Astaxanthin enhances autophagy, amyloid beta clearance and exerts anti-inflammatory effects in in vitro models of Alzheimer’s disease-related blood brain barrier dysfunction and inflammation

Defective degradation and clearance of amyloid-β as well as inflammation per se are crucial players in the pathology of Alzheimer's disease (AD). A defective transport across the blood-brain barrier is causative for amyloid-β (Aβ) accumulation in the brain, provoking amyloid plaque formation. Using primary porcine brain capillary endothelial cells and murine organotypic hippocampal slice cultures as in vitro models of AD, we investigated the effects of the antioxidant astaxanthin (ASX) on Aβ clearance and neuroinflammation. We report that ASX enhanced the clearance of misfolded proteins in primary porcine brain capillary endothelial cells by inducing autophagy and altered the Aβ processing pathway. We observed a reduction in the expression levels of intracellular and secreted amyloid precursor protein/Aβ accompanied by an increase in ABC transporters ABCA1, ABCG1 as well as low density lipoprotein receptor-related protein 1 mRNA levels. Furthermore, ASX treatment increased autophagic flux as evidenced by increased lipidation of LC3B-II as well as reduced protein expression of phosphorylated S6 ribosomal protein and mTOR. In LPS-stimulated brain slices, ASX exerted anti-inflammatory effects by reducing the secretion of inflammatory cytokines while shifting microglia polarization from M1 to M2 phenotype. Our data suggest ASX as potential therapeutic compound ameliorating AD-related blood brain barrier impairment and inflammation.


Introduction
Alzheimer's disease (AD) is a multi-factorial degenerative disease of the brain with an estimated prevalence of 1 in 9 at the age of 65 and older. The global population burden is projected to triple by 2050 (Rajan et al., 2021;Alzheimer's Dement, 2022;GBD 2019Dementia Forecasting Collaborators, 2022. Aggregation and misfolding of the amyloid-β (Querfurth and LaFerla 2010) and tau proteins (Ballatore et al., 2007;Ittner and Gotz 2011), a leaky blood-brain barrier (BBB) (Zlokovic,2005(Zlokovic, , 2010de la Torre, 2010;Marchesi 2011), synaptic degeneration (Terry et al.,1991, Selkoe, 2002, neuroinflammation (McGeer, 2001(McGeer, , 2002a(McGeer, , 2002bMcGeer and Rogers, 1992), and neuronal cell death (Niikura et al.,2006) are crucial players in AD pathophysiology. Risk factors like obesity (Tsai et al., 2019) and stress, (Machado et al., 2014) but mainly age contribute to the development of this neurodegenerative disorder (van der Flier and Scheltens, 2005). Three main hypotheses namely, cholinergic hypothesis, Aβ cascade hypothesis, tau protein hypothesis have been reported for AD but the exact pathogenesis of the disease still remains unknown. The cholinergic hypothesis states that a possible cause of AD is a loss of central cholinergic neurons and ensuing deficiency of acetylcholine, a neurotransmitter involved in memory and learning (Francis et al., 1999). The tau hypothesis proposed that AD may result from abnormal aggregation (excessive or abnormal phosphorylation) of tau proteins that leads to the formation of tangles within nerves cells in the brain (Mudher and Lovestone, 2002;Iqbal and Grundke, 2008), the amyloid cascade hypothesis postulates that AD may be caused by accumulation of abnormally folded β-amyloid proteins generated by the proteolytic cleavages of APP (Hardy and Selkoe, 2002).
According to the Aβ cascade hypothesis, Aβ is formed by the amyloidogenic cleavage of the amyloid precursor protein (APP) by β-and γ-secretases (Rajendran et al., 2006). Aβ generation is initiated by proteolysis of amyloid precursor protein (APP) by the γ-secretase enzyme BACE1 (Cole and Vassar, 2008). The amyloidogenic peptides are either rerouted to lysosomal compartments for degradation, or secreted into the extracellular space for removal ). An imbalance between production and clearance of Aβ is known to be the initiating factor in AD (Selkoe et al., 2001). Inadequate clearance of protein aggregates that deposit in the brain of proteinopathies, including AD, as well as defects in aggregated and misfolded protein clearance systems are associated with disease pathogenesis (Dakkak et al., 2022). Amyloidogenic Aβ peptides are known to aggregate and oligomerize, and thus prompt misfolded protein-induced endoplasmic reticulum stress (Fonseca et al., 2013). The two key processes, autophagy and proteasomal degradation, are essential for regulating degradation of cell proteins and maintaining proteostasis (Boland et al., 2018;Evrard et al., 2018). Both processes can degrade fully folded as well as misfolded and aggregated proteins to reduce protein stress. Autophagy is a well-known surveillance system that contributes to proteostasis through lysosomal degradation (Bourdenx et al., 2021). Loss of proteostasis is known to increase with age as well as in neurodegenerative diseases such as AD (Ashkavand et al., 2020). Additionally, defective autophagy has been proposed to actively participate in the accumulation of protein aggregates in the elderly as well as in degenerative processes of the brain (Metaxakis et al., 2018).
Apart from autophagy, the inflammatory response is one of the important pathways involved in the cellular response to stress. It has been reported that inflammation can play either beneficial or detrimental roles in response to endogenous or exogenous stressors (Qian et al., 2017). Autophagy is known to regulate inflammation and oxidative stress response (Netea-Maier et al., 2016;Zhong et al., 2016). Direct interactions between autophagy proteins and immune signaling molecules have been reported (Saitoh and Akira, 2010). A defective autophagy machinery often manifests as mis-regulated inflammation in models of human and animal diseases (Deretic and Levine, 2018).
Current treatment strategies of AD such as Acetylcholinesterase inhibitors (AChEI) (Tacrine, galantamine, rivastigmine and donepezil) (Summers et al., 1981;1986, Birks,2006 and Memantine, a low affinity NMDA receptor antagonist (McShare et al., 2006;Schmidt et al., 2015) focus on the increasing cholinergic activity in the brain of AD patients to address the cholinergic hypothesis which, however, only brings about symptomatic relief without addressing the supposed main and underlying causes of AD involving tau proteins and β-amyloid proteins. In addition, anti-tau antibodies have failed in clinical trials (Mullard et al., 2021;Dam et al., 2021;Höglinger et al., 2021;Shulman et al., 2021;Monteiro et al., 2021;Slomski,2022), and two recently approved antiamyloid monoclonal antibodies; Aducanumab and Lecanemab are expensive and have associated side effects (Mintun et al., 2021;Swanson et al., 2021;Lacorte et al., 2022;Cummings, 2023). It is, therefore imperative to look into other potential interventions to devise new experimental drugs and/or repurposing approved drugs targeting tau proteins and β-amyloid proteins that are relatively cheap and easily accessible. Astaxanthin (ASX) is a lipid-soluble keto carotenoid synthesized by several microorganisms and various types of marine organisms (Choi CC, 2019). ASX is one of the most potent antioxidants found in nature (Ambati et al., 2014). It was reported to increase the expression of peroxisome proliferator-activated receptor-alpha (PPARα), to downregulate the peroxisome proliferator-activated receptor-gamma (PPARγ), to stimulate the bile acid synthesis pathway and to inhibit cholesterol biosynthesis (Jia et al., 2012). In addition, ASX exhibits potential pharmacological activities, like immunomodulation, antioxidant, anti-apoptotic, anti-inflammatory, anti-cancerous and antidiabetic properties (Ambati et al., 2014, Zhang et al., 2014a, 2014bZhou et al., 2015). The neuroprotective effects of ASX have been corroborated previously in vitro (Lobos et al., 2016), in vivo (Che et al., 2018, Rahman et al., 2019 and in clinical studies (Ito et al., 2018). Yet, the mechanisms of how ASX exerts its neuroprotective action are poorly understood. Blood brain barrier dysfunction is an early event in the pathophysiology of AD. Moreover, AD is thought to be due to imbalance between amyloid beta production and clearance. Hence, in part, this study aims to investigate the amyloid beta clearance of Astaxanthin at the blood brain barrier even in the presence of exogenous amyloid beta peptides. At the same time, we evaluated its anti-inflammatory function in an in vitro model containing the neurons, astrocytes and microglia cells using two in vitro models of AD. The effects of ASX on autophagy, Aβ clearance and neuroinflammation were evaluated in a BBB model composed of primary porcine brain capillary endothelial cells and in mouse organotypic hippocampal slice cultures (OHSC).

Purity of isolated pBCECs
The isolation of highly purified brain capillary endothelial cells is key for studying the mechanisms of endothelial function. qPCR was carried out to validate the purity of isolated pBCEC. mRNA expression of marker genes such as platelet endothelial cell adhesion molecule (PECAM1 = CD31) for endothelial cells, β-type platelet-derived growth factor receptor (PDGFRβ) for pericytes, glial fibrillary acidic protein (GFAP) for astrocytes, microtubule-associated protein 2 (MAP2) for neurons as well as zonula occludens-1 (ZO-1) for tight junctions were performed (Fig. 1A, 1B). These results confirmed that isolated pBCEC fraction was highly enriched in CD31-positive endothelial cells.

Astaxanthin enhances autophagy in Aβ 1-40 -treated pBCECs
Defective autophagy has been proposed to be one of the dysregulated processes causing accumulation of protein aggregates in the elderly and/ or diseased brain (Metaxakis et al., 2018). To evaluate the effect of ASX on Aβ-induced autophagy, pBCECs were incubated with ASX for 22 h and/or Aβ 1-40 for 6 hr under serum-free conditions. Treatment of pBCECs with ASX alone (p < 0.001) or together with Aβ 1-40 (p = 0.006) increased protein levels of LC3B compared to the vehicle control (VC) ( Fig. 2A, 2B). Similarly, ASX alone (p = 0.003) or together with Aβ 1-40 resulted in a significant increase of LC3B-II as compared to VC (p = 0.01; Fig. 2A, 2C). Increased LC3B-II protein levels are considered to be a marker of autophagy induction (Kabeya et al., 2000, Tanida et al., 2005. Interestingly, treatment of pBCECs with Aβ 1-40 resulted in increased mTOR phosphorylation (p = 0.08) which was significantly reduced with ASX treatment (p = 0.05; Fig. 2A, 2D). Furthermore, activated mTORC1 is known to phosphorylate S6 ribosomal kinase (S6K) on Thr389, followed by activation of S6K to phosphorylate ribosomal protein S6 (S6RP), a component of the 40S ribosomal subunit which is often considered as a readout of mTORC1/S6K activity (Biever et al., 2015). In line, a significant increase in phosphorylation of S6RP protein was observed in Aβ 1-40 -treated pBCECs compared to the VC (p = 0.02). Phosphorylation of S6RP protein was significantly reduced when pBCECs were treated with ASX alone (p = 0.002; Fig. 2A, 2E) and tentatively lower when co-incubated with ASX albeit differences lacked statistical significance.
Incubation of pBCEC with Aβ and ASX increased expression of total LC3B and lipidated LC3B-II while reducing the phosphorylation of mTOR protein. Serum-free Aβ-treated pBCECs were pre-incubated with 10 μM ASX for 16 hr and further incubated with 240 nM Aβ 1-40 for 6 hr.
ASX enhances autophagic flux in Aβ-treated pBCECs in the presence of the lysosomal inhibitor bafilomycin (BAF) A1. Serum-free Aβ-treated pBCECs were pre-incubated with 10 μM ASX for 16 hr further incubated with 240 nM Aβ 1-40 for 6 hr. 100 nM BAF A1 was added during the last 4 hr of the experiment. Western blot showing LC3B in pBCECs treated or co-treated with Aβ 1-40 , ASX and bafilomycin A1 (A). Densitometric analyses of LC3B (B) protein expression levels in pBCECs treated or cotreated with Aβ 1-40 , ASX and bafilomycin A1. Solvent was used as vehicle control (VC). Data are mean + SEM; n = 6-7. p-Values were calculated by one-way ANOVA followed by Dunnett's post hoc test as compared to Aβ 1-40 and vehicle control.

Astaxanthin enhances Aβ clearance in Aβ 1-40 -treated pBCECs
Increased production as well as impaired clearance of Aβ contributes to the development of AD (Wildsmith et al., 2013). To evaluate the effect of ASX on Aβ production and impaired clearance of Aβ, pBCECs were incubated with ASX for 22 h and Aβ 1-40 for 6 h under serum-free conditions. Interestingly, co-incubation of pBCECs with Aβ and ASX significantly reduced the expression levels of amyloid precursor protein (APP)/Aβ compared to VC (p = 0.003) and Aβ (p < 0.001, Fig. 4A, 4B). This was corroborated by apparently lower protein levels of APP/Aβ in the incubation media of Aβ and ASX co-treated cells as compared to that of Aβ treated cells (Fig. 4C, Fig. 4D).
Activation of liver X receptor α (LXRα) increases the expression of apolipoprotein E (ApoE) as well as cholesterol transporters ABCA1 and ABCG1, leading to increased clearance of Aβ (Teemu et al., 2013). Incubation of pBCECs with ASX or co-incubation with Aβ and ASX significantly increased the mRNA expression of ABCA1 compared to VC (p = 0.04 and 0.01, respectively, Fig. 4E). pBCECs treated with ASX significantly increased the expression of ABCG1 mRNA relative to the VC (p = 0.009, Fig. 4F).
Low density lipoprotein receptor-related protein 1 (LRP 1) in the brain endothelial cells is known to be a critical regulator of BBB integrity and function (Storck et al., 2021). In addition, LRP1 has been reported to be important for the rapid removal of Aβ from the brain . In our pBCECs when treated with Aβ, we observed a significant downregulation of LRP1 mRNA compared to the VC (p = 0.04). Coincubation of pBCECs with Aβ and ASX ameliorated mRNA expressions of LRP1 when compared to Aβ control (p = 0.04, Fig. 5G). Taken together, increased expression levels of ABC transporter as well as LRP1 suggest that ASX enhances Aβ clearance in pBCECs.

Astaxanthin enhances Aβ clearance, with PPARα activation being a likely mechanism
ASX has been reported to act as PPARα agonist (Choi CI, 2019). To investigate the mechanism of action of ASX, pBCECs were incubated with ASX, fenofibrate (PPARα activator), GW6417 (PPARα antagonist), and solvent as vehicle control (VC) for 22 hr and Aβ 1-40 for 6 hr under serum-free conditions. Co-incubation of pBCECs with Aβ and either ASX (p = 0.01) or fenofibrate (p = 0.03, Fig. 5A) significantly increased the mRNA expression of ABCA1 relative to VC. Double immunofluorescent labeling of pBCECs co-incubated with Aβ and ASX, fenofibrate or GW6417 showed reduced LRP1 signals in Aβtreated cells and of cells co-incubated with Aβ and GW6417 (Fig. 5B). In addition, reduced LC3B staining were observed in pBCECs co-incubated with Aβ and ASX, fenofibrate, or GW6417 (Fig. 5C). Taken together, these results suggest that ASX enhances Aβ clearance in Aβ-treated pBCECs probably via PPARα activation.

Astaxanthin reduces the secretion of inflammatory cytokines in LPSstimulated organotypic hippocampal brain slices
Dysregulation of the neuronal immune system is emerging as a key player rather than simply as passenger in the pathology of AD. To investigate the anti-inflammatory function of ASX, hippocampal organotypic brain slices were pre-incubated with ASX and then treated with LPS for 24 h. LPS stimulation of brain slices resulted in a significant increase of all cytokines analyzed including TNF-α, IL-6, IL-10, IL-1β and  KC/GRO but not of IL-4 (Fig. 6). Treatment of hippocampal brain slices with LPS in combination with ASX resulted in a significant reduction of the inflammatory cytokines TNF-α (p < 0.001), IL-6 (p < 0.001), IL-10 (p = 0.03), and KC/GRO (p = 0.02; Fig. 6A -C and E), but not of IL-1β and IL-4 ( Fig. 6D and F). As expected, co-treatment with LPS and dexamethasone reduced all investigated cytokine levels even further, except for IL-4 ( Fig. 6A -F). Furthermore, we also determined mRNA levels of the pro-and anti-inflammatory cytokines IL-6 and IL-10, respectively, and observed similar patterns showing that treatment of LPS-stimulated brain slices with ASX resulted in decreased mRNA expression levels for IL-6 (p = 0.03) but not of IL-10 (p = 0.61) (Fig. 6G, H), while dexamethasone treatment almost blunted mRNA levels of IL-6 and IL-10 ( Fig. 6G, H).
Organotypic hippocampal brain slices were incubated with 10 ng/ mL LPS as well as with 10 ng/mL LPS in combination with 50 µM ASX and 10 µM dexamethasone (Dexa) for 24 hr, followed by detection of cytokine release in the supernatant and mRNA expression levels in brain slices by qRT-PCR. Solvent was used as vehicle control (VC). p-Values were calculated by one-way ANOVA with Turkey's post hoc test. Data are mean + SEM, n = 5-10 for Figure

Astaxanthin decreases M1 while increasing M2 polarization in LPSstimulated organotypic hippocampal brain slices
Microglia polarize in two directions producing two different phenotypes. These phenotypes can either be of the pro-inflammatory M1 or anti-inflammatory M2 phenotype (Zhang et al., 2019). Since an organotypic hippocampal slice culture (OHSC) comprises different CNS cell types including glia cells, we investigated the effects of LPS stimulation on glia cell activation. As expected, treatment of brain slices with LPS and co-treatment of LPS and ASX led to an increase in mRNA expression of ITGAM, a marker for activated microglia, as compared to the vehicle control treated brain slices (p = 0.08 and p < 0.001, respectively) (Fig. 7A). Co-treatment with ASX led to an even further increase in ITGAM mRNA expression compared to LPS stimulated brain slices (p = 0.03, Fig. 7A), while dexamethasone downregulated LPS induced ITGAM expression.
ASX induced ITGAM (microglial marker) expression was obviously accompanied by the afore described reduced secretion of inflammatory cytokines (Fig. 6A, 6B and 6D). Thus, together this might indicate a shift from the M1 to the M2 microglia phenotype. To investigate if microglia activation led to an M1 or M2 microglia phenotype, we evaluated the expression levels of Arginase 1 (ARG1), a marker for the M2 microglia phenotype at both mRNA and protein levels. Treatment with LPS or cotreatment with LPS and ASX induced ARG1 mRNA expression levels (p = 0.0013 and p = 0.0001, respectively) (Fig. 7B), while treatment with dexamethasone reduced ARG1 mRNA expression levels comparable to that of VC (p = 0.0081, Fig. 7B). Similarly, also at the protein level, LPS and even more so co-treatment with ASX induced Arg1 protein expression (p = 0.002, Fig. 7C, 7D) compared to the vehicle control, while dexamethasone downregulated LPS induced Arg1 expression (Fig. 7D).
Our results indicate that ASX may likely shift the microglia phenotype from the detrimental M1 type to the protective M2 type in LPS stimulated brain slices.

Discussion
BCECs are integral components of both the BBB and the neurovascular unit (Graves and Baker, 2020). Defective transport across the BBB is an important mediator of Aβ accumulation in the brain and a contributing factor in the pathogenesis of AD (Shackleton et al., 2016). OHSCs represent a physiologically relevant 3D model of the brain. It

Fig. 4. ASX reduces protein expression of APP/Aβ as well as increases expression of genes involved in Aβ clearance in Aβ-treated pBCECs. ASX alters Aβ
processing while enhancing its clearance in Aβ-treated pBCEC. Proteins were extracted from cells and TCA-precipitated from supernatants. Protein levels of Aβ species (normalized to Tubulin levels) were detected by immunoblot using 6E10 as primary antibody. Western blot showing intracellular 6E10 reactive APP/Aβ (A). Densitometric analyses of intracellular 6E10 reactive APP/Aβ (B). Western blot showing secreted 6E10 reactive APP/Aβ in the culture media (C). Densitometric analyses of secreted 6E10 reactive APP/Aβ in the culture media normalized to ponceau (D). Quantitative analyses of ABCA1 (E), ABCG1 (F) and LRP1 (G) mRNA expression levels by qPCR of Aβ-treated pBCECs. Data are mean + SEM, n = 4-8; significant differences between the groups were calculated using two-tailed, unpaired students t-test.
comprises different cell types such as neurons, astrocytes, microglia, and oligodendrocytes of the central nervous system almost in their original architecture, allowing to study neuro-inflammatory and pathological processes (Croft et al., 2019).
In this study, we investigated the protective effects of ASX on Aβ clearance and brain inflammation, common features in proteinopathies like AD by using induced pBCEC and OSHC as AD models.
Aβ clearance from the brain is achieved through various mechanisms including blood-brain clearance, interstitial fluid bulk-flow clearance, the perivascular and paravascular system, autophagic lysosomal degradation, and many more (Marchi et al. 2016). Our study showed that ASX alone or when co-treated with Aβ induced autophagy as evidenced by increased expression of LC3B and LC3B-II in pBCECs, thereby enhancing the clearance of Aβ. This is consistent with other studies by Wani et al., (2000) and Li et al., (2019) where they showed that autophagy induction enhanced Aβ clearance.
Numerous studies have demonstrated an increase of mTOR signaling following Aβ administration (Vander Haar et al., 2007, Ito et al., 2006, Zhang et al., 2009. Further support linking Aβ accumulation with the upregulation of mTOR signaling comes from a study showing that primary neurons exposed to different concentrations of synthetic Aβ oligomers results in increased phosphorylation of mTOR proteins (Bhaskar et al., 2009). In agreement with previous reports by others (Vander Haar et al., 2007, Ito et al., 2006, Zhang et al., 2009, Aβ accumulation in pBCECs upregulated the phosphorylation of S6RP and mTOR proteins indicating elevated mTORC1 activation. Notably, cotreatment of pBCECs with Aβ and ASX reduced the phosphorylation of both proteins probably by inhibiting mTORC1 signaling pathway. Aβ is produced from the APP through the amyloidogenic pathway by β-and γ-secretase (Chun et al., 2020) and APP expression has been shown to reflect the level of Aβ production (Xue et al., 2022). In this study, ASX significantly reduced the expression of intracellular and secreted APP/Aβ species implying decreased Aβ production. Impaired functions of the ABCA and ABCG families of transporters have been implicated in brain degenerative disorders, due to the importance of balanced cholesterol concentration for the maintenance of CNS functions (Pereira et al., 2018). Using a mouse model of AD, Fitz et al. reported that treatment with LXR agonists elevated levels of ApoE and ABCA1, which correlated with cognitive improvements and lowered Aβ deposition (Fitz et al., 2010). In addition, the clearance function of LRP1 in different cells of the brain seems crucial as multiple cell-specific knockout mouse models have shown that LRP1 ablation leads to an accumulation of Aβ in the brain (Van et al., 2019). In the brain endothelium, LRP1 and the receptor for advanced glycation end products (RAGE) facilitate efflux and influx of Aβ from the brain to the periphery and vice versa. Inhibition of LRP1 expression have been shown to cause vascular endothelial damage, significantly reducing Aβ clearance, and increasing the Aβ burden in brain tissue (Jaeger et al., 2009). Gali et al. (2019) reported a reduction in cerebral and cerebrovascular LRP1 levels in brains of 9-month-old 3XTg-AD mice. In our study, we observed a down-regulation of ABCG1 and LRP1 mRNA expression in Aβ-treated pBCECs suggesting increased Aβ production. ASX reversed the Aβmediated mRNA downregulation of ABCG1 and LRP1 while enhancing  Serum-starved Aβ-treated pBCECs were pre-incubated with 10 μM ASX, 1 μM fenofibrate and 1 μM GW6417 for 16 hr further incubated with 240 nM Aβ 1-40 for 6 hr. Quantitative analyses of ABCA1 (A) mRNA expression level by qPCR in Aβ-treated pBCECs. pBCECs were pre-incubated with 10 μM ASX, 1 μM Fenofibrate and 1 μM GW6417 for 16 hr and further incubated in the absence of serum with 240 nM Aβ 1-40 for 6 hr, fixed with 4% PFA, probed with antibodies against 6E10 (green) and LRP1 (red) to visualize Aβ-LRP1interaction (Blue) (B) or against 6E10 (green) and LC3B (red) to visualize Aβ-LC3B interaction (blue) (C). Solvent was used as vehicle control (VC). Data are mean + SEM, n = 5-6. p-Values were calculated by one-way ANOVA followed by Turkey's post hoc test as compared to vehicle control and Aβ 1-40 . ABCA1, ABCG1 and LRP1 expression decreases γ-secretase activity, suppresses Aβ production, and enhances clearance of Aβ from cerebrovascular endothelial cells.
PPARα is a member of the nuclear receptor PPAR family. Natural ligands like fatty acid and synthetic ligands such as hypolipidemic fibrates activate PPARα, resulting in a stimulation of target gene transcription via formation of heterodimer complexes with retinoid × receptor (RXR) (Fruchart et al., 1999). Astaxanthin activates PPARα and antagonizes PPARγ, another member of the PPAR family (Jia et al., 2012;Chai et al., 2019) while fenofibrate is a well-known activator of PPARα (Deplanque et al., 2003;Harano et la.,2006;Villavicencio-Tejo et al, 2021). In co-incubations of pBCECs with Aβ and either ASX or fenofibrate, mRNA expression of ABCA was significantly increased while a decrease was observed for pBCECs co-incubated with Aβ and GW6417, a known antagonist of PPARα. In fact, incubation of pBCECs with Aβ led to a reduced protein expression of ABCA1 which was ameliorated by cotreatment with ASX ( Supplementary Fig. 1). Conversely, treatment of pBCECs with Aβ increased the expression of ABCA1 at mRNA level. This agrees with other reports showing that the relative distribution of ABCA1 mRNA in tissues may significantly diverge from protein expression patterns (Albrecht et al., 2004;Wellington et al., 2002), due to the fact that ABCA1 is regulated at the transcriptional and posttranscriptional level (Schmitz & Orso, 2001). Treatment of pBCECs with Aβ alone or in combination with GW6417 reduced the expression of LRP1. Co-incubation of pBCECs with Aβ and ASX or fenofibrate resulted in increased expression of LRP1, another receptor crucial for Aβ efflux from the brain to the blood.
The levels of pro-inflammatory cytokines in AD patients' serum and post-mortem brains are reported to be elevated (Wang et al., 2015, Stamouli andPolitis, 2016). Pro-inflammatory cytokines like IL-6, TNFα and IL-1β, secreted by the M1 microglia phenotype, induce inflammation (Zhang et al., 2016) whereas the M2 microglia phenotype releases IL-4, Arg1, IL-10 and neurotrophic factors that are antiinflammatory (Turtzo et al., 2014). In the present study we showed that ASX impacts inflammation in hippocampal brain slices. In agreement with previously published reports (Papageorgiou et al., 2016;Delbridge et al., 2020), our data show that stimulation of hippocampal brain slices with LPS induced IL-1β, TNF-α, IL-6, IL-10, KC/GRO, IL-2 and IL-12p70 secretion into the supernatant. Furthermore, we confirmed the anti-inflammatory effect of ASX in LPS-stimulated hippocampal brain slices. This anti-inflammatory effect was mediated by reduced expressions of pro-inflammatory cytokines. Co-treatment of LPS stimulated brain slices with ASX resulted in a reduction of the secretion of most of these cytokines into the supernatant (TNF-α, IL-6, IL-10 and KC/GRO) and reduced mRNA expression of IL-6.
The pro-inflammatory M1 type (killer cells) and the anti- inflammatory M2 type (repair type cells) microglia functions have been reported to be regulated by cytokines and microbial-derived products including LPS (Orihuela et al., 2016). To determine whether the antiinflammatory activities of ASX are due to M2 microglia polarization, we evaluated the expression of ITGAM and ARG1. We showed increased mRNA expression of ITGAM, a marker for activated macrophages and an increase in the mRNA and protein expression of ARG1, a marker for M2 microglia polarization after co-treatment of LPS-stimulated brain slices with ASX. These results suggest that ASX shifts microglia polarization from M1 type to the M2 type which is consistent with a similar report by Wen et al., (2017).
Although we could demonstrate that ASX enhances both autophagy and Aβ clearance while abating inflammation, a major limitation of our study is the use of in vitro model systems. Further studies are needed to validate our results using preclinical in vivo models and investigate potential interventions to address the involvement of tau proteins in AD patients.  Fig. 7. Microglial marker expression in lipopolysaccharide-stimulated brain slices. Organotypic hippocampal brain slices were incubated with 10 ng/mL LPS as well as with 10 ng/mL LPS in combination with 50 µM ASX and 10 µM dexamethasone (Dexa) for 24hr followed by qRT-PCR mRNA quantification of ITGAM (A) and arginase 1 (B) and immunoblotting for ARG1 protein relative to GAPDH (C and D). Solvent was used as vehicle control (VC). p-Values were calculated by oneway ANOVA with Turkey's post hoc test. Data are shown as mean + SEM, n = 5-8.

Conclusion
Our study demonstrates that co-incubation of ASX with Aβ induces autophagy, reduces Aβ production and enhances Aβ clearance in pBCECs. Our results could have important implications for Aß production and clearance at the BBB. Furthermore, ASX's anti-inflammatory potentials impact inflammation by reducing the secretion of proinflammatory cytokines and activating anti-inflammatory proteins. Thus, ASX is an interesting potential therapeutic candidate, which might be relevant due to its ability to ameliorate impaired Aβ clearance, inflammation and eventually other pathophysiological processes associated with AD.

Isolation and culture of primary porcine brain capillary endothelial cells
Primary porcine brain capillary endothelial cells (pBCEC) were isolated from 3 pooled hemispheres of freshly slaughtered pigs (~6 months old, male and female) according to the protocol previously described by Franke and colleagues with minor modifications (Chirackal et al., 2014;Franke et al., 2000) from 3 pooled hemispheres of freshly slaughtered pigs (~6 months old, male and female). Porcine brains were obtained from the local slaughterhouse and meninges and blood vessels were removed. The gray and white matter of the brain cortex were then chopped using a cutter with rolling plates. To isolate the capillaries, dispase, a protease (70 mg/brain, Life Technologies, USA) was dissolved in 40 mL of Preparation Medium (Earle's medium M199 1X, 1% P/S, 1% gentamycin and 1 mM L-glutamine), chopped brains added and incubated for 1 h in the water bath at 37 • C with gentle stirring. After incubation, 150 mL dextran (VWR, USA) solution (200 g dextran, 2.4 g NaHCO 3 and 109.1 mL MEM (10x) were dissolved in 1.2 L of d2H 2 O, stirred overnight at 4 • C to dissolve, density was adjusted to 1.0612 g/l) was added, mixed, and the suspension was centrifuged (8,000 × g, 10 min, 4 • C). The pellet was resuspended in Medium A (Earle's medium M199 1X, 1% P/S, 1% gentamycin, 1 mM L-glutamine and 10% porcine serum) and the capillaries were isolated by filtering the suspension through a nylon mesh. Then, capillaries were disrupted enzymatically by adding 350 μL collagenase/dispase (10 mg/mL). The suspension was carefully pipetted onto a percoll bi-phase gradient (15 mL of 1.07 g/mL percoll solution on the bottom, 20 mL of 1.03 g/mL percoll solution on top) and centrifuged (1,300 × g, 10 min, RT) in a swinging bucket rotor. Endothelial cells were aspirated from the interphase and washed once with Medium A. Cells were plated onto collagen-coated 75 cm2 cell culture flasks in Medium A and incubated at 37 • C in humified air containing 5% CO2. After 24 h of incubation, pBCEC were washed twice with 1x PBS and cultured in Medium B (Earle's medium M199 1X, 1% P/ S, 1 mM L-glutamine and 10% porcine serum) until reaching confluency. After 3 days, confluent pBCEC were trypsinized using 0.5% trypsin solution and split onto collagen-coated 6-well (60 μg/mL collagen) or chamber slides (120 μg/mL collagen) and incubated for 3 days in Medium B. Media were changed to serum-free on the day of treatment and pBCEC were incubated with ASX (10 μM), GW6417 (1 μM) and fenofibrate (1 μM) for 16 h prior to Aβ 1-40 (240 nM) for another 6 h  or bafilomycin A1 (100 nM) for 4 h. Media were removed, cells were washed twice with cold 1x PBS and lysed in either protein lysis buffer (50 μL/well) or TRIzol (500 µL/well) for protein and RNA isolation, respectively.

Culture and treatment of organotypic hippocampal slice cultures (OHSC)
Organotypic hippocampal slice cultures were done according to the protocol previously described by Croft and others (Croft et al., 2019). Briefly, P9/P10 mouse pups were decapitated, skin and skull gently removed, and brains immersed in slicing medium (Opti-MEM 1, 20 µM glucose). Brains were hemisected and the hippocampus was isolated. The hippocampus was placed on the cutting disc of a McIlwain Tissue Chopper (Cavey Laboratory Engineering Co, Surrey, UK). Hippocampal slices of 300 µm thickness were chopped transversely. Nine slices per hippocampus were placed onto porous (0.4 µm), transparent membrane inserts (30 mm in diameter) and incubated for 1 h on ice in HBSS medium containing 10 mM glucose. Afterwards inserts were transferred to fresh 6-well plates containing 1.2 mL culture medium (50% MEM/EBSS, 25% horse serum, 25 %CMF-HBSS, 25 mM glucose, 0.5% pen/strep). Slices were maintained at 37 • C and 5% CO 2 . The full medium was changed between days in vitro (DIV) 3 and 5. On DIV8, culture medium was replaced with serum-reduced treatment medium (MEM/EBSS, 5% horse serum, 25 mM glucose) and slices were maintained in this medium for the remaining treatment period. After changing to serum-reduced medium, ASX (Sigma-Aldrich; 7542-45-2; 50 μM) and the control compound dexamethasone (Sigma-Aldrich; L6529; 10 μM) were added 2 h before LPS stimulation (Sigma-Aldrich; L6529; 10 ng/mL in medium). Cells treated with vehicle and cells treated with LPS alone served as controls. Twenty-four hours after LPS stimulation, cell supernatants were collected for cytokine measurements. Slices were lysed with 500 µL of TRIzol for RNA and protein isolation.

SDS-PAGE and immunoblotting
For isolation of secreted proteins, pBCEC medium was collected and stored at − 80 • C. Proteins were precipitated by adding 30% (v/v) trichloroacetic acid. After 1 h incubation on ice, the suspension was centrifuged at 10,000 × g for 10 min at 4 • C. The pellet was washed twice with ice-cooled acetone, centrifuged at 10,000 × g for 10 min at 4 • C, and dissolved in 1x sample buffer (Biorad) containing 1x reducing agent .
Protein lysates obtained from pBCEC and murine organotypic hippocampal slices were subjected to immunoblotting. Lysates were mixed with sample buffer, followed by denaturation of proteins for 5 min at 95 • C in a thermocycler. Equal amounts of protein (10--20 μg) were loaded onto Criterion XT pre-cast gel (Bio-Rad, California, USA) and subjected to SDS-PAGE under reducing conditions. Proteins were electrophoretically transferred to 0.20 μm nitrocellulose membranes (Biorad). 5% non-fat dry milk (Carl Roth, Karlsruhe, Germany) in Trisbuffered saline containing 0.05% Tween 20 (TBST) was used for blocking for 1 h. Membranes were probed with primary antibodies diluted in TBST containing 5% milk powder. Membranes were further incubated in Clarity Western ECL Substrate (Bio-Rad) and the chemiluminescent signals was detected using a ChemiDoc imager (Bio-Rad). ImageLab software (version 5.2.1, Bio-Rad) was applied for quantification of signals. All primary and secondary antibodies used are listed in Supplementary Table 1.

RNA isolation & quantitative real-time PCR (qPCR)
For the quantification of mRNA levels, RNA was extracted from cells using TriReagent RT (Molecular Research Centre, Cat.No. RT111) according protocol recommended by manufacturer. Cells were washed with PBS and lysed cells incubated with 500 µL/well of TRIzol for 5 min at RT. 100 µL of chloroform was added, the mixture shaken for 15 sec and incubated for 3 to 10 min at RT. Each tube was centrifuged for 5 min at 12,000×g and RT. Thereafter, the aqueous phase was pipetted into a 1.5 mL tube. 100 µL of 2-propanol was added to the aqueous phase and incubated for 10 min on ice. The solution was centrifuged again for 10 min at 12,000×g and 4 • C. The supernatant was discarded after centrifugation and the pellet was washed twice with 500 µL 75% ethanol. The pellet was air-dried and resuspended in aqua bidest. The amount and quality of total extracted RNA was evaluated by UV-VIS spectrometry using a NanoDrop 1,000 spectrophotometer. One µg RNA of each sample was reverse transcribed using High-Capacity Reverse Transcriptase Kit (Life Technologies) following the manufacturer's protocol. Quantification of mRNA expression levels was performed using SYBR Green Master Mix (Biorad) on a CFX96 Real time system (Biorad) and the housekeeping genes tubulin and hypoxanthine guanine phosphoribosyl transferase (HPRT). The difference in Ct values of genes of interest were normalized to HPRT or tubulin (ΔCt). Sequences of primers used are listed in Supplementary Table 2.

Cytokine measurement
Levels of cytokines were measured in supernatants of LPS-stimulated hippocampal brain slices. Cytokines were measured by an immunosorbent assay (V-PLEX Proinflammatory Panel 1 Mouse Kit, Cat.No K15048D, Mesoscale Discovery) according to the instructions of the manufacturer and evaluated in comparison to calibration curves provided with the kit. Results are presented as pg per mL.

Immune-fluorescent staining (IF)
Immunocytochemistry was done according to the protocol previously described by Zandl-lang and colleagues (Zandl-Lang et al., 2018), pBCEC were cultured on Lab-Tek chamber slides (Thermo Fisher Scientific, NY, USA). Cells were fixed with 4% PFA and air dried. Cells were rinsed and blocked with donkey serum prior to primary antibodies incubation for 1 h. Negative controls were incubated with the appropriate IgG fractions as isotope controls. After washing with TBST, secondary antibodies were applied. DAPI was used to visualize nuclei. Slides were rinsed with TBST before mounting with Vectashield mounting medium (Vector Lab, Inc., Burlingame, CA, USA). To acquire computerized images of sections and cells, a Leica DM4000 B microscope (Leica Cambridge Ltd.) equipped with Leica DFC 320 Video camera (Leica Cambridge Ltd.) was used. All primary and secondary antibodies used are listed in Supplementary Table 3.

Statistical analysis
All statistical analyses were carried out using the Prism 10.0 (GraphPad Software Inc, USA). Two tailed un-paired t-test was used to compare between groups in Figs. 2 and 4 while One-way ANOVA followed by Turkey's Multiple Comparison test was used in Fig. 3,5, 6 and 7.

Declaration of Competing Interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: I. I., D.L., T.L., I.S., S.F., M.P. and B.H.P. are presently or formerly affiliated with QPS Austria..

Data availability
Data will be made available on request.

Acknowledgement
Much gratitude to colleagues at Division of Immunology and Pathophysiology, Otto Loewi Research Center, Medical University of Graz and QPS Austria GmbH for their unflinching support.

Funding
This work was funded by the Austrian Science Fund (FWF) in the doctoral program Metabolic and Cardiovascular Disease (DK-MCD W1226) at the Medical University of Graz with support from QPS Austria GmbH.