener solvents of the future. Mol. Divers 17, 151–196. Das, N.R., Roy, K., 2014. Predictive modeling studies for the ecotoxicity of ionic liquids towards the green algae Scenedesmus vacuolatus. Chemosphere 104, 170–176. Deng, Y., Besse-Hoggan, P., Sancelme, P., Delortet, A.M., Husson, P., Costa Gomes, M. F., 2011. Influence of oxygen functionalities on the environmental impact of imidazolium based ionic liquids. J. Hazard. Mater. 198, 165–174. Egorova, K., Annanikov, V.P., 2014. Toxicity of ionic liquids: Eco(cyto)activity as complicated, but unavoidable parameter for task-specific optimization. ChemSusChem 7, 336–360. Fatemi, M.H., Izadiyan, P., 2011. Cytotoxicity estimation of ionic liquids based on their effective structural features. Chemosphere 84, 553–563. Fawcett, T., 2006. An introduction to ROC analysis. Pattern Recognit. Lett. 27, 861–874. Fisher, R.A., 1936. The use of multiple measurements in taxonomic problems. Ann. Eugen. 7, 179–188. Frade, R.F.M., Alfonso, C.A.M., 2010. Impact of ionic liquids in environmental and humans: an overwiev. Hum. Exp. Toxicol. 29, 1038–1054. Garcia-Lorenzo, A., Tojo, E., Tojo, J., Teijeira, M., Rodriguez-Berrocal, F.J., Gonzalez, M.P., Martinez-Zorzano, V.S., 2008. Cytotoxicity of selected imidazolium-derived ionic liquids in the human Caco-2 cell line. Sub-structural toxicological interpretation through a QSAR study. Green Chem. 10, 508–516. Helguera, A.M., Cordeiroa, M.N.D.S., Pérez, M.A.C., Combes, R.D., González, M.P., 2008. Quantitative structure carcinogenicity relationship for detecting structural alerts in nitroso-compounds. Toxicol. Appl. Pharmacol. 231, 197–207. Latała, A., Nędzi, M., Stepnowski, P., 2010. Toxicity of imidazolium ionic liquids towards algae. Influence of salinity variations. Green Chem. 12, 60–64.

Luis, P., Garea, A., Irabien, A., 2010. Quantitative structure–activity relatioships (QSARs) to estimate ionic liquids ecotoxicity EC50 (Vibrio fischeri). J. Mol. Liq. 152, 28–33. Luis, P., Ortiz, I., Aldaco, R., Irabien, A., 2007. A novel group contribution method in the development of a QSAR for predicting the toxicity (Vibrio fischeri EC50) of ionic liquids. Ecotoxicol. Environ. Saf. 67, 423–429. Matzke, M., Stolte, S., Thiele, K., Juffernholz, T., Arning, J., Ranke, J., Welz-Biermann, U., Jastorff, B., 2007. The influence of anion species on the toxicity of 1-alkyl-3methylimidazolium ionic liquids observed in an (eco)toxicological test battery. Green Chem. 9, 1198–1207. Pham, T.P., Cho, C.W., Yun, Y.-S., 2010. Environmental fate and toxicity of ionic liquids: a review. Water Res. 44, 352–372. Popelier, P.L.A., 1999. Quantum molecular similarity. 1. BCP space. J. Phys. Chem. A 103, 2883–2890. Radošević, K., Tonković, T., Slivac, I., Kniewald, Z., Gaurina Srček, V., 2011. Comparison of cytotoxicity induced by 17α-ethynylestradiol and diethylstilbestrol in fish CCO and mammalian CHO-K1 cell lines. Bull. Environ. Contam. Toxicol. 86, 252–257. Radošević, K., Cvjetko, M., Kopjar, N., Novak, R., Dumić, J., Gaurina Srček, V., 2013. In vitro cytotoxicity assessment of imidazolium ionic liquids: Biological effects in fish Channel Catfish Ovary (CCO) cell line. Ecotoxicol. Environ. Saf. 92, 112–118. Ranke, J., Molter, K., Stock, F., Bottin-Weber, U., Poczobutt, J., Hoffmann, J., Ondruschka, B., Filser, J., Jastorff, B., 2004. Biological effects of imidazolium ionic liquids with varying chain lengths in acute Vibrio fischeri and WST-1 cell viability assays. Ecotoxicol. Environ. Saf. 58, 396–404. Ranke, J., Müller, A., Bottin-Weber, U., Stock, F., Stolte, S., Arning, J., Störman, R., Jastroff, B., 2007. Lipophilicity parameters for ionic liquid cations and their correlation to in vitro cytotoxicity. Ecotoxicol. Environ. Saf. 67, 430–438. Roy, K., Das, R.N., Popelier, P.L.A., 2014. Quantitative structure–activity relationship for toxicity of ionic liquids to Daphnia magna: aromaticity vs lipophilicity. Chemosphere 112, 120–127. Samori, C., Malferrari, D., Valbonesi, P., Montecavalli, A., Moretti, F., Galletti, P., Sartor, G., Tagliavini, E., Fabbri, E., Pasteris, A., 2010. Introduction of oxygenated side chain into imidazolium ionic liquids: evaluation of the effects at different biological organization levels. Ecotoxicol. Environ. Saf. 73, 1456–1464. Samori, C., Pasteris, A., Galleti, P., Tagliavini, E., 2007. Acute toxicity of oxygenated and nonoxygenated imidazolium-based ionic liquids to Daphnia magna and Vibrio fischeri. Environ. Toxicol. Chem. 26, 2379–2382. Schirmer, K., 2006. Proposal to improve vertebrate cell cultures to establish them as substitutes for the regulatory testing of chemicals and effluents using fish. Toxicology 224, 163–183. Stephens, M.A., 1976. Asymptotic results for goodness-of-fit statistics with unknown parameters. Ann. Stat. 4, 357–369. Stepnowski, P., Skladanowski, A.S., Ludwiczak, A., Laczyńska, E., 2004. Evaluating the cytotoxicity of ionic liquids using human cell line HeLa. Hum. Exp. Toxicol. 23, 513–517. Stolte, S., Arning, J., Bottin-Weber, U., Matzke, M., Stock, F., Thiele, K., Uerdingen, M., Welz-Biermann, U., Jastorff, B., Ranke, J., 2006. Anion effects on the cytotoxicity of ionic liquids. Green Chem. 8, 621–629. Stolte, S., Arning, J., Bottin-Weber, U., Müller, A., Pitner, W.-R., Welz-Biermann, U., Jastorff, B., Ranke, J., 2007. Effects of different head groups and functionalised side chains on the cytotoxicity of ionic liquids. Green Chem. 9, 760–767. Swatloski, R.P., Holbrey, J.D., Memon, S.B., Caldwell, G.A., Caldwell, K.A., Rogers, R.D., 2004. Using Caenorhabditis elegans to probe toxicity of 1-alkyl-3-methylimizolium chloride based ionic liquids. Chem. Commun., 668–669. Tan, F., Wang, M., Wang, W., Lu, Y., 2008. Comparative evaluation of the cytotoxicity sensitivity of six fish cell lines to four heavy metals in vitro. Toxicol. In Vitro 22, 164–170. Ternjej, I., Gaurina Srček, V., Kopjar, N., Mihaljević, Z., 2013. Cytotoxic and genotoxic effects of water and sediment samples from gypsum mining area in channel catfish ovary (CCO) cells. Ecotoxicol. Environ. Saf. 98, 118–127. Torrecilla, J.S., Palomar, J., Lemus, J., Rodrígueza, F., 2010. A quantum-chemicalbased guide to analyze/quantify the cytotoxicity of ionic liquids. Green Chem. 12, 123–134. Ventura, S.P., Marques, C.S., Rosatella, A.A., Afonso, C.A., Gonçalves, F., Coutinho, J.A., 2012. Toxicity assessment of various ionic liquid families towards Vibrio fischeri marine bacteria. Ecotoxicol. Environ. Saf. 76, 62–68. Wang, X.F., Ohlin, A.C., Lu, Q., Fei, Z., Hu, J., Dyson, P.J., 2007. Cytotoxicity of ionic liquids and precursor compounds towards human cell line HeLa. Green Chem. 9, 1191–1197. Wang, L.S., Wang, L., Wang, L., Wang, G., Li, Z.H., Wang, J.J., 2009. Effect of 1-butyl-3methylimidazolium tetrafluoroborate on the wheat (Triticum aestivum L.) seedlings. Environ. Toxicol. 24, 296–303. Yap, C.W., 2011. PaDEL-Descriptor: an open source software to calculate moleculardescriptors and fingerprints. J. Comput. Chem. 32, 1466–1474. Yu, M., Wang, S.H., Luo, Y.R., Han, Y.W., Li, X.Y., Zhang, B.J., Wang, J.J., 2009. Effects of the 1-alkyl-3-methylimidazolium bromide ionic liquids on the antioxidant defense system of Daphnia magna. Ecotoxicol. Environ. Saf. 72, 1798–1804. Zhang, B., Li, X., Chen, D., Wang, J., 2013. Effects of 1-octyl-3-methylimidazolium bromide on the antioxidant system of Lemna minor. Protoplasma 250, 103–110. Zhu, S., Yu, P., Lei, M., Tong, Y., Zheng, L., Zhang, R., Ji, J., Chen, Q., Wu, Y., 2013. Investigation of the toxicity of the ionic liquid 1-butyl-3-methylimidazolium chloride to Saccharomyces cerevisiae AY93161 for lignocellulosic ethanol production. Pol. J. Chem. Technol. 15, 94–98.