The animal experiments were conducted according to the national regulations of the usage and welfare of laboratory animals and approved by the Animal Experiment Committee in State Provincial Office of Southern Finland. The mice were group-housed under a 12-h light-dark cycle and allowed free access to standard rodent chow and water. Astrocyte cultures were prepared from C57Bl/ 6j tg mice expressing enhanced green fluorescence protein, eGFP [20], purchased from Jackson Laboratories (Maine, USA) and maintained in C57Bl/6j strain in the facilities of the Lab Animal Center in Kuopio, University of Eastern Finland.

The mouse brain sections used for ex vivo assays were prepared from tg APdE9 mouse line (created by co-injection of chimeric mouse/human APPSwe and PS1-dE9 vectors, both controlled by their own mouse prion protein promoter element (APdE9 mice [21] provided by Prof. Heikki Tanila) and their wt C57Bl/6j littermates. In the tg APdE9 mouse line, the first Aβ deposits appear at 4 - 5 months of age.

Hippocampi and cortices were isolated either from 2 to 3-day-old or 6 to 8-week-old tg eGFP mice and the tissue was suspended in DMEM/F12 (3:1, Gibco BRL, NY, USA) containing 10% heat-inactivated fetal bovine serum (Gibco BRL) and 100 U/ml penicillin-streptomycin (Gibco BRL). The suspension was triturated 10 × and centrifuged at 1500 rpm for 5 min at RT. After adding 0.25% trypsin-EDTA (Gibco BRL) the suspension was incubated at 37˚C for 30 min. Fresh culture medium was added and the suspension was centrifuged at 1500 rpm for 5 min. The cells were treated with Percoll (SigmaAldrich, St. Louis, USA), centrifuged at 1500 rpm for 10 min at RT or until the phases separated clearly and washed once with fresh culture media. The cells were plated onto poly-L-lysine coated flasks in DMEM/F12 (3:1) containing 10% heat-inactivated fetal bovine serum, 100 U/ml penicillin-streptomycin and G5 supplement (Gibco BRL) at the density of 4 × 10⁴ cells/ml. Before the experiments these mixed glial cell cultures were shaken to remove microglia at 200 rpm for 2 h. For the neonatal astrocyte cultures the culture medium was DMEM (+4500 mg/l glucose, + L-glutamine, -pyruvate, Gibco BRL) without G5 supplement and the isolation method included no Percoll treatment.

The cell density for purity analyses was 2 × 10⁴/well in a 48-well plate. Standard anti-GFAP (anti-glial fibrillary acidic protein, 1:500; Dako Cytomation, Glostrup, Denmark) immunocytochemistry was used to confirm the purity of astrocyte cultures. The possible microglial contamination of adult and neonatal eGFP expressing cells used for ex vivo experiments was evaluated using anti-Iba (2 µg/ml; Wako, Japan) and anti-CD11b (1:500 dilution; Serotec, UK). The cells were incubated with biotinylated goat anti-rabbit IgG (1:200 dilution; Vector Laboratories, CA) secondary antibody and Vectastain ABC Elite kit was used (Vector Laboratories) according to manufacturer’s protocol. Hydrogen peroxide (H₂O₂) and nickel-enhanced diaminobenzidine (Ni-DAB) were used to visualize the immunoreactivity. Three fields per well and at least three wells per cell type were analyzed for the percentage of Ni-DAB stained cells of all cellular profiles in the field. Both adult and neonatal astrocyte cultures contained on average 99.6% ± 0.4% GFAP-immunoreactive astrocytes. The cultures contained less than 0.4% of microglial cells in all experiments.

We examined the possible differences in gene expression profiles in eGFP expressing neonatal and adult astrocytes as depicted in Figure 1. Isolated neonatal or adult astrocytes (4 × 10⁵ cells/well) were cultured on top of APdE9 or wt mouse brain sections. Six sections per cell and section type were processed and before the mRNA isolation process, the astrocyte samples from two sections were pooled to yield n = 3. After 22 h culture co-incubation of plated astrocytes and brain sections the medium was

removed from the wells and the brain sections with cultured eGFP astrocytes were incubated with 1 × trypsin-EDTA (500 µl/well) for up to 15 min at 37˚C. The detachment (at this point, the astrocytes do not lose the grip completely) of the astrocytes from the underlying brain section was carefully monitored with a standard light microscope. To inactivate trypsin, the brain sections with cultured astrocytes were carefully removed into a small Petri dish filled with fetal bovine serum containing culture medium. Subsequently, the brain sections were immediately examined with a fluorescence microscope (Olympus IX-71, Japan). The eGFP expressing astrocytes were separated from the underlying brain section using a special suction hose modified for this purpose from suction hoses typically used for the microinjection techniques. One end of the tube contained a pipet tip as a mouthpiece, the other end a stretched glass capillary (prepared with a Bunsen burner) and in the middle of the tube there was a 0.22 µm filter. The eGFP expressing astrocytes were gently picked up from the underlying brain section using wavelength 468 nm of the fluorescence microscope and simultaneous light microscopy. Two parallel samples were collected into one 15 ml tube and the cells were centrifuged at 1500 rpm for 5 min. The supernatant was removed and the remaining pellet was resuspended in 1 ml of culture medium. The cell suspension was removed into an Eppendorf tube and centrifuged again at 12,000 rpm for 3 min.

Ribopure^TM Kit (Ambion, Applied Biosystems, USA) was used for the total mRNA isolation according to the manufacturer’s instructions. The cell pellet was lysed by triturating with 100 µl TRI reagent.

The microarray procedure was performed at Biomedicum Genomics (BMGen), University of Helsinki, Finland. The qualities of mRNA samples were evaluated using Bioanalyzer/Nano RNA assessments. After the samples passed the quality control criteria, BMGen performed gene expression analyses using 12 × Affymetrix GeneChip Mouse Genome 430 2.0 hybridizations (2-Cycle method). Three replicate arrays for each sample group were used and scanned with an Affymetrix GeneChip scanner. All data from arrays passed the quality control (Chipster v1.4.7; CSC, Finland; [22]) performed according the manual of the software (http://chipster.csc.fi/manual/tools. html^#quality). The rest of data analysis was made with Genespring GX 11.0 (Agilent, Santa Clara, Calif., USA) by preprocessing the raw data with the robust multi-chip algorithm [23], normalizing per chip to the median and filtering the differentially expressed data of all 12 microarrays containing 45,101 probes by fold change >2.0 or <0.5. The change in the astrocyte gene expression was identified according to the sections the astrocytes were incubated with (wt or tg; Figure 1). The differenttially expressed probe sets were further processed using statistical analysis with Welch unpaired t-test assuming unequal variances (p < 0.05; fold change >2.0 or <0.5). Benjamini-Hochberg false-discovery rate (FDR) correction for multiple testing was used to control the number of false-positives results. By using MS Excel, replicate probes for individual genes were combined by taking averages of values and probes without gene annotations were excluded. Differentially expressed genes were grouped with their molecular function and biological process terms of Gene Ontology (GO) annotations by using visualization program Cytoscape [24] and its plugin BiNGO 2.4.2 (p < 0.05, Benjamini-Hochberg FDR, whole GO annotation as reference set) [25]. Molecular function groups were categorized with CateGOrizer 3.2.18 [26] by making the classification with modified Generic GO_slim-list (http://www.geneontology.org/GO.slims.shtml).

The expression profiles of the astrocytes were compared to publicly available datasets [27], created by using the same Affymetrix GeneChip Mouse Genome 430 2.0 Array (GEO accession: GSE9566). The additional data consisted of expression data of highly purified brain cells included astrocytes from early postnatal age at postnatal day 1, and later postnatal age at day 30 when astrocytes are morphologically fully differentiated. In addition, the datasets included expression profiles from neurons and oligodendrocytes. One dataset was a mixed sample of mouse forebrain. The additional datasets were loaded as original CEL files and normalized with the same protocol as the present data. The combined datasets were clustered using standard correlation coefficient.

Synthesis of cDNA was made by using 400 ng of total RNA, Maxima reverse transcriptase and random hexamer primers (Fermentas life sciences). Six samples were prepared for each sample group. The relative expression levels of mRNA encoding several genes were measured according to manufacturer’s protocol by quantitative RT-PCR (StepOnePlus™ Real-Time PCR System; Applied Biosystems, USA) by using specific assays-on-demand (Applied Biosystems, USA) target mixes. The mean expression levels were normalized to Beta-2 microglobulin (B2m) and the four relative values representing the four sample groups (neonatal or adult astrocytes on wt or tg APdE9 brain sections) for each gene were scaled to range 0 - 100 ± SEM.

A statistical probability of p < 0.05 using MS Excel (two-sample t-test) and Genespring GX 11.0 (Welch unpaired t-test) was considered significant. The results of the qRT-PCR are represented as the mean ± SEM and other results are shown as basic mean values.

To verify that the GFAP-immunoreactive cells analyzed in microarray experiment were indeed astrocytes, we compared our data (GEO accession number: GSE29317) to published microarray data sets of highly purified astrocytes, neurons and oligodendrocytes [27]. Our data sets clustered well with control data of in vivo purified adult astrocytes and in vitro cultured neonatal astroglia (Figure 2). Minor deviations observed possibly indicated difference in culture conditions and ages of tissues used for cell isolation.

By comparing expression profiles between the astrocytes cultured on top of wt and tg APdE9 brain sections, 877 genes in adult and 132 genes in neonatal astrocytes were differentially expressed two fold or greater (Figure 3(a)). Up-regulation here means that gene expression in astrocytes cultured on tg APdE9 brain section was higher than in astrocytes cultured on wt section, down-regulation conversely. While one third of the genes showing altered expression when cultured on tg APdE9 brain sections were up-regulations both in adult and neonatal astrocytes, only 13 genes of the up-regulated and 30 genes of the down-regulated genes were the identical. To illustrate the differences in gene expression related to certain cellular functions, the genes with altered expression were classified with Gene Ontology (GO) annotations (Figures 3(b), (c); 4(a), (b)). Adult astrocytes upand down-regulated grossly equal number of genes belonging to categories of catalytic activity and hydrolase activity when cultured on top of APdE9 brain sections. However, there was a clear shift towards up-regulation of peptidase activity, scavenger receptor activity and glutathione transferase activity (Figures 3(b), 5(a)) which are cellular functions that could enhance clearance of Aβ or reduce its toxic effects. Correspondingly, several peptidase inhibitors were specifically down-regulated by adult as-

trocytes grown on tg APdE9 brain section (Figure 5(a)). Even though gene expressions of peptidase inhibitors were down-regulated also in neonatal astrocytes, GO analysis did not generate any enriched groups of peptidases or proteases in these cells when cultured on top of APdE9 brain sections (Figures 4(b), 5(b)). GO class transferase activity was highly enriched among the genes down-regulated in neonatal astrocytes (Figure 4(b)) and most of those genes were related to cholesterol synthesis (Figure 5(b)). Based on microarray results, neonatal astrocytes down-regulated but also up-regulated several genes associated with endocytosis when cultured on top of APdE9 brain sections (Figure 5(b)). For example, the array revealed an increased expression level of macrophage scavenger receptor 1 (Msr1), which has been shown to mediate endocytosis of Aβ in microglia [28].

To confirm the results obtained from the microarray analysis, we selected 13 genes and analyzed their expression by qRT-PCR. Selection was made by evaluating the differences and similarities between adult and neonatal expression profiles as well as selecting genes possibly related to clearance of Aβ based on GO analysis and literature. The genes with very low expression levels in microarray data were excluded from further study, for example Cathepsin S (Ctss) had very high fold change value but expression levels were extremely low.

Adult astrocytes grown on tg APdE9 brain sections had very high, over 200 fold up-regulation of Scavenger receptor class A, member 5 (Scara5), which has been found to mediate phagocytosis of bacterial material and ferritin bound iron [29,30]. Scara5 was expressed at extremely low levels in neonatal astrocytes and adult astrocytes grown on wt brain section (Figure 6). Apolipoprotein E (apoE) has been shown to be the major susceptibility gene for AD [reviewed in 31-34]. Our microarray results demonstrated that adult up-regulated apoE over 3 fold, but only down-regulation in neonatal astrocytes was confirmed by qRT-PCR. However, expression of apoE was 10 times higher in neonatal than in adult astrocytes grown on tg APdE9 brain sections (Figure 6). Also, neonatal astrocytes seemed to down-regulate two genes encoding enzymes of cholesterol synthesis: 3-hydroxy-3- methylglutaryl-CoA synthase 1 (Hmgcs1) and 24-dehydrocholesterol reductase (Dhcr24) in same extent as apoE (Figure 6).

Phosphate regulating gene with homologies to endopeptidases on the X chromosome (Phex) has been proved to degrade synthetic Aβ₄₀, even though only to minor extent [35]. When grown on tg APdE9 brain sections, adult astrocytes up-regulated Phex over 10 fold whereas expression by neonatal astrocytes was very low (Figure 6). A member of high temperature requirement family of serine proteases, Htra1, has also been shown to decrease Aβ levels in vitro and cell cultures (Figure 6) [36]. Both adult and neonatal astrocytes expressed moderate levels of HtrA serine peptidase 1 (Htra1), but only adult astrocytes up-regulated this protease on tg ApdE9 brain sections compared to astrocytes grown on wt mouse brain sections. Based on microarray results, among all matrix metallopeptidases and well-known Aβ degrading enzymes, neonatal astrocytes expressed very high level of matrix metallopeptidase 9 (Mmp9; data not shown). Quantitative RT-PCR confirmed significant up-regulation of Mmp9 in adult astrocytes cultured on tg APdE9 mouse brain section compared to astrocytes cultured on wt mouse brain section (Figure 6).