lable at ScienceDirect

Neurobiology of Aging xxx (2014) 1e13

Contents lists avai

Neurobiology of Aging

journal homepage: www.elsevier .com/locate/neuaging

Amyloid-b pathology is attenuated by tauroursodeoxycholic acidtreatment in APP/PS1 mice after disease onset

Pedro A. Dionísio a, Joana D. Amaral a,*, Maria F. Ribeiro a, Adrian C. Lo b, Rudi D’Hooge b,Cecília M.P. Rodrigues a,*

aResearch Institute for Medicines (iMed.ULisboa), Faculty of Pharmacy, Universidade de Lisboa, Lisbon, Portugalb Laboratory of Biological Psychology and Leuven Institute of Neuroscience and Disease (LIND), University of Leuven, Leuven, Belgium

a r t i c l e i n f o

Article history:Received 4 June 2014Received in revised form 31 July 2014Accepted 12 August 2014

Keywords:Amyloid-bAmyloid precursor proteinAPP/PS1 miceGlycogen synthase kinase bGliosisTauTUDCA

* Corresponding authors at: Research Institute foFaculty of Pharmacy, Universidade de Lisboa, Lisbon 121 794 6490; fax: þ351 21 794 6491.

E-mail addresses: jamaral@ff.ul.pt (J.D. Amaral), cRodrigues).

0197-4580/$ e see front matter � 2014 Elsevier Inc. Ahttp://dx.doi.org/10.1016/j.neurobiolaging.2014.08.034

a b s t r a c t

Alzheimer’s disease (AD) is a neurodegenerative disorder hallmarked by the accumulation of extracel-lular amyloid-b (Ab) peptide and intraneuronal hyperphosphorylated tau, as well as chronic neuro-inflammation. Tauroursodeoxycholic acid (TUDCA) is an endogenous anti-apoptotic bile acid with potentneuroprotective properties in several experimental models of AD. We have previously reported thetherapeutic efficacy of TUDCA treatment before amyloid plaque deposition in APP/PS1 double-transgenicmice. In the present study, we evaluated the protective effects of TUDCA when administrated after theonset of amyloid pathology. APP/PS1 transgenic mice with 7 months of age were injected intraperito-neally with TUDCA (500 mg/kg) every 3 days for 3 months. TUDCA treatment significantly attenuated Abdeposition in the brain, with a concomitant decrease in Ab1-40 and Ab1-42 levels. The amyloidogenicprocessing of amyloid precursor protein was also reduced, indicating that TUDCA interferes with Abproduction. In addition, TUDCA abrogated GSK3b hyperactivity, which is highly implicated in tauhyperphosphorylation and glial activation. This effect was likely dependent on the specific activation ofthe upstream kinase, Akt. Finally, TUDCA treatment decreased glial activation and reduced proin-flammatory cytokine messenger RNA expression, while partially rescuing synaptic loss. Overall, our re-sults suggest that TUDCA is a promising therapeutic strategy not only for prevention but also fortreatment of AD after disease onset.

� 2014 Elsevier Inc. All rights reserved.

1. Introduction

Alzheimer’s disease (AD) is the most common form of dementia,leading to memory impairment and progressive cognitive decline(Reitz and Mayeux, 2014). AD is pathologically hallmarked by thepresence of extracellular senile plaques composed of amyloid-b(Ab) peptide aggregates, intraneuronal neurofibrillary tangles(NFTs) composed of hyperphosphorylated tau protein, and chronicneuroinflammation (Selkoe, 2001). Senile plaques occur mainly asdeposits of amyloid fibrils surrounded by dystrophic neurites(Selkoe, 2001). Ab peptides, containing 39-43 residues, derivefrom the sequential proteolysis of the amyloid precursor protein(APP), a single-pass transmembrane glycoprotein with a largeextracellular domain (Haass et al., 2012; Selkoe, 2001). The first stepof the amyloidogenic processing of APP is mainly catalyzed by beta-site APP cleaving enzyme 1 (BACE1), the canonical b-secretase,

r Medicines (iMed.ULisboa),649-003, Portugal. Tel.: þ351

mprodrigues@ff.ul.pt (C.M.P.

ll rights reserved.

leading to the shedding of the ectodomain (soluble APP-b fragmentor sAPP-b) and the production of a small membrane-spanning C-terminal fragment b (CTF-b). CTF-b is further cleaved by g-secretasein the transmembrane domain, generating Ab and releasing C-ter-minal fragment g into the cytoplasm (Haass et al., 2012). Mutationsin the genes coding for APP and presenilin 1 (PS1), the catalytic coreof g-secretase, are respectively associated with increased amyloi-dogenic processing of APP and preferential production of longer Abspecies with higher amyloidogenic propensity, such as Ab1-42(Selkoe, 2001). Several of these mutations have beenwidely used inthe establishment of a plethora of mouse models of Ab pathology(Elder et al., 2010). Nevertheless, other AD-associated phenotypesare differentially observed among the available transgenic models,with none of them fully recapitulating the cognitive declinedescribed in AD patients (Webster et al., 2014). In fact, over-expression of APP and/or PS1 using artificial promoters posesseveral drawbacks such as overproduction of APP and its cleavagefragments besides Ab, which also impact on AD-associated phe-notypes (Born et al., 2014; Saito et al., 2014). Furthermore, the vastmajority of mouse models manifest global neuronal death onlyseveral months after most AD-associated features are already firmly

P.A. Dionísio et al. / Neurobiology of Aging xxx (2014) 1e132

established (Kokjohn and Roher, 2009). Still, APP/PS1 double-transgenic mice are among the most successful models, promptlydeveloping Ab deposition and a robust neuroinflammatoryresponse toward Ab plaques, along with synaptic integrity loss, andmemory and cognitive impairments (Nunes et al., 2012; Raddeet al., 2006). Ab aggregation can elicit the development of inflam-matory processes and increased gliosis, culminating in the abun-dant presence of activated microglia and reactive astrocytes in thevicinity of Ab plaques (Selkoe, 2001). During the progression ofAD, sustained glial activation increases the levels of secretedproinflammatory molecules that, in turn, can induce their ownexpression in a positive feedback loop. This mechanism furtherexacerbates inflammation, ultimately contributing to neuro-degeneration and dysregulation of signaling pathways that favorthe amyloidogenic processing of APP and tau hyperphosphorylation(Meraz-Rios et al., 2013).

Glycogen synthase kinase 3 (GSK3) is a partially constitutivelyactive serine/threonine kinase ubiquitously expressed as 2 highlyhomologous isoforms, GSK3a and GSK3b (Soutar et al., 2010).Although both isoforms exert similar functions, they are differen-tially expressed among cell types, with GSK3b being preferentiallyexpressed in the nervous tissue (Soutar et al., 2010). Importantly,GSK3 hyperactivity has been extensively correlated with AD pa-thology. GSK3b is known to phosphorylate tau and APP, whichprobably contributes to NFT formation and increased amyloido-genic processing of APP (Aplin et al., 1996; Durairajan et al., 2012;Lee et al., 2003). GSK3b overactivation is also thought to impairlearning and memory (Zhu et al., 2007). Moreover, GSK3 can acti-vate proapoptotic pathways in neurons exposed to insults usuallyobserved in AD (Mines et al., 2011). Finally, GSK3 is an importantregulator of the immune response, promoting the shift from anti-toproinflammatory phenotypes in several cell types, includingmicroglia and astrocytes (Beurel and Jope, 2010; Koistinaho et al.,2011; Yuskaitis and Jope, 2009). Inhibition of GSK3 in vivo hasalready been proven to ameliorate all major AD-associated pa-thologies (Durairajan et al., 2012; Sereno et al., 2009).

Tauroursodeoxycholic acid (TUDCA) is an endogenous hydro-philic bile acid with strong anti-apoptotic properties in severalexperimental models of neurodegeneration. In vitro studieshave determined that TUDCA inhibits Ab-induced apoptosis bypreventing caspase-2 activation and modulating p53-dependentpathways (Ramalho et al., 2004, 2006; Sola et al., 2006; Vianaet al., 2010). Furthermore, TUDCA activates the prosurvival PI3K/Akt signaling cascade, decreasing Ab-mediated apoptosis in pri-mary rat cortical neurons (Sola et al., 2003). More importantly,TUDCA enters the systemic circulation by oral administration andis able to cross the blood-brain barrier (Keene et al., 2002). In thisrespect, the neuroprotective effects of TUDCA have been demon-strated in several animal models of neurodegenerative diseases,including Huntington (Keene et al., 2001, 2002) and Parkinson’sdiseases (Castro-Caldas et al., 2012; Duan et al., 2002), andischemic and hemorrhagic stroke (Rodrigues et al., 2002, 2003).We have previously reported the therapeutic efficacy of TUDCA inAPP/PS1 double-transgenic mice fed with a diet containing 0.4%TUDCA for 6 months (Nunes et al., 2012). Notably, TUDCA treat-ment significantly decreased Ab levels and deposition in micebrain, markedly ameliorating memory deficits (Lo et al., 2013;Nunes et al., 2012).

Given the promising results of TUDCA treatment before theonset of amyloid pathology, we hypothesized that administration ofTUDCA at later stages is beneficial and ameliorates Ab pathology. Inthis study, we used APP/PS1 mice to evaluate the protective effectsof TUDCA when administrated at 7 months of age. The results ob-tained here further confirm the robust protective effects of TUDCAas a potential therapeutic option for AD.

2. Methods

2.1. Transgenic mice and treatment

APP/PS1 double-transgenic mice, maintained on a C57BL/6J geneticbackground, coexpress KM670/671NL “Swedish” mutated humanAPP695andL166PmutatedhumanPS1under the regulatorycontrol ofa murine Thy-1 minigene promoter, which is a neuron-specific pro-moter that restricts transgene expression to the postnatal period(Radde et al., 2006). Mice were genotyped by polymerase chain reac-tion (PCR) analysis of tail DNA. All animals were housed in standardcages with ad libitum access to food and water in a temperature-controlled environment with a 12-hour light/dark cycle. Male APP/PS1 mice and wild-type littermates were randomly assigned to 4groups: TUDCA-treated and untreated (control) wild-type mice andTUDCA-treated and untreated (control) APP/PS1 mice. At 7 months ofage, animals were injected with either TUDCA (500 mg/kg of bodyweight) or vehicle every 3 days for 3 months. The dosage of TUDCAused was calculated based on previous studies developed in rodents,includingAPP/PS1mice (Keene et al., 2002;Nunes et al., 2012). Animalswith 7 months of age were used to evaluate the protective effectsof TUDCA after the onset of amyloid pathology, because amyloiddeposition in APP/PS1 mice starts at 6 weeks in the neocortex and2e3 months of age in the hippocampus (Radde et al., 2006).

2.2. Morris water maze

Spatial learning was evaluated in the Morris water maze aspreviously described (Goddyn et al., 2006; Lo et al., 2013). Briefly,mice were trained in a total of 10 days (specifically in 2 series of 5training days, with 2 days of rest between each series) to find asubmerged platform. Four trials starting from 4 different startingpositions were performed each day with a trial interval of 30 mi-nutes. When mice failed to find the hidden platform within 2 mi-nutes, they were guided to the platform and were left there for 15seconds, before being returned to their cages. Latency to find thehidden platform was recorded with Ethovision (Noldus Bv, Wage-ningen, the Netherlands). To evaluate retention memory, probetrials were introduced on days 6 and 11. During these probe trials,the platform was removed, and the swimming path was recordedduring 100 seconds. Time spent in each quadrant was measured.

2.3. Immunohistochemistry

Saline-perfused brains were excised, and one hemisphere wassnap frozen for protein extraction. The other hemisphere was fixedin 4% paraformaldehyde for 48 hours and stored in 30% sucrose andphosphate-buffered saline at 4 �C. The treated hemispheres werefurther dehydrated and embedded in paraffin. Sequential coronalbrain sections (4-mm thick) were obtained and mounted onSuperFrost-Plus glass slides (Thermo Scientific, Rockford, IL, USA).For immunostaining, brain sections were deparaffined and rehy-drated, and antigen retrieval was performed by boiling the sectionsfor 20 minutes in 10 mM citrate buffer, pH 6.

The sections were then blocked for 1 hour in Tris-buffered saline(TBS) containing 10% (vol/vol) normal donkey serum (JacksonImmunoResearch Laboratories Inc, West Grove, PA, USA) and 0.1%(vol/vol) Triton X-100 (Sigma-Aldrich) and subsequently incubatedin appropriately diluted primary antibodies overnight at 4 �C. Afterwashing with TBS/0.025% Tween20, the primary antibodies weredeveloped with diluted (1:200) Alexa Fluor 568 (anti-mouse) orAlexa Fluor 594 (anti-rabbit) conjugated secondary antibodies(Invitrogen) for 2 hours at room temperature. After rinsing, thesections were counterstained with Hoechst 33,258 (Sigma-Aldrich)and mounted on Fluoromount (Sigma-Aldrich). The following

Table 1Primer sequences used to amplify indicated mouse cDNAs

Sense primer (50-30) Antisense primer (50-30)

HPRT CAGTCCCAGCGTCGTGATTA TGGCCTCCCATCTCCTTCATBACE1 TCCTTCCTCAGCAATACCTACG GGATGACTGTGAGACAGCGATNF-a GCCTCTTCTCATTCCTGCTTG CTGATGAGAGGGAGGCCATTIL-1b TGCCACCTTTTGACAGTGATG TGATGTGCTGCTGCGAGATTIL-6 GAGGATACCACTCCCAACAGACC AAGTGCATCATCGTTGTTCATACA

Key: BACE1, beta-site APP cleaving enzyme 1; cDNA, complementary DNA;HPRT, hypoxanthine-guanine phosphoribosyltransferase; IL-1b, interleukin-1b;IL-6, interleukin-6; TNF-a, tumor necrosis factor-a.

P.A. Dionísio et al. / Neurobiology of Aging xxx (2014) 1e13 3

primary antibodies were used: Ab deposits were stained with amouse monoclonal anti-Ab antibody (6E10; Covance; 1:1000); as-trocytes were stained with a mouse monoclonal anti-glial fibrillaryacidic protein (GFAP) antibody (GA5; Millipore Corporation,Temecula, CA, USA; 1:200); microglia were stained with a rabbitpolyclonal anti-Iba-I antibody (Wako Pure Chemicals, Richmond,VA, USA; 1:100); p-APP was stained with a rabbit polyclonal anti-p-APP (Thr668) antibody (Cell Signaling; 1:200); APP was stainedwith a rabbit polyclonal anti-APP, C-terminal antibody (Sigma-Aldrich; 1:200); p-tau was stained with a rabbit polyclonal anti-p-tau (Ser396) antibody (101815; Santa Cruz Biotechnology; 1:50);and presynaptic terminals were stained with a mouse monoclonalanti-synaptophysin (SYN) antibody (SY38; Millipore; 1:100).

2.4. Histochemistry

After deparaffinization and rehydration, sections were stained withThioflavin T, a highly sensitive marker of Ab deposits. Staining was per-formed with 0.05% Thioflavin T (Sigma-Aldrich) solution in phosphate-buffered saline for 8 minutes at room temperature (Nunes et al., 2012).

2.5. Image analysis

All images were captured using an Axioskop fluorescence mi-croscope (Carl Zeiss GmbH, Hamburg, Germany). At least 8 imagesper hippocampal and cortical regions were acquired for each animaland converted to gray scale with an 8-bit format. Semiquantitativeanalysis of GFAP, Iba-I, SYN, and p-tau were performed with ImageJversion 1.46r software (National Institute of Health, Bethesda, MD,USA). The background of each set of images was subtracted, and athreshold optical density was determined and held constant. Meangray values obtained for GFAP, Iba-I, SYN, and p-tau immunostainingin the 4 mice groups were normalized to the mean gray values unitsof control wild-type mice and are presented as percentage of wild-type control mice. The number of Ab-plaques stained with Thio-flavin T or 6E10 antibody were counted and presented as plaquenumber per square millimeter for both hippocampus and frontalcortex. Measurement of amyloid (Thioflavin) and Ab (6E10) burdenwere also performed in thresholded images by applying an unbiasedcomputer-assisted image analyzer available in the ImageJ software,whichwas used to quantify the areas occupied by positive staining inthe regions of interest. Amyloid and Ab burden was then calculatedby normalizing the reactive area to the total area of the regions ofinterest and is presented as percentage of the total area.

2.6. Real-time PCR

Dissected hippocampus and frontal cortex were homogenized inTRIzol (Invitrogen) using a motor-driven Bio-vortexer (No1083;Biospec Products, Bartlesfield, OK, USA). After homogenization,total RNAwas isolated according to themanufacturer’s protocol andquantified using a Qubit 2.0 fluorometer (Invitrogen). Total RNAwas

converted into complementary DNA using Superscript II ReverseTranscriptase (Invitrogen) according to the manufacturer’s in-structions. Quantitative real-time PCR analyses were performed inthe 7300 Real-Time PCR System (Applied Biosystems, Foster City,CA, USA), using SYBR Green PCR master mix (Fermentas Interna-tional Inc, Glen Burnie, Maryland, USA). The expression levels of thegenes of interest relative to the housekeeping gene hypoxanthine-guanine phosphoribosyltransferase were calculated using the DDCtmethod. Control wild-type mice were used as the calibrator, andthe relative changes in gene expression were calculated accordingto the formula 2�DDCt. Primer sequences are presented in Table 1.

2.7. Western blot analysis

Dissected hippocampus and frontal cortex were homogenized inTRIzol, and total protein extracts were obtained from TRIzol-chloroform fractions (Simoes et al., 2013). Protein concentrationswere calculated using the Bio-Rad protein assay kit, according to themanufacturer’s recommendations. Equal amounts of protein (60 mg)were electrophoretically resolved on denaturing 8% or 12% poly-acrilamide gels. To evaluate Ab and APP levels with the corre-sponding cleavage products (sAPP-b and CTF-b), 60 mg of totalprotein extracts were electrophoretically separated in 10%e20%TriseTricine gels (Bio-Rad). The resolved proteins were transferredonto nitrocellulose membranes, and blocking was performed with a5% milk solution. Membranes were then incubated with thefollowing primary antibodies: rabbit polyclonal anti-p-APP (Thr668)antibody (3823; Cell Signaling); rabbit polyclonal anti-APP, C-ter-minal antibody (A8717; Sigma-Aldrich); mouse monoclonal anti-Abantibody (6E10; Covance), used to detect both total Ab peptide andCFT-b; mouse monoclonal antibody anti-sAPP-b fragment with theSwedish mutation (sAPP-b; 6A1; Immuno-Biological Laboratories,Inc, Minneapolis, MN, USA); mouse monoclonal anti-GFAP antibody(GA5; Millipore Corporation, Temecula, CA, USA); rabbit polyclonalanti-p-Akt (Ser473) antibody (7985; Santa Cruz Biotechnology);rabbit polyclonal anti-Akt (8312; Santa Cruz Biotechnology); rabbitpolyclonal anti-p-GSK3a/b (Ser21/9) (9331; Cell Signaling); mousemonoclonal anti-GSK3 a/b antibody (7219; Santa Cruz Biotech-nology); rabbit polyclonal anti-p-tau (Ser396) (101815; Santa CruzBiotechnology); and mouse monoclonal anti-tau (tau5; 58,860;Santa Cruz Biotechnology). The membranes were then incubatedwith goat secondary antibodies conjugated with horseradishperoxidase anti-mouse or anti-rabbit (BioRad Laboratories, Hercules,CA, USA) for 2 hours at room temperature. After rinsing with TBS/0.2% Tween 20, the immunoreactive proteins were visualized withImmobilonWestern (Millipore) or SuperSignalWest Femto substrate(Thermo Scientific). GAPDH (32,233; Santa Cruz Biotechnology) or b-actin (AC-15; Sigma-Aldrich) were used as loading controls. Densi-tometric analyses were performed with the Image Lab softwareversion 5.1 Beta (Bio-Rad). For APP cleavage fragments, Akt andp-tau, each sample was analyzed twice to confirm the results.

2.8. Sandwich enzyme-linked immunosorbent assay

Sandwich ELISA kits (Millipore) were used to determine totalAb1-40 and Ab1-42 concentrations from total protein extracts ofhippocampus and frontal cortex obtained from TRIzol-chloroformfractions. TRIzol contains guanidine isothiocyanate, a strong dena-turing agent, which allows for the recovery of total Ab present inboth soluble and insoluble fractions.

2.9. Statistical analysis

Data comparisons were conducted with 1-way analysis of vari-ance (ANOVA) followed by post hoc Bonferroni test. Differences

Fig. 1. TUDCA treatment reduces amyloid plaque pathology in the brains of APP/PS1mice. (A) Representative images of thioflavin staining in the hippocampus of controlwild-type mice and hippocampus and frontal cortex of TUDCA-treated and untreatedAPP/PS1 mice. Scale bar, 100 mm. (B) Quantification of amyloid plaque number persquare millimeter in hippocampus and frontal cortex of TUDCA-treated and controlAPP/PS1 mice. (C) Amyloid plaque burden presented as percentage of positive amyloidstaining area relative to the total area of interest in hippocampus and frontal cortex ofTUDCA-treated and untreated APP/PS1 mice. Values represent mean � SEM of 6e7mice per group. yp < 0.01 and *p < 0.05 from control APP/PS1 mice. Abbreviations:SEM, standard error of the mean; TUDCA, tauroursodeoxycholic acid. (For interpreta-tion of the references to color in this Figure, the reader is referred to the web version ofthis article.)

P.A. Dionísio et al. / Neurobiology of Aging xxx (2014) 1e134

between 2 groups were analyzed by Student 2-tailed unpaired ttest. Behavioral data obtained with the Morris water maze wereanalyzed with 2-way ANOVA for repeated measures (RM), followedby Tukey post hoc test. Analyses and graphical presentation wereperformed with the GraphPad Prism software version 5 (GraphPadSoftware, Inc, San Diego, CA, USA). Results are presented as mean �standard error of the mean.

3. Results

3.1. Short-term treatment with TUDCA attenuates Ab deposition inAPP/PS1 mice

APP/PS1 mice show increased Ab production and parenchymalAb deposition beginning at 6 weeks of age, in the cortex, and at2e3 months, in the hippocampus (Radde et al., 2006). In a previousstudy, we demonstrated that TUDCA treatment started at initialphases of Ab accumulation significantly prevented amyloid plaquedevelopment (Nunes et al., 2012). In the present study, we evaluatedthe effect of TUDCA on Ab deposition when administrated after theonset of amyloid pathology in APP/PS1 mice, at 7 months of age.Thioflavin T staining (Fig. 1A) and 6E10 immunolabeling (Fig. 2A)revealed the presence of extensive amyloid deposits in the brains ofAPP/PS1 mice, which were not observed in wild-type mice. Impor-tantly, amyloid plaque number was significantly decreased in thehippocampus of TUDCA-treated transgenic mice when comparedwith control transgenic mice (p < 0.01), whereas no difference wasobserved in the frontal cortex (Fig. 1B). Ab immunohistochemistryfurther confirmed these results, showing approximately 50%reduction in Ab plaque number in the hippocampus (p < 0.05)(Fig. 2B). Moreover, amyloid burden, as evaluated by Thioflavin T(Fig. 1C) and 6E10 staining (Fig. 2C) was decreased in the hippo-campus and frontal cortex of TUDCA-treated APP/PS1 mice byapproximately 65% and 40%, respectively (p < 0.05). The cortex isthe first brain region affected by Ab deposition in this APP/PS1mouse model (Radde et al., 2006). It has been demonstrated thatcortical formation of newamyloid plaques in thismodel occurs until4e5 months of age and greatly decreases after this period, whereasboth newly formed and existing plaques appear to grow at a similarrate (Hefendehl et al., 2011). Conversely, in the hippocampus, am-yloid plaque formation only starts in the dentate gyrus at2e3 months of age and in CA1 at 4�5 months of age (Radde et al.,2006), which might explain the differences observed in overallplaque burden and number between the frontal cortex and hippo-campus with short-term TUDCA treatment (Figs. 1 and 2). Takentogether, these results demonstrate that TUDCA treatment attenu-ates brain Ab deposition after the onset of amyloid pathology,particularly in the hippocampus of APP/PS1 mice.

3.2. TUDCA reduces amyloidogenic processing of APP and Abgeneration in APP/PS1 mice

We have reported that preventive TUDCA treatment diminishesamyloidogenic processing of APP in APP/PS1 mice, suggesting thatAb reduction by TUDCA is the main mechanism underlying atten-uated amyloid pathology (Nunes et al., 2012). To evaluate the effectof short-term treatment with TUDCA on the amyloidogenic pro-cessing of APP, we analyzed protein levels of APP b-secretasecleavage products, sAPP-b, and CTF-b, by Western blot (Fig. 3).Notably, sAPP-b fragment was significantly decreased both in thehippocampus and frontal cortex (p < 0.05) of TUDCA-treated APP/PS1 mice as compared with control APP/PS1 mice, while full-lengthAPP levels remained unchanged. CTF-b levels were also reduced in

Fig. 2. TUDCA treatment reduces Ab plaque pathology in the brains of APP/PS1mice. (A) Representative images of Ab immunostaining (6E10) in the hippocampusof control wild-type mice and hippocampus and frontal cortex of TUDCA-treatedand untreated APP/PS1 mice. Scale bar, 100 mm. (B) Quantification of Ab plaquenumber per square millimeter in hippocampus and frontal cortex of TUDCA-treatedand control APP/PS1 mice. (C) Amyloid burden presented as percentage of positiveAb immunoreactive area relative to the total area of interest in hippocampus andfrontal cortex of TUDCA-treated and untreated APP/PS1 mice. Values representmean � SEM of 6e7 mice per group. *p < 0.05 from control APP/PS1 mice. Ab-breviations: Ab, amyloid-b; TUDCA, tauroursodeoxycholic acid. (For interpretationof the references to color in this Figure, the reader is referred to the web version ofthis article.)

P.A. Dionísio et al. / Neurobiology of Aging xxx (2014) 1e13 5

the hippocampus (p< 0.01) and frontal cortex (p< 0.05) of TUDCA-treated transgenic mice when compared with controls. In agree-ment, total Abwas also significantly decreased in the hippocampus(p < 0.05) and, to a lesser extent, in the frontal cortex (p < 0.05) ofTUDCA-treated APP/PS1 mice, indicating that short-term treatmentwith TUDCA reduces the amyloidogenic processing of APP even atlater stages of amyloid pathology.

We have also used sandwich ELISA to evaluate Ab1-42 and Ab1-40levels in both brain regions of the 4 mice groups studied. As ex-pected, wild-type mice presented undetectable levels of either Ab1-42 or Ab1-40 (data not shown). Furthermore, we determined thatboth Ab1-42 and Ab1-40 levels were highly elevated in the brains ofAPP/PS1 mice, with Ab1-42 exceeding Ab1-40 by several fold (Fig. 4).Importantly, TUDCA treatment resulted in a significant reduction ofboth Ab species in the hippocampus and frontal cortex of APP/PS1mice, when compared with untreated transgenic mice (p < 0.05).These findings indicate that TUDCA interferes with Ab generationby reducing the amyloidogenic processing of APP after the onset ofthe disease, thus representing a potential alternative for the exist-ing therapeutic options.

3.3. TUDCA supplementation alters Akt/GSK3b activities andprevents tau hyperphosphorylation in APP/PS1 mouse brains

The Akt/GSK3 pathway is known to be deregulated in AD(Durairajan et al., 2012; Lee et al., 2009; Malm et al., 2007; Ryderet al., 2004). The serine/threonine kinase Akt is fully activated byphosphorylation at serine 473, after other downstream effectsmediated by the phosphatidylinositol signaling pathway (Beaulieuet al., 2009). In turn, GSK3b is a substrate of Akt that is inhibited byphosphorylation of serine 9 (Beaulieu et al., 2009). In line withthese findings, we determined the levels of phosphorylated Akt(p-Akt) and GSK3b (p-GSK3b) in brain lysates from TUDCA-treatedand untreated APP/PS1 mice. Consistent with others (Durairajanet al., 2012), phosphorylation of Akt was significantly decreasedin the frontal cortex of control APP/PS1 mice relative to wild-typemice (p < 0.05) (Fig. 5A). Noteworthy, the reduction in p-Aktlevels was reverted in TUDCA-treated transgenic mice (p < 0.05),indicating increased activity of Akt with TUDCA treatment. Unex-pectedly, our results showed no significant differences between Aktphosphorylation levels in the hippocampus of control APP/PS1 andcontrol wild-type mice. In contrast, a trend to increase wasobserved in p-Akt levels of TUDCA-treated wild-type mice relativeto control wild-type littermates, whereas a significant increase wasdetected between TUDCA-treated and control APP/PS1 mice (p <

0.05), indicating that TUDCA can specifically activate Akt in thehippocampus of wild-type mice.

Regarding p-GSK3b levels, they were significantly reduced inboth the hippocampus (p < 0.01) and frontal cortex (p < 0.005) ofcontrol transgenic mice relative to wild-type controls (Fig. 5B).Importantly, TUDCA treatment reverted the reduction observed inp-GSK3b levels in APP/PS1 mice for both brain regions analyzed(p < 0.005). Our findings suggest that TUDCA may abrogate AD-associated GSK3b dysregulation, probably by specifically acti-vating the upstream Akt signaling pathway.

Interestingly, it has recently been reported that specific GSK3binhibition resulted in decreased BACE1 messenger RNA (mRNA)expression and protein levels in an ADmouse model (Ly et al., 2013).Other authors have also described increased levels of BACE1 in ADpatient brains (Yang et al., 2003). Taking this into account, we eval-uated if TUDCA modulation of the amyloidogenic processing of APPwas correlated with the bile acid effect on GSK3b activity and, sub-sequently, BACE1 expression levels. However, this was not the case

Fig. 3. TUDCA treatment decreases the amyloidogenic processing of APP in the brains of APP/PS1 mice, as evaluated by the production of sAPP-b, CTF-b, and Ab. Representativeimmunoblots of hippocampus and frontal cortex from control (n ¼ 5) and TUDCA-treated (n ¼ 5) wild-type mice, and control (n ¼ 6) and TUDCA-treated (n ¼ 7) APP/PS1 mice withthe respective densitometric analyses. b-Actin was used as loading control. Values are expressed as mean � SEM. yp < 0.01 and *p < 0.05 from control APP/PS1 mice. Abbreviations:Ab, amyloid-b; APP, amyloid precursor protein; CTF-b, C-terminal fragment b; sAPP-b, soluble APP-b fragment; SEM, standard error of the mean; TUDCA, tauroursodeoxycholic acid.

P.A. Dionísio et al. / Neurobiology of Aging xxx (2014) 1e136

because no changes were detected in BACE1 levels after TUDCAtreatment, at both mRNA and protein levels (Supplementary Fig. 1).Alternatively, other studies in vitro and in vivo suggest that GSK3b isable to phosphorylate APP at threonine 668 (p-APP), which increasesAb production (Aplin et al., 1996; Durairajan et al., 2012; Lee et al.,2003) Therefore, we hypothesized that the effect of TUDCA onGSK3b activity impacts on APP phosphorylation levels. Immunoblotanalyses revealed that p-APP levels were increased in the frontalcortex of APP/PS1 mice comparing with wild-type mice (p < 0.05);however, theywere unaffected by TUDCA treatment (SupplementaryFig. 2A). We further analyzed APP phosphorylation by immunohis-tochemistry and observed extensive accumulation of p-APP indystrophic neurites closely associated with amyloid plaques,whereas total APP immunoreactivity was detected in both dystro-phic neurites and neuronal perikaria (Supplementary Fig. 2B).However, no changes were detected with TUDCA (SupplementaryFig. 2B). These results are in agreement with previous studies inAD patient brains and in mouse models (Lee et al., 2003) (Shin et al.,2007). Curiously, p-APP immunoreactivity around amyloid plaqueswas more extensively detected in the frontal cortex than in thehippocampus of APP/PS1 mice (Supplementary Fig. 2B).

In respect to tau, it is known that the APP/PS1 mice used in thisstudy present hyperphosphorylated tau-positive neuritic structureslocated in the proximity of amyloid plaques at 8 months of age, withno significant NFT formation (Radde et al., 2006). Because GSK3b isone of the major kinases implicated in tau hyperphosphorylation(Leroy et al., 2010; Li and Paudel, 2006), we evaluated whetherTUDCA-dependent inhibition of GSK3bwas affecting p-tau levels. Tauphosphorylation was determined by immunoblotting and immuno-histochemistry using an antibody against tau phosphorylated atSer396 (corresponding to Ser936 of mouse origin), a residue alreadyidentified as being specifically phosphorylated by GSK3b in theabsence of priming events (Leroy et al., 2010; Li and Paudel, 2006). Asdepicted in Fig. 6A, p-tau levels were increased approximately 2-foldin both hippocampus and frontal cortex (p < 0.05) of APP/PS1 mice

relative to control wild-type mice. In contrast, TUDCA-treatedtransgenic animals presented a strong reduction in p-tau levels inboth brain regions, showing a decrease of approximately 40% relativeto control APP/PS1 mice (p < 0.05). Immunohistochemical analysesfurther revealed the presence of phosphorylated tau associated withdystrophic neurites surrounding amyloid plaques (Fig. 6B). Quanti-fication of phosphorylated tau immunoreactivity further confirmedthese results, supporting the view that the inhibitory effect of TUDCAon GSK3b activity positively impacts on tau hyperphosphorylation.

3.4. TUDCA treatment ameliorates astrocytosis and microgliosis inAPP/PS1 mice

Amyloid deposition leads to extensive microgliosis and astro-cytosis surrounding the affected areas in AD patients and mousemodels (Meraz-Rios et al., 2013; Selkoe, 2001). We have alreadydescribed that TUDCA supplementation mitigates the activation ofglial cells in APP/PS1 mice (Nunes et al., 2012). Interestingly, GSK3bis emerging as a key enzyme involved in the regulation of pathwaysand transcription factors involved in microglial and astrocyticactivation (Beurel and Jope, 2010; Koistinaho et al., 2011; Yuskaitisand Jope, 2009). Because of the multiple lines of evidence showinga TUDCA-dependent effect on both GSK3b phosphorylation levelsand glial activation, we assessed astrocytosis by GFAP immuno-staining and immunoblot analyses, and microgliosis by Iba-I im-munostaining. As expected, GFAP immunofluorescence showed aprominent increase of reactive astrocytes in the hippocampus (p <

0.01) and frontal cortex (p < 0.005) of control APP/PS1 micecompared with control wild-type mice (Fig. 7A and B). Noteworthy,GFAP-reactive astrocytes were significantly reduced in both brainregions of TUDCA-treated APP/PS1 mice compared with APP/PS1controls (p < 0.05). Additionally, GFAP protein levels were alsoevaluated by immunoblot, further confirming the dramatic increasein GFAP levels in the hippocampus (p < 0.005) and frontal cortex(p< 0.001) of control transgenicmice, when comparedwith control

Fig. 4. TUDCA reduces Ab1-42 and Ab1-40 levels in the brains of APP/PS1 mice, asdetermined by ELISA. Quantification of Ab1-42 and Ab1-40 levels in the hippocampusand frontal cortex of control and TUDCA-treated APP/PS1 mice. Data are expressed asmean � SEM of 6e7 mice per group. *p < 0.05 from control APP/PS1 mice. Abbrevi-ations: Ab, amyloid-b; ELISA, enzyme-linked immunosorbent assay; SEM, standarderror of the mean; TUDCA, tauroursodeoxycholic acid.

Fig. 5. TUDCA treatment increases Akt and GSK3b phosphorylation levels in the brainsof APP/PS1 mice. (A) Representative immunoblots of p-Akt (Ser473) and total Akt in thehippocampus and frontal cortex of control and TUDCA-treated wild-type mice andAPP/PS1 mice, and respective densitometric analyses of the p-Akt to Akt ratio. (B)Representative immunoblots of p-GSK3b (Ser9) and total GSK3b in the hippocampusand frontal cortex of control (n ¼ 5) and TUDCA-treated (n ¼ 6) wild-type mice, andcontrol (n ¼ 6) and TUDCA-treated (n ¼ 7) APP/PS1 mice with the respective densi-tometric analyses of the p-GSK3b to GSK3b ratio. Values are expressed as mean � SEM.*p < 0.05, yp < 0.01, and zp< 0.005 from control wild-type mice and xp < 0.05 and {p <

0.01 from control APP/PS1 mice. Abbreviations: SEM, standard error of the mean;TUDCA, tauroursodeoxycholic acid.

P.A. Dionísio et al. / Neurobiology of Aging xxx (2014) 1e13 7

wild-type mice (Fig. 7C). Moreover, our results revealed approxi-mately 30% decrease in GFAP levels in the hippocampus and 45% inthe frontal cortex of TUDCA-treated APP/PS1 mice relative to APP/PS1 controls (p < 0.05), further corroborating the results obtainedfor GFAP immunostaining (Fig. 7C).

Microgliosis was significantly elevated in the hippocampus andfrontal cortex of control transgenic mice relative to wild-type mice(p < 0.005) (Fig. 8A and B). Importantly, TUDCA treatment of APP/PS1 animals decreased Iba-I immunofluorescence in both hippo-campus and frontal cortex (p< 0.01). Because reactive microglia areusually tightly clustered around amyloid deposits, we performeddouble immunohistochemical staining with Iba-I and thioflavin toevaluate the effect of TUDCA in amyloid-dependent microgliosis. Inaccordance with previous reports, clusters of hypertrophic micro-glia were detected in close proximity to Ab plaques in the brains ofcontrol APP/PS1 mice, whereas less reactive microglia were visu-alized surrounding amyloid plaques in the parenchyma of TUDCA-treated transgenic mice (Fig. 8C).

To further characterize the effect of TUDCA on the proin-flammatory response, we analyzed by quantitative real-time PCRthe expression levels of TNF-a, IL-1b, and IL-6, proinflammatorycytokines extensively described as being augmented in AD (Meraz-Rios et al., 2013). As anticipated, the mRNA levels of all 3 cytokines

were significantly elevated in both hippocampus and frontal cortexof control APP/PS1 mice when compared with control wild-typemice (Fig. 9), with TNF-a mRNA in the frontal cortex presentingthe highest increase (approximately 11-fold, p < 0.0001). TNF-amRNA levels were significantly decreased by TUDCA treatment bothin the hippocampus (p<0.05) and in the frontal cortex (p< 0.005) ofAPP/PS1 mice relative to control APP/PS1 mice. IL-1b and IL-6 mRNAlevels showed no significant differences or a trend to decrease be-tween these 2 groups, respectively. These results further indicatethat TUDCA reduces neuroinflammation in APP/PS1 mice.

3.5. TUDCA treatment reduces synaptic loss in APP/PS1 mice

Ab species and extensive gliosis have been widely implicated inthe neurodegenerative processes observed during AD progression.However, transgenic mouse models of AD only manifest globalneuronal loss at very advanced ages, with our APP/PS1 modelpresenting neuronal loss solely in the dentate gyrus of the hippo-campus at 17 months of age (Radde et al., 2006; Wirths and Bayer,2010; Rupp et al., 2011). On the other hand, soluble Ab oligomers

Fig. 7. TUDCA treatment attenuates astrogliosis in the brains of APP/PS1 mice. (A)Representative images of GFAP immunostaining in the hippocampus of controlwild-type mice and hippocampus and frontal cortex of control and TUDCA-treatedAPP/PS1 mice. Scale bar, 25 mm. (B) Quantification of GFAP mean gray values in thehippocampus and frontal cortex of control and TUDCA-treated wild-type mice andAPP/PS1 mice, presented as percentage of control wild-type mice (n ¼ 6e7 miceper group). (C) Representative immunoblots of GFAP in hippocampus and frontalcortex lysates from control (n ¼ 5) and TUDCA-treated (n ¼ 6) wild-type mice, andcontrol (n ¼ 6) and TUDCA-treated (n ¼ 7) APP/PS1 mice with the respectivedensitometric analyses. GAPDH was used as loading control. Values are expressedas mean � SEM. yp < 0.01, zp < 0.005, and xp < 0.001 from control wild-type miceand *p < 0.05 from control APP/PS1 mice. Abbreviations: GFAP, glial fibrillary acidicprotein; SEM, standard error of the mean; TUDCA, tauroursodeoxycholic acid. (Forinterpretation of the references to color in this Figure, the reader is referred to theweb version of this article.)

Fig. 6. TUDCA treatment reduces tau phosphorylation in the brains of APP/PS1 mice.(A) Representative immunoblots of p-tau (Ser396) and total tau in the hippocampusand frontal cortex of control (n ¼ 6) and TUDCA-treated (n ¼ 6) wild-type mice, andcontrol (n ¼ 6) and TUDCA-treated (n ¼ 7) APP/PS1 mice with the respective densi-tometric analyses of the p-tau to tau ratio. (B) Double staining of p-tau (red) andamyloid plaques (thioflavin T, green) in the hippocampus and frontal cortex of controland TUDCA-treated APP/PS1 mice. Scale bar, 25 mm. (C) Quantification of p-tau meangray values in the hippocampus and frontal cortex of control and TUDCA-treated APP/PS1 mice, presented as percentage of control APP/PS1 mice (n ¼ 6e7 mice per group).Values are expressed as mean � SEM. *p < 0.05 from control wild-type mice and yp <

0.05 and xp < 0.01 from control APP/PS1 mice. Abbreviations: SEM, standard error ofthe mean; TUDCA, tauroursodeoxycholic acid. (For interpretation of the references tocolor in this Figure, the reader is referred to the web version of this article.)

P.A. Dionísio et al. / Neurobiology of Aging xxx (2014) 1e138

are widely accepted as the main neurotoxic species, directlycontributing to loss of synaptic integrity and subsequently tomemory impairment and cognitive deficits (Selkoe, 2002). Inter-estingly, several animal models of AD present loss of presynapticterminals, as determined by SYN immunostaining and protein

Fig. 8. TUDCA treatment reduces microgliosis in the brains of APP/PS1 mice. (A)Representative images of Iba-I immunostaining in the hippocampus of control wild-type mice and hippocampus and frontal cortex of control and TUDCA-treated APP/

P.A. Dionísio et al. / Neurobiology of Aging xxx (2014) 1e13 9

levels (Ding et al., 2008; Nunes et al., 2012; Radde et al., 2006). Thisevent starts as early as at 5 months of age in the hippocampaldentate gyrus (Ding et al., 2008), indicating that APP/PS1 micemodels mimic the progression of synaptic dysfunction that isobserved early in patients with very mild to mild AD (Masliah et al.,2001; Selkoe, 2002). The hippocampal dentate gyrus is highlyenriched in synaptic connections between the mossy fibers of thedentate gyrus and the dendrites of the pyramidal neurons;furthermore, the synaptic connections of this brain region arecommonly affected in APP/PS1mice in an age-related manner, evenin regions devoid of amyloid plaques, which is thought to relate tothe development of memory deficits usually observed in thesemodels (Ding et al., 2008; Rutten et al., 2005). Consequently, weproceeded to evaluate synaptic integrity in the dentate gyrus bySYN immunostaining. As anticipated, semiquantitative analysis ofSYN immunostaining revealed an accentuated decrease in SYNimmunoreactivity in the polymorphic layer of the dentate gyrus ofcontrol APP/PS1mice relative to control wild-typemice (p< 0.005),whereas TUDCA treatment partially reverted this phenotype inAPP/PS1 mice (p < 0.05) (Fig. 10). These data indicate that TUDCAadministration even after the onset of AD pathology can stillpartially protect from loss of synaptic function.

APP/PS1 double-transgenic mice undergo memory deteriorationdue to the age-associated progression of Ab pathology (Radde et al.,2006). To evaluate whether TUDCA administration ameliorates mem-orydeficits in agedAPP/PS1mice, all animalswere trained in theMorriswater maze. During training, we measured no significant differencesbetween groups. All groups learned to locate the hidden platformposition over consecutive training days (2-way RM ANOVA: F9, 351 ¼82.96, p< 0.001) and showed similar learning curves (Fig.11A). Duringthe second probe trial, we removed the platform and measured thetime mice spent in each of the different quadrants. We found a sig-nificant quadrant (2-way RM ANOVA: F3, 117 ¼24.59, p < 0.001) andquadrant by group effect (F9,117¼ 2.144, p< 0.05). Post hoc test showsthat both wild-type groups spent a significant amount of time in thetarget quadrant (p< 0.001), whereas such a target quadrant preferenceis lacking in control APP/PS1 mice (Fig. 11B). Nevertheless, a trend toimproved memory was detected, which probably correlates withTUDCA-dependent attenuation of Ab pathology.

4. Discussion

AD is the most prevalent neurodegenerative disorder in agingpopulations worldwide. Available treatments are mostly symptom-atic andunable to arrest or revert theprogression of the disease (Reitzand Mayeux, 2014). In this regard, we have recently described aremarkable rescue of different memory types, as well as improvedsynaptic efficiency in APP/PS1 transgenic mice long-term treatedwith neuroprotective TUDCA (Lo et al., 2013; Nunes et al., 2012;Ramalho et al., 2013). Now, we extended our studies and furtherdemonstrated the therapeutic efficacy of TUDCAwhen administratedat 7 months of age, after the onset of amyloid pathology. The rele-vance of using transgenic mouse models to study AD has raisedseveral concerns because most of the clinical trials using AD thera-peutic agents tested in mice result in negative therapeutic outcomes

PS1 mice. Scale bar, 25 mm. (B) Quantification of Iba-I mean gray values in the hip-pocampus and frontal cortex of control and TUDCA-treated wild-type mice and APP/PS1 mice, presented as percentage of control wild-type mice. Values are expressed asmean � SEM of 6e7 mice per group. yp < 0.005 from control wild-type mice and *p <

0.01 from control APP/PS1 mice. (C) Double staining of Iba-I and amyloid plaques(thioflavin T staining) in the hippocampus of control and TUDCA-treated APP/PS1 mice.Scale bar, 25 mm. Abbreviations: SEM, standard error of the mean; TUDCA, taur-oursodeoxycholic acid. (For interpretation of the references to color in this Figure, thereader is referred to the web version of this article.)

Fig. 10. TUDCA treatment partially rescues loss of presynaptic terminals. (A) Repre-sentative images of SYN immunostaining in the hippocampus of control and TUDCA-treated wild-type mice and APP/PS1 mice. Scale bar, 25 mm. (B) Quantification ofSYN mean gray values in the hippocampus of control and TUDCA-treated wild-typemice and APP/PS1 mice, presented as percentage of control wild-type mice. Values aremean � SEM of 6e7 mice per group. yp < 0.005 from control wild-type mice and *p <

0.05 from control APP/PS1 mice. Abbreviations: SEM, standard error of the mean; SYN,synaptophysin; TUDCA, tauroursodeoxycholic acid. (For interpretation of the refer-ences to color in this Figure, the reader is referred to the web version of this article.)

Fig. 9. TUDCA treatment decreases TNF-a mRNA expression in the hippocampus andfrontal cortex of APP/PS1 mice. Quantification of TNF-a, IL-1b, and IL-6 mRNA levels inthe hippocampus and frontal cortex of control and TUDCA-treated wild-type mice andAPP/PS1 mice. Data are expressed as mean � SEM of 6e7 mice per group as fold changetowards control wild-type mice. *p< 0.01, yp< 0.005, zp< 0.001 from control wild-typemice and xp < 0.05 and {p < 0.005 from control APP/PS1 mice. Abbreviations: mRNA,messenger RNA; SEM, standard error of the mean; TUDCA, tauroursodeoxycholic acid.

P.A. Dionísio et al. / Neurobiology of Aging xxx (2014) 1e1310

(Kokjohn and Roher, 2009). However, animal models, including APP/PS1mice, remain instrumental for dissecting themolecular pathwaysinvolved in AD and evaluating new therapeutic avenues.

Notably, our results suggest that although TUDCA treatmentdoes not completely revert amyloid pathology after its establish-ment, it can still attenuate the progression of Ab deposition, asconfirmed by the diminished production of APP cleavage fragmentsand decrease of total Ab levels in the hippocampus and frontalcortex of APP/PS1 mice.

We have previously shown that TUDCA can activate the PI3K/Aktsignaling pathway, resulting in neuronal protection against Abtoxicity (Sola et al., 2003). Interestingly, the Akt/GSK3 axis isconsistently deregulated in AD. The brains of AD patients demon-strate decreased Akt activity that correlates to GSK3 hyperactivity, afeature widely mimicked in vivo by transgenic mouse models usedto study AD, and in vitro by cell cultures harboring familial ADmutations or exposed to Ab (Durairajan et al., 2012; Jia et al., 2013;Lee et al., 2009; Ryder et al., 2004). GSK3b has also been extensivelyassociated with tau hyperphosphorylation (Leroy et al., 2010; Li andPaudel, 2006).

We showed that Akt activity was reduced in the frontal cortex ofAPP/PS1 mice relative to control littermates although no significantchanges were observed in the hippocampus. However, GSK3bwas hyperactivated in both brain regions, which indicates thathippocampal GSK3b dysregulation is probably independent of theupstream Akt pathway. Although unexpected, the absence of dif-ferences in Akt phosphorylation levels in the hippocampus, with aconcomitant decrease in GSK3b phosphorylation between wild-type and APP/PS1 mice has also been demonstrated in APP/PS1mice at 16 months of age (Malm et al., 2007). Controversially, otherstudies have reported a reduction in both Akt and GSK3b phos-phorylation levels in the hippocampus of 6- and 13-month-old APP/PS1 mice (Hu et al., 2013; Jia et al., 2013). Interestingly, AD patientsappear to present reduced Akt phosphorylation in hippocampalneurons only at end stage disease, when altered Akt activity isalready established in the cortex (Griffin et al., 2005; Lee et al.,2009). Taken together, despite some controversy regarding Aktactivity levels in the hippocampus, it is plausible that hippocampalAkt dysregulation correlates with affected neuronal integrity in thisbrain region, which is manifest in transgenic mice only severalmonths after the development of other AD-related features.

Nevertheless, the role of TUDCA treatment in the modulation ofAkt/GSK3b signaling was highlighted by increased Akt activity in

Fig. 11. Cognitive performance in the Morris water maze in vehicle and TUDCA-treatedwild type mice and APP/PS1 mice. (A) During training, no significant group differenceswere measured in time to find the hidden platform. All groups showed similar learningcurves. (B) During the second probe trial, we found a significant target quadrantpreference in both control and TUDCA-treated wild-type mice. Such preference wasnot found in control APP/PS1 mice. TUDCA-treated APP/PS1 mice showed a trend totarget quadrant preference, albeit not significant (n ¼ 10 control wild-type mice; n ¼ 8TUDCA-treated wild-type mice; n ¼ 11 control APP/PS1 mice; and n ¼ 14 TUDCA-treated APP/PS1 mice). Data are expressed as means � SEM. *p < 0.001 betweentarget quadrant and all other quadrants, yp < 0.05 between target quadrant and A1quadrant, and zp < 0.01 between target quadrant and opposite quadrant. Abbrevia-tions: A1 and A2, adjacent quadrants; O, opposite quadrant, SEM, standard error of themean; T, target quadrant; TUDCA, tauroursodeoxycholic acid.

P.A. Dionísio et al. / Neurobiology of Aging xxx (2014) 1e13 11

the frontal cortex and hippocampus of APP/PS1 mice. Importantly,this increase was correlated with inhibition of GSK3b activity andtau hyperphosphorylation, suggesting that inactivation of GSK3b byTUDCA positively modulates AD-associated tau pathology.

Although GSK3b has been also widely accepted as a major playerin several pathologic mechanisms associated with AD, the mecha-nism by which this kinase regulates cerebral amyloidosis is stillsomewhat controversial. Some authors have reported that GSK3binhibition downregulates BACE1 gene expression in APP23/PS45double transgenic mice (Ly et al., 2013), whereas others did notdetected significant alterations in BACE1 protein levels (Ding et al.,2008; Durairajan et al., 2012). Alternatively, GSK3b may be involvedin the phosphorylation of APP at the cytoplasmic residue Thr668,targeting the protein for fast axonal transport to nerve terminals. Co-localization of p-APP with b- and g-secretases within axonal or

presynaptic vesicles culminates in enhanced Ab generation(Durairajan et al., 2012; Lee et al., 2003, 2005). This prompted us toinvestigate whether TUDCA regulates APP processing and subse-quent Ab production via GSK3b. Surprisingly, no significant changeswere detected in BACE1 levels between APP/PS1 mice and wild-typelittermates, treated or untreated with TUDCA. Similarly, TUDCAadministration didnot impact onAPPphosphorylation. Still, there areother kinases involved in AD pathology, such as cyclin-dependentkinase 5 and c-Jun N-terminal kinase, that are postulated to phos-phorylate APP, which might explain the unaltered p-APP levelsdespite GSK3b inhibition (Durairajan et al., 2012).

Other aspect of GSK3 inhibition is the associated decrease ofglial activation and overall neuroinflammatory markers (Beurel andJope, 2010; Koistinaho et al., 2011; Yuskaitis and Jope, 2009). In thiscontext, GSK3b appears to upregulate the expression of severalproinflammatory mediators, such as TNF-a (Wang et al., 2010;Yuskaitis and Jope, 2009), while decreasing anti-inflammatorymolecules (Koistinaho et al., 2011). Moreover, GSK3b inhibitionreduces microglial activation and migration (Yuskaitis and Jope,2009), and increases inflammatory tolerance in astrocytes onrepeated inflammatory stimuli, suggesting a paramount role forthis kinase during chronic neuroinflammation (Beurel and Jope,2010). As expected, a marked neuroinflammatory phenotype wasevident in the brains of APP/PS1, with a significant increase ingliosis and upregulation of proinflammatory TNF-a, IL-1b, and IL-6.Importantly, TUDCA administration after the onset of amyloid pa-thology significantly ameliorated astrocytosis and microgliosis inboth hippocampus and frontal cortex of APP/PS1 mice and specif-ically inhibited TNF-a mRNA expression. These results are inagreement with others showing decreased levels of TNF-amRNA inactivated astrocytes treated with glycoursodeoxycholic acid but noalterations in IL-1b expression (Fernandes et al., 2007). Moreimportantly, anti-TNF-a therapeutic strategies were reported toreduce amyloid deposition, gliosis, and tau hyperphosphorylation,rescuing cognitive and memory deficits in transgenic mice (Shiet al., 2011; Tweedie et al., 2012). In addition, intrinsic anti-inflammatory properties have also been attributed to TUDCA,whichmay explain the striking results obtained for astrocytosis andmicrogliosis in our model. TUDCA has been shown to decrease glialactivation and microglial migration in acute neuroinflammationmodels both in vivo and in vitro (Yanguas-Casas et al., 2014). Finally,TUDCA has also been reported to inactivate GSK3b and reduceinflammation in liver tissue obtained after partial hepatectomy(Ben Mosbah et al., 2010).

Other studies also showed the anti-inflammatory properties ofbile acids similar to TUDCA, such as ursodeoxycholic acid and gly-coursodeoxycholic acid, in astrocytes and microglia exposed toproinflammatory stimuli including Ab (Fernandes et al., 2007; Jooet al., 2003). Because Ab accumulation is the main mechanisminvolved in glial activation during AD (Meraz-Rios et al., 2013;Radde et al., 2006), and GSK3b is also potentially involved, wesuggest that TUDCA-mediated effects on Ab load, GSK3b activityand TNF-a expression are responsible for the reduced neuro-inflammatory conditions in TUDCA-treated APP/PS1 mice. In thisregard, we are currently elucidating the molecular mechanisms bywhich TUDCA protects from glial activation.

Activation of PI3K/Akt signaling is also crucial for the expressionof late-phase long-term potentiation (LTP), a form of synapticplasticity that is closely related with learning and memory, andspecifically affected in AD (Selkoe, 2002). This effect is particularlyimportant in the dentate gyrus of the hippocampus (Karpova et al.,2006). Impairments in LTP may eventually culminate in synapticdepression that in turn is correlated with structural changes insynaptic morphology (Selkoe, 2002). Ab oligomers also inhibit LTPand destabilize synaptic connection (Selkoe, 2002). Importantly,

P.A. Dionísio et al. / Neurobiology of Aging xxx (2014) 1e1312

treatment with TUDCA resulted in a reduction of presynaptic ter-minals loss, as evidenced by an increase in the presynaptic markersynaptophysin in the dentate gyrus of APP/PS1 mice. This effectprobably derives from the decrease in Ab load and activation ofPI3K/Akt by TUDCA. However, only a trend to improved spatialmemory was detected in TUDCA-treated APP/PS1 mice comparedwith control APP/PS1 mice. Because this mouse model alreadypresents extensive cognitive deficits and memory impairment at8 months of age (Radde et al., 2006), it is likely that TUDCA treat-ment started at 7 months is not sufficient to revert emerging orpreestablished cognitive deficits. Still, mouse models do notdevelop the same type of cognitive decline usually observed in ADpatients (Webster et al., 2014), which may also influence thebehavioral data.

In conclusion, our results demonstrate the therapeutic efficacyof TUDCA in APP/PS1 mice with established amyloid pathology byattenuating Ab production and deposition, tau pathology, glialactivation, and loss of synaptic function. Most of these effects arelikely related to the activation of the Akt/GSK3b signaling pathway.However, modulation of Ab deposition may influence severalpathways that impact on tau hyperphosphorylation and neuro-inflammation, which implicates that the reduction observed in Abload after TUDCA treatment may alter these AD-phenotypic traitsby GSK3b-independent pathways. Finally, because chronic neuro-inflammation is also strongly associated with accelerated AD pro-gression, the anti-inflammatory properties of TUDCA furtherhighlight its therapeutic potential. These evidences, allied to itssafety and brain bioavailability, point TUDCA as a promising ther-apeutic strategy to attenuate AD progression.

Disclosure statement

The authors declare that they have no conflict of interest.

Acknowledgements

The authors thank Prodotti Chimici e Alimentari S.p.A. (Basaluzzo,Italy) for the supply of TUDCA. APP/PS1mice were kindly donated byBart De Strooper. They also thank VéroniqueHendrickx for assistancein genotyping themice. Thisworkwas supported bygrant PTDC/SAU-NMC/117877/2010 from Fundação para a Ciência e a Tecnologia,Portugal. Joana D. Amaral was a recipient of postdoctoral fellowship(SFRH/BPD/47376/2008) fromFundaçãopara a Ciência e a Tecnologia,Portugal. Rudi D’Hooge and Adrian C Lo were funded by 7FP grantMEMOSAD and the Federal Science fund FWO-Vlaanderen (grantnumber G.0327.08).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.neurobiolaging.2014.08.034.

References

Aplin, A.E., Gibb, G.M., Jacobsen, J.S., Gallo, J.M., Anderton, B.H., 1996. In vitrophosphorylation of the cytoplasmic domain of the amyloid precursor protein byglycogen synthase kinase-3beta. J. Neurochem. 67, 699e707.

Beaulieu, J.M., Gainetdinov, R.R., Caron, M.G., 2009. Akt/GSK3 signaling in the actionof psychotropic drugs. Annu. Rev. Pharmacol. Toxicol. 49, 327e347.

Ben Mosbah, I., Alfany-Fernandez, I., Martel, C., Zaouali, M.A., Bintanel-Morcillo, M.,Rimola, A., Rodés, J., Brenner, C., Roselló-Catafau, J., Peralta, C., 2010. Endo-plasmic reticulum stress inhibition protects steatotic and non-steatotic livers inpartial hepatectomy under ischemia-reperfusion. Cell Death Dis. 1, e52.

Beurel, E., Jope, R.S., 2010. Glycogen synthase kinase-3 regulates inflammatorytolerance in astrocytes. Neuroscience 169, 1063e1070.

Born, H.A., Kim, J.Y., Savjani, R.R., Das, P., Dabaghian, Y.A., Guo, Q., Yoo, J.W.,Schuler, D.R., Cirrito, J.R., Zheng, H., Golde, T.E., Noebels, J.L., Jankowsky, J.L.,

2014. Genetic suppression of transgenic APP rescues hypersynchronousnetwork activity in a mouse model of Alzheimer’s disease. J. Neurosci. 34,3826e3840.

Castro-Caldas, M., Carvalho, A.N., Rodrigues, E., Henderson, C.J., Wolf, C.R.,Rodrigues, C.M., Gama, M.J., 2012. Tauroursodeoxycholic acid prevents MPTP-induced dopaminergic cell death in a mouse model of Parkinson’s disease.Mol. Neurobiol. 46, 475e486.

Ding, Y., Qiao, A., Wang, Z., Goodwin, J.S., Lee, E.S., Block, M.L., Allsbrook, M.,McDonald, M.P., Fan, G.H., 2008. Retinoic acid attenuates beta-amyloid depo-sition and rescues memory deficits in an Alzheimer’s disease transgenic mousemodel. J. Neurosci. 28, 11622e11634.

Duan, W.M., Rodrigues, C.M., Zhao, L.R., Steer, C.J., Low, W.C., 2002. Taur-oursodeoxycholic acid improves the survival and function of nigral transplantsin a rat model of Parkinson’s disease. Cell Transpl. 11, 195e205.

Durairajan, S.S., Liu, L.F., Lu, J.H., Chen, L.L., Yuan, Q., Chung, S.K., Huang, L., Li, X.S.,Huang, J.D., Li, M., 2012. Berberine ameliorates beta-amyloid pathology, gliosis,and cognitive impairment in an Alzheimer’s disease transgenic mouse model.Neurobiol. Aging 33, 2903e2919.

Elder, G.A., Gama Sosa, M.A., De Gasperi, R., 2010. Transgenic mouse models ofAlzheimer’s disease. Mt. Sinai J. Med. 77, 69e81.

Fernandes, A., Vaz, A.R., Falcao, A.S., Silva, R.F., Brito, M.A., Brites, D., 2007.Glycoursodeoxycholic acid and interleukin-10 modulate the reactivity of ratcortical astrocytes to unconjugated bilirubin. J. Neuropathol. Exp. Neurol. 66,789e798.

Goddyn, H., Leo, S., Meert, T., D’Hooge, R., 2006. Differences in behavioural testbattery performance between mice with hippocampal and cerebellar lesions.Behav. Brain Res. 173, 138e147.

Griffin, R.J., Moloney, A., Kelliher, M., Johnston, J.A., Ravid, R., Dockery, P.,O’Connor, R., O’Neill, C., 2005. Activation of Akt/PKB, increased phos-phorylation of Akt substrates and loss and altered distribution of Aktand PTEN are features of Alzheimer’s disease pathology. J. Neurochem.93, 105e117.

Haass, C., Kaether, C., Thinakaran, G., Sisodia, S., 2012. Trafficking and proteolyticprocessing of APP. Cold Spring Harb. Perspect. Med. 2, a006270.

Hefendehl, J.K., Wegenast-Braun, B.M., Liebig, C., Eicke, D., Milford, D., Calhoun, M.E.,Kohsaka, S., Eichner, M., Jucker, M., 2011. Long-term in vivo imaging of beta-amyloid plaque appearance and growth in a mouse model of cerebral beta-amyloidosis. J. Neurosci. 31, 624e629.

Hu, Y.S., Long, N., Pigino, G., Brady, S.T., Lazarov, O., 2013. Molecular mechanisms ofenvironmental enrichment: impairments in Akt/GSK3beta, neurotrophin-3 andCREB signaling. PLoS One 8, e64460.

Jia, N., Han, K., Kong, J.J., Zhang, X.M., Sha, S., Ren, G.R., Cao, Y.P., 2013. (-)-Epi-gallocatechin-3-gallate alleviates spatial memory impairment in APP/PS1 miceby restoring IRS-1 signaling defects in the hippocampus. Mol. Cell Biochem. 380,211e218.

Joo, S.S., Kang, H.C., Won, T.J., Lee, D.I., 2003. Ursodeoxycholic acid inhibits pro-inflammatory repertoires, IL-1 beta and nitric oxide in rat microglia. Arch.Pharm. Res. 26, 1067e1073.

Karpova, A., Sanna, P.P., Behnisch, T., 2006. Involvement of multiple phosphatidy-linositol 3-kinase-dependent pathways in the persistence of late-phase longterm potentiation expression. Neuroscience 137, 833e841.

Keene, C.D., Rodrigues, C.M., Eich, T., Chhabra, M.S., Steer, C.J., Low, W.C., 2002.Tauroursodeoxycholic acid, a bile acid, is neuroprotective in a transgenicanimal model of Huntington’s disease. Proc. Natl. Acad. Sci. U.S.A. 99,10671e10676.

Keene, C.D., Rodrigues, C.M., Eich, T., Linehan-Stieers, C., Abt, A., Kren, B.T., Steer, C.J.,Low, W.C., 2001. A bile acid protects against motor and cognitive deficits andreduces striatal degeneration in the 3-nitropropionic acid model of Hunting-ton’s disease. Exp. Neurol. 171, 351e360.

Koistinaho, J., Malm, T., Goldsteins, G., 2011. Glycogen synthase kinase-3beta: amediator of inflammation in Alzheimer’s disease? Int. J. Alzheimers Dis. 2011,129753.

Kokjohn, T.A., Roher, A.E., 2009. Amyloid precursor protein transgenic mousemodels and Alzheimer’s disease: understanding the paradigms, limitations, andcontributions. Alzheimers Dement. 5, 340e347.

Lee, E.B., Zhang, B., Liu, K., Greenbaum, E.A., Doms, R.W., Trojanowski, J.Q., Lee, V.M.,2005. BACE overexpression alters the subcellular processing of APP and inhibitsAbeta deposition in vivo. J. Cell Biol. 168, 291e302.

Lee, H.K., Kumar, P., Fu, Q., Rosen, K.M., Querfurth, H.W., 2009. The insulin/Aktsignaling pathway is targeted by intracellular beta-amyloid. Mol. Biol. Cell. 20,1533e1544.

Lee, M.S., Kao, S.C., Lemere, C.A., Xia, W., Tseng, H.C., Zhou, Y., Neve, R.,Ahlijanian, M.K., Tsai, L.H., 2003. APP processing is regulated by cytoplasmicphosphorylation. J. Cell Biol. 163, 83e95.

Leroy, A., Landrieu, I., Huvent, I., Legrand, D., Codeville, B., Wieruszeski, J.M.,Lippens, G., 2010. Spectroscopic studies of GSK3{beta} phosphorylation of theneuronal tau protein and its interaction with the N-terminal domain of apoli-poprotein E. J. Biol. Chem. 285, 33435e33444.

Li, T., Paudel, H.K., 2006. Glycogen synthase kinase 3beta phosphorylates Alz-heimer’s disease-specific Ser396 of microtubule-associated protein tau by asequential mechanism. Biochemistry 45, 3125e3133.

Lo, A.C., Callaerts-Vegh, Z., Nunes, A.F., Rodrigues, C.M., D’Hooge, R., 2013.Tauroursodeoxycholic acid (TUDCA) supplementation prevents cognitiveimpairment and amyloid deposition in APP/PS1 mice. Neurobiol. Dis. 50,21e29.

P.A. Dionísio et al. / Neurobiology of Aging xxx (2014) 1e13 13

Ly, P.T., Wu, Y., Zou, H., Wang, R., Zhou, W., Kinoshita, A., Zhang, M., Yang, Y., Cai, F.,Woodgett, J., Song, W., 2013. Inhibition of GSK3beta-mediated BACE1 expres-sion reduces Alzheimer-associated phenotypes. J. Clin. Invest. 123, 224e235.

Malm, T.M., Iivonen, H., Goldsteins, G., Keksa-Goldsteine, V., Ahtoniemi, T.,Kanninen, K., Salminen, A., Auriola, S., Van Groen, T., Tanila, H., Koistinaho, J., 2007.Pyrrolidine dithiocarbamate activates Akt and improves spatial learning in APP/PS1 mice without affecting beta-amyloid burden. J. Neurosci. 27, 3712e3721.

Masliah, E., Mallory, M., Alford, M., DeTeresa, R., Hansen, L.A., McKeel Jr., D.W.,Morris, J.C., 2001. Altered expression of synaptic proteins occurs early duringprogression of Alzheimer’s disease. Neurology 56, 127e129.

Meraz-Rios, M.A., Toral-Rios, D., Franco-Bocanegra, D., Villeda-Hernandez, J., Cam-pos-Pena, V., 2013. Inflammatory process in Alzheimer’s disease. Front. Integr.Neurosci. 7, 59.

Mines, M.A., Beurel, E., Jope, R.S., 2011. Regulation of cell survival mechanisms inAlzheimer’s disease by glycogen synthase kinase-3. Int. J. Alzheimers Dis. 2011,861072.

Nunes, A.F., Amaral, J.D., Lo, A.C., Fonseca, M.B., Viana, R.J., Callaerts-Vegh, Z.,D’Hooge, R., Rodrigues, C.M., 2012. TUDCA, a bile acid, attenuates amyloidprecursor protein processing and amyloid-beta deposition in APP/PS1 mice.Mol. Neurobiol. 45, 440e454.

Radde, R., Bolmont, T., Kaeser, S.A., Coomaraswamy, J., Lindau, D., Stoltze, L.,Calhoun, M.E., Jaggi, F., Wolburg, H., Gengler, S., Haass, C., Ghetti, B., Czech, C.,Holscher, C., Mathews, P.M., Jucker, M, 2006. Abeta42-driven cerebralamyloidosis in transgenic mice reveals early and robust pathology. EMBO Rep. 7,940e946.

Ramalho, R.M., Borralho, P.M., Castro, R.E., Sola, S., Steer, C.J., Rodrigues, C.M., 2006.Tauroursodeoxycholic acid modulates p53-mediated apoptosis in Alzheimer’sdisease mutant neuroblastoma cells. J. Neurochem. 98, 1610e1618.

Ramalho, R.M., Nunes, A.F., Dias, R.B., Amaral, J.D., Lo, A.C., D’Hooge, R.,Sebastião, A.M., Rodrigues, C.M., 2013. Tauroursodeoxycholic acid suppressesamyloid beta-induced synaptic toxicity in vitro and in APP/PS1 mice. Neurobiol.Aging 34, 551e561.

Ramalho, R.M., Ribeiro, P.S., Sola, S., Castro, R.E., Steer, C.J., Rodrigues, C.M., 2004.Inhibition of the E2F-1/p53/Bax pathway by tauroursodeoxycholic acid in am-yloid beta-peptide-induced apoptosis of PC12 cells. J. Neurochem. 90, 567e575.

Reitz, C., Mayeux, R., 2014. Alzheimer disease: epidemiology, diagnostic criteria, riskfactors and biomarkers. Biochem. Pharmacol. 88, 640e651.

Rodrigues, C.M., Sola, S., Nan, Z., Castro, R.E., Ribeiro, P.S., Low, W.C., Steer, C.J., 2003.Tauroursodeoxycholic acid reduces apoptosis and protects against neurologicalinjury after acute hemorrhagic stroke in rats. Proc. Natl. Acad. Sci. U.S.A. 100,6087e6092.

Rodrigues, C.M., Spellman, S.R., Sola, S., Grande, A.W., Linehan-Stieers, C., Low, W.C.,Steer, C.J., 2002. Neuroprotection by a bile acid in an acute stroke model in therat. J. Cereb. Blood Flow Metab. 22, 463e471.

Rupp, N.J., Wegenast-Braun, B.M., Radde, R., Calhoun, M.E., Jucker, M., 2011. Earlyonset amyloid lesions lead to severe neuritic abnormalities and local, but notglobal neuron loss in APPPS1 transgenic mice. Neurobiol Aging 32, 2324.e1ee6.

Rutten, B.P., Van der Kolk, N.M., Schafer, S., van Zandvoort, M.A., Bayer, T.A.,Steinbusch, H.W., Schmitz, C., 2005. Age-related loss of synaptophysin immuno-reactive presynaptic boutons within the hippocampus of APP751SL, PS1M146L,and APP751SL/PS1M146L transgenic mice. Am. J. Pathol. 167, 161e173.

Ryder, J., Su, Y., Ni, B., 2004. Akt/GSK3beta serine/threonine kinases: evidence for asignalling pathway mediated by familial Alzheimer’s disease mutations. CellSignal. 16, 187e200.

Saito, T., Matsuba, Y., Mihira, N., Takano, J., Nilsson, P., Itohara, S., Iwata, N.,Saido, T.C., 2014. Single App knock-in mouse models of Alzheimer’s disease. Nat.Neurosci. 17, 661e663.

Selkoe, D.J., 2001. Alzheimer’s disease: genes, proteins, and therapy. Physiol. Rev. 81,741e766.

Selkoe, D.J., 2002. Alzheimer’s disease is a synaptic failure. Science 298, 789e791.Sereno, L., Coma, M., Rodriguez, M., Sanchez-Ferrer, P., Sanchez, M.B., Gich, I.,

Agullo, J.M., Perez, M., Avila, J., Guardia-Laguarta, C., Clarimon, J., Lleo, A.,

Gomez-Isla, T., 2009. A novel GSK-3beta inhibitor reduces Alzheimer’s pathol-ogy and rescues neuronal loss in vivo. Neurobiol. Dis. 35, 359e367.

Shi, J.Q., Shen, W., Chen, J., Wang, B.R., Zhong, L.L., Zhu, Y.W., Zhang, Q.Q.,Zhang, Y.D., Xu, J., 2011. Anti-TNF-alpha reduces amyloid plaques and tauphosphorylation and induces CD11c-positive dendritic-like cell in the APP/PS1transgenic mouse brains. Brain Res. 1368, 239e247.

Shin, R.W., Ogino, K., Shimabuku, A., Taki, T., Nakashima, H., Ishihara, T.,Kitamoto, T., 2007. Amyloid precursor protein cytoplasmic domain withphospho-Thr668 accumulates in Alzheimer’s disease and its transgenicmodels: a role to mediate interaction of Abeta and tau. Acta Neuropathol.113, 627e636.

Simoes, A.E., Pereira, D.M., Amaral, J.D., Nunes, A.F., Gomes, S.E., Rodrigues, P.M.,Lo, A.C., D’Hooge, R., Steer, C.J., Thibodeau, S.N., Borralho, P.M., Rodrigues, C.M.,2013. Efficient recovery of proteins from multiple source samples after TRI-zol((R)) or TRIzol((R))LS RNA extraction and long-term storage. BMC Genomics14, 181.

Sola, S., Amaral, J.D., Borralho, P.M., Ramalho, R.M., Castro, R.E., Aranha, M.M.,Steer, C.J., Rodrigues, C.M., 2006. Functional modulation of nuclear steroid re-ceptors by tauroursodeoxycholic acid reduces amyloid beta-peptide-inducedapoptosis. Mol. Endocrinol. 20, 2292e2303.

Sola, S., Castro, R.E., Laires, P.A., Steer, C.J., Rodrigues, C.M., 2003. Tauroursodeox-ycholic acid prevents amyloid-beta peptide-induced neuronal death via aphosphatidylinositol 3-kinase-dependent signaling pathway. Mol. Med. 9,226e234.

Soutar, M.P., Kim, W.Y., Williamson, R., Peggie, M., Hastie, C.J., McLauchlan, H.,Snider, W.D., Gordon-Weeks, P.R., Sutherland, C., 2010. Evidence that glycogensynthase kinase-3 isoforms have distinct substrate preference in the brain.J. Neurochem. 115, 974e983.

Tweedie, D., Ferguson, R.A., Fishman, K., Frankola, K.A., Van Praag, H.,Holloway, H.W., Luo, W., Li, Y., Caracciolo, L., Russo, I., Barlati, S., Ray, B.,Lahiri, D.K., Bosetti, F., Greig, N.H., Rosi, S., 2012. Tumor necrosis factor-alphasynthesis inhibitor 3,6’-dithiothalidomide attenuates markers of inflamma-tion, Alzheimer pathology and behavioral deficits in animal models of neuro-inflammation and Alzheimer’s disease. J. Neuroinflammation 9, 106.

Viana, R.J., Ramalho, R.M., Nunes, A.F., Steer, C.J., Rodrigues, C.M., 2010. Mod-ulation of amyloid-beta peptide-induced toxicity through inhibition of JNKnuclear localization and caspase-2 activation. J. Alzheimers Dis. 22,557e568.

Wang, M.J., Huang, H.Y., Chen, W.F., Chang, H.F., Kuo, J.S., 2010. Glycogen synthasekinase-3beta inactivation inhibits tumor necrosis factor-alpha production inmicroglia by modulating nuclear factor kappaB and MLK3/JNK signaling cas-cades. J. Neuroinflammation 7, 99.

Webster, S.J., Bachstetter, A.D., Nelson, P.T., Schmitt, F.A., Van Eldik, L.J., 2014.Using mice to model Alzheimer’s dementia: an overview of the clinicaldisease and the preclinical behavioral changes in 10 mouse models. Front.Genet. 5, 88.

Wirths, O., Bayer, T.A., 2010. Neuron loss in transgenic mouse models of Alzheimer’sdisease. Int. J. Alzheimers Dis. 2010, pii: 723782.

Yang, L.B., Lindholm, K., Yan, R., Citron, M., Xia, W., Yang, X.L., Beach, T., Sue, L.,Wong, P., Price, D., Li, R., Shen, Y., 2003. Elevated beta-secretase expressionand enzymatic activity detected in sporadic Alzheimer disease. Nat. Med. 9,3e4.

Yanguas-Casas, N., Barreda-Manso, M.A., Nieto-Sampedro, M., Romero-Ramirez, L.,2014. Tauroursodeoxycholic acid reduces glial cell activation in an animal modelof acute neuroinflammation. J. Neuroinflammation. 11, 50.

Yuskaitis, C.J., Jope, R.S., 2009. Glycogen synthase kinase-3 regulates microglialmigration, inflammation, and inflammation-induced neurotoxicity. Cell Signal.21, 264e273.

Zhu, L.Q., Wang, S.H., Liu, D., Yin, Y.Y., Tian, Q., Wang, X.C., Wang, Q., Chen, J.G.,Wang, J.Z., 2007. Activation of glycogen synthase kinase-3 inhibits long-termpotentiation with synapse-associated impairments. J. Neurosci. 27,12211e12220.