Selenium (Se) is an essential trace element in mammals, of which both deficiency and excess can have detrimental effects on health (1). Se supplementation within a narrow range optimizes activity of Se-dependent antioxidant enzymes that incorporate Se co-translationally in the form of selenocysteine. Se also counteracts the toxicity of certain heavy metals, such as arsenic, lead, and mercury (2, 3). Deficient Se intake impairs thyroid hormone metabolism and reduces activity of the antioxidant enzymes, glutathione peroxidase and thioredoxin reductase (4). In contrast, Se can be detrimental at high doses, with documented neurotoxic effects (5).

Se-deficiency is estimated to occur in roughly 10% of the world's population and is observed predominantly in regions with low soil Se-content, such as Scandinavia, New Zealand, and Northeast China (6). Furthermore, future climate change is predicted to decrease soil Se content in agricultural regions and augment the prevalence of Se-deficiency worldwide (7). Common symptoms associated with Se-deficiency include hypothyroidism, cardiomyopathy, compromised immunity, fatigue, and cognitive deficits (8–10). Also, altered serum Se levels have been documented in both autism and schizophrenia (11, 12), and it is hypothesized that redox imbalance during neurodevelopment increases risk for these neuropsychiatric conditions (13, 14).

On the opposite end of the spectrum, Se-excess can lead to toxicity and increased oxidative stress. In rodents, acute Se overexposure elicits motor deficits, catalepsy-like behavior and increased levels of dopamine, with inorganic selenium compounds being significantly more toxic than organic counterparts (15, 16). Moreover, in humans, rare cases of chronic Se-excess have been associated with elevated incidences of amyotrophic lateral sclerosis (17–19). Additionally, elevated selenium intake has been linked to higher incidences of type 2 diabetes (20), as have heightened levels of the selenium transport protein, selenoprotein P (21).

Whereas, the influence of Se has been extensively studied in many contexts, the developmental in vivo effects of chronic Se-deficiency and Se-excess upon measures of neurobehavior and energy metabolism have not been comprehensively characterized. Thus, we performed an expansive assessment of various behavioral and metabolic indices in young adult mice receiving dietary supplementation at levels corresponding to Se-deficient, Se-supplemented, and Se-excess upon weaning.

For this study, newly weaned male mice were allocated into three groups devised to represent conditions of Se-deficiency (Se-def), Se-supplementation (Se-sup), and Se-excess (Se-exc). All mice were fed Se-deficient laboratory chow (~0.08 ppm Se), with the Se-sup and Se-exc groups receiving additional Se supplementation in their drinking water at doses of 10 μM and 100 μM, respectively. As anticipated, we observed no differences among groups for food intake (Figure 1A), although water consumption did vary (Figure 1B) [F_{(2, 17)} = 4.918, p = 0.0206], with significant differences between the Se-def and Se-exc groups (p = 0.018). We also calculated Se intake based on water and food consumption, and mean values corresponded to 0.26, 2.14, and 13.94 μg/days for the Se-def, Se-sup, and Se-exc groups, respectively (Figure 1C).

Upon sacrifice, tissue was harvested for determination of Se content and additional molecular analyses. For kidney [F_{(2, 9)} = 78.91, p < 0.0001], liver [F_{(2, 9)} = 100.5, p < 0.0001], and serum samples [F_{(2, 9)} = 8.578, p = 0.0082], Se levels differed among groups in a dose-dependent manner, whereas in brain [F_{(2, 8)} = 0.7685, p = 0.4951] and testes [F_{(2, 9)} = 3.430, p = 0.0781], levels were comparable (Figure 1D). Parallel analyses of glutathione peroxidase (GPx) activity were conducted on liver, serum, and brain samples. Surprisingly, liver GPx activity [F_{(2, 9)} = 2.256, p = 0.1606] was similar between groups, whereas serum GPx activity [F_{(2, 9)} = 0.5429, p = 0.5937] showed similar non-significant trends as observed for Se analysis (Figure 1E). For brain, we detected a significant main effect of Se group upon GPx activity [F_{(2, 9)} = 4.263, p = 0.0498], with differences between the Se-def and Se-sup groups attaining significance (p = 0.0441). Western blotting was also performed on liver and brain samples to assess various markers of Se status. We probed for GPX1 and TXNRD2, two abundant selenoproteins known to be responsive and non-responsive to alterations in Se supply (28), respectively, and the selenium binding protein (SELENBP1), a putative factor protecting against Se-toxicity (29). For liver samples, levels of GPX1, TXNRD2, and SELENBP1 were not impacted by Se group (Figures 1F,G). Brain levels of TXNRD2 and SELENBP1 were comparable between groups, but we did observe altered levels of GPX1 [F_{(2, 9)} = 15.67, p = 0.0012], as levels were significantly reduced in the Se-def group (vs Se-sup: p = 0.0038; vs. Se-exc: p = 0.0016) (Figures 1H,I).

Prior to harvesting of tissue, mice were subjected to a battery of neurobehavioral and metabolic tests. Cognition was evaluated using the Barnes maze, a widely utilized paradigm for spatial learning in rodents. Mice were trained to find a hidden escape tunnel located beneath one of 40 holes on the periphery of the circular maze. As anticipated, we observed a main effect of time on spatial learning, as indicated by less primary errors [F_{(5, 100)} = 88.6, p < 0.001] and a faster primary latency (F_{(5, 100)} = 66.7, p < 0.0001] when locating the escape tunnel (Figures 2A,B). We also detected a main effect of Se group upon the number of primary errors [F_{(2, 25)} = 4.907, p = 0.0159], but not upon the primary latency [F_{(2, 25)} = 1.44, p = 0.2560]. For both primary errors [F_{(8, 100)} = 1.697, p = 0.1083] and primary latency [F_{(8, 100)} = 1.015, p = 0.4295], the time × Se group interaction was non-significant. Post-hoc analyses revealed higher levels of primary errors in the Se-def group during trial block one (vs. Se-exc: p = 0.0151) and two (vs Se-sup: p =0.0008; vs. Se-exc: p = 0.0233). Likewise, primary latencies were significantly higher in the Se-def group during trial block 2 (vs Se-sup: p = 0.0146; vs. Se-exc: p = 0.0144). No significant differences were observed between groups for these measures during the remaining trial blocks nor during a probe trial conducted after trial block 5 (data not shown). As a whole, these results indicate that spatial learning is impaired by Se-deficiency.

Motor coordination, as determined by the latency to fall off a rotating rod of increasing speed, was examined at 8, 12, and 16 weeks of age (Figure 2C). Two-way ANOVA analyses revealed a main effect of time [F_{(2, 50)} = 7.826, p = 0.0011) and a significant time x Se group interaction effect [F_{(4, 50)} = 2.779, p = 0.0367], whereas the influence of Se group was non-significant [F_{(2, 25)} = 1.208, p = 0.3156]. In our initial test at 8 weeks of age, we observed differences between the Se-sup and Se-exc groups, with Se-exc mice performing significantly worse (p = 0.0252). Surprisingly, motor coordination improved over time in the Se-exc group, whereas performance declined in both the Se-def and Se-sup groups.

To assess sensorimotor gating, mice were tested for acoustic startle reactivity and prepulse inhibition. For acoustic startle (Figure 2D), we observed a main effect for stimulus intensity [F_{(6, 150)} = 53.31, p < 0.0001), whereas both Se group [F_{(2, 25)} = 2.241, p = 0.1273] and the stimulus intensity × Se group interaction [F_{(12, 150)} = 1.610, p = 0.0942] were not significant. Across the vast majority of stimulus intensities, startle magnitude was most pronounced in the Se-exc group, with statistically significant differences detected at 95 dB (vs. Se-sup: p = 0.0410) and 110 dB (vs. Se-def: p = 0.0025). In testing for prepulse inhibition (Figure 2E), a main effect of prepulse intensity [F_{(2, 50)} = 16.07, p < 0.0001] was found, while the effects of Se group [F_{(2, 25)} = 1.062, p = 0.3608] and the prepulse intensity × Se group interaction [F_{(4, 50)} = 2.185, p = 0.0841] failed to reach significance. Moreover, at the highest prepulse intensity (16 dB), the Se-def group exhibited significantly reduced inhibition relative to the Se-exc group (p = 0.0361).

Mice were also tested for glycemic control and body weight was regularly monitored. For glucose tolerance testing (Figure 3A), a main effect of time [F_{(4, 76)} = 63.18, p < 0.0001] and a significant time x Se group interaction effect [F_{(8, 76)} = 2.454, p = 0.0203] was detected, whereas the effect of Se group was non-significant [F_{(2, 19)} = 1.680, p = 0.2129]. Post-hoc tests revealed significantly elevated blood glucose levels 120 min after glucose injection in the Se-def group relative to the Se-exc group (p = 0.0148). With respect to body weight, two-way ANOVA analysis revealed a main effect of time [F_{(8, 200)} = 316.1, p < 0.0001], with non-significant effects observed for Se group [F_{(2, 25)} = 2.307, p = 0.1203] and the time × Se group interaction [F_{(16, 200)} = 0.4277, p = 0.9740]. Levels gradually diverged over time between the Se-def and Se-exc groups, with differences reaching significance at 20 wks (Figure 3B) (p = 0.0443). Upon sacrifice, we also found that relative levels of inguinal white adipose tissue (iWAT) differed between groups (Figure 3C) [F_{(2, 9)} = 6.227, p = 0.0201], with statistically significant differences between the Se-def and Se-exc groups (p = 0.0165). Similar non-significant trends were also observed for gonadal white adipose tissue (gWAT) (Figure 3D) [F_{(2, 9)} = 2.119, p = 0.1762] and serum leptin (Figure 3E) [F_{(2, 21)} = 2.563, p = 0.1009]. Finally, average lipid droplet size in brown adipose tissue (BAT) significantly differed between groups (Figures 3F,G) [F _{(2, 7, 750)} = 77.77, p < 0.0001], as droplets were larger in the Se-def group (p < 0.0001).

At 14–16 weeks of age, mice were placed in metabolic chambers for 48-hrs to evaluate activity, respiratory metabolism, and ingestive behavior. For locomotion (Figures 4A,B), two-way ANOVA analysis detected main effects for both light cycle [F_{(1, 17)} = 46.81, p < 0.0001] and Se-group [F_{(2, 17)} = 7.126, p = 0.0057], in conjunction with a non-significant light cycle x Se group interaction effect [F_{(2, 17)} = 0.809, p = 0.4617]. During both the light (p = 0.0164) and dark cycles (p = 0.0018), the Se-exc group exhibited elevated locomotion relative to the Se-sup group. For measures of energy expenditure (Figure 4C) [EE: F_{(1, 17)} = 1,238, p < 0.0001] and the respiratory quotient (Figure 4D) [RQ: F_{(1, 17)} = 140.2, p < 0.0001], we also observed a main effect for the light cycle, but not for Se group [EE: F_{(2, 17)} = 0.5002, p = 0.6151; RQ: F_{(2, 17)} = 3.126, p = 0.0698]. Moreover, we also detected a significant light cycle x Se group interaction effect for energy expenditure [EE: F_{(2, 17)} = 4.133, p = 0.0344], but not for the respiratory quotient [RQ: F_{(2, 17)} = 1.212, p = 0.3222].

In summary, these results detail the negative consequences of juvenile Se-deficiency upon measures of behavior and metabolism in early adulthood. Se-deficient mice displayed delayed learning and altered sensorimotor gating, and these deficits coincided with reduced GPx activity in brain. Moreover, Se-deficiency also resulted in impaired glycemic control, elevated body weight, and increased adiposity. Finally, Se-excess, at levels known to be toxic to humans, was surprisingly well-tolerated in mice and exerted beneficial effects on energy metabolism.

Our study corroborates prior findings that Se-deficiency hinders spatial learning (30, 31) and, to the best of our knowledge, represents the first association of Se-deficiency with impairments in sensorimotor gating. Deficits in cognition and sensorimotor gating are hallmarks of many neurodevelopmental disorders, including schizophrenia and autism. For both schizophrenia (11, 32, 33) and autism (12, 34, 35), reduced Se levels have been chronicled in the literature, albeit there are many exceptions (36–38), and it is unclear whether this represents a cause or consequence of these conditions. Of particular significance to our results is a recent report examining the Se status of 287 Polish children, which were divided into four groups, corresponding to: (1) autism spectrum disorder (ASD) with obesity, (2) ASD without obesity, (3) non-ASD with obesity, and (4) non-ASD without obesity (12). Observed Se levels were lowest in ASD patients with obesity and highest in non-ASD patients without obesity, with differences between groups being highly significant (p < 0.001) for serum, urine, toenail samples. Moreover, across groups, Se levels were inversely correlated with body mass index (p < 0.001) for all sample types.

The influence of Se supplementation upon energy metabolism is hotly debated and nuanced in the existing literature. An unanticipated corollary of the Nutritional Prevention of Cancer (NPC) trial was the observation that Se-supplementation (200 μg daily as Se-yeast) increased risk of type 2 diabetes for participants with baseline plasma Se levels within the upper tertile (20). Since these findings were documented, excess Se supplementation has been shown to adversely impact insulin signaling in multiple rodent models (39, 40). In contrast, reduced serum Se levels have been observed in morbidly obese patients (41), and supranutritional Se supplementation (240 μg/day) in the form of selenomethionine (SeMet) was recently shown to decrease both fat mass and circulating leptin levels in a 3-months dietary intervention study of obese individuals (42). Of potential relevance, we previously reported increased adiposity and elevated leptin levels in mice lacking SELENOM (43), an ER-resident selenoprotein that is highly expressed in brain and regulated by Se levels. Further studies showed that leptin upregulates SELENOM in hypothalamic neurons and that SELENOM, in turn, promotes leptin signaling (44). More recently, supranutrional Se supplementation (2.25 ppm SeMet in chow) was found to facilitate selenocysteine incorporation at sites canonically encoding cysteine, promote thermogenesis, and protect against diet-induced obesity (45).

One unexpected outcome of this study was the beneficial influence of Se at a dosage (8 ppm in water) originally hypothesized to elicit toxic effects. We chose to use chow that was mildly Se-deficient and provide further Se supplementation in the drinking water as selenite to the Se-sup and Se-exc groups. Inorganic Se species (selenite, selenate) are less readily absorbed by the intestine than organic counterparts (SeMet) (46–48), and are also significantly more toxic. For instance, the toxicity of selenite was found to be 53-fold greater than that of SeMet when administered intracerebroventricularly to rats (15). With specific regard to supplementation of inorganic Se species in drinking water, increased mortality was previously reported at levels >6 ppm, although lower Se dosages (2–3 ppm) did lead to decreased body weights (22). Similarly, chow containing Se at >5 ppm, has been shown to adversely impact growth and mortality in rodents (49) and pigs (50), with effects being more severe when selenite was the predominant Se species. Although relatively rare in humans, Se intoxication leads to loss of hair and nails, skin lesions, and nervous system abnormalities. A case study of Se toxicity in the Enshi district of China reported neurological defects in 18 of 22 subjects displaying signs of selenosis, and symptoms included hyperreflexia, convulsions, motor weakness, and hemiplegia (51). Moreover, blood Se levels in affected patients were observed to be roughly 100 times greater than subjects receiving a Se-adequate diet. Chronic Se overexposure has also been associated with an elevated risk of neurodegenerative disease, specifically amyotrophic lateral sclerosis (ALS). This linkage was first noted in 1977, when a cluster of ALS cases was reported in a seleniferous area of South Dakota (19) and further substantiated by increased incidences of ALS in an Italian population chronically exposed (1974–1988) to drinking water containing high levels of selenate (18). Of further significance, elevated levels of selenite have been reported in the cerebrospinal fluid (CSF) of newly diagnosed ALS patients (52).

Another unanticipated finding was that Se supplementation modulated GPx activity to a greater extent in brain than liver. Brain Se levels are typically lower than other organs and blood (53), with Se homeostasis in the nervous system being tightly regulated by the blood-brain barrier (BBB) (54). Se transport to brain is regulated by endothelial cell-mediated uptake of SELENOP via the lipoprotein-related receptor, ApoER2, at the BBB (54, 55). SELENOP is also expressed in astrocytes (56–58), especially those lining the BBB, and astrocyte-derived SELENOP is speculated to supply ApoER2-expressing neurons with Se within the parenchyma (59). In cases of severe Se-deficiency, it is known that the brain and testes preferentially retain Se at the expense of other organs, and this phenomenon is dependent upon SELENOP and ApoER2 (55, 60–62). It should be duly noted that our Se-deficient chow contained 0.08 ppm Se, several-fold higher than that of many Se deprivation studies, but still well below the 0.15 ppm minimum recommended for rodent diets by the AIN (63). Given that liver and kidney represent the primary sites of Se metabolism and excretion (64), respectively, the fact that supplementation most impacted Se content in these tissues was expected. The effect of supplementation on serum Se was less robust, suggesting that most Se was converted to excretory metabolites in liver, with a small fraction being incorporated into SELENOP. Moreover, it appears that our Se-deficient diet did not affect liver GPx activity, in line with prior findings by Sunde and colleagues showing that hepatic levels of GPx activity plateau when dietary Se is 0.09 ppm or greater (65). Furthermore, prior evidence suggests that supplementation at levels similar to our study can significantly impact brain GPx activity. For example, Whanger and colleagues reported that increasing dietary Se content from 0.1 to 4 ppm raised GPX activity by 32 and 77% in cortex and cerebellum, respectively (66).

It is imperative to note this study has several caveats that merit consideration. First, to reduce cost and animal usage, only male mice were used. Sex-specific differences in the biological effects of Se are well-documented in the literature (67–69), with males typically being more adversely impacted by deviations in Se intake. Second, experiments were conducted on young adult mice (3–5 months) and the possibility that long-term Se overexposure elicits neurodegenerative effects at later time points cannot be ruled out. Finally, given that our dietary intervention began shortly after weaning, it is probable that Se supplementation triggered developmental epigenetic adaptations to cope with Se-excess. Interestingly, Se-excess mice performed significantly worse in the initial rotarod test, but their performance improved over time, while that of the other two groups (Se-def, Se-sup) declined. It is quite possible that providing adult mice with Se at our chosen excessive dose may elicit detrimental toxic effects not observed in juveniles.

Nevertheless, these results detail the adverse effects of mild Se-deficiency and suggest that juvenile Se status is critical for optimal neurodevelopment. These findings may have important implications for future prevention and treatment of neurodevelopmental disorders where redox imbalance is a key characteristic.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by University of Hawaii's Institutional Animal Care and Use Committee.

VK and MP designed the experiments. VK, AS, DT, CC, JP, and MP performed research. CW contributed reagents/analytic tools. MP analyzed data and wrote the paper. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.