[1] 20w The choroid plexus transcriptome reveals changes in type I and II interferon responses in a mouse model of Alzheimer's disease

[1] 265w a b s t r a c t 34 Alzheimer's disease (AD) is a neurodegenerative disease characterized by a marked decline in cognition 35 and memory function. Increasing evidence highlights the essential role of neuroinflammatory and 36 immune-related molecules, including those produced at the brain barriers, on brain immune surveillance, 37 cellular dysfunction and amyloid beta (Ab) pathology in AD. Therefore, understanding the response at the 38 brain barriers may unravel novel pathways of relevance for the pathophysiology of AD. Herein, we 39 focused on the study of the choroid plexus (CP), which constitutes the blood-cerebrospinal fluid barrier, 40 in aging and in AD. Specifically, we used the PDGFB-APPSwInd (J20) transgenic mouse model of AD, 41 which presents early memory decline and progressive Ab accumulation, and littermate age-matched 42 wild-type (WT) mice, to characterize the CP transcriptome at 3, 5-6 and 11-12 months of age. The most 43 striking observation was that the CP of J20 mice displayed an overall overexpression of type I interferon 44 (IFN) response genes at all ages. Moreover, J20 mice presented a high expression of type II IFN genes in 45 the CP at 3 months, which became lower than WT at 5-6 and 11-12 months. Importantly, along with a 46 marked memory impairment and increased glial activation, J20 mice also presented a similar overexpres-47 sion of type I IFN genes in the dorsal hippocampus at 3 months. Altogether, these findings provide new 48 insights on a possible interplay between type I and II IFN responses in AD and point to IFNs as targets for 49 modulation in cognitive decline.50

[1] 154w Alzheimer's disease (AD) is the most prevalent form of dementia and is predicted to affect eighty million people worldwide by 2040 (Ballard et al., 2011;Querfurth and LaFerla, 2010). The major constraint of AD patients is the severe loss of cognitive abilities (Ballard et al., 2011). Moreover, this neurodegenerative disease is characterized by two main brain pathological hallmarks: the extracellular deposition of amyloid beta (Ab) peptides, in the form of senile plaques, and the formation of intracellular neurofibrillary tangles, as a result of increased aggregation of hyperphosphorylated Tau protein (Gotz et al., 2001;Hardy and Selkoe, 2002;Lewis et al., 2001;Scheuner et al., 1996). The amyloidogenic hypothesis of AD considers that amyloid pathology may result 67 the blood-CSF barrier and in the brain parenchyma, at different stages of Ab pathology, particularly before and after Ab plaque formation. The value of achieved power (1Àb) of the sample was calculated using the IBM SPSS Statistics for Windows (Version 23.0.

[2] 44w Armonk, NY: IBM Corp, USA), on the basis of the sample size (N = 80) and effect size (partial eta squared = 0.044) associated with the Morris water maze (MWM) test, which was analyzed by a parametric repeated measures two-way ANOVA (a = 0.05).

[3] 40w Taking that the Mauchly's Test revealed a violation of sphericity for the analyzed groups, the Huynh-Feldt correction was applied and a final value of 1Àb = 0.841 was obtained, which is considered to be statistically powerful (Mazen et al., 1985).

[4] 307w Memory was evaluated during the light phase of the diurnal cycle using the MWM paradigm. The MWM was performed in a white circular pool (170 cm in diameter and 50 cm in height), filled with tap water (23 ± 1 °C; 30 cm of depth) and placed in a poorly lit room. The water tank was divided into four imaginary quadrants (north, east, south, and west), each corresponding to a fixed extrinsic clue visible to the mouse. A transparent escape platform (14 cm in diameter and 30 cm high), invisible to the mouse, was placed in the center of one of the quadrants and was maintained in that same position during the five days of the acquisition. For spatial reference-memory acquisition, a HPC-dependent task, the mice were placed randomly in one of the quadrants in each trial, and allowed to search for the hidden platform. During the five days of the acquisition phase, each mouse performed four trials per day. Each trial was concluded when the platform was reached within the time-limit of 120 s. If failing to reach the platform within this time-period, the mouse was guided to the platform and allowed to stay in it for 30 s. The four random consecutive trials of each day were video-captured with a video-tracking system (Viewpoint, Champagne au Mont d'or, France) and the mean values of the time (latency to platform) and speed to reach the platform were calculated. To analyze the MWM performance, mice were divided into groups according to their age and genotype. After two independent experiments with identical results, the final number of mice per group was the following: at 3 months, N (WT) = 16 and N (J20) = 8; at 5-6 months, N (WT) = 8 and N (J20) = 10; at 11-12 months, N (WT) = 23 and N (J20) = 15.

[1] 92w The MWM was used to test the performance of WT and J20 mice in a spatial reference-memory task at 3, 5-6 and 11-12 months of age (Fig. 1A and B). At 3 months, WT mice took significantly less time to find the hidden platform, when compared to age-matched J20 mice (Fig. 1B). At the ages of 5-6 months and 11-12 months, the differences between WT and J20 mice became smaller; still, at 11-12 months, J20 mice presented a significantly worst performance than WT mice at day 5 of the MWM (Fig. 1B).

[2] 479w No statistical significant differences were observed in terms of swimming speed in the MWM, when comparing mice of different genotypes at each age (Supplementary Fig. 1). To assess brain amy-454 loid pathology in J20 mice, we measured the levels of Ab in the CSF 455 and evaluated the degree of Ab aggregation and deposition in the 456 dHPC (Fig. 1C and D). There were no statistically significant differ-457 ences between the levels of Ab in the CSF of J20 mice at different 458 ages (Fig. 1C); yet, there was a progressive accumulation of Ab 459 and increased plaque burden in the dHPC of J20 mice with aging 460 that was more obvious at 11-12 months (Fig. 1D). No Ab was 461 detected in the CSF or in the brain of WT mice (Fig. 1C and D). to the changes observed in age-matched WT mice (Fig. 1A). The 467 highest number of up-and down-regulated genes in the CP of 468 J20 mice was observed at 11-12 months (Fig. 1E). Of notice, we found no commonly altered genes after overlapping the CP transcriptomic changes of J20 mice at 3, 5-6 and 11-12 months (Fig. 1F). Moreover, we found few commonly altered genes when comparing the transcriptomic changes in the CP of J20 mice two ages at a time (Fig. 1F). The ten most up-or down-regulated genes in the CP of J20 mice (Table 1) revealed alterations in genes that are involved in the functions classically attributed to the CP, such as the modulation of the CSF's composition, and that regulate the survival of CP cells and the maintenance of the barrier integrity (Table 1). We also analyzed the effect of aging, per se, in the CP transcriptome, by normalizing the changes observed in WT or J20 mice at 11-12 months to that of 3 months-old mice of the same genotype (Supplementary Figs. 2 and 3). Interestingly, we found a consistent alteration in the expression of genes that regulate the cellular circadian rhythm in 11-12 months-old mice, which is describe in detail in the Supplementary data (Supplementary Figs 2 and 3 and Supplementary Tables 1 and 2). Surprisingly, when considering all the analyzed data sets and comparisons performed, 512 beta 1 (Ifnb1) and signal transducer and activator of transcription 513 1 (Stat1) (Fig. 2F). The gene networks 2-5, affected at 11-514 12 months, showed again alterations in processes such as small 515 molecule biochemistry, molecular transport and cellular signaling, 516 movement and survival (Fig. 2E). 517 3.4. Type I and II IFN responses were altered in the CP of J20 mice 518 Taking into account the previous alterations, particularly those 519 regarding gene pathways involved in the regulation of the inflam-520 matory/immune responses in the CP of J20, we next pursued the 521 underlying transcriptional regulators that could be responsible 522 for such alterations in gene expression (Fig. 3 and Supplementary 523

[3] 47w Table 3). For that analysis, we used the IPA Upstream Regulator 3), this analysis allowed the 530 assembly of gene interactomes (Fig. 3B-D). At 3 months, this 531 approach discriminated the genes Toll-like receptor 9 (Tlr9), inter-532 feron regulatory factor 7 (Irf7) and interferon (alpha and beta)

[4] 87w receptor 1 (Ifnar1), all of them predicted to be up-regulated in the CP of J20 mice, when compared with age matched WT mice (Fig. 3A and B). Surprisingly, at the ages of 5-6 and 11-12 months, interferon gamma (Ifng) presented a low activation z-score that suggested a possible down-regulation of this gene in the CP of J20 mice (Fig. 3A, C and D). Noticeably, the inactivation of the Ifng-regulated pathway was predicted to affect the expression of 18 target genes at 5-6 months (Fig. 3C and Supplementary

[5] 22w Table 3), an effect that was even more marked at 11-12 months, affecting 25 downstream genes/molecules (Fig. 3D and Supplementary Table 3).

[6] 227w To confirm and complement the array data, the expression of type I IFN-related genes [Tlr9, Ifnar1, Irf7, interferon alpha 1 (Ifna1) and interferon beta 1 (Ifnb1)] and of type II IFN-related genes [Ifng and intercellular adhesion molecule 1 (Icam1)] was measured by qRT-PCR (Fig. 4A-G). Once again, we observed an overexpression of Tlr9, Ifnar1, Irf7 and Ifnb1, in the CP of J20 mice at 3 months, when compared with WT mice; together with Ifna1, these genes remained highly expressed in the CP of J20 mice at 5-6 months when compared to 3 months-old WT mice (Fig. 4A-E). In the CP of WT mice, despite the unaltered expression of Irf7, Ifna and Ifnb1 (Fig. 4C-E) at different ages, we observed a significant overexpression of Tlr9 and Ifnar1 at 11-12 months, when compared to the 3 months-old mice of the same genotype (Fig. 4A and B). The genes Ifng and Icam1, which are related with a type II IFN response, were also significantly overexpressed at 3 months in the CP of J20 mice, when compared to age-matched WT mice (Fig. 4F and G). On the other hand, at 5-6 months, there was a significant down-regulation of Ifng in the CP of J20 mice, when compared to the expression levels at 3 months in J20 and at 5-6 months in WT mice, which was maintained at 11-12 months (Fig. 4F).

[7] 105w Next, to discriminate the specific contribution of CP epithelial cells to the expression of type I or type II IFN genes in AD, we performed primary cultures of CP epithelial cells on a transwell system (to mimic the in vivo tight and polarized arrangement that these epithelial cells present in the blood-CSF barrier, Fig. 4H), and stimulated these cells at the apical membrane, which mimics the CSF-side, with 1 lM Ab 1-42. We observed that in response to Ab 1-42 , the CP epithelial cells significantly overexpressed the type I IFN genes Tlr9 and Ifna1, but not the type II IFN gene Ifng (Fig. 5I-K).

[8] 68w To further investigate the IFN response at the brain parenchyma, we measured the expression of the same type I and II IFN genes in the dHPC of WT and J20 mice at different ages (Fig. 5A-G). As observed before in the CP of J20 mice, there was was statistically significant at 11-12 months, when compared to 587 the expression levels of WT mice at 3 months (Fig. 5A-E).

[9] 48w Regarding the expression of type II IFN genes in the dHPC 589 (Fig. 5F and G), even though there was an increased expression of Icam1 in 11-12 months-old WT mice (Fig. 5G), no changes were observed on the levels of Ifng expression at the different ages (Fig. 5F).

[10] 166w In an attempt to unravel the specific contribution of different brain cell populations to the overexpression of type I and II IFN genes, we measured the expression of Tlr9, Ifna1 and Ifng in primary cultures of astrocytes and of neurons, at different time points upon stimulation with vehicle or 1 lM Ab 1-42 . However, no statistical significant differences were observed when comparing vehicle-or Ab 1-42 -treated cells at 24 and 72 h (Supplementary 600 Fig. 4). 601 3.6. Early activation of astrocytes and microglia was observed in the 602 dHPC of J20 mice 603 To have a different readout of the inflammatory response in the 604 dHPC of WT and J20 mice at different ages, we measured the 605 expression of Gfap and Iba1, as well as the morphology of astro-606 cytes and the number of microglia/macrophages from the DG of 607 the dHPC, based on the staining for GFAP and IBA1, respectively 608 (Fig. 5H-K). There was a significant overexpression of Gfap and 609

[11] 235w Iba1 in the dHPC of J20 mice, at 3 months, when compared to the 610 expression levels in age-matched WT mice (Fig. 5H and J). The 611 expression of Gfap was also significantly increased in the dHPC of 612 WT mice at 11-12 months (Fig. 5H). Regarding the morphological 613 characterization of astrocytes, overall, we observed that astrocytes 614 from the dHPC of J20 mice presented a more ramified phenotype at 615 3 months, whereas, in the WT mice, these morphological alter-616 ations were observed at the age of 5-6 months (Fig. 5I). In that 617 sense, except for the number of processes, a significant increase 618 in the values of the analyzed astrocytic morphological parameters (length of individual processes, total surface and number of endings) was observed in astrocytes from the DG of J20 mice at 3 months, when compared to 3 months-old WT mice and to 5-6 months-old J20 mice (Fig. 5I). Noticeably, at 5-6 months, the astrocytes from the DG of WT mice presented more processes and were larger and more ramified than those of 5-6 months-old J20 mice and of 3 months-and 11-12 months-old mice of the same genotype (Fig. 5I). Concerning the quantification of microglia/macrophages, there was a significant increase in the number of IBA1 + cells per mm 2 in the DG of the dHPC of J20 mice at 11-12 months, when compared to all other groups (Fig. 5K).

[12] 17w 3.7. Increased level of IFN-a in the CSF was correlated with a worst performance in the MWM

[13] 120w The level of IFN-a was measured in the CSF of WT and J20 mice, at different ages (Fig. 6A). Although not significant, we observed a tendency for an increased level of IFN-a in the CSF of J20 mice, when compared to 3 months-old WT mice, which was sustained over time (Fig. 6A). Interestingly, we observed a significant increase in the level of IFN-a in the CSF of WT mice at 5-6 and at 11-12 months, when compared to 3 months-old WT mice (Fig. 6A), which was positively correlated with an increased time to 640 find the hidden platform in the MWM (calculated as the average 641 time of latency to platform in the five days of the test) (Fig. 6B).

[14] 141w 642 However, no significant correlations were observed between the 643 levels of IFN-a and the latency to platform, when analyzing mice 644 at each age separately (Supplementary Fig. 5). Table 2). The genotype and the levels of 650 IFN-a in the CSF, alone, were significantly associated with alter-651 ations in the performance in the MWM (model 1, Table 2). 652 However, when the interactions between the independent vari-653 ables were considered, the genotype significantly moderated the 654 relation between age and latency (model 2, Table 2). Of relevance, 655 when comparing the two models, the inclusion of the interaction 656 effects only represented an increment of 4.65% in the explained 657 variance of latency which was of 67.12% when no interactions were 658 included in the analysis. 659 4. Discussion 660 A progressive dysfunction of the blood-CSF barrier, which is 661

[15] 142w formed by the CP epithelial cells, is described in AD (Johanson 662 et al., 2004;Marques et al., 2013;Serot et al., 2012). To better along with a decreased inflammatory gene expression in the dHPC, the behavioral impairment of J20 mice seems to become milder, when compared to 3 months-old mice of the same genotype. Once again, whether there is a causal relationship between these events in the brain of J20 mice and the deficits in memory is something that needs to be addressed in the future. In addition an age-dependent increased level of IFN-a in the CSF, observed only in WT mice, is correlated with a poor performance in a spatial reference-memory task. On the other hand, the J20 mice, presented no significant alterations in the level of IFN-a in the CSF, which is in agreement with their early and sustained memory impairment.

[16] 63w In fact, it is still controversial whether an early induction of a type I IFN inflammatory response in the brain, followed by a reduced type II IFN response, is beneficial or detrimental for AD presented by these mice, without any obvious inflammatory side-effects (Scholtzova et al., 2009(Scholtzova et al., , 2014)). Also, increased production of IFN-b by astrocytes, upon specific TLR3 stimulation, was

[17] 45w shown to affect the spontaneous activity of neurons from the CA1 region of the HPC. This effect of IFN-b on neuronal excitability was shown to be mediated by IFNAR1, due to the lack of changes in the HPC of Ifnra1-null mice (Costello and Lynch, 2013).

[18] 279w Additionally, 9 months-old transgenic APP/PS1 mice present higher levels of IFN-a in the brain, when compared to age-matched WT mice (Taylor et al., 2014). Interestingly, our results also suggest that, along with a possible impact on behavior, an activation of a type I IFN response in the dHPC, in younger J20 mice and in 11-12 months-old WT mice may be linked with changes in the astrocytes' morphology. It is widely accepted that increased levels of inflammatory molecules, whether in the circulation or in specific regions of the CNS, can affect the morphology and function of brain cells, especially of glial cells (Baron et al., 2014;Ghosh et al., 2013). Astrocytes in particular seem to play a central role in the inflammatory response and are good indicators of the level of brain inflammation (Medeiros and LaFerla, 2013;Wyss-Coray, 2006). Concurrently, the activation of astrocytes and microglia was shown to occur before Ab plaque formation in the HPC of J20 mice (Baron et al., 2014;Beauquis et al., 2014;Wright et al., 2013) and that an increased microglial activation and clustering, observed at 4 weeks of age in the HPC of these mice, precedes neuronal atrophy, decreased neurogenesis and a decrease in the volume of the HPC (Fu et al., 2014). However, it is necessary to further investigate whether the early induction of type I IFN genes in the dHPC of J20 mice is directly promoting changes in glial activation and Ab pathology. Moreover, it will be important to explore the impact of increased level of type I IFN cytokines, such as IFN-a, on the premature memory deficits observed in this AD mouse model and on the memory and cognitive impairments in older WT mice.

[19] 38w Overall, this study reveals that structures such as the CP and the dHPC may be key elements in the inflammatory signaling in the CNS, by modulating the levels of IFNs, with a possible impact on memory and cognition.

[1] 38w During the light phase of the diurnal cycle, specifically in the morning, mice were anesthetized with an intraperitoneal injection of a mixture of ketamine hydrochloride (150 mg/kg, Imalgene Ò 1000) and medetomidine hydrochloride (0.3 mg/kg, Dorben Ò ).

[2] 157w Under deep anesthesia, CSF samples were collected from the cisterna magna and checked for blood contamination using a haemocytometer. Samples were stored at À80 °C until further use. After this, mice were transcardially perfused with 0.9% saline and the brains were removed from the skull. For gene expression analysis by microarray and qRT-PCR, the CP samples from each brain ventricle of the same mouse were rapidly removed under a conventional light stereomicroscope (SZX7, Olympus, Hamburg, Germany), pooled, snap-frozen and stored at À80 °C. Specific brain areas, namely the dHPC, were obtained by macrodissection, snap-frozen and stored at À80 °C, for RNA extraction and gene expression analysis. Whole brains were fixed in 4% paraformaldehyde (PFA, Panreac Química S.L.U., Barcelona, Spain) for 48 h and kept in paraffin blocks until further sectioning. Alternatively, brains were immediately embedded in Richard-Allan Scientific™ Neg-50™ Frozen Section Medium (ThermoFisher Scientific, Waltham, MA, USA), snap-frozen and kept frozen at À20 °C until further sectioning.

[3] 11w Values are reported as mean ± standard error of mean (SEM).

[4] 19w The number of biological replicates (N) of the representative independent experiments is specified in the legend of each figure.

[5] 87w Statistical significant differences between groups were determined using the parametric repeated measures two-way ANOVA with Bonferroni post hoc test (two-tailed) or the non-parametric two-tailed Mann Whitney test and Kruskal-Wallis with Dunns' multiple comparison test (two-tailed). Values were considered to be statistically significant for p < 0.05 ( ⁄ ), p < 0.01 ( ⁄⁄ ) and p < 0.001 ( ⁄⁄⁄ ). Simple correlations were obtained by regression analysis, using the Pearson's correlation test, and results were considered to be statistically significant for p < 0.05 (two-tailed).

[6] 62w Regression models (two-tailed) were designed and run in Matlab R2009b (7.9.0.529, Mathworks, Natick, MA, USA). To avoid collinearity, the independent variables age and IFN-a were centered, considering for that the respective mean values. To verify the statistical significance of the interactions between the different predictors, four interaction terms were created (age*genotype, age*IFN-a, genotype*IFN-a and age*genotype*IFN-a). No animals were excluded from the analysis.

[1] 127w Total RNA was extracted from CP tissue using the RNeasy Ò Plus Micro Kit (Qiagen, Hamburg, Germany), following the manufacturer's instructions. After quality assessment using the Agilent Bioanalyzer (Agilent Technologies, CA, USA), 500 ng of total RNA were used for T7-based mRNA amplification using the manufactured kit TargetAmp™ 2 -Round Biotin aRNA Amplification Kit (www.bioconductor.org) packages as described before (Coppola, 2011;Coppola et al., 2008). The raw data was deposited in the GEO database, accession number GEO: GSE66598. Gene pathway and functional analysis was performed using the Ingenuity Pathway Analysis 6.0 (IPA; Ingenuity Ò Systems, www.ingenuity. com). The CP transcriptome of WT mice at 3 (N = 3), 5-6 (N = 4) and 11-12 (N = 6) months, and of J20 mice at 3 (N = 4), 5-6

[2] 13w (N = 4) and 11-12 (N = 3) months, was analyzed and compared.

[3] 26w The individual data sets consisted on: (1) the comparison between the CP transcriptome of J20 mice and the CP transcriptome of WT mice, at different ages;

[4] 197w (2) the comparison between the CP transcriptome of 11-12 months-old mice and that of 3 months-old mice, for each genotype independently. The data sets containing the accession numbers, the log ratio and the p-value for each gene, whose expression was significantly changed (p-value <0.05), were uploaded into the IPA database. Using IPA and its applications, each gene identifier was mapped and the genes were used to build networks associated with a specific canonical pathway or with a relevant biological function or disease, according to the IPA database library. The Ingenuity Ò Upstream Regulator Analysis computed an activation z-score and a Fisher's exact p-value (overlap p-value) for the most altered genes. Furthermore, we used the IPA Upstream Regulator application to assemble an interactome using the five upstream regulators with the highest activation z-scores and the five upstream regulators with the lowest activation z-scores, along with the corresponding downstream target genes/molecules whose expression was significantly altered in the CP of J20 mice at each age. To be included in the interactome, the networks had to present at least one connection with an upstream regulator or with a downstream gene of a different adjacent network; orphan networks were not presented.

[5] 104w Primary cultures of CP epithelial cells were prepared from the CP tissue of rat pups due to the higher cellular yield of the isolation procedure when compared to the mouse CP. This procedure was performed as described before (Marques et al., 2009), with minor modifications. Briefly, neonates (P5-P7) were sacrificed and the CP dissected from the four brain ventricles, under a conventional light stereomicroscope (SZX7, Olympus). The tissue was rinsed in phosphate buffer saline (PBS, Invitrogen, Carlsbad, CA, USA) without Ca 2+ and Mg 2+ , followed by a 25 min digestion with 0.1 mg/mL pronase (Sigma-Aldrich, St. Louis, MO, USA) at 37 °C.

[6] 16w Predigested tissue was recovered by sedimentation and briefly shaken in 0.025% trypsin-EDTA (Invitrogen) containing 12.5 lg/ml

[7] 917w DNAse I (Roche, Amadora, Portugal). The supernatant was then 271 withdrawn and kept on ice with 10% fetal bovine serum (FBS, 272 Invitrogen). This step was repeated five times. Cells were collected 273 by centrifugation and re-suspended in culture media consisting of 274 Ham's F-12 and Dulbecco's modified Eagle's medium (DMEM-F12, 275 Invitrogen) supplemented with 10% FBS, 2 mM glutamine, 276 50 mg/mL gentamycin, 5 mg/mL insulin, 5 mg/mL transferrin, 277 5 ng/mL sodium selenite, 10 ng/mL epidermal growth factor, 278 2 mg/mL hydrocortisone and 5 ng/mL basic fibroblast growth fac-279 tor (all from Sigma-Aldrich). For further enrichment, cells were 280 incubated on plastic dishes for 2 h at 37 °C. The supernatant con-281 taining the CP epithelial cells was collected and placed for seeding 282 on 0.4 lm pore size transwells (Corning Life Sciences, Lowell, MA, 283 USA) coated with laminin (BD Biosciences, Bedford, MA, USA) at a 284 density of 1 Â 10 5 cells/cm 2 . The culture medium was changed 285 every two days and CP epithelial cells were maintained in culture 286 for 5-7 days, at 37 °C in a humid atmosphere (5% CO 2 ), before 287 performing the experiment. To assess the cellular purity and con-288 fluence after 7 days in culture, an immunofluorescent staining 289 was performed, using a primary antibody for transthyretin (TTR) 290 (kindly provided by Dr. Maria Joao Saraiva, Institute for 291 Molecular and Cell Biology, Porto, Portugal), a specific marker of 292 CP epithelial cells (Sousa et al., 2007). Cell counting revealed that 293 P95% of the cells stained positive for TTR. Stock Ab 1-42 294 (American Peptide Company, Sunnyvale, CA, USA) was prepared 295 by dissolving the peptides in Tris-NaCl buffer (150 mM NaCl and 296 50 mM Tris-HCl pH 7.4, all from Sigma-Aldrich) to a final concen-297 tration of 221.5 lM and kept at À80 °C. After reaching confluence, 298 CP epithelial cells were incubated for 60 h with DMEM-F12 alone 299 (vehicle) or a solution of 1 lM monomeric/dimeric Ab 1-42 , by dis-300 solving stock Ab 1-42 in DMEM-F12 medium. 301 2.7. Primary cultures of astrocytes and neurons 302 This procedure was performed as described before (Mesquita 303 et al., 2014). Details are provided in Supplementary methods 304 (Appendix A). 305 2.8. Gene expression analysis by qRT-PCR 306 The quality assessment and quantification of the total RNA 307 extracted from CP samples was performed in the NanoDrop Ò 308 ND-1000 and 350 ng of RNA from each sample were amplified 309 using the MessageAmp™ II aRNA Amplification Kit (Ambion, 310 Carlsbad, CA, USA) according to the manufacturer's instructions. 311 Total RNA was also extracted from primary cultures of CP epithelial 312 cells, using the RNeasy Ò Plus Micro Kit (Qiagen), or from samples of 313 dHPC, using the Trizol reagent (Invitrogen), following the manufac-314 turer's instructions. Then, after quantification in the NanoDrop Ò , 315 500 ng of total or amplified RNA from each sample was reverse 316 transcribed into cDNA using the iScript™ cDNA Synthesis Kit 317 (Bio-Rad Laboratories, Hercules, CA, USA) following the manufac-318 turer's instructions. Primers used to measure the expression levels 319 of selected mRNA transcripts by qRT-PCR were designed using the 320 Primer-BLAST tool of the National Center for Biotechnology 321 Information (Bethesda, MD, USA) on the basis of the respective 322 GenBank accession numbers. The reference gene hypoxanthine 323 guanine phosphoribosyl transferase (Hprt) was used as internal 324 standard for the normalization of the expression of selected tran-325 scripts. All GenBank accession numbers for Mus musculus (in vivo 326 studies) or Rattus norvegicus (in vitro experiments) gene transcripts 327 and primer DNA sequences are provided in Supplementary data 328 (Appendix A). Annealing temperatures are available on request. 329 The qRT-PCR was performed on a CFX 96TM real-time system 330 instrument (Bio-Rad), with the SsoFast™ EvaGreen Ò Supermix 331 (Bio-Rad) according to the manufacturer's instructions, using equal 332 amounts of cDNA from each sample. The cycling parameters were 1 cycle at 95 °C for 15 min, followed by 40 cycles at 94 °C for 15 s, 334 annealing temperature (primer specific) for 30 s and 72 °C for 30 s, 335 finishing with 1 cycle at 65 °C to 95 °C for 5 s (melting curve). 336 Product fluorescence was detected at the end of each elongation 337 cycle. All melting curves exhibited a single sharp peak at the 338 expected temperature. 339 2.9. Quantification of Ab and IFN-a in the CSF 340 The quantification of Ab in the CSF was performed by a direct enzyme-linked immunosorbent assay (ELISA) performed in Nunc 342 MaxiSorp Ò flat-bottom 96-well plates (ThermoFisher Scientific). 343 The wells were coated with 1 lL of CSF diluted in 99 lL of a 344 KH 2 PO 4 /K 2 HPO 4 buffer (pH 8.0) solution (1:100 dilution factor), 345 for 2 h at 37 °C. After washing with PBS 0.05% Tween Ò 20 346 (Sigma-Aldrich) and PBS, a blocking step with PBS 10% skim milk 347 (Nestle S.A., Linda-a-Velha, Portugal) was performed for 1 h at 348 room-temperature (RT). Then, consecutive incubations for 1 h at 349 RT were performed: first with rabbit anti-human Ab (1:500, 350 D54D2, Cell Signaling Technology, Danvers, MA, USA), second with 351 biotinylated horse anti-rabbit (1:500, Vector Laboratories, 352 Burlingame, CA, USA) and third with streptavidin-horseradish per- Olympus) and with the NeuroLucida software (Olympus). The samples were given a code name so that the morphology could be assessed by an investigator who was blinded to the experiment.

[8] 8w Data was analyzed with the NeuroExplorer software (Olympus).

[9] 41w The number of microglia/macrophages in the DG of the dHPC was assessed in serial brain sections and the number of cells per mm 2 was calculated (N = 4/group at 3 and 5-6 months, and N = 5/group at 11-12 months).