Dietary Sargassum fusiforme improves memory and reduces amyloid plaque load in an Alzheimer’s disease mouse model

Activation of liver X receptors (LXRs) by synthetic agonists was found to improve cognition in Alzheimer’s disease (AD) mice. However, these LXR agonists induce hypertriglyceridemia and hepatic steatosis, hampering their use in the clinic. We hypothesized that phytosterols as LXR agonists enhance cognition in AD without affecting plasma and hepatic triglycerides. Phytosterols previously reported to activate LXRs were tested in a luciferase-based LXR reporter assay. Using this assay, we found that phytosterols commonly present in a Western type diet in physiological concentrations do not activate LXRs. However, a lipid extract of the 24(S)-Saringosterol-containing seaweed Sargassum fusiforme did potently activate LXRβ. Dietary supplementation of crude Sargassum fusiforme or a Sargassum fusiforme-derived lipid extract to AD mice significantly improved short-term memory and reduced hippocampal Aβ plaque load by 81%. Notably, none of the side effects typically induced by full synthetic LXR agonists were observed. In contrast, administration of the synthetic LXRα activator, AZ876, did not improve cognition and resulted in the accumulation of lipid droplets in the liver. Administration of Sargassum fusiforme-derived 24(S)-Saringosterol to cultured neurons reduced the secretion of Aβ42. Moreover, conditioned medium from 24(S)-Saringosterol-treated astrocytes added to microglia increased phagocytosis of Aβ. Our data show that Sargassum fusiforme improves cognition and alleviates AD pathology. This may be explained at least partly by 24(S)-Saringosterol-mediated LXRβ activation.

www.nature.com/scientificreports www.nature.com/scientificreports/ or LXRβ. In contrast, a lipid extract of the edible brown seaweed Sargassum fusiforme which contains large amounts of 24(S)-Saringosterol potently activated LXRβ. In a mouse model for AD, dietary supplementation with Sargassum fusiforme or its lipid extract not only increased the expression of LXR response genes in the CNS, but also improved cognition without inducing hepatic steatosis. Improved memory performance in these mice was paralleled by a strong reduction in CNS Aβ plaque load. In contrast, the selective LXRα agonist AZ876 did not counteract cognitive decline in AD mice or reduce Aβ plaque load, and induced liver steatosis. These findings indicate that Sargassum fusiforme is an attractive option for add-on treatments in the emerging field of nutritional neuroscience.
Results phytosterols present in a Western diet do not activate LXRs. First, we determined the capacity of phytosterols typically present in a Western diet to activate LXRs. To define cellular specificity in LXR activation in vitro, cells derived from peripheral tissues (HEK293.T and COS7) and the CNS (CHME3, MO3.13, N2a/APPswe) were tested. Physiologically relevant concentrations of stigmasterol, fucosterol, brassicasterol, β-sitosterol, or a mix of phytosterols did not activate LXRα or LXRβ in any of the cell lines used (Fig. S2). An increased incubation period promoted the capacity of the synthetic LXR agonist T0901317, but not of phytosterols, to activate LXRα and LXRβ (Fig. S3a,b). Although cellular uptake of phytosterols and their capacity to activate nuclear receptors improves when complexed to proteins 45,46 , pre-incubation of phytosterols with BSA did not significantly increase their ability to activate LXRα and LXRβ (Fig. S3c,d).

An extract of Sargassum fusiforme activates LXRβ in vitro.
In addition to phytosterols present in a Western diet, crude extracts of Eastern plants were screened for their capacity to activate LXRα or LXRβ. Plants were chosen based on their application in Asian traditional medicine and their presence in an Eastern diet. Extracts from Asparagus racemosus, Azadirachta indica, Cassia fistula, Curcuma aromatica, Datura metel, Piper retrofractum, Senna tora, and Terminalia chebula did not significantly activate either LXRα or LXRβ in microglial CHME3 cells (Fig. 1). However, at a dose of 5 µg/ml an extract of Sargassum fusiforme containing the recently identified LXRβ agonist 24(S)-Saringosterol 34 significantly activated LXRβ but not LXRα (Fig. 1). Higher doses of Sargassum fusiforme induced cell death (data not shown). As Sargassum fusiforme showed the highest capacity to activate LXRβ, it was selected for further in vivo testing.
Dietary supplementation with Sargassum fusiforme results in 24(S)-Saringosterol accumulation in the cerebellum and activation of LXR-response genes in AD mice. To assess in vivo effects  www.nature.com/scientificreports www.nature.com/scientificreports/ of Sargassum fusiforme APPswePS1ΔE9 mice were used as a model of AD 47 . These mice begin to develop Aβ plaques at 4 months of age and cognitive decline occurs from 6 months onwards 48,49 . From the age of 5 months APPswePS1ΔE9 mice and WT littermates were fed either standard chow or chow supplemented with 50% (w/w) dried crude Sargassum fusiforme. Ten weeks of Sargassum fusiforme dietary supplementation resulted in LXR activation in the CNS, evidenced by a cerebral induction of LXR response genes (Abcg1, Scd1) (Fig. 2a,b). ApoE, Srebp-1c, and Abcg1 expression was not altered in animals treated with Sargassum fusiforme (Fig. 2c-e). Interestingly, 24(S)/(R)-Saringosterol was detectable in serum and in the cerebellum of animals that were fed Sargassum fusiforme, but not in those fed normal chow (Fig. 3a,b). As Sargassum fusiforme contains predominantly 24(S)-Saringosterol 34 , we postulate that this isoforms is the most abundant isoform present in animals fed Sargassum fusiforme. Animals fed a diet supplemented with Sargassum fusiforme did not show differences in total cholesterol content in the cerebellum (Fig. 3c), but showed a marked reduction in circulating cholesterol levels compared to chow fed animals (Fig. 3d). Levels of cholesterol precursors, cholesterol metabolites, and phytosterols were decreased in serum and cerebellum of animals supplemented with Sargassum fusiforme (Fig. 3e-j). On a standard chow diet, WT mice displayed significantly higher serum cholesterol levels than AD mice (Fig. 3d), as described previously 10,50 .
Sargassum fusiforme improves memory and reduced Aβ plaque load in AD mice. Next, we investigated the impact of dietary supplementation with crude Sargassum fusiforme on cognition and neuropathology in APPswePS1ΔE9 mice. To assess the influence of selective LXRα activation on cognition, animals were fed with the synthetic selective LXRα agonist AZ876. The open field test revealed no effect of genotype or diet on general locomotor activity, and anxiety (Table 1). Dietary supplementation with Sargassum fusiforme significantly improved object memory in APPswePS1ΔE9 mice in the object recognition task (ORT) at a 1 h but not a 24 h inter-trial interval (Fig. 4a,b). In contrast, selective LXRα activation did not impact object memory at either the 1 h or 24 h inter-trial interval. Sargassum fusiforme, but not AZ786, markedly decreased Aβ plaque load in the cortex (70% decrease; Fig. 4c,e) and hippocampus (81% decrease; Fig. 4d,e) in APPswePS1ΔE9 mice. In concordance, Sargassum fusiforme supplementation reduced Aβ40 protein and APP mRNA expression in the CNS of AD animals (Fig. 4f,h). Sargassum fusiforme did not significantly reduce insoluble Aβ42 (Fig. 4g). Reduced Aβ plaque load was not associated with an increased expression of phagocytic receptors such as Axl and MerTK (Fig. S4a,b) 51 . Yet, the reduced Aβ plaque load was associated with a reduced TREM2 expression, reflecting the expelled need for further Aβ clearance (Fig. S4c).
To determine whether the lipid moiety of Sargassum fusiforme was sufficient to modulate the AD-related pathology, a purified lipid extract of Sargassum fusiforme was administered daily by oral gavage to APPswePS1ΔE9 and WT littermates for 45 days. Working memory was significantly improved in APPswePS1ΔE9 mice, resulting in the prevention of the spatial working deficit (Fig. 5a,b). In line with these findings, a significant reduction in insoluble Aβ 42 was observed in the cortex (99% decrease; Fig. 5c) and hippocampus (57% decrease; Fig. 5d) of the treated APPswePS1ΔE9 mice. Notably, treatment with the Sargassum fusiforme lipid extract resulted an increased expression of ApoE in the CNS ( Fig. 5e-g), which indicates active LXR signaling.
Sargassum fusiforme feeding does not lead to liver steatosis. While synthetic LXR agonists are well-known to induce liver steatosis in mice 52 , Sargassum fusiforme supplementation did not induce liver steatosis in APPswePS1ΔE9 mice, as evidenced by the absence of lipid droplets within the liver of these animals (Fig. 6a,b). Supplementation with AZ876 resulted in a marked increase in lipid droplets in the liver (Fig. 6a,b). Likewise, while AZ876 increased circulating triglyceride levels, Sargassum fusiforme did not impact the level of triglycerides (Fig. 6c). Similar findings were observed in WT mice fed chow supplemented with Sargassum fusiforme or AZ876 (data not shown). Sargassum fusiforme is known to contain relatively high levels of (in)organic arsenic that can accumulate in the body 53 . Although relatively high levels of arsenic were measured in dried Sargassum fusiforme (40.4 mg/kg), no arsenic accumulation was detectable in lung tissue of the animals that consumed Sargassum fusiforme (Fig. 6d).

Discussion
Our data show that dietary supplementation with the brown seaweed Sargassum fusiforme improves cognitive function and reduces Aβ load in APPswePS1ΔE9 mice without inducing liver steatosis. Our data further suggest the involvement of the stereospecific oxyphytosterol 24(S)-Saringosterol, a selective LXRβ activator, in the observed effects on cognition and Aβ plaque load. Considering that selective LXRα activation did not improve cognitive decline or Aβ plaque deposition in APPswePS1ΔE9 mice, our findings suggest that natural and possibly also synthetic LXRβ agonists are attractive options for the treatment of neurodegenerative disorders such as AD.
Phytosterols, which can cross the blood-brain barrier, have been reported to activate LXRs without inducing hypertriglyceridemia and hepatic steatosis, rendering them potentially suitable candidates for treatment of neurodegenerative disorders 11 . Yet, in contrast with reported data on LXR activation by phytosterols common in a Western diet [34][35][36][37][38][39] , we observed very little if any effect of stigmasterol, brassicasterol, β-sitosterol, fucosterol, or a commercial mix of phytosterols, on activation of LXRα and LXRβ in vitro. The discrepancy between our data and the literature may be explained by differences in the experimental design, such as phytosterol concentration, incubation time, and cell type and assay used to measure LXR activation. For instance, we used a cell-based luciferase reporter assay and CNS-derived cell lines, as well as HEK293.T and COS-7 cells, to define LXR activation and cellular specificity in LXR activation, while other studies used cell-free assays 38 . Furthermore, we used phytosterol concentrations in a physiological range that can be reached through dietary supplementation (1-25 µM), whereas other studies used concentrations far exceeding this physiological range up to 200 µM 36 . Of note, compared to synthetic LXR agonists, the reported ability of phytosterols commonly present in a Western diet to activate LXRs is relatively limited [34][35][36][37][38][39] . In line with this, no alterations were observed in the expression of LXR-response genes in the brain of ABCG5-knockout mice despite high cerebral levels of common phytosterols 26 . In summary, our data indicate that phytosterols present in a Western type diet do not activate LXRs at concentrations that can be reached through dietary supplementation 11 .
Out of the tested extracts of Asian plants, the extract of Sargassum fusiforme most potently activated LXRβ in vitro. This is in line with the recently reported LXRβ-activating effects of Sargassum fusiforme-derived 24(S)-Saringosterol 34 . In Sargassum fusiforme-fed animals, 24(S)-Saringosterol was detected in the cerebellum in concentrations equal to common oxysterols, suggesting a good CNS bioavailability. The amount of 24(S)-Saringosterol detected in the CNS was comparable to previously reported concentrations of other dietary phytosterols in the CNS 26,32,33 . With respect to these findings, we found that cerebral expression of LXR-response genes was upregulated in animals fed Sargassum fusiforme-enriched chow, further supporting the ability of Sargassum fusiforme to activate LXRβ. The route of entry of 24(S)-Saringosterol to the CNS remains unclear. Further research is warranted to define the role of ApoE and SR-B1 in facilitating the transport of 24(S)-Saringosterol into the CNS 11,32 . Moreover, since 24(S)-Saringosterol is oxydated plant sterol, we hypothesize that 24(S)-Saringosterol can enter the CNS via transcellular diffusion, comparable to other oxidized cholesterol derivatives such as 24(S)-Hydroxycholesterol 54 . Interestingly, in contrast to 24(S)-Saringosterol, all plant sterol concentrations decreased in serum and CNS upon feeding Sargassum fusiforme. It can be hypothesized that 24(S)-Saringosterol activates intestinal and hepatic Abcg5/g8 transporters and thereby drives the enterohepatic secretion of plant sterols, thereby lowering circulating levels of sterols 55 . The latter hypothesis is in line with the strong reduction in circulating cholesterol levels we found in this study.
Our data indicate that the LXRα selective agonist AZ876 does not significantly counteract cognitive decline or reduce Aβ plaque load in APPswePS1ΔE9 mice. However, no active LXR signaling was found in the CNS of AZ876 treated animals, which suggest poor CNS bioavailability of AZ876 and consequently does not allow us to conclude that LXRα activation does not impact AD pathology. On the other hand, while both LXRα and LXRβ are expressed in the murine CNS, expression of LXRβ is 15-fold higher compared to LXRα 56,57 . Hence, the relative contribution of LXRα activation upon full-LXR agonist treatment is therefore expected to be limited. The latter studies may suggest that the beneficial impact of pan-LXR agonists on cognitive decline in AD models likely relies on LXRβ activation.
Synthetic LXR agonists that reduce AD pathology and improve cognitive performance induce hypertriglyceridemia and liver steatosis 2,13,14,24,25,[27][28][29][30] . Sargassum fusiforme supplementation counteracted cognitive decline of APPswePS1ΔE9 transgenic mice without inducing hepatic steatosis. This may be explained by selective hepatic LXRβ activation, since animals fed AZ786 (2 mg/kg body weight) displayed a marked increase of liver lipid droplets, indicative of liver steatosis. Alternatively, the presence of other protective compounds present in Sargassum fusiforme such as fucosterol may prevent hepatic steatosis 11 . This is supported by the observation that activation of LXRβ by synthetic agonists was also reported to induce plasma and hepatic hypertriglyceridemia 58 .  www.nature.com/scientificreports www.nature.com/scientificreports/ We show that dietary supplementation with crude or lipid extract of Sargassum fusiforme reduces Aβ load in AD mice. Our in vitro data indicate that factors released by 24(S)-Saringosterol-treated astrocytes promote microglial clearance of Aβ1-42. Within the brain, astrocytes are the major source of the LXR response gene Apoe. ApoE is well-known to increase microglial clearance of Aβ 59 , possibly via inhibiting cellular uptake of Aβ 60 . Of interest, 24(S)-Saringosterol and a lipid extract of Sargassum fusiforme increased ApoE protein expression in astrocytes and the CNS respectively, which suggests a role for ApoE in the observed increase in microglial Aβ clearance. In addition, we provide evidence that 24(S)-Saringosterol decreases the release of Aβ by neurons in vitro. Together, these findings indicate that 24(S)-Saringosterol impacts both the clearance and generation of Aβ 42 . These findings are of particular interest, as beneficial effects of LXR agonists on cognition are not necessarily accompanied by a reduced Aβ plaque load 14,41 . Recently, it has been reported that other constituents of Sargassum fusiforme, such as fucosterol, fucoidan, and fucoxantin improve learning and memory deficiencies in pharmacological models for cognitive impairment [61][62][63] . Although synergism among constituents of Sargassum fusiforme is likely, future studies should determine whether 24(S)-Saringosterol is sufficient to improve cognitive performance and AD pathology.
Although relatively high levels of arsenic were measured in extracts of Sargassum fusiforme, no arsenic was detectable in lung tissue, which generally represents a major reservoir for this element 64 . Arsenic present in Sargassum fusiforme may therefore be rapidly excreted, degraded or absorbed in low amounts in the intestine.
In summary, our data indicate that phytosterols present in a Western type diet do not activate LXRs at concentrations that can be reached through dietary supplementation. However, a crude lipid extract of the edible seaweed Sargassum fusiforme did selectively activate LXRβ in vitro. Here, we show for the first time that dietary supplementation with Sargassum fusiforme significantly improves memory performance and reduces Aβ plaque load in a well-established AD model. The accumulation of the natural LXRβ agonist 24(S)-Saringosterol in the CNS suggests its involvement in the neuroprotective impact of Sargassum fusiforme. Collectively, our findings point to 24(S)-Saringosterol-containing Sargassum fusiforme as being a novel candidate for dietary supplementation to prevent or modulate neurodegenerative disorders such as Alzheimer's disease. Luciferase-based nuclear receptor reporter assay. To determine the ability of plant extracts and phytosterols to bind LXRα and LXRβ, a luciferase-based reporter assay was performed using the ONE-Glo TM Luciferase Assay System kit (Promega, Leiden, the Netherlands), according to manufacturer's instructions. Cell lines were transfected with bacterial plasmid constructs expressing luciferase under control of the promotor region of the ligand-binding domain for LXRα or LXRβ 68 . Cells were grown to 50-60% confluency in 60 mm plates, transfected with 1.8 μg of plasmid DNA including 0.2 μg pGAL4hLXRα or pGAL4hLXRβ, 1 μg pG5-TK-GL3, and 0.6 μg of pCMV-β-galactosidase, using JetPEI (Polyplus-transfection SA, Illkirch, France) as transfection reagent. Following treatment, cells were lysed in lysis buffer (25 mM Glycyl-Glycine, 15 mM MgSO 4 , 4 mM EGTA, and 1x Triton; all from Sigma). To correct for transfection efficacy, β-galactosidase activity was measured using lysate diluted 1:10 in B-gal buffer, consisting of 20% 2-Nitrophenyl β-D-galactopyranoside (ONGP; Sigma) and 80% effect in AD genotype: p < 0.0001). Representative IHC staining of all groups is shown (e). (f,g) Sargassum fusiforme treated APPswePS1ΔE9 mice show a significant decrease in Aβ 40 ((f); U = 6, n ctrl = 11, n extract = 5, p = 0.0133, Mann-Whitney) but not Aβ 42 levels (g). (h) Sargassum fusiforme treated APPswePS1ΔE9 mice show a significant decrease in the mRNA expression of APP (U = 0, n ctrl = 8, n extract = 4, p = 0.0040, Mann-Whitney).
The second batch of animals was divided into two groups and received either the Sargassum fusiforme-derived lipid extract (APPswePS1ΔE9 n = 9; C57Bl6/J n = 8), or the vehicle (APPswePS1ΔE9 n = 11; C57Bl6/J n = 12) from 6 until 7.5 months of age. Lipids of Sargassum fusiforme were extracted using the Folch-method 65 . Briefly, dried Sargassum fusiforme was pulverized and soaked in a 2:1 (v/v) chloroform:methanol (VWR, Leuven, www.nature.com/scientificreports www.nature.com/scientificreports/ Belgium) mixture for 16 hours at room temperature. Next, the mixture was sonicated (10 min), filtered, and subsequently evaporated by heating the extract until 60 °C. The remaining extract was washed in ethanol, sonicated (10 min), and evaporated. Finally, the extract was dissolved in corn oil (Vita D' or, the Netherlands) as vehicle and sonicated prior to in vivo supplementation. Using gas chromatography/mass spectrometry, 24(S)/ (R)-Saringosterol content was measured to be 332 mg/dl (7,744 mM). The extract was administered by daily gavage (200 µl/25 g body weight) for a period of 45 days, corresponding to a daily lipid intake of 3.5 g seaweed/day (664 µg 24(S)/(R)-Saringosterol/day). In a previous study we defined that Sargassum fusiforme contains predominantly 24(S)-Saringosterol 34 . Hence, we expect that this isoforms is the most abundant isoform present in our in vivo experiments.
One week prior to the start of the behavioral experiments all animals were housed individually. Animals were kept at an inverse 12 h light/12 h dark cycle, and behavioral experiments were performed during the dark phase of the cycle. Cognitive performance was scored blind. All animal procedures were performed in accordance with institutional guidelines and approved by the ethical committee for animal experiments of Hasselt University.
Behavioral tasks. The object recognition task (ORT) was conducted after 9 weeks of Sargassum fusiforme supplementation when the mice were ~8 months old, as described previously 70,71 . In brief, all animals were habituated to the arena over 2 days, in one trial of 4 min per day. In each habituation trial, two different objects were placed symmetrically in the center of the arena, about 10 cm from the wall. Four different objects were available and at the end of the habituation period each animal had encountered all four objects once. After one resting day, the experiment started. During the first trial (T1) two identical objects (samples) were placed in the arena. After exploring the samples for 4 min, the animal was placed back in its home cage. Subsequently, after a predetermined delay interval (1 or 24 hours), the second trial of 4 minutes (T2) was performed using one familiar object from trial 1 and one new object. Exploration time for each object during T1 and T2 was recorded manually using a personal computer. Sitting on the object or biting in the object was not considered exploratory behavior. Discrimination index d2 in T2 (d2 = [(exploration time for the novel object) − (exploration time for the familiar object)]/(total exploration time in T2)) was calculated as measure for object memory.
The Y-maze spontaneous alternation test 72 was conducted at baseline and after 6 weeks of extract treatment. The maze used for the Y-maze spatial alteration task consisted of three arms of grey Perspec of equal length (40 × 10 × 15 cm) (labeled A, B and C), separated at an angle of 120°. The animals were free to explore the maze for 6 minutes. Afterwards, the working memory is calculated in order of percentage of alternations defined as: (number of triads/(total number of entries −2)) * 100. The number of triads was recorded manually and was defined as subsequently entering the three different arms, while entering an arm was defined as placing both hind paws in that arm. After each trial, the maze was cleaned with 70% ethanol to prevent olfactory cues. www.nature.com/scientificreports www.nature.com/scientificreports/ The open field test (OF) was conducted as described previously 73 , after 9.5 weeks of Sargassum fusiforme treatment. In short, animals were placed in a square, Perspex box (50 cm × 50 cm × 30 cm), with an open top, grey walls, and a white floor. Video tracking with a computerized system (Etho Vision TM , Noldus, Wageningen, The Netherlands) was used to record and analyse movements and position of the animals, in order to determine time in zone (TIZ) and distance moved (DM) over the 10-minute trial 14 . tissue sample preparation. Mice were anaesthetized with Nembutal (100 mg/kg i.p., CEVA Logistics, Brussels, Belgium) prior to transcardial perfusion with Ringer's solution. Blood was collected and centrifuged for 10 min at 200 g to obtain serum, which was snap-frozen in liquid nitrogen. Liver samples were divided into two parts and either directly snap-frozen or frozen in Tissue-Tek O.C.T embedding compound (Sakura Finetek, Berchem, Belgium). Brains were divided into three parts prior to snap-freezing; the cerebellum and two hemispheres. The right hemisphere was preserved for immunohistochemistry and frozen in Tissue-Tek O.C.T. embedding compound. The rostral half of the left hemisphere was preserved for mRNA expression analyses with quantitative PCR. The cerebellum was preserved for sterol analyses.
Aβ quantification. The right hemisphere of the crude Sargassum fusiforme-treated mice was cut in the coronal plane with a Leica CM1900UV cryostat (Leica Microsystems, Wetzlar, Germany) to obtain 10 μm sections. Sections were mounted on glass slides, air-dried overnight, and stored at room temperature until used. After blocking of endogenous peroxidases by incubation for 30 min with 3% H 2 O 2 in methanol, the sections were incubated for 30 min with blocking solution (5% bovine serum albumin in 1x TBS) followed by overnight 4 °C primary antibody incubation in blocking solution (clone 3D6, Amyloid-β Antibody 1:8000). The sections were then incubated in www.nature.com/scientificreports www.nature.com/scientificreports/ biotinylated goat anti-human IgG (BA-3000, 1:1000, Vector Laboratories, Burlingame, CA), and the immunoreactivity was developed by the ABC avidin-biotin peroxidase KIT (PK-6100, Vector) for 1 hour, and 5 min with diaminobenzidine (DAB, ImmPACT SK-4105, Vector). Between every other step, the slides were rinsed (3×) with TBS/0.3% Triton-X 100. The sections were counterstained with haematoxylin for 30 seconds and coverslipped. Digital images of the sections were obtained using a Leica DM 2000 LED microscope (Leica Microsystems, Diegem, Belgium) equipped with hardware and software from Leica Application suite software (Leica Microsystems). The Aβ plaque load was quantified in 3 sections per brain in 5 animals per group using Fiji ImageJ, by defining the pixel intensity of Aβ plaques in the total cortical or hippocampal area at Bregma −1.5 to −2.5.
Next, left hippocampus and cortex were dissected from the brains of the Sargassum fusiforme extract-treated APPswePS1ΔE9 mice and homogenated in a 2% SDS TBS-T buffer. Aβ 42 levels were quantified and related to the protein content in the starting homogenate using an Aβ 42 ELISA (Invitrogen, USA), according to the manufacturer's instructions 74 . In brief, the homogenates were sonicated twice and mixture centrifuged to generate an SDS-insoluble fraction (21000 g, 10 min). The SDS-insoluble pellet was sonicated twice in 70% formic acid (FA), yielding the FA-extracted insoluble Aβ 42 fraction.
Determination of astrocytic Apoe secretion. To generate 24(S)-Saringosterol-astrocyte conditioned medium, mouse primary astrocytes were seeded in a 24-well plate with a density of 250000 cells/well. Next, astrocytes were treated with either 10 µM purified 24(S)-Saringosterol or ethanol for 2 consecutive days after which medium was replenished and collected 2 days afterwards. Next, ApoE secretion was determined using western blot. Equal amounts of cellular supernatant (40 µl) were loaded and separated via electrophoresis through a 10% sodium dodecyl sulfate polyacrylamide gel (SDS-PAGE). Afterwards, proteins were transferred onto a PVDF membrane (VWR, Belgium), and blots were blocked (5% non-fat dry milk TBS-Tween solution) for 1 h at room temperature. Next, membranes were incubated with rabbit anti-ApoE primary antibody (ABBIOTEC, USA). After washing the membranes with TBS-Tween, the blots were incubated with swine anti-rabbit HRP conjugated antibody (WAKO, Japan). Finally, membranes were washed in TBS and protein bands were detected using Pierce TM ECL Plus Western Blotting Substrated and ImageQuant TM LAS 4000 mini. ImageJ (http://imagej.nih. gov/ij/) was used to quantify the protein bands.
Aβ phagocytosis assay. Primary murine cells were isolated and cultured as described previously 75 . In short, mixed glial cultures were prepared from postnatal d0 mouse cerebral cortices of C57Bl6/J mice (Harlan Netherlands B.V.) and cultured in high glucose DMEM medium supplemented with 10% FCS and 100 U pen/100 µg strep/ml. Mixed glial cultures were used to generate microglia-enriched glial cultures by separating the microglia from the astrocyte monolayer using orbital shaking followed by purification via differential adhesion to plastic after 14 days in culture. Purified microglia were seeded on poly-L-lysine (5 µg/ml; Sigma-Aldrich) coated 96-well plates with a density of 50000 cells/well. Primary microglia were cultured in high glucose DMEM medium supplemented with 10% FCS, 100 U pen/100 µg strep/ml and 15% L929 conditioned medium.  www.nature.com/scientificreports www.nature.com/scientificreports/ The capacity of primary microglia to phagocytose FAM-labeled Aβ 1−42 (Eurogentec, Seraing Belgium) was assessed with a plate-based assay. Twenty hours after seeding the microglia, cells were treated with 24(S)-Saringosterol-astrocyte conditioned medium (40%) and FAM-labeled Aβ 1−42 was added for 1 h at a final concentration of 500 nM. Next, medium was removed and 100 µl 0.2% trypan blue in PBS (pH 4.4) was added to quench extracellular Aβ. After aspiration, fluorescence was measured at 485/535 nm excitation/emission. Finally, to normalize for cell number, 100 µl DAPI in PBS was added, incubated for 10 min and medium was aspirated. Fluorescence was subsequently measured in a plate reader using 360 nm/465 nm excitation/emission wavelengths.
Sterol profile and triglyceride content determination. Sterol profiles were determined in serum and cerebellum. Preceding the analysis, samples were spun in a speed vacuum dryer (Savant AES 1000) at 12 mbar for 24 h to relate individual sterol concentrations to their dry weight. The sterols were extracted from the dried tissue by placing them in a mixture of chloroform:methanol (2:1) for 24 h at 4 °C. Sterol levels were determined by gas chromatography/mass spectrometry (GC/MS) as described previously 32,76 . Triglyceride contents of serum were determined using the GOD-PAP method with enzymatic reagent kits according to manufacturer's instructions (DiaSys Diagnostic Systems, Holzheim, Germany).
Determination of arsenic concentration. Arsenic halogenide concentrations in lung tissues were analyzed using inductively coupled plasma-optical emission spectrometry (ICP-OES, Agilent Technologies, 700 Series, Belgium). Tissues were oven-dried and digested with 70-71% HNO 3 in a heat block. Prior to analysis, lung samples and dried Sargassum fusiforme were diluted 1:10 in 1% nitric acid. Concentrations were measured using a standard calibration curve (NIST Spinach 1570a).
RNA isolation and Rt-Q-pCR. Brain tissue was homogenized in Qiazol (Qiagen, Venlo, The Netherlands) preceding RNA isolation, and total RNA was prepared using the RNeasy mini kit (Qiagen), according to manufacturer's instructions. RNA concentration and quality was determined with a NanoDrop spectrophotometer (Isogen Life Science, IJsselstein, The Netherlands). RNA was reverse transcribed to cDNA using the qScript TM cDNA synthesis kit (Quanta Biosciences, Gaithersburg, USA). As previously described 77,78 , quantitative PCR was subsequently conducted on a StepOnePlus detection system (Applied biosystems, Gaasbeek, Belgium). Relative quantification of gene expression was performed using the comparative Ct method. Data were normalized to the most stable reference genes (Cyca and Hmbs). Primers were chosen according to literature or designed using Primer-Express (http://www.ncbi.nlm.nih.gov/tools/primer-blast), and details of primers used are shown in Table 2. experimental design and statistical analyses. All statistical analyses were performed using GraphPad Prism 7 TM and are reported as mean ± SEM. D' Agostino and Pearson omnibus normality test was used to test normal distribution. Unless specified differently, one-way ANOVA (post-hoc: Tukey) or two-tailed unpaired Student t-test was used for normally distributed data sets. The Kruskal-Wallis (Dunns post hoc comparison) or Mann-Whitney analysis was used for data sets assessed not to be normally distributed. The ORT discrimination index d2 (compared to the 0) and spatial alteration Y-maze performance (compared to the 50% chance level) was analyzed using the one-sample t-test. Animals that did not reach the minimum of 4 s exploration in T1 or T2 were excluded from further analysis 71,79 . Data from RT-Q-PCR was analyzed using two-way ANOVA with treatment and genotype as independent variables. In all data sets, extreme values were excluded by means of Dixon's principles of exclusion of extreme values 80,81 . Significance levels are denoted as follows: $ p < 0.10 *p < 0.05, **p < 0.01, ***p < 0.001 or ****p < 0.0001. ethical approval. All animal procedures were performed in accordance with institutional guidelines and approved by the ethical committee for animal experiments of Hasselt University.

Data Availability
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.