Excess selenium has been shown to accumulate as selenomethionine (SeMet) in proteins of hair and hooves/nails, in tissues including heart, liver, muscle, and skin, and to affect regions of the central nervous system (Kim and Mahan, 2001; Panter et al., 1996; Tiwary et al., 2006; Yang et al., 1983). The substitution of SeMet for methionine is known to alter protein stability (Jackson and Combs, 2008) and was recently demonstrated to affect target protein binding of the calcium regulatory protein, calmodulin (Yamniuk et al., 2009). Selenocysteine can be generated from SeMet and is site specifically incorporated into selenoproteins, such as thioredoxin reductases (TRXR) and glutathione peroxidases (GPX), proteins that function as enzymatic antioxidants (Battin and Brumaghim, 2009; Lu et al., 2009). The GPX proteins reduce hydrogen peroxides and lipid hydroperoxides, an action leading to the oxidation of glutathione (GSH), a cofactor for this reaction, which also functions as the major intracellular antioxidant (Forman et al., 2009). Depletion of intracellular pools of GSH during selenium-induced oxidative stress is furthered by the formation of hydrogen selenide, a product of selenium metabolism highly reactive with GSH and other intracellular thiols. This loss of GSH antioxidant activity is believed to allow damaging reactive oxygen species (ROS) levels to rise in cells and has been proposed as one of the mechanisms contributing to the overall oxidative damage and cell death observed with toxic selenium exposures (Misra and Niyogi, 2009; Spallholz and Hoffman, 2002).

The glutaredoxins (GLRX or GRX) are a group of oxidoreductases that catalyze the reversible reduction of protein disulfides through the cysteinyl residues of their active site (Ghezzi and Di Simplicio, 2009) and are in turn maintained in their reduced active form through the nonenzymatic oxidization of GSH (Lillig et al., 2008). GLRXs are classified into two major groups: dithiol or monothiol based on the consensus sequences found within their active sites (CPYC and CGFS, respectively) (Lillig et al., 2008). Deletion of both yeast dithiol GLRXs led to a selenium-induced growth inhibition that was reversible in an oxygen-free environment (Lewinska and Bartosz, 2008) and more recently shown to protect against selenium-induced cell death (Izquierdo et al., 2010). In contrast, small-interfering RNA knockdown of human GRX-1 decreased selenium-induced cytotoxicity in a human lung cancer cell line (Wallenberg et al., 2010). These studies suggest GLRX proteins play an essential role in selenium-induced toxicity.

Recent advances in understanding the effects of both acute high-dose and long-term low-dose exposures to selenium have been sparked by interest in selenium as an anti-tumorigenic agent (Batist et al., 1986; Facompre and El-Bayoumy, 2009; Jackson and Combs, 2008). This interest coupled with the commercialization of selenium because of its antioxidant capabilities has led to increased supplementation of selenium in food products (Boosalis, 2008; Navarro-Alarcon and Cabrera-Vique, 2008). In addition, large amounts of selenium are released into the atmosphere through mining operations, processing, and consumption of fossil fuel, and into the soil through agricultural irrigation in areas with high selenium content (Lemly, 1997; Lenz and Lens, 2009; Palmer et al., 2010). These factors have made elucidating the mechanisms of selenium toxicity more pressing as the sum of their effects increases the overall risk that a toxic selenium exposure will occur. The aim of this study was to evaluate the effects of selenium exposure on both lethality and movement behaviors in the nematode Caenorhabditis elegans and to determine the role of the worm glutaredoxin system, exemplified by the gene glrx-21, in selenium-induced toxicosis.

Removal of hydrogen peroxide from cells occurs through the activities of one of three enzymes, glutathione peroxidase, catalase, or peroxiredoxins (Nordberg and Arnér, 2001) but has also been shown to be reduced by a thioredoxin reductase–dependent pathway (Björnstedt et al., 1995). TRXR-1, an ortholog of the human enzymatic antioxidant thioredoxin reductase-1 has been demonstrated to be the only selenoprotein in C elegans (Gladyshev et al., 1999; Taskov et al., 2005). Our observation that at the midrange concentrations (1–2mM) selenium is protective against H₂O₂-induced toxicity, whereas at the lowest ranges (0.001–0.01mM) selenium is toxic (Figure 1) suggests that in C elegans selenium incorporation into enzymatic antioxidant selenoproteins may be dosage dependent. In previous studies examining the effects of dietary selenium levels on the activity of glutathione peroxidase and thioredoxin reductase, the activity of both enzymes was shown to gradually increase as the concentration of selenium in the feed increased (Hadley and Sunde, 2001; Sunde and Hadley, 2010). Yet, in both cases, selenoprotein activity reached a plateau after which increasing selenium concentration did not increase the activity of the enzymes. In our studies, as selenium concentration neared the Se-LC₅₀ (3.47mM), it ceased to have any measurable antioxidant effect, and increasing lethality was observed within the population as exposure levels increased (Figure 1). With increasing Na₂SeO₃ concentration, hydrogen selenide (H₂Se) produced by the excess selenium would generate ROS causing oxidative stress rather than becoming incorporated into antioxidant selenoproteins (Letavayová, et al., 2006). Our findings of elevated ROS levels in animals exposed to excess selenium (Figure 2) lend support to this theory. Additionally, selenium-dependent oxidation of GSH would in turn further sensitize cells to hydrogen selenide–induced superoxide anions leading to even greater oxidative stress. The dose-dependent ability of glutathione to suppress the ROS formation (Figure 2), lethality (Figure 4B), and motility impairment (Figure 5B) induced by excess selenium is consistent with the predicted depletion of GSH levels by the thiol oxidizing properties of hydrogen selenide and suggests that decreased GSH levels play a key role in the cytotoxic effects of selenium. This is consistent with a recent finding showing that GSH depletion in cells led to the increased sensitivity of cells to selenite (Wallenberg et al., 2010).

The glutaredoxin proteins (GLRXs) are not selenoproteins and thus are not directly affected or regulated by the levels of selenium in cells, but function in the reduction of proteins and low–molecular weight mixed disulfides. Under conditions of oxidative stress, the glutaredoxins can reverse this role and instead catalyze the formation of mixed disulfides, thus preventing further oxidation of thiols such as GSH (Fernandes and Holmgren, 2004; Holmgren, 2000; Lillig et al., 2008). During selenium-induced oxidative stress, the glutaredoxins have been demonstrated to be an essential component in the cellular response of yeast and cyanobacteria (Lewinska and Bartosz, 2008; Marteyn et al., 2009). The acceleration of the timing of motility impairment observed with the loss of the C elegans GLRX-21 protein (Figure 5C and Supplementary fig. 2B) is consistent with a role for glutaredoxin in the cellular responses of C elegans to selenium-induced oxidative stress. In addition, our finding that the loss of functional GLRX-21 protein in worms reversed the suppressive effects of GSH on lethality (Figure 4B), suggests that GLRX-21 is required for this GSH function. GLRX-21 is an ortholog of the human GLRX2 protein (Figure 3A and 3B; Sagemark et al., 2007), which has been shown to reduce mitochondrial apoptotic cell death induced by ROS-producing agents (Holmgren, et al., 2005; Lundberg et al., 2001). Like all GLRX proteins, GLRX2 is maintained in its reduced active form by GSH, which binds to specific sites within the GLRX2 protein that are highly conserved across glutaredoxins (Figure 3A). The data presented here suggest similarly that activation of GLRX-21 by GSH may be required for the reduced lethality observed with increasing GSH concentration in our selenium-treated animals (Figure 4B).

The lethality rate and the LC₅₀ value obtained from our liquid assays provided a useful general measure of the relative toxicity of selenium compared with other metals previously assayed in this model. Yet, as demonstrated in Figure 6A, a 5mM Na₂SeO₃ exposure of animals on agar plates caused death in only 22.5 (±12.14)% of the animals after 24-h exposure versus 79.2 (±18.01)% of the animals exposed to 5mM Na₂SeO₃ in our liquid assay with only a 12-h exposure period (Figure 1). This reduction in lethality of animals exposed on agar media allowed the additional observation of a selenium-induced motility phenotype that occurred prior to death (Figure 5). Locomotion in C elegans requires the coordination of three main groups of neurons: sensory neurons, interneurons, and the motor neurons that innervate the body wall muscles (Whittaker and Sternberg 2004). If selenium exposure causes progressive damage to the energy-producing functions of the muscles and/or neurons controlling movement, the resultant reduced energy availability could potentially lead to the observed gradual slowing of movement over time. Yet, our later observations showed that the motility phenotype could be further broken down into two distinct movement defects: impaired backing in response to head tap and paralysis (Figure 6A). In our studies examining single animals chronically exposed to selenium (Figure 6B), the observation that these phenotypes occurred progressively from a period of impaired backing before paralysis and paralysis before dying suggests that there is a stereotypic sequence of cellular damage caused by selenium. Because the ability to coordinate backward movement occurs prior to the more generalized paralysis phenotype, the cells involved in backward movement would appear to be more sensitive to selenium-induced toxicity, whereas those required for forward movement would seem less sensitive. Our observation that a subset of animals that had impaired backing were able to resume normal movement after removal from selenium (Figure 7), suggests that the damage that had been done to those cells was reversible. Thus, the progression of phenotypes could represent the induction of a multistep pathway leading to cellular dysfunction and cell death that could become irreversible by the time enough damage has accumulated. This hypothesis was supported by the fact that paralyzed animals did not recover (Figure 7).