Introduction

Society is facing a variety of challenges in environmental risk assessment (ERA): growing concerns about the effects of multiple stressors (both chemical and non-chemical); risks associated with exposure to complex mixtures; and demands to quantify local site-specific risks. At the same time, risk assessors are seeking to provide a more efficient framework on which to address these emerging problems and questions in a manner that reduces cost and the use of laboratory animals. This workshop assessed the applicability of using the thermodynamic chemical activity concept for organic chemicals in the interpretation of effects data within the context of environmental risk assessment.

Chemical activity and effect concentrations

The concept of thermodynamic chemical activity has been shown to be a useful approach for relating exposure to acute toxicity endpoints (Mackay et al., 2011; Ecetoc, 2013; Mackay et al., 2014; Ferguson, 1939; Mayer and Holmstrup, 2008; Smith et al., 2010) but can also be used to help understand the environmental fate and distribution of chemicals, analogous to the use of fugacity (Mackay 1979; Di Toro et al., 1991; Franco et al., 2011; Trapp et al., 2010; Reichenberg and Mayer, 2006; Mackay and Arnot, 2011).

The most common approach to estimate chemical activity in the aqueous phase is as the fraction of the water solubility (liquid or sub-cooled liquid, if the substance is a solid at room temperature), i.e.

where C_W is the concentration of the chemical in the aqueous phase (e.g., mg/L) and is the water solubility (liquid or sub-cooled liquid). Equation 1 thus results in a dimensionless metric of between 0 and 1, which provides a quantitative measure of the fraction of saturation in the aqueous phase observed in the environment/test system. If Raoult’s law holds, activity will equal the mole fraction. Alternatively, toxicity data (e.g., LC₅₀s, EC₅₀s) can also be expressed in terms of chemical activity by replacing C_W with the selected endpoint concentration, i.e.

Considerable effort has recently been invested towards defining the chemical activity domain of non-polar neutral organic chemicals that act as acute lethal baseline toxicants, where La₅₀ values are >0.01 (Reichenberg and Mayer, 2006; Schmidt and Mayer, 2015; Mackay et al., 2014). Fewer studies, however, have addressed chemicals with excess toxicity, La₅₀ are <0.01 (Reichenberg and Mayer, 2006; Schmidt and Mayer, 2015; Mackay et al., 2014). Consequently, current understanding and application of the chemical activity concept imply that the tool could be readily used to assess the environmental risk of non-polar organic chemicals that act as baseline toxicants. Additionally, by estimating chemical activity using equation 2, based on data obtained from toxicity studies, it might be possible to differentiate between baseline and excess toxicity. Excess toxicity (Te) is commonly defined as the ratio of the effect concentration that represent base line toxicity and the observed effect concentration (Lipnick et al., 1987a, b).

The chemical activity approach for toxicity illustrated in equation 2 assumes external concentrations (i.e. in water, sediment, soil, air) approximate concentrations in organisms and at the target site. For baseline toxicity this has proved beneficial in reducing variability in external effect concentrations as a result of normalising against the sub-cooled liquid water solubility of the chemical, for organic chemicals that are solids at the temperature of the test, or simply the water solubility of liquid organic chemicals. The approach thus provides a method for comparing effect concentrations for non-polar neutral organics exhibiting baseline toxicity across (1) compounds; (2) species; and (3) environmental media (Smith et al., 2010; Reichenberg and Mayer, 2006) Furthermore, given the additive nature associated with baseline toxicity, it is also possible to sum the chemical activities associated with mixtures of non-polar neutral organic chemicals to assess the potential risks associated with mixture exposure (Smith et al., 2013; Schmidt et al., 2013a).

Whereas there are numerous examples that relate chemical activity to acute baseline toxicity, there are limited studies that have attempted to assess the relationship between chemical activity and excess toxicity and chronic effects. Additionally, the approach has also seen limited application to miscible and ionisable organic chemicals. These limitations were identified as important data gaps within the ECETOC task force report (Ecetoc, 2013) and represent an important driver for initiating discussions aimed at addressing approaches for possibly expanding the applicability domain of chemical activity.

1. Modes of action in ecotoxicology and classification schemes

Numerous classes of compounds have specific modes of action (Lipnick et al., 1987a, b; Escher and Hermens, 2002; McCarty and Mackay. 1993). Several reporter gene assays are available to study specific mechanisms and modes of action, including for example genotoxicity, oxidative stress and hormonal effects (Scholz et al., 2013; van der Linden et al., 2014). Examples of experimental research in the area of ecotoxicology include: uncouplers of the oxidative phosphorylation (Escher and Schwarzenbach, 2002), acetylcholine esterase inhibitors (de Bruijn and Hermens, 1993) and alkylating agents (reactive compounds) (Hermens, 1990) Also polar narcosis is sometimes regarded as a mode of action different from “non-polar narcosis” (Roberts and Costello, 2003; McCarty et al., 1992). Systematic studies into mode of action (MOA) in in vivo fish toxicity studies are scarce. The early work of McKim and co-workers from the US EPA based in Duluth, Minnesota, is a good example of detailed and pioneering studies into modes of action (McKim et al., 1987).

Classifying compounds according to their mode of action is not a trivial exercise. An example of an attempt to develop a clear classification system is the approach proposed by the US EPA Duluth laboratory (Russom et al., 1997). In their classification system a number of requirements for the assignment of a MOA to a specific compound are defined, including: (a) results from fish acute toxicity syndrome studies, (b) literature data on mechanistic studies, (c) joint toxicity data, (d) behaviour syndromes, (e) excess toxicity (Te)[1] and (f) similarity in chemical structure or chemical properties.

Most other classification systems adopt simpler approaches, which are typically based on items e) and f) from the list of requirements above. Structural alerts or rules are then applied to assign a MOA to a chemical (Verhaar et al., 1992; Enoch et al., 2008). Recently there has been considerable effort in developing an adverse outcome pathway (AOP) framework, related to developing greater mechanistic insight regarding specific modes of action by studying the chain of events that occur following a molecular initiating event (MIE) with a target up to the whole organism effect level (Ankley et al., 2010; Russom et al., 2014). It is believed that the chemical activity approach could provide a complementary approach towards an improved understanding of an AOP, as it has the potential to link exposure with the MIE in a single metric. However, defining where it is and where it isn’t useful would be very helpful in effectively illustrating the added value of chemical activity. Conversely, the concepts of AOP and MIE could be very useful in classification of chemicals, which could lead to improved ability to study relationships with chemical activity.

2. Modes of action, interactions, target sites

The target site for narcosis is the cell membrane. For specifically acting compounds also, the target can be located in the cell membrane (for example a specific protein). However, the location of the target site can also be in a more aqueous environment such as the cytosol (aqueous phase inside the cell) or blood. Differences in internal distribution may also lead to a shift in the mode of action for chemicals within a certain class, for example a shift from a specific mode of action to narcosis (Freidig et al., 1999). Interaction of compounds with a target may vary from reversible van der Waals interactions, hydrogen bonding or irreversible covalent binding (Escher and Hermens, 2002).

3. Dose metrics, dynamic aspects and modelling

As described above, chemical activity represents a measure of effective exposure, which can also be a useful metric in toxicokinetic studies. It is notable that, for specifically acting chemicals, the dynamic aspects (interaction with a target) are additional factors that can influence the final effect concentration. Consequently the influence of exposure time on effect concentrations is often related to a kinetic parameter, and represents a non-equilibrium scenario. For specifically acting chemicals, the effect of time may also be related to toxicodynamics, for instance, in the case of an irreversible interaction of a compound with its target (Legierse et al., 1999). A simple dose related parameter may thus represent an inappropriate metric for quantifying the response in a dose-response relationship, and exposure time may be needed as an additional parameter. A nice example is a study from Gülden et al. (2010) where “area under the curve” was successfully applied in an analysis of cytotoxic potency of H2O2 in cell cultures. An additional challenge is with respect to assessing relationships between an incipient effect concentration and the concept of chemical activity. It is generally acknowledged that modelling of both the kinetic as well as the dynamic aspects will lead to a better understanding of the effects of compounds with more specific modes of action and also of differences in species sensitivity (Ashauer and Brown, 2008; Jager and Kooijman, 2005; Kretschmann et al., 2012; Nyman et al., 2014). It is anticipated that these modelling approaches may provide insight on appropriate dose parameters in ecotoxicology of compounds with modes of action other than baseline toxicity.

4. Uncertainty in key physicochemical property data

Although calculating chemical activity in the aqueous phase appears to be a relatively straight-forward exercise, it is important to recognise that uncertainty exists in both the toxicity data (e.g., LC₅₀s) and the water solubility data. For chemicals which are solids at the system temperature, an additional consideration is that the sub-cooled liquid water solubility is estimated from the water solubility of the solid using the Fugacity Ratio (F):

A simplified approach to estimate the Fugacity Ratio (F) at 25 °C is as shown below,

where T_M is the melting point of the chemical and T is the system temperature. Implicit to this simplified approach is the applicability of Walden’s Rule, which states that the entropy of melting (ΔS_M) is 56.5 J/K•mol. The equation for estimating the Fugacity Ratio in a more expanded form is as follows.

Consequently, an improved understanding related to the uncertainty surrounding the assumption of ΔS_M, as well as the propagation of error associated with uncertainty and variability in both the water solubility value used and the effect concentrations that have either been measured or are based on nominal concentrations, and the uncertainty associated with the melting point temperature of solids are required (Muller and Klein, 1992, Yalkowsky et al., 1994; Ran et al., 2002; Lian and Yalkowsky, 2014; Jain et al., 2004a; Jain and Yalkowsky, 2006, 2007; Tetko et al., 2014). A greater appreciation of the influence of uncertainty is believed to be useful in helping to align chemical activity values to specific modes of action.

1. Estimating chemical activity in non-aqueous media (i.e. biota)

Because acute toxicity data may be reported in terms of internal concentration (i.e., critical body burden) and non-equilibrium conditions may necessitate the estimation of internal concentrations and chemical activity in biota, the reliability of methods to calculate chemical activity in non-aqueous phases also requires careful consideration. Mackay et al. (2011) suggest that chemical activity in non-aqueous phases can be calculated in an analogous fashion to aqueous phases, i.e.,

where CBR is the Critical Body Residue (i.e., internal LC₅₀) and  is the solubility of the chemical in the organism, which is estimated as:

where K_BW is the biota-water partition coefficient. Adding to the uncertainty discussed above, are thus the challenges associated with assessing the uncertainties aligned with equation 6. Alternatively, the internal lethal (or effect) concentration can also be estimated from external LC₅₀ using toxicokinetic (TK) models. Given the various methods that can be used in obtaining chemical activity, guidance is thus needed if the concept is to be routinely and transparently applied within a risk assessment framework.

Feasibility of applying the chemical activity concept to miscible organic chemicals (MOCs)

Whereas the concept of chemical activity has been widely applied to non-polar neutral organics, a key challenge is assessing how to apply the concept to very hydrophilic chemicals (i.e., organic chemicals that are miscible at the temperature of the system of interest). In the case of MOCs there is no quantifiable limit to the solubility of the chemical in water. ‘Empirical’ water solubilities of miscible chemicals (e.g., methanol, ethanol, acetone), however, may still be reported, for instance, as 10⁶ mg/L in the EPISUITE database.

For neutral organic chemicals, chemical activity can be calculated using the following alternative expression:

where χ_i is the concentration of the chemical in the aqueous phase expressed as a mole fraction and γ_W is the (dimensionless) activity coefficient of the chemical in water (at the given mole fraction). γ_W for neutral organic chemicals span multiple orders of magnitude, which follows from the inverse relationship between the activity coefficient and water solubility. For example, the activity coefficients at infinite dilution for 1-butanol and benzo(a)pyrene are 50 and 10⁸ respectively (Schwarzenbach et al., 2003). For more hydrophobic chemicals, γ_W in dilute solution and at saturation are typically within 30% and any concentration-dependence of the activity coefficient can be ignored (Schwarzenbach et al., 2003). For more hydrophilic (miscible) compounds, γ_W will exhibit a stronger concentration-dependence. However, as per the definition of γ_W, γ_W tends towards a value of 1 as the mole fraction of the chemical in aqueous solution increases.

There are numerous experimental techniques available for measuring γ_W at infinite dilution and empirical data are available for some miscible chemicals e.g., γ_W = 1.6, 3.7 and 7.0 for methanol, ethanol and acetone respectively (Schwarzenbach et al., 2003). As both the upper bound (γ_W at infinite dilution) and lower bound (γ_W = 1) is known for these miscible chemicals, LC₅₀s can be expressed using chemical activity (i.e., converted to Ea₅₀s), at least as a bounded first approximation, i.e.,

where χ_i(LC₅₀) is the LC₅₀ expressed as a mole fraction. Equation 8 thus represents a possible approach to enable an estimate of chemical activity in relation to an effect concentration. Further assessment, however, is required to better understand the feasibility of the approach and to quantify uncertainties and define potential limitation (Yalkowsky et al., 1994; Lian and Yalkowsky, 2014; Schwarzenbach et al., 2003; Sherman et al., 1996; Brockbank et al., 2013; Ingram et al., 2011; Hilal et al., 2004).

1. Feasibility of applying the chemical activity concept to ionisable organic chemicals (IOCs)

The applicability of the chemical activity concept to ionisable organic chemicals (IOCs) represents an additional challenge. Firstly, the water solubility of IOCs is a function of the solubility of the compound and the degree of ionisation (He and Yalkowsky, 2004; Jain et al., 2006) where the degree of ionisation is a function of pH and pK_a. The type and concentration of counterion(s) present in solution are also considerations for determining the apparent solubility limit (Chowan, 1978; Serajuddin, 2007). Even if the apparent water solubility can be estimated, the relevance of using this estimate in chemical activity calculations (such as Equation 1) has not yet been fully addressed. While methods to estimate activity coefficients for IOCs have been proposed (Franco and Trapp, 2010), it is unclear if these approaches are congruent with the methods for neutral organic chemicals and hence can be used in the same manner (i.e. Equation 7). IOCs also exhibit different partitioning behaviour compared to neutral organic chemicals (Avdeef et al., 1998; Escher et al., 2000a,b; Armitage et al., 2013) complicating the estimation of the biota-water partition coefficient (or more appropriately, the biota-water distribution ratio, D_BW). Furthermore, in addition to challenges associated with estimating a chemical activity for IOCs, there is the complexity associated with aligning a derived chemical activity to an effect concentration. The key challenge here is that IOCs, examples of which include active pharmaceutical ingredients and pesticides, are unlikely to act strictly as baseline toxicants, but will have one or more specific modes of action.

_{[1] Excess toxicity is defined as the ratio of the predicted effect concentration for base line toxicity and the observed effect concentration.}