Toxicological study of some ionic liquids

2.1 Chemicals

B, BM₂, BM₃ and BM₄ with purities of 99% and P with a purity of 98% were provided by IoliTec (Heilbronn, Germany), dried for 24 h under vacuum (ca. 0.05 kPa), and stored in a desiccator.

2.2 Inhibition of bioluminescence in V. fischeri

Lyophilized V. fischeri (strain NRRL-B-11177) were purchased from Macherey-Nagel (Düren, Germany). The experimental bioluminescence inhibition tests were performed in triplicate with re-activated bacteria according to UNE-EN ISO 2009 [33] norm, with positive (phenol, 40 mg l⁻¹) and negative controls [34]. A stock solution of each of the five ILs at concentrations of 2% NaCl were prepared and the pH adjusted to 7–7.5 with 0.1 mol l⁻¹ HCl or 0.1 mol l⁻¹ NaOH.

Aliquots of 0.5 ml of the reactivated bacterial suspension was transferred to each cuvette and kept for 10 min at 15°C. The first measurements were then taken to obtain the initial luminescence with a luminometer (Biofix^® Lumi-10, Macherey-Nagel, Düren, Germany), equipped with an ultra-fast single photon counter detector covering the 380–660 nm spectral range, using the acute mode (Biotox B). Subsequently, 0.5 ml of each of the tested dilutions was added to the cuvette. After 30 min, the inhibition of luminescence was measured again and the percentages of inhibition were calculated.

2.3 Daphnia magna acute immobilization test

Daphnia magna vials (ref. DM090812, Vidrafoc, Barcelona, Spain) were stored at 4°C. OCDE 202 protocols were employed [35], [36] according to the test conditions. First, the media for the ephippia were prepared in accordance with the supplier specifications. Next, the eggs were incubated for 72 h at 20°C–22°C with 6000 lux in an incubator (model CH-0120D-AC/DC, TOXKIT, ECOTEST, Valencia, Spain). The obtained neonates were fed 2 h prior to the bioassay by adding a Spirulina vial.

Solutions of the studied ILs were prepared in stock solution with inorganic salts provided in vials by ECOTEST (Valencia, Spain). Positive controls were tested with K₂Cr₂O₇ (EC₅₀ values range between 0.6 and 2.1 mg l⁻¹). Negative controls were accepted if <10% of the individuals were immobilized after the assay. The pH levels of the solutions were adjusted to fall between 7 and 7.5 using 0.1 mol l⁻¹ NaOH or 0.1 mol l⁻¹ HCl solutions. A total of 20 neonate organisms (aged <24 h) divided into four groups of five organisms were exposed to each concentration after 2 h of feeding. Daphnids were incubated in complete darkness for 24 h at 20°C–22°C.

The immobilization of the daphnids was measured according to the operating protocol. The daphnids were considered immobilized when they were unable to swim for 15 s after gentle stirring.

2.4 Statistics and graphical representation

The experimental results were fitted using the least squares method to EC₅₀ values and standard deviations (SD):

where %I denotes % bioluminescence inhibition, c is the concentration (in mg l⁻¹) and a and b are adjustable parameters.

One-way ANOVA Tukey’s multiple comparison tests were performed to compare between pairs of ILs in each biondicator. As for ethical approval, the biotests were carried out according to internationally accepted guidelines.

3.1 Toxicity of ILs in V. fischeri

All of the studied ILs slightly affected the bioluminescence of the bacteria, and differences in the EC₅₀ values between most of the ILs could be detected. The action mechanism of the luminescence emission is related to the modification of protein and lipid biosynthetic pathways, which are relevant in cellular respiration [37].

The most toxic chemical studied was BM₂ followed by BM₃, P, B and BM₄. All the EC₅₀ values were significantly different except BM₂ and BM₃, BM₂ and P, BM₃ and P, BM₄, and B (ANOVA, Tukey’s multiple comparison test). The toxicity increased as the length of the substituent at nitrogen decreased. The presence of an extra methyl substituent in position 4 did not seem to substantially affect the toxicity; however, an extra methyl substituent in positions 2 or 3 increased the toxicity. Clearly, the closer the methyl substituent is to the nitrogen atom, the more toxic the IL.

According the Passino and Smith classification [38], which evaluates the acute toxicity of chemical compounds in the aquatic environment, BM₂, BM₃ and P could be considered as practically harmless for V. fischeri as their EC₅₀ values with this biondicator are in the range of 100–1000 mg l⁻¹. Furthermore, B and BM₄ are clearly harmless because their EC₅₀ values are higher than 1000 mg l⁻¹.

Table 2 provides V. fischeri toxicity data of some conventional solvents for comparison [39] and Figure 3 shows this information in V. fischeri. There are solvents that are more harmful for this biondicator than the studied ILs, i.e. o-xylene, toluene or phenol. These chemicals can be classified as slightly or moderately toxic, according the Passino and Smith classification [38]. However, there are also VOCs that are much less harmful, such as methanol, acetone or dichloromethane, which are harmless for V. fischeri, such as BM₄ or B. Regarding other ILs [21], [22] against the studied ones, just 1-propyl-3-methylimidazolium tetrafluoroborate is less toxic for V. fischeri than the selected ILs. The rest of the exposed ILs are more harmful for this biondicator. According to the Passino and Smith classification [38], their aquatic toxicity includes several ranges of toxicity, i.e. 1-decyl-3-methylimidazolium tetrafluoroborate and 1-octyl-3-methylimidazolium chloride could be considered as highly toxic for D. magna. However, 1-hexyl-3-methylimidazolium tetrafluoroborate or 1-pentyl-3-ethylimidazolium tetrafluoroborate can be considered practically harmless.

Figure 3:

Log EC₅₀ graphical comparison of toxicity in Vibrio fischeri.

The codification is shown in Table 2. Pointed bars correspond with imidazolium ILs, squared bars correspond with traditional solvents and grey bars correspond with the studied ILs. The dashed line shows the limit between slightly toxic and practically harmless compounds according to Passino and Smith classification.

Furthermore, according to the literature, pyridinium ILs in bioluminescence assay with V. fischeri are less toxic than the corresponding imidazolium, with the same anion, in QSPR studies [43]. This statement can be confirmed with the EC₅₀ values of the Table 2 [21], [22], where almost all of the imidazolium ILs have lower EC₅₀ values than the studied pyridinium compounds. For example, comparing the toxicity of the cations 1-butyl-3-methylimidazolium and 1-butyl-3-methylpyridinium with the same anion tetrafluoroborate, the imidazolium is more toxic than the pyridinium IL. It is also interesting to note that, although a direct correlation between the length of the alkyl chain of methylimidazolium-based ILs and toxicity has been previously shown [44], our results do not follow this trend. Thus, extrapolations of behavior in this sense are a matter of concern.

3.2 Toxicity of ILs in Daphnia magna

Each of the studied ILs strongly affected the mobility of D. magna. Although the mechanisms of action are still unknown, several authors have noted that ILs have the potential to cause enzyme inhibition, disruption of membrane permeability, structural damage and oxidative stress [9], [45].

The most toxic IL studied was BM₄, followed by BM₂/BM₃, B and P. Only the BM₂ and BM₃ EC₅₀ values are not significantly different from each other (ANOVA, Tukey’s multiple comparison test). In general, the presence of an extra methyl substituent increased the observed toxicity. As it pertains to the length of the substituent of at nitrogen, B is moderately more toxic than P. All of these ILs are included in the category of “slightly toxic” for D. magna, according the Passino and Smith classification [38].

The D. magna toxicities of conventional solvents are shown in Table 2 [21], [40], [41], [42] and compared to one another in Figure 4. The EC₅₀ values of the studied ILs are lower to chloroform and dichloromethane, although higher than toluene. In comparison with other ILs, as shown in Table 2 [21], [29], pyridinium compounds are, at least, as toxic for the aquatic environment as imidazolium ILs, i.e. 1-butyl-3-methylimidazolium tetrafluoroborate, 1-butyl-3-methylimidazolium chloride and 1-butyl-3-methylimidazolium bromide. These are in the same range of aquatic toxicity than the studied pyridinium chemicals according the Passino and Smith classification [38]. Others, like 1-hexyl-3-methylimidazolium chloride and 1-octyl-3-methylimidazolium chloride, are considered as moderately toxic or highly toxic, respectively. These results confirmed the Couling QSAR modeling [43] as those equations predicted that the toxicity increases with the number of aromatic atoms in the cation ring.

Figure 4:

Log EC₅₀ graphical comparison of toxicity in Daphnia magna.

The codification is shown in Table 2. Pointed bars correspond with imidazolium ILs, squared bars correspond with traditional solvents and grey bars correspond with the studied ILs. The dashed line shows the limit between slightly toxic and practically harmless compounds according to Passino and Smith classification.

In general, our results agree with the reported toxicity behavior of ILs in D. magna. There is a well-established link between toxicity and alkyl-chain length [8]. Furthermore, our results also agree with quantitative structure–toxicity relationship (QSTR) modelling, which predicts an increase in toxicity due to the methylation of the aromatic carbons [46].

3.3 Toxicity of ILs according other biomodels

There are not much available data about these ILs with biomodels corresponding to other trophic levels. Ranke et al. [47] analyzed BM₂, BM₃, BM₄ and B in IPC-81 rat cell model during 48 h, obtaining their EC₅₀ values. According to these results, the distance of the methyl group to the nitrogen in the pyridinium cation does not have a direct relationship with the toxicity, with BM₄ being the most cytotoxic IL (221 mg l⁻¹), followed by B (322 mg l⁻¹), BM₂ (421 mg l⁻¹) and BM₃ (473 mg l⁻¹). On the other side, there are also previous data for the acetylcholinesterase inhibition assay for BM₂, BM₃ and B [8]. In this case, the progression of the toxicity is in full agreement with our results in V. fischeri. Stock et al. [48] also performed the acetylcholinesterase inhibition assay with several ILs, including two pyridinium-based ILs, 1-butyl-3-methylpyridinium tetrafluoroborate (BM₃) and hexafluorophosphate. The EC₅₀ values found for this biomarker are very similar to each other (8 mg l⁻¹ and 8.3 mg l⁻¹, respectively), and are also lower than in our studied biomodels. Thus, there is no clear relationship among biomodels with this type of ILs. These observations indicate the high disparity of the results in this type of solvents and, therefore, remarking the relevance to evaluate the toxicity in several biomodels to ensure the harmlessness of future industrial solvents [49].

3.4 QSAR studies

The development of QSAR helps in understanding how the toxicities, properties, or activities of chemicals vary with structural composition. We have applied the QSAR concept to provide a mathematical model derived from the available physicochemical properties [50], [51], with respect to two specific endpoints in V. fischeri and D. magna. As a first approximation, the relationships between several physicochemical properties (viscosity, surface tension and heat capacity, at 298.15 K) and the acute effective concentrations for the V. fischeri bioindicator and D. magna were investigated. Then, a multiregression expression was derived. For this study, we have included both the toxicological information obtained in this work and previous bibliographic data of other ILs in order to increase the number of input data and make the model reliable. To select the properties that give higher signals in the multiregression analysis, we performed an initial screening using different combinations of the physicochemical properties. Finally, the selected properties were surface tension (mN·m⁻¹) and viscosity (mPa·s). All of the data required for the calculations are presented in Table 1 for the studied compounds [15], [16], [17], [18], [19], and Tables 3 (for V. fischeri) and 4 (for D. magna) for the previous referenced ILs [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32].

The simple relationships between the selected properties and acute effective concentrations for the V. fischeri bioindicator and D. magna, as well as the R coefficient, are shown in the following individual structure-projected toxicity-parameter expressions:

The global multiregression expressions, along with the statistical indicator R² are as follows:

In Figure 5, plots of the experimental values versus those calculated with the regression models are shown. It may be noted that small correlations for direct relationships between each physicochemical property are found, except for surface tension that seems to carry the strongest signal. The property is closely related to intermolecular interactions between the ions. With regards the global multiregression analysis, despite the small sample size and the very different molecular structures tested, the overall correlation using basic QSAR displays acceptable values for the coefficients of determination, with better results found for the bioindicator V. fisheri. The IL showing the highest deviations in the correlation study are P in the case of V. fischeri and 1-butyl-3-methylimidazolium hexafluorophosphate in the case of D. magna.

Figure 5:

Plots of predicted vs. experimental values of log EC₅₀ as calculated through multiregression analysis in Vibrio fischeri (A) and Daphnia magna (B).