**4. Metal subcellular fates and partitioning**

#### **4.1. Subcellular partitioning**

Bioaccumulation is a prerequisite, but not necessarily a reliable indicator of metal toxicity due to the species-specific capacity of organisms to detoxify the metals accumulated [9, 59, 82]. The detoxifying mechanisms have been suggested to account for the tolerance of organisms [83]. Besides metal uptake, metal toxicity is determined by the subcellular partitioning of metals in organisms [83]. Metals accumulated are distributed to different cellular components and sequestered by binding to proteins or peptides (e.g., metallothionein and glutathione) and granules [10, 83–86]. Therefore, only parts of meals are accumulated in sensitive cellular fractions [9]. From a toxicology perspective, metals accumulated have been divided into two fractions, that is, metal-sensitive fraction (MSF) and biologically detoxified metal (BDM) [83, 87, 88]. The former consists of metals in mitochondria, associated with heat-denaturable proteins (HDPs), lysosomes, and microsomes. The latter includes metals in association with heat-stable proteins (HSPs) or metallothionein (MT)-like proteins and metal-rich granules.

It has been suggested that sub-lethal toxicity is accompanied with changes in subcellular partitioning, especially when the threshold is exceeded, that is, saturation of detoxification mechanisms [83]. Eyckmans et al. [89] showed the relationship between the tolerance of three freshwater fish species to excess Cu and the subcellular partitioning of Cu. The subcellular partitioning of metals may divulge potential mechanisms of toxicity as well as the fate of accumulated metals [90]. For instance, the binding of metals to cytosolic proteins such as MT has widely been demonstrated to alleviate toxic effects [91]. In contrast, the association of metals such as Cd to mitochondria modulates oxidative phosphorylation, followed by decreases in ATP production [92]. Similarly, the binding of Cd to the nucleus is potentially toxic because it can lead to DNA damage and stimulate mutagenesis [93, 94]. Different ligands are included in the HSP fraction, that is, amino acids, glutathione, and metallothioneins [95]. This fraction determines the tolerance and resistance of organisms to metals [83, 84, 87]. The relationships between subcellular partitioning and metal toxicity are complicated because of the different physiological functions of each fraction. The accumulation of metals in the lysosomes and microsomes might reflect the storage for eventual elimination if metals are mainly in the lysosomal fraction [85, 96]. In contrast, metals associated with microsomes can indicate toxicity because of the presence of fragmented endoplasmic reticulum, which is involved in the synthesis and transport of proteins [85, 87]. In addition, mitochondria is the most sensitive fraction [85]. Because of these factors, subcellular partitioning has recently been included in the assessment of metal toxicity.

#### **4.2. Assessment based on metallothionein induction**

The induction of MT has been integrated in various monitoring programmes and ecotoxico‐ logical assessment. However, recent reviews on the use of MT induction as biomarkers of metal exposure and toxicity raise questions on the validity of this method [11, 97]. The main concern comes from the lack of well-described time- and dose-dependent MT induction, while the induction of MT widely varies, depending on metals, species, and environmental conditions [97]. Further concern results from the lack of well-described relationships between MT induction and metal exposure, metal accumulation, and biological effects [11]. The validity of the use of MT induction as a biomarker is questionable because of the lack of enhancement in MT induction in response to metal exposure and/or the insignificant relationships between metal and MT concentrations [11]. As explained in the reviews by these authors, these results are attributed to different factors determining the cytosolic free metal ion concentrations, the presence of different MT isoforms in different tissues with various physiological functions, and turnover kinetics of MT as well as the high variability of MT induction and the basal level [11]. These factors should be taken into consideration in the assessment of metal exposure and toxicity based on MT induction. Another concern is related to the selection of organs/tissues as the target in the assessment. The digestive gland has widely been used as the target tissue [98, 99] from the point of view that this is the long-term storage organ. In addition, this organ plays an important role in different processes such as immune defence, homeostasis, xenobiotic mechanisms, elimination, and detoxification [100–103]. Another reason for the use of the digestive gland is that this organ has the highest amount of proteins, including MT [104]. However, the suitability of the digestive ligand is doubtful because of the influence of natural factors and physiological changes on the induction of MT in this organ [104, 105]. As a major tissue for metal uptake, the induction of MT in gills is expected to represent the response of organisms to the ongoing exposure or accidental pollution [106]. Because of these issues, a single value of MT concentrations in the whole soft tissue or in a specific tissue is not always good indicator of metal exposure or effects on biota. As suggested by Le et al. [11], the induction of MT needs to be considered in relation to a number of factors as briefly mentioned above.

#### **4.3. Assessment based on subcellular partitioning**

Binding to MT is not the only mechanism for organisms to detoxify and to sequester excess metals. Metals can be associated with insoluble complexes in granules or lysosomes as mentioned above. The proportion of metals bound to MT is not always the dominant fraction and so MT induction is not necessarily a reliable indicator of metal exposure and biological effects. According to Adams et al. [107], the kinetic partitioning of metals into MSF and BDM is the key factor for an understanding of the exposure-toxicity relationship. This has been demonstrated by the shift of Cu accumulated in gills of common carp from MSF to BDM in the first period of the Cu exposure [89]. The 'Spillover' hypothesis has been used to simulate the relationship between metal bioaccumulation and toxicity. This term indicates the situation when the capacity of organisms to detoxify excess metals is overwhelmed, thereby resulting in adverse effects [66]. From a broad perspective, spillover is assumed to occur when the uptake rate exceeds a combination of the detoxification and elimination rates, leading to the accumu‐ lation of metals in the MSF fractions [9, 108]. This hypothesis has been used in some approaches for assessing metal toxicity.

The first approach is based on the ratio between MSF and MDP fractions or the relative contribution of the MSF and MDF fractions in other words [89, 90]. This approach is simplistic and does not provide a full description of toxicity [90]. There is not constant partitioning between MSF and BDM. Metals can be exchanged between the two fractions, and only a small fraction of metals occurs as free ions [90]. The use of the relative distribution between these two pools may lead to misinterpretation of spillover [90, 92, 109]. For instance, the increase in the amount of metals bound to MSF might be hidden by the lack of changes in the relative proportion, while the accumulation in all compartments increases. This has been illustrated by increases in metal concentrations in both BDM and MSF fractions in yellow eels [110], mummichogs [111], or yellow perch [85].

The second approach is to develop a relationship between MSF and toxic effects [112], assuming that adverse impacts on organisms increase with increasing amounts of metals in MSF. This method is based on the hypothesis that spillover occurs when certain detoxification mechanisms are overwhelmed and excess metals are accumulated in MSF, leading to adverse effects.

The third approach is based on the kinetics of metal accumulation in MSF [108, 113, 114]. The accumulation of metals in MSF can be described as a balance of uptake (e.g., from the dis‐ solved), losses via elimination, and the partitioning of the metals to BDM, expressed by the detoxification rate:

$$\text{MIIT} = \left(k\_u \times C\_w\right) = \left(k\_{\text{dran}} + k\_e\right) \times C\_{IT} \tag{7}$$

where MIT (mol/g/d) is the metal influx threshold ; *k*<sup>u</sup> (L/g/d) is the uptake rate constant; *C*<sup>w</sup> (mol/L) is the dissolved metal concentration; *k*detox (1/d) is the detoxification rate constant; ke (1/d) is the elimination rate constant; *C*IT (mol/g) is the metal accumulation in MSF at the influx threshold.

This approach is supported by a negative correlation between elimination and detoxification [113, 115]. However, disadvantages are inherent in the assumption of this method that spillover occurs when the uptake rate exceeds combination of the elimination and detoxification rates, that is, no metal is accumulated in MSF below the threshold. Previous studies have indicated that spillover does not happen at low-exposure concentrations. In other words, there is no threshold below which the accumulation of metals in the sensitive fractions does not occur [84, 85].

#### **5. Biological responses to metal exposure**

Available approaches based on the induction of MT or taking into account subcellular partitioning of metals have shown some limitations in predicting metal toxicity as mentioned in the previous section. Moreover, the target organ or tissue is species specific, depending on the kinetics of metal accumulation, that is, the differences in absorption, distribution, and excretion [6]. A method that has been demonstrated to be more applicable to different species is to predict metal toxicity based on the accumulation of metals at biological surfaces, which allows for taking into account interactions between organisms and metals at biological surfaces.

#### **5.1. Biotic Ligand Model**

Interactions at the water-organism interface have been integrated in the Biotic Ligand Model (BLM). The conceptual framework of the BLM originates from two models: the gill surface interaction model and the free ion activity model (FIAM) [116–118]. The FIAM model assumes that free ions are the main reactive species of metals, determining metal bioavailability and toxicity [116, 117]. The FIAM model has then been extended to take into account the interac‐ tions of metals with organisms. One example is the fish gill surface interaction model [119], which has been developed by integrating conditional metal-gill surface binding constants to a geochemical speciation model. On the basis of the fish gill surface interaction model, the BLM has been developed to facilitate the application to various species, metals, and exposure conditions as well. According to the concept of the BLM, environmental geochemistry as well as toxicology principles are taken into consideration in determining the fraction of metals that provokes effects on biota [119–121]. Initial toxicology bases of the model were effects of trace metals on ionoregulation. Apical (e.g., Na+ and Ca2+ channels) and basolateral (e.g., Na+ /K+ and Ca2+-ATPase) transport channels, exchangers, and enzymes are negatively charged proteins that potentially bind to metal cations [122].

With the principles mentioned above, the BLM has shown substantial advantages in assess‐ ment of metal bioavailability. By including interactions of metal ions with both biotic and abiotic ligands, the BLM might provide more reliable estimates of metal bioavailability and toxicity. Moreover, the accumulation of metal ions at the physiological active sites of toxic actions, which determines toxic effect, is, in principle, distinguished from the total body burden [6]. Another advantage of the BLM is the inclusion of environmental conditions in estimating metal bioavailability and subsequent application to site-specific assessment [6]. Furthermore, the BLM provides a mechanistic understanding of metal-organism interactions as well as a mechanistic interpretation of metal toxicity, as illustrated by the results in the study of Le et al. [123]. The BLM has potential for assessing the toxicity of metal mixtures [123, 124]. Norwood et al. [124] suggested that metal-metal interactions can be predicted based on known stability constants. If two metals compete for the same binding site, the total amount of the metals bound to the biotic ligands determines combined effects [123]. Alternatively, the accumulation of individual metals at the biotic ligands can be used as inputs to the response addition model for estimating toxicity of metal mixtures [124, 125]. This suggestion has been demonstrated by the increasing application of the BLM for predicting joint toxicity of metals. Liu et al. [126] predicted the toxicity of metal mixtures based on the simple sum of the fraction of biotic ligands occupied by individual metals. This approach allows for taking into account the metal-specific affinity for binding sites of biotic ligands, but not the metal-specific toxic potency. In other studies, the BLM concepts are usually combined with conventional concepts of mixture toxicity such as concentration addition. As such, the specific toxic potency of metals is taken into account in the estimates. For instance, the toxicity of metal mixtures has been related to the toxic equivalent quotient (TEQ), which is based on the accumulation of metal ions at the binding sites of biotic ligands while giving consideration to the metal-specific toxic potency [123, 126]. The toxic unit (TU) is the concept mostly integrated in the BLM-based approaches for estimating joint toxicity of metals.

#### **5.2. The electrostatic model**

The electrostatic model originates from the reported differences between the ion concentration at the root plasma membrane surface and the ion concentration in the external medium, which is induced by the negative charge at the plasma membrane surface [127, 128]. The surface potential affects the activity of ions at the plasma membrane surface via electrostatic attraction or repulsion. In addition, the potential influences the difference in the electrical potential across the membrane, which stimulates the transport of ions through the membrane. A number of studies have shown the effects of electrical potential at the plasma membrane surface on cation uptake [127, 128]. The principle effect of the electrical potential at the plasma membrane surface is to control the ion activity at the surface [129].

The model allows for integrating plant-ion interactions in predicting metal toxicity [130–132]. In addition, the model gives consideration to the interactions among ions while estimating metal toxicity [132]. The electrostatic model has been applied to simulate effects of major cations on the toxicity of trace metal ions [131–134]. According to the principle of the electro‐ static model, additions of cations to the bulk medium reduce the negativity of the electrical potential at the plasma membrane surface, thereby leading to decreases in the negativity of the electrical potential at the plasma membrane surface accompanied by reduced accumulation of trace metal ions at the plasma membrane surface and subsequent alleviation of toxic effects [135]. The electrostatic approach therefore might provide additional explanation for interpret‐ ing ion-ion interactions as well as effects of major cations on the toxicity of trace metal ions besides the competitive binding assumed in the BLM. According to the electrostatic principles, three mechanisms have been suggested to account for ameliorative effects of Ca2+ on metal toxicity [128]. The first mechanism is the electrostatic displacement of trace metal ions at the plasma membrane surface by Ca2+. The second is the restoration of Ca2+ at the cell surface in response to low levels of surface Ca2+. This mechanism is mediated in order to avoid the inhibition of the plant growth induced by the low contents of Ca2+ at the cell surface. The third mechanism covers other types of interactions between Ca2+ and trace metal ions, such as the blockade of the ion channel [136]. Despite the toxicant- and major cation-specific relative importance of these mechanisms, mechanism 1 generally occurs in all cases while the contri‐ bution of mechanism 2 is minor [128]. The significance of mechanism 3 is trace metal ion specific. The relative influence of major cations such as Ca2+ and Mg2+ on the toxicity of trace metal ions varies, depending on the trace metal ions and species [128]. For instance, Ca2+ has stronger effects on the toxicity of Al3+ to soybean than Mg2+ [137, 138]. In wheat, Mg2+ is more effective than Ca2+ in inhibiting Zn2+ toxicity, whereas Mg2+ does not have effects on the toxicity of Al3+ [128].

#### **5.3. WHAM-based bioavailability model**

Originally developed for determining and quantifying chemical forms and species of metals in different phases in the environment, speciation models such as WHAM have recently been used for estimating metal bioavailability and toxicity (hereafter referred to as the WHAMbased model). In this method, humic acid is considered a surrogate to biological surfaces [139– 144]. As such, the amount of metals bound to humic acid computed by WHAM is used for estimating metal toxicity for different species. In the application of the WHAM-based model to plants, humic acid is used as a surrogate for the root surface [144]. In other words, the interactions of metal ions and the root surface are simulated by the interactions with humic acid.

The WHAM-based model was initiated from the similarities in terms of chemical structure between the root surface and humic acid. The root surface is a heterogeneous mixture of various metal-binding functional groups [145–147]. Similarly, humic acid possesses a hetero‐ geneous mixture of functional groups, mainly carboxylic and phenolic acids [141]. Further‐ more, the ratio of 1:2 between the site densities of phenolic and carboxylic groups assumed in WHAM [40, 41] lies in the range from 0.49:1 to 1:1 reported for root cell walls of different plant species [148]. Because of this similar structure of the root surface and humic acid, metal accumulation on the root surface and metal binding to humic acids are influenced by similar factors. These factors include electrostatic interactions and chemical heterogeneity [127, 128, 149, 150]. The relevance of using metal binding to humic acid to represent metal accumulation at biological surfaces is further substantiated by the nature of the sorption of metal cations to the biological surfaces [151].

The WHAM-based model allows for the interactions between metal ions to be integrated in modelling metal toxicity [144]. Specifically, interactions between ions are considered in modelling metal binding to humic acid in WHAM and therefore accounted for in estimating metal accumulation at the biological surfaces. Another advantage of the WHAM-based model is related to the availability of binding constants in WHAM, which facilitates a wide application to different metals [144]. Moreover, previous studies have shown small variations in the binding constants among different species [145, 147, 152, 153]. For instance, the logarithm of the binding constants of Cd to the carboxylic group on the membrane surface of bacteria, fungi, and plants were in the range 3.3–3.5. Moreover, the cell walls of algae, maize, soybeans, and higher plants have similar titration curves [154–156]. These results indicate the applicability of a single set of binding constants to different organisms.

In the study of Le et al. [144], the amount of metals bound to humic acid was computed with WHAM in which metal sorption to humic substances is simulated by using a structured formulation of discrete, chemically plausible binding sites for protons and metals. This allows the creation of regular arrays of bidentate and tridentate binding sites for metals. Metal aquo ions compete with their first hydrolysis products, protons, and other metals for binding sites. In addition to the intrinsic metal binding strength to uncharged molecules, electrostatic effects were taken into consideration while predicting metal binding to humic acid. Le et al. [144] showed the potential of the WHAM-HA model for predicting both total and internalised metal concentrations in roots. Another example of the approaches based on the geochemical equilibrium in WHAM is the WHAM-*F*TOX model for estimating metal toxicity [140, 143]. In this approach, mixture toxicity was simulated as a function of the metal-specific toxic potency and the amount of metal ions bound to the biological surfaces computed with WHAM.
