3.1. The Working System
Usually, the classical assessment of new industrial chemicals is
inflexible, being directed by regulations and standardized procedures, while biological test systems are very expensive and the costs increase with the number of new chemicals released. So, a
flexible test strategy is needed in correlating the chemical compounds with the systems they act upon. Although qualitative (eco) toxicological algorithms for selection of test systems have been proposed in terms of identification of individual effects of different head-groups (with identical side chains and anions) and different anions (with identical cations) [
7], a quantitative
theoretical prediction algorithm for tested compounds in biological systems is still need be proposed due the lack of readily accessible ecotoxicological data [
8].
The principles of green chemistry say that one has to consider the whole process (life cycle analysis) rather than individual components of reaction (single issue sustainability), based on the fact that the acute toxicity measurements do not provide a complete characterization of the full impact of a substance release into environment, but are only part of the environmental impact assessment [
6,
8].
In this context, ionic liquids with cations like pyridinium, imidazolium and pyrrolidinium have already been nominated in the United States National Toxicology Program (NTP) for toxicological testing based upon their potential of new solvents but also due they ability to enter in aquatic system [
22]; if an accidentally discharge of ionic liquids into water occur, many of them being water soluble, they may be an environmental risk to aquatic plants and animals [
23].
For this reason the current application will develop the S-SAR complete algorithm up to the mechanistic prediction of the correlated structural causes and ecotoxicological effects for the ionic liquids of
Figure 2, containing ammonium, pyridinium, phosphonium, choline, and imidazolium cations, on aquatic bacteria
Vibrio fischeri[
24]. Whereas the
Vibrio fisheri species was previously found to be one of the most resistant species to ordinary phenol compounds toxicity [
21], the present ionic liquids show quite a wide structural variety to furnish useful information of their environmental action. The cationic and anionic structural properties, the lipophylicity, the electronic polarizability and the total energy, where computed with the HyperChem computational environment [
25] and displayed in
Table 2 together with the reported measured activity of their containing ionic liquids, respectively.
The data in
Table 2 are suitable arranged so that the S-SAR analysis is performed successively at the cationic and anionic level and then at the ionic liquid levels based on
equations (4)–
(10) and
(15)–
(17) in a Hansch type expansion:
when accounting for different combinations between the lipophylicity (hydrophobic character), polarizability (electronic character) and total energy (steric character) factors, respectively. The spectral hierarchy of the predicted activities with respect to different concerned endpoints would lead to the mechanistic scheme according with the cationic and anionic sides as well as the overall ionic liquids influences the environmental (Vibrio fischeri) toxicity.
3.2. Results and Discussion
As earlier mentioned, the first step in our analysis consists in deriving the cationic and anionic QSARs that link the structural lipophilic-electronic-steric parameters of the ionic liquids of
Figure 2 and
Table 2 with the observed activities of the whole containing ionic liquids upon the
Vibrio fischeri species. The results are presented in
Tables 3 and
4 for the cationic and anionic subsystems for all main combinations, i.e. generating the uni-modes
Ia-to-
Ic when only one structural parameter is correlated, the two-modes
IIa-to-
IIc when two combined structural factors are taken into account and for the three-mode correlation
III with all structural factors involved, respectively. For each such mode of action, the associated endpoint norm, the statistic and algebraic correlation factors were reported, computed with the
equations (6),
(7), and
(8), respectively.
As a general observation, in all cases there was recorded a systematic increase of the correlation factor when computed in spectral space, i.e. using the algebraic definition, as compared with the standard statistical values. We would like to take this opportunity to advocate the use of the algebraic definition instead of the statistical one since the first one has the physical meaning of the “length of action” respecting the old-fashioned dispersion analysis. Nevertheless, dispersion being a consequence of the appropriateness of the fit, the vectorial norm or the “length of action” accounts merely for the degree with which a certain model approaches the observed, or measured or manifested (chemical-biological) interaction. In this respect, it is also worth noting that both cationic and anionic predicted activities poorly resemble the experimentally expected activities with a smooth increase on the anionic influence, for all computed models except two cases based on the lipophylicity correlation (Ia and IIa).
Apparently, this behaviour disagrees with the previously reported studies in which the anionic effect was only marginal in cytotoxicity [
1,
4,
26], but in some special cases of certain ionic liquids tests [
7].
This situation was based on the observation that, for instance, since imidazolium ring in cations is a delocalized aromatic system with high electron acceptor potential, the nitrogen atoms are not capable to form any hydrogen bonds and the result is a very rigid and sterically inflexible system, while the elongation of R2 residue in side chain leads to a continuous increase of flexibility implying more conformational freedom [
19]. The reason for this actions relies on the fragmental hydrophobicity of each carbon connected to a quaternary amine which, combining the geometric bond factor (that applies to the neutral solute) with a negative electronic bond factor, decreases its magnitude with the square of the distance from the central nitrogen atom [
26,
27].
Moreover, the systematic variation of R1, R2, etc., at identical head groups and anions, in all published data, from the molecular to individual organism level, leads to the conclusion that the shorter the chain lengths of side chains the lower the cytotoxicity (higher EC50 values) [
7,
9]. The electronic portion of the bond factor extends along a chain of no more than 5 alkane carbons causes a decrease in overall hydrophobicity so the chains longer than 5–6 atom carbons will have a greater permeability through the cell membrane (see for example [DMIM][BF4]) [
26]. This, probably because the chemical transformation of the side chains of ionic liquids may reduce toxicity as far as the metabolites are less toxic compared to their parent chemicals [
7].
Next, aiming to combine the two somewhat separate effects of cations and anions in the ionic liquid activities, the spectral-SAR results are given in the
Table 5, showing the working ionic liquid QSAR equation as well the actual vectorial norm, statistic and algebraic correlation factors for each mode of action envisaged so far.
The data in
Table 5 clearly demonstrate that the ionic liquid S-SAR models always predict higher norms in endpoint activities thus providing the considerable increase in the algebraic correlation factors with respect to the cationic and anionic subsystem effects. It is very interesting to see that the statistical values not only furnish lower values than the algebraic outputs for the ionic liquids, but often lie even below the corresponding statistical values of the cationic and anionic subsystems. This situation gives us a chance to establish the limits of using the dispersion based correlation factor definition since it does not properly reproduced the addition effect of the two interacting subsystems as the anionic-cationic interaction in the ionic liquid structures. On the other hand, the mixture effect of the cationic and anionic vectorial actions in (eco)toxicological studies is well established as far as the single substances of a mixture acts in similar and close quantified manner in all considered modes.
The fact that this is the present case can be firstly visualized by the close inspection of
Tables 3 and
4 where the spectral norms of the predicted activities feature close values in relatively narrow domain of actions. Moreover, the reinforcement of this idea comes from the data of
Table 6 in which the angle of interactions between the cationic and anionic subsystems are computed, as given by the formula (
15), for all considered endpoints with almost constant values around the 0.600 cosines of the angle, while the only higher fluctuation deviation appears in the
Ia and
IIa cases – the same previously evidenced for the cationic dominancy over the anion effects.
Overall, once the electronic and steric mechanisms have also been considered, the anion could play a central role as technicophore because it exhibits a high potential for change technological properties (solubility, viscosity) or due its peculiarity to partially decompose itself in the ion pair interaction [
26,
28], leading to its dominant effect in the chemical-biological engaged activity.
With these consideration we can safely assume that the Spectral-SAR in general and the algebraic correlation factors in particular are especially suited for modelling the (eco)toxicological activities of ionic liquids from its anionic and cationic component effects.
The final part of discussion is devoted for picturing a mechanistically mode of action for cationic, anionic and of their summed effects in containing ionic liquids on the considered
Vibrio fischeri species. That is, the minimum path procedure among all possible ways connecting endpoints from each category of models (i.e. with one, two or three factors dependency) is to be considered. The path lengths were computed employing the
equation (10) to all cationic, anionic and ionic liquids data of
Tables 3,
4, and
5, and the results are displayed in the
Table 7, respectively.
However, in order to identify the shortest paths in each category of endpoint connections, the following rules are applied: the first choice is the overall minimum path in a certain column of
Table 7 (i.e. a system with a specific way of correlation factor); if the overall minimum belongs to many equivalent paths (as is the case of cationic with algebraic factor column in
Table 7, for instance) the minimum path will be considered the one that links the starting endpoint with the closest endpoint in the sense of norms (as is the norm of
IIa mode the closest to the norm of
Ia mode in cationic-algebraic column of
Table 7, for example); the overall minimum path will set the dominant hierarchy of the mechanistically mode of action towards the experimentally observed activity and will be called
the alpha path (
α); once the alpha path has been set the next minimum path will be looked for such that the starting endpoint should be different from the one involved in the alpha path (that is, if in the alpha path the starting end point was
Ia, the next path to be identified will begin either from the
Ib or the
Ic mode); the next minimum paths are chosen on the same rules as before and will be called as beta and gamma paths,
β and
γ, respectively. At the end of this procedure each mode of action is “touched” only once, except for the final endpoint, here
III (as a computational substitute of EXP), so that all the methodology being regarded as searching minimum path throughout variation of paths with the fixed final endpoint. Now, the alpha, beta and gamma path can be easily identified in
Table 7 and there are accordingly marked.
With the aim of giving the complete results of the analysis in a single shot,
Figure 3 is a “spectral” representation of the data, with the norms for each cationic, anionic, and resulted cationic liquid being drawn, linked by the major hierarchical paths against the considered mode of actions toward the observed “length” (norm) of the toxicity activity of the
Vibrio fischeri species, for both statistical and algebraic correlation pictures. Figure 7 clearly underlines the fact that while anionic and cationic activity tendencies are somewhat complementary, they do not cancel each other in the ionic liquid that contains them but add up to attain the overall observed toxicity, in the spectral norm – correlation factor space. Besides this, some other useful information can be extracted from Figure 7 concerning the major path of structural causes in manifested toxicological action as revealed bellow.
While in cationic case the statistical and algebraic path does not coincide at all (e.g. what the alpha Ia-IIc-III path in statistic differs than the alpha Ia-IIa-III path in algebraic views), in the anionic case they are identically predicted (excepting the fact that in the algebraic case the shortened paths are registered), together providing a mixed behavior for the resulted ionic liquid (only the beta path Ic-IIc-III is overlapping between statistic and algebraic views). While in cationic and anionic subsystems the path hierarchies are reversed, as α→β→γ and γ→β→α, in the resulting ionic liquid again the mixture effect is recorded since the succession α→γ→β, against the successions of starting endpoints Ia→Ib→Ic, respectively. It is worth pointing out here that, indeed, the cationic alpha path is started on the lipophylicity causes (Ia), the same as for the containing ionic liquid, a different situation arising for the alpha anionic path that is beginning with the steric influence (Ic). This way the previously noted dominance of cationic influence when correlated with lipophylicity as well the recorded anionic influence related with steric effects are in this picture theoretically confirmed.
Other interesting features about ecotoxicological paths rely on the fact that, while for cationic and anionic subsystems the algebraic paths are systematically lower that the corresponding statistical ones, in the ionic liquid case the situation is reversed. The interpretation of this fascinating result is that it also confirms that the chemical-biological ionic liquid dispersive (not specific) actions in environment are merely through its subsystem components than from itself as a whole. This result further motivates the design of ionic liquids by tailoring the properties of its containing anionic and cationic components for prevent their hazardous toxicity.
Finally, let us also note that the ecotoxicological paths in cationic and anionic subsystems are summed up in the paths of the corresponding ionic liquids in a not trivial ways:
From these S-SAR equations some conceptual, however computationally based, ecotoxicological path rules for ionic liquids can be concluded, namely:
- ■
the anionic gamma path effect is marginal over the cationic alpha path –
equation (19);
- ■
the cationic and anionic beta paths decay into the gamma ionic liquid path when met together so that recording a sort of reciprocal cancellation of their effects –
equation (20);
- ■
the anionic alpha path effect is reinforcing over the cationic gamma path averaging both at the beta path level of the resulted ionic liquid –
equation (21).
This way, the S-SAR model presented seems to provide a unitary picture of the anionic-cationic interaction in ionic liquids as conciliating the anionic and cationic effects observed so far. However, further studies on different species with diverse computation schemes and parameters are required in order to conceptually asses a definitely [a definitive] theory of ionic liquid inter- and intra- mode of action.