Lipophilicity parameters for ionic liquid cations and their correlation to in vitro cytotoxicity

Abstract

Regarding the great structural variability of the currently expanding group of ionic liquids, it is highly desirable to understand the basic factors affecting their toxicity in different biological systems. The present study of a set of 74 ionic liquids with imidazolium, pyrrolidinium, pyridinium, quinolinium, quaternary phosphonium and quaternary ammonium cations and the comparatively small anions ${\mathrm{Cl}}^{-}$, ${\mathrm{Br}}^{-}$, ${\mathrm{BF}}_4^{-}$, or ${\mathrm{PF}}_6^{-}$ demonstrates the influence of the cation lipophilicity on the cytotoxicity in IPC-81 leukemia cells from rats. The scope of this correlation is limited to ionic liquids with these or similarly small anions that are sufficiently nonreactive under physiological and chromatographic conditions and whose cation lipophilicity does not exceed a certain threshold.

Introduction

Because of the importance to gain knowledge about possible environmental and human health impacts of newly developed industrial chemicals already during their design phase (Anastas and Farris, 1994), the increasingly popular group of chemicals termed ionic liquids (IL) is of present interest with respect to their potential to cause environmental and human health risks (Jastorff et al., 2003, Scammels et al., 2005).

In attempts to generate toxicity profiles of chemicals more efficiently and to avoid the use of laboratory animals, integrated test strategies have been developed, based amongst others on in vitro testing and physicochemical parameters (Blaauboer, 2002). Our study is the result of such an integrated approach.

Concerning in vitro cytotoxicity, the term basal cytotoxicity has been coined (Ekwall, 1983, Ekwall, 1995), addressing toxicity of chemicals to basal functions of all cells of an organism, as opposed to organ specific toxicity and extracellular toxicity. Basal cytotoxicity in mammalian cells has been reported to be similar in and therefore relevant to such different organisms as plants (Kristen, 1997) and fish (Castano and Gómez-Lechón, 2005) and under certain conditions, correlations to in vivo data have been found to be of predictive value (Garle et al., 1994).

We suggest that cytotoxicity tests in largely dedifferentiated cancer cell lines such as the one used in this study provide a convenient screening method for obtaining first rough estimates for the toxic potential of relatively large sets of substances.

In a previous paper (Ranke et al., 2004, Ranke et al., 2005), we have found a good correlation of the alkyl chain lengths in 1,3-dialkylimidazolium salts with their cytotoxicity in two different cell cultures and the marine bacteria V. fischeri. The present study sets out to generalize this relationship for different headgroups in ionic liquid cations by correlating cytotoxicity and retention behavior of the ionic components of IL in reversed phase high-performance liquid chromatography (RP-HPLC).

Since the pioneering publications of Meyer (1899) and Overton (1901), the correlation between lipophilicity and toxicity of chemical substances has been described and reviewed many times (e.g. Franks and Lieb, 1994, Antkowiak, 2001). While the term narcosis has been established in the field of anaesthesia, describing unspecific reversible effects of mainly apolar compounds on mammals, in aquatic toxicology the term baseline toxicity has been coined (e.g. Könemann, 1981) for the correlation between lethal effect concentrations and lipophilicity. The common pattern of relating toxicity to lipophilicity suggests that baseline toxicity and narcosis are closely related, which is generally accepted in aquatic toxicology, and further supported by a study on the acute lethal effect of narcotics on rats (DeJongh et al., 1998).

A distinction between the baseline toxicity of apolar narcotics (narcosis type I, general or apolar narcosis) and polar narcotics (narcosis type II, polar narcosis) has been proposed (Saarikoski and Viluksela, 1982, Veith and Broderius, 1990), and there is an ongoing debate about the underlying mechanisms (Franks and Lieb, 1994, Vaes et al., 1998, Roberts and Costello, 2003b).

The quantification of lipophilicity in narcosis related studies is most often based on the partitioning coefficient $\log K_{\mathrm{ow}}$ between water and 1-octanol, which can be directly measured, indirectly determined by reversed phase HPLC, or estimated by various computational methods.

In surfactant science (Roberts and Costello, 2003a), and recently also in the IL literature (Stepnowski and Storoniak, 2005), $\log K_{\mathrm{ow}}$ values are in many cases predicted from their structure, based on the group contribution method of Hansch and Leo (Hansch et al., 1995). This works well within groups of IL sharing the same basic ionic structure and the same counter ion (e.g. benzalkonium chlorides, 1-alkyl-3-methylimidazolium tetrafluoroborates). However, an absolute prediction for different ionic headgroups is not possible, since the necessary fragment constants for the headgroups are not available.

Different methods have been proposed to calculate lipophilicity parameters from a single gradient HPLC run (Snyder and Dolan, 1996, Krass et al., 1997, Valko et al., 1997). The approaches of Krass et al. (1997) and Valko et al. (1997) are designed to predict $\log K_{\mathrm{ow}}$ values. It has been shown however, that under some circumstances, chromatographic parameters are even better descriptors of bioavailability, including toxicity, than octanol/water partitioning coefficients (Hsieh and Dorsey, 1995).

Therefore, a quick chromatographic method for characterizing the lipophilicity of cations and maybe for anions was sought. While there have been extensive efforts to characterize IL as stationary phases by chromatography as reviewed by Poole (2004), the use of gradient HPLC for the characterization of IL as solutes is to our knowledge unprecedented.

The retention behavior in reversed phase chromatography is often described by the linear solvent strength (LSS) model

$$\log k=\log k_w-S\phi \text{,}$$

describing the dependence of the capacity factor $k=(t_R-t_0)/t_0$ on the fraction of organic solvent ($\phi$) in the mobile phase, with $k_w$ ideally describing the capacity factor at 0% organic solvent, i.e. an aqueous phase. $t_R$ is the retention time and $t_0$ is the dead time of the system. The slope factor $S$ shows the influence of the organic modifier on $k$ and is generally also substance dependent. $k$, $S$ and $\phi$ are dimensionless numbers and $\phi$ ranges from 0 to 1 (inclusive).

$k_w$ can be derived from many observations of $\log k$ at different values of $\phi$ by linear regression according to Eq. (1). The good correlation of $\log k_w$ values for reversed phase HPLC with $\log K_{\mathrm{ow}}$ within a group of sufficiently similar neutral organic substances is long-established (e.g. Braumann and Jastorff, 1986, Ritter et al., 1995). For eight ionic liquid cations, $\log k_w$ values have been successfully correlated to theoretically predicted $\log K_{\mathrm{ow}}$ values by Stepnowski and Storoniak (2005).

For an estimation of $k_w$ for a set of substances with widely varying retention behavior, two or more gradient runs can be used (Snyder and Dolan, 1996). If two different gradient runs are used, the two unknown $k_w$ and $S$ values can be calculated from the data using the equations

$$t_R=\frac{t_0}{b}\log (2.3k_{0}b(1-x)+1)+t_0+t_D\text{,}$$

and

$$b=\frac{V_m\mathrm{\Delta }\phi S}{t_{G}F}\text{,}$$

where $t_0$ is the column dead time, $t_D$ is the equipment dwell time, $b$ is the dimensionless gradient slope, $k_0$ is the capacity factor at the beginning of the gradient, $V_m$ is the column dead volume, $t_G$ is the gradient time, $F$ is the flow velocity, $\mathrm{\Delta }\phi$ is the change in the fraction of organic solvent during the gradient, and $x=t_D/(t_0{k}_0)$ is the fractional migration through the column during pre-elution. In the following, the gradient always starts with 100% aqueous phase (${\phi }_0=0$) i.e. $k_0$ is equal to $k_w$. Note that Eq. (2) is only valid if $t_R<t_G+t_0+t_D$ (Snyder and Dolan, 1998).

According to a method initially established by de Galan and co-workers (Schoenmakers et al., 1981) and later refined by Snyder and Dolan (1996), $k_0$ can also be estimated from a single gradient run if a correlation between $k_w$ and $S$ is established according to the equation

$$S=p+q\log k_w$$

for a subset of the substances in question with $S$ and $k_w$ derived either from Eq. (1) or from two or more gradient runs and Eq. (2). Here, the coefficients $p$ and $q$ can be obtained as intercept and slope of a linear regression of $S$ against $\log k_w$ for different substances. This then leads to the possibility to calculate $k_0$ directly from $t_R$ obtained in one gradient run by inserting Eq. (4) into Eq. (2).

While in our former study (Ranke et al., 2004), a homologous series of 1-methyl-3-alkylimidazolium salts was studied, in this paper a diverse set of cations was chosen in order to test the generality of the correlation between their lipophilicity and cytotoxicity. As evident from Table 1, the cations are mainly based on substructures containing quaternary nitrogen (alkyl-substituted imidazolium, quinolinium, pyrrolidinium and pyridinium compounds), alternatively containing quaternary phosphorus (tetraalkyl phosphonium compounds), while the anions are limited to ${\mathrm{Cl}}^{-}$, ${\mathrm{Br}}^{-}$, ${\mathrm{BF}}_4^{-}$, and ${\mathrm{PF}}_6^{-}$. This limitation was introduced in order to tentatively isolate the influence of the cation, since a variation within this set of anions has not resulted in a significant influence on cytotoxicity in our previous study (Ranke et al., 2004).

These attempts to understand the behavior of the components of ionic liquids on a molecular basis are undertaken to serve a more sustainable design of new industrial chemicals, where development is closely accompanied by risk research (Jastorff et al., 2003).