Reversal of memory and neuropsychiatric symptoms and reduced tau pathology by selenium in 3xTg-AD mice

Accumulation of amyloid-β plaques and tau contribute to the pathogenesis of Alzheimer’s disease (AD), but it is unclear whether targeting tau pathology by antioxidants independently of amyloid-β causes beneficial effects on memory and neuropsychiatric symptoms. Selenium, an essential antioxidant element reduced in the aging brain, prevents development of neuropathology in AD transgenic mice at early disease stages. The therapeutic potential of selenium for ameliorating or reversing neuropsychiatric and cognitive behavioral symptoms at late AD stages is largely unknown. Here, we evaluated the effects of chronic dietary sodium selenate supplementation for 4 months in female 3xTg-AD mice at 12–14 months of age. Chronic sodium selenate treatment efficiently reversed hippocampal-dependent learning and memory impairments, and behavior- and neuropsychiatric-like symptoms in old female 3xTg-AD mice. Selenium significantly decreased the number of aggregated tau-positive neurons and astrogliosis, without globally affecting amyloid plaques, in the hippocampus of 3xTg-AD mice. These results indicate that selenium treatment reverses AD-like memory and neuropsychiatric symptoms by a mechanism involving reduction of aggregated tau and/or reactive astrocytes but not amyloid pathology. These results suggest that sodium selenate could be part of a combined therapeutic approach for the treatment of memory and neuropsychiatric symptoms in advanced AD stages.

combinated with antioxidants in multiple human diseases 13 . Randomized clinical trials with Se plus antioxidants are currently in progress in AD but without reported outcome data yet 14,15 .
Sodium selenite and selenate protect hippocampal neurons against Aβ42 toxicity and reduce amyloid burden in APP/PS1 transgenic mice [16][17][18] . Chronic sodium selenate in drinking water mitigates tau pathology in AD and traumatic brain injury murine models 19,20 . Sodium selenate-treated tau transgenic mice exhibit improved spatial learning and memory and show lower levels of phosphorylated and insoluble tau levels in the hippocampus and amygdala 21,22 . More recently, selenomethionine, supplemented before appearance of pathology, was shown to prevent spatial learning deficits, and reduced total and phosphorylated tau levels by promoting autophagy in mix sex groups of young 3xTg-AD mice 20,23 . Whether Se supplemented as a natural salt reverses emotional and memory symptoms specifically in AD female mice at late stages remain unclear. In this study, we investigated the effects of chronic dietary sodium selenate supplementation on both cognitive and neuropsychiatric-like domains, and studied the pathological mechanisms involved, in old 3xTg-AD mice.

Results
Chronic sodium selenate treatment stabilizes idiopathic behaviors in 3xTg-AD mice. 3xTg-AD mice show age-and sex-related differences in brain AD-like pathology and oxidative stress and behavior, which include enhanced amyloid pathology in females 24,25 , and immunological and behavioral alterations in males [26][27][28] . To evaluate specifically the effects of Se in females, the sex more affected by the disease in humans, we evaluated the effects of 4 months-chronic Se supplementation diet in 12-14 month-old female control (WT) and 3xTg-AD mice, which started treatment at 8 months of age. No significant differences were found in rotarod test between genotype and treatment groups [F (3,44) = 0.512, P > 0.05; data not shown]. During ethological assessment, we found no significant differences among groups for rearing seated or against the wall, nor regular or intense grooming and chewing. By contrast, we found significant differences between groups in locomotion [F (3,44) Fig. 1A). Bonferroni test revealed that 3xTg-AD mice showed reduced locomotion (t = 3.926, P = 0.002), free rearings (t = 3.103, P = 0.021) and sifting (t = 3.472, P = 0.007), whereas Se significantly reversed these alterations in locomotion (t = 3.451, P = 0.008), sifting (t = 7.24, P = 0.001) and stillness (t = 4.197, P < 0.001). Corner index measures revealed no group differences on corner visits [F (3,44) = 2.056, P > 0.05] but significant rearings differences among groups [F (3,44) = 49.546, P < 0.001] (Fig. 1B). 3xTg-AD mice visited slightly fewer corners and performed less rearings compared to WT mice (P < 0.001), a phenotype that was significantly ameliorated by Se treatment (rearings: t = 4.276, P = 0.001) (Fig. 1B).
Rescue of hippocampus-dependent spatial learning and memory in 3xTg-AD mice by sodium selenate treatment. In the Morris water maze (MWM), a spatial memory task that depends on the hippocampus, all experimental groups improved significantly during the training days but with significant learning differences between the groups [Two-way RM ANOVA; Escape latencies: group effect, F (3,40) = 7.74, P < 0.001; day effect, F (4,160) = 43.514, P < 0.001; Pathlengths: group effect, F (3,40) = 15.66, P < 0.001; day effect: F (4,160) = 33.5752, P < 0.001; Fig. 2A]. Post hoc comparisons revealed that untreated 3xTg-AD mice displayed longer escape latencies in comparison to other groups at days 3 to 5 (all P < 0.001). Swimming velocities were not different between the groups [F (3,40) = 2.305, P > 0.05], but all mice became faster and better swimmers during the days (F (4,160) = 3.887, P < 0.01). In the probe trial, both WT groups displayed a clear preference searching in the target quadrant (WT Veh: 39% time; WT Se: 36%) compared with 3xTg-AD mice (22%) (Fig. 2B,C). After Se treatment, 3xTg-AD mice spent significantly more time in the target quadrant (39%) indicating rescue of reference memory [F (3,43) = 5.121, P = 0.004]. Indeed, heat plots reveals that control groups and Se-treated 3xTg-AD mice mostly searched close to the designated platform position, whereas vehicle 3xTg-AD mice seemed to be circling more aimlessly (Fig. 2C).
During acquisition training, different spatial search strategies can be used (Fig. 2D). When counting the number of trials categorized in three main search strategies, we observed that both vehicle and Se-supplemented WT mice increasingly applied spatial strategies to locate the hidden platform (Fig. 2D). On day 5, vehicle-and Se-supplemented WT mice almost exclusive used hippocampal-dependent spatial strategies (72% and 65%, respectively), compared to vehicle 3xTg-AD mice (17%), which tended to use scanning and random strategies (36%). Notably, the use of spatial strategies increased after Se treatment in 3xTg-AD mice (69%).

Sodium selenate does not affect accumulation of amyloid plaques nor intraneuronal Aβ.
To determine whether Se could be beneficial in cognition by affecting AD-related pathological hallmarks, we next performed Aβ histopathological analyses in the above treated groups. 3xTg-AD mice accumulate abundant intracellular Aβ and amyloid plaques in the hippocampus, especially in the subiculum, and only sporadically in  (E) Evaluation of fear associative memory in the passive avoidance test. The reduced latencies to enter to the dark compartment were reversed in 3xTg-AD mice after Se treatment. (F,G) Assessment of wellbeing and social collaboration in the nest building test. 3xTg-AD mice showed a significant less nest-building ability than WT groups, while Se treatment increases nesting activity in 3xTg-AD mice (F) as observed in representative pictures of nests (G). Data represent mean ± s.e.m. Number of mice: WT vehicle (n = 11), WT Se (n = 10), 3xTg-AD vehicle (n = 9) and 3xTg-AD Se (n = 15). Statistical analysis was determined by one-way ANOVA followed by Bonferroni post hoc test. *P < 0.05, **P < 0.001, ***P < 0.001. the neocortex both in vehicle-and Se-treated conditions (Fig. 3A). Quantitative analysis revealed no significant differences in the number of dense, sparse or total amyloid plaques in the hippocampus between vehicle-and Se-treated 3xTg-AD mice (P > 0.05; Fig. 3B). Since intraneuronal Aβ is one of the earliest neuropathological hallmarks of 3xTg-AD mice that coincides with memory deficits 4 , we next quantified Aβ-containing neurons in the hippocampus. Quantitative analyses showed no significant changes in the number of Aβ-positive CA1 pyramidal neurons between vehicle-and Se-treated 3xTg-AD mice (P > 0.05; Fig. 3A,B).
Sodium selenate reduces accumulation of aggregated tau in 3xTg-AD mice. 3xTg-AD mice develop age-dependent tau pathology starting in the basolateral amygdala at 6-9 months and cortex and hippocampus at 12 months of age 3,6 . Staining of 3xTg-AD brains with anti-phosphorylated (CP13, PHF1) and conformational (MC1) tau antibodies show abundant tau accumulation in soma and axons of neurons in the hippocampus (CA1, CA2, CA3, DG), amygdala and cortex (parietal, auditory and entorrhinal cortices) (Fig. 4A). Quantitative analyses revealed no major changes in the number of phosphorylated tau-positive neurons in all   Table 1). The number of MC1-positive neurons was, however, significantly reduced in CA1 hippocampus and auditory cortex in 3xTg-AD mice after Se treatment (P < 0.05; Fig. 4A, Table 1). Western blot analysis showed a significant increase of total (TG5 and 17025 antibodies) and phosphorylated tau at Ser202 (CP13), Thr231 (AT180), Ser396/404 (PHF1) residues in hippocampal lysates of vehicle-treated 3xTg-AD mice compared with WT groups (P < 0.05; Fig. 4B). In 3xTg-AD mice, Se treatment slightly decreases phosphorylated tau levels but even more significantly reduced total tau levels, as analyzed with two distinct antibodies (TG5 and 17025) (P < 0.05; Fig. 4B,C). Surprisingly, biochemical analysis showed a significant increase of phosphorylated (Ser9)/total GSK3β ratio in the 3xTg-AD groups (P < 0.001), which inversely reflects its enzymatic activity (Fig. 4B,C). Finally, immunostaining revealed that 3xTg-AD treated with Se show a global significant reduction of GFAP staining in CA1, CA3 and DG regions (P < 0.05) but not in corpus callosum (Fig. 5A). Biochemical analysis confirmed a significant increase of GFAP protein levels in hippocampal lysates of 3xTg-AD mice, which were significantly reduced after Se treatment (P < 0.001; Fig. 5B). Together, these results indicate that chronic sodium selenate treatment reduces accumulation of aggregated tau and astrogliosis without globally affecting phosphorylated tau in the hippocampus of 3xTg-AD mice.

Discussion
There is growing evidence that besides genetic and non-genetic risk factors dementia is influenced by a number of environmental factors. A recent review of sixty studies reveals only a limited number of environmental risk factors for dementia, including selenium 9 , an inorganic element whose deficiency is associated with increased risk of developing AD 10,29,30 (see 11 for review). Importantly, our study shows that chronic treatment with sodium selenate, a common natural source of Se, efficiently reverses neuropsychiatric and cognitive symptoms in old 3xTg-AD female mice with advanced amyloid and tau pathologies. Importantly, memory improvement is associated with reduced total/aggregated tau and astrogliosis but unchanged Aβ pathology in the hippocampus. These results suggest for the first time that reducing pathological tau with Se may be useful for the treatment of memory and neuropsychiatric symptoms in AD.
Our results indicate that chronic Se treatment efficiently ameliorates neuropsychiatric-like symptoms of dementia, whose effects contribute significantly to the clinical profile in mild cognitive impairment and AD 31,32 . As previously reported 5 , 3xTg-AD mice display less exploratory behavior, as revealed by reduced locomotor activity, free rearings, sifting behaviors and visiting fewer corners and increased stillness. This behavioral phenotype could reflect a subtle change in idiopathic behaviors, since walling, grooming and chewing are not altered. Interestingly, Se treatment has positive effects on fine motoric properties, and anxiety as evaluated by corner visit behaviors, and does not affect gross motor functions nor working memory. In addition, Se improves nest building, an ethological behavior involving high executive function, and therefore useful to assess progressive loss of executive function seen in AD 33 . Besides, nesting behavior is sensitive to apathy and neophobia 34 , two common neuropsychiatric symptoms in individuals with AD 31,35 . Se also reduces neophobic, anxiety and fear-related behaviors in 3xTg-AD mice, a result consistent with increased locomotor and rearing activities. This result is relevant since emotional symptoms, including anxiety and fear, are early clinical features of mild cognitive impairment and AD 36 , although the cellular mechanisms involved are largely unknown. Importantly, accumulation of intraneuronal Aβ and phosphorylated tau in the basolateral amygdala seems to contribute to these symptoms 6 , whereas Se treatment causes a slight non-significant reduction in tau pathology in this region in 3xTg-AD mice ( Table 1). While tau pathology in the amygdala could be involved in the emotional disturbances observed in AD, understanding the molecular mechanisms by which Se ameliorates AD-like neuropsychiatric symptoms will be relevant to treat the psychiatric and behavioral symptoms of dementia.
AD mouse models develop memory deficits associated with disrupted synaptic function in the hippocampus. We observed impaired spatial learning and memory in old 3xTg-AD mice likely due to use of less directed search navigation strategies 4,6,37 . This is in concordance with studies showing that hippocampus-impaired animals use spatial search strategies less frequently than controls [38][39][40] . Se treatment increases the use of direct spatial strategies   (e.g., direct swimming, focal search…) resulting likely in memory improvement in 3xTg-AD mice. Indeed, sodium selenate and selenomethionine increase spatial learning in the water maze in young AD transgenic mice 20,22,41 , and a multi-nutrient diet containing Se alleviates cognitive deficits in APP/PS1 mice by increasing the use of navigation search strategies 39 . This effect was mediated by reduction of Aβ and amyloid plaques in young APP/PS1 mice, while it was independent of brain amyloid pathology in old mice 39,42 . Importantly, our results indicate that sodium selenate does not affect the number of amyloid plaques nor intracellular Aβ-containing neurons in old 3xTg-AD mice. Differences in the inorganic (e.g. sodium selenate) and organic (e.g. selenomethionine) source and metabolic fate of the distinct Se compounds may result in different and/or additional therapeutic properties. Nonetheless, the hippocampal-dependent memory recovery by Se is reinforced by rescue of associative memory in the passive avoidance task, a similar result than that found in tau transgenic mice 21 . Although the pathological mechanisms underlying Se-mediated hippocampal-dependent memory recovery are still unclear, the fact that Se reduces total and aggregated tau in the hippocampus strongly suggests a tau-mediated mechanism. In 3xTg-AD mice, tau pathology begins in the basolateral amygdala and then extends to the hippocampus and cortex, whereas hippocampal synaptic transmission and plasticity deficits are evident at 6 months of age or even earlier 3,6,43 . Besides that treatment was extended during the progression of pathological and behavioral features (i.e., from 8 to 12-14 months of age), Se reduced aggregated tau-positive neurons in CA1 hippocampus and temporal auditory cortex in 3xTg-AD mice.
Notably, decreasing endogenous tau or reducing tau pathology independently of Aβ ameliorate memory deficits in AD mice 44,45 . By contrast, phosphorylated tau was not globally affected by Se treatment, a result in accordance with unchanged inactive phosphorylated GSK3β levels. Recent investigations show, however, that Se mitigates tau hyperphosphorylation and aggregation by modulating Akt/GSK3β and/or PP2A activities and autophagy pathways [20][21][22][23] .
Our results reinforce the view that tau pathology contributes to AD-like neuropsychiatric and memory deficits, and that reducing aggregated tau without affecting amyloid pathology is beneficial in 3xTg-AD mice. The exact mechanism(s) mediating functional recovery of memory neural circuits by Se is unclear. Age-dependent mitochondrial dysfunction and reduced metal homeostasis and antioxidant defense directly affect synaptic function leading to memory loss, whereas synaptic failure is an early pathological feature that plays an important role in cognitive impairment in AD [46][47][48] . Whether sodium selenate reverts hippocampal-dependent memory deficits by affecting synaptic function remains unclear. One possibility is that seleno-derivatives, acting as antioxidants may affect synaptic transmission and/or plasticity 49 . Indeed, 3xTg-AD mice develop neuronal oxidative stress caused by reduced glutathione and increased reactive oxygen species 50 , whereas selenomethionine increases glutathione levels in these mice 20 . Moreover, disrupted oxidative stress and energy metabolism affect the production and accumulation of Aβ and tau pathologies, which can exacerbate mitochondrial and synapse dysfunction 46 . It is tentative to speculate that Se improves synaptic function leading to memory and psychiatric recovery by reducing directly or indirectly oxidant and tau toxic species. Since tau is also involved in synaptic pathology in AD 51 , investigation of synaptic mechanisms regulated by Se may be important for developing future therapeutic strategies.
In summary, chronic Se supplementation efficiently ameliorates neuropsychiatric-like and memory symptoms and reduces tau aggregation at advanced AD pathological stages. Considering that Se and antioxidant compounds progressively decrease in brain and plasma in mild cognitive impairment and AD patients 52-54 , our results may have important diagnostic and clinical implications. In this context, a signature of at least six essential inorganic elements in serum that includes Se was proposed recently to monitor AD progression 55 . Furthermore, a sodium selenate diet is safe, tolerated and improves neuroimaging measures in mild-moderate AD patients 14 . Nutritional supplementation with Se may represent a promising therapeutic strategy to ameliorate and/or reverse neuropsychiatric and cognitive symptoms in dementia.

Materials and Methods
Experimental design and transgenic mice. Triple-transgenic AD mice (3xTg-AD; C57BL/6 × 129 background) harboring the human PS1 (M146V), APPSwe (K670N/M671L) and tau (P301L) transgenes and wild-type (WT) mice 3 , were bred at the Universitat Autònoma de Barcelona (UAB). All mice were females (3xTg-AD = 24 and WT = 21) housed in groups of 3-6 per cage in standard laboratory conditions, including a 12 h light/dark cycle and ad libitum access to water and food pellets (Brand: ssniff ® R/M-H, containing 0.3 mg Se/Kg; ssniff Spezialdiäten GmbH, Germany). We treated 8 month-old mice with vehicle (3xTg-AD, n = 9; WT, n = 11) or 12 μg/ml sodium selenate (Se; Sigma-Aldrich/Merck, Darmstadt, Germany) (3xTg-AD, n = 15, one died during testing; WT, n = 10) added to the drinking water ad libitum, and continued throughout behavioral testing. Chronic treatment with Se-Met (6 μg/ml) in drinking water was shown previously to elevate Se in the cortex and hippocampus (1.2-1.5 mg/ kg) indicating incorporation of Se in the brain 20 . The mice were approximately 12 months old at behavioral testing, and 13-14 months old at physiological, biochemical and pathological analyses.
Behavioral experiments. Behavioral testing successively included the following tests (in this order): rotarod, ethogram, nesting behavior, corner index, dark/light box, Morris water maze and passive avoidance. Motor coordination and equilibrium were tested on an accelerating rotarod (MED Associates Inc., St. Albans, Vermont, USA). Mice were first trained on a constant speed (4 rpm, 2 min), before starting with four test trials (inter-trial interval, 10 min). During these trials, mice had to balance on the rotating rod that accelerated from 4 to 40 rpm within 5 min. The latency until the mice fell from the rotating rod was recorded, up to a maximum of 5 min 56,57 . Ethological assessments were carried out using a rapid time-sampling behavioral checklist technique, as described previously 58 . Each mouse was observed individually for 5 s periods at 1 min intervals over 15 consecutive min, using an ethologically based behavioral checklist (occurring alone or in any combination) that included: sniffing, locomotion, rearing seated, rearing to wall, rearing free, sifting, grooming, intense grooming, chewing and stillness. This cycle of assessment was repeated twice over an initial exploratory period of 60 min, total counts for each behavior were analysed. Nesting behavior was evaluated by introducing a piece of paper tissue inside the home cage. The presence and quality of the nest were rated two days later on a 1 to 5 point scale 34  shredded but often no identifiable site, 4 identifiable but flat, 5 perfect or nearly. Pictures were taken prior to evaluation for documentation. Two experimenters scored the nest blind and independently of each other. Neophobia was measured using the corner index recorded in standard transparent home-cages. In a 30 s period, the number of visited corners, as well as the rearings in the corners were counted as described 5 . Anxiety behavior was examined in the dark/light box (0,50 m × 0,50 m) divided into same size dark and light sides. The mouse was placed in the dark side of the box for 10 sec acclimatization. A door separating the two compartments was opened and the amount of time that the mouse took to emerge fully (all four paws) into the open area, total time in the light area, and number of transitions were measured for 5 min. Hippocampus-dependent spatial memory abilities were examined in the hidden-platform version of the water maze as described 59 . Time required to locate the hidden platform (escape latency), distance traveled (path length), and swimming speed were recorded using the Ethovision video tracking system (Noldus Information Technology, Wageningen, The Netherlands) as described 56,60 . Swim paths for each mouse in each trial were plotted and categorized into eight mutually exclusive categories: thigmotaxis, chaining, random search, scanning, directed search, focal incorrect, focal search and direct swimming to the platform, of which the latter three are considered true spatial strategies. Two days following the acquisition phase, a 100 sec probe trial, in which the platform was removed, was performed. We visualized swimming paths using a custom-made MATLAB protocol. Single-trial passive avoidance learning was examined in a grid-floor apparatus (MED Associates Inc., St. Albans, Vermont, USA) as described 40,60 . Briefly, mice were dark adapted for 30 min, and then placed in the small illuminated compartment. The sliding door to the dark compartment was opened after 5 sec, and the entry latency was recorded. When mice entered the dark compartment a foot shock (0.3 mA, 1 sec) was delivered. Retention was tested 24 h later without delivering the shock. Immunohistochemical analysis. For immunohistochemical analysis, vehicle-and Se-treated WT and 3xTg-AD mice (n = 6-7/group) were perfused with PBS followed by 4% paraformaldehyde. Left brain hemispheres were dehydrated in ethanol before paraffin embedding. Coronal brain sections (5 μm) were deparaffinized in xylene, rehydrated and incubated with 3% hydrogen peroxide in methanol. For Aβ staining, sections were treated with 60% formic acid (6 min), a protocol that allows specific labeling of Aβ over full-length APP in APP transgenic mice as described 6 . For tau and GFAP staining, sections were pretreated with citrate buffer and microwave heated (10 min) to allow antigen retrieval. Sections were washed in TBS, incubated overnight with antibodies against human Aβ1-16 (6E10; 1:1,000; Covance), hyperphosphorylated tau (PHF1, 1:50; CP13, 1:50), conformational aggregated tau (MC1, 1:20) or GFAP (1:200), and processed for avidin-biotin immunoperoxidase staining using the Vectastain Elite ABC kit (Vector Laboratories, Burlingame, CA) 6 . Sections corresponding to bregma −2, −2.5, −3, −4 and −5 mm (n = 5/mouse) were stained and imaged (10×) using a Nikon Eclipse 90i microscope (Nikon Instruments Europe, Amsterdam, The Netherlands). We quantified GFAP-stained area and the number of dense, sparse and total amyloid plaques, intraneuronal Aβ-positive neurons and phosphorylated tau-positive neurons in multiple regions/ fields (680 × 840 μm) of distinct sections (≥4) of 3xTg-AD Veh (n = 6) and 3xTg-AD Se (n = 7) mice using ImageJ.

Statistical analysis.
Behavioral comparisons were conducted with one-way analysis of variance (ANOVA) followed by Bonferroni post-hoc test. Morris water maze results were analyzed with two-way ANOVA for repeated measures (RM), followed by Bonferroni post-hoc test by using SigmaStat software (Systat Software, San Jose, CA). Statistical analyses of pathological and biochemical studies were performed using t-test (immunohistochemistry) or two-way ANOVA followed by Bonferroni post-hoc test for multiple comparisons (biochemical studies) by using Prism software (GraphPad, La Jolla, CA). *P < 0.05, **P < 0.01, ***P < 0.0001.
Ethical experimental statement. This study was performed in accordance with the experimental European Data availability statement. All data generated during this study are included in this published article or are available from the corresponding author upon reasonable request.