Microscope round cover glasses #1.5 (12 mm Menzel-Gläser 1000 Deckgläser/cover slips) (cat# 11846933, Fisher Scientific, Waltham, MA, United States), staining rack (cat# 8542E40, Thomas Scientific, Swedesboro, NJ, United States); DNase I grade II, from bovine pancreas (cat# 10104159001, ROCHE, Basel, Switzerland); Dumont no. 5 forceps (cat# 11252-23, Fine Science Tools, Heidelberg, Germany); Natriumtetraborat (cat# 11648, RdH Laborchemikalien GbH and Ko KG, Selze, Germany); 2.5% Trypsin (10x), 100 ml, (cat# 15090-046); 0.05% Trypsin-EDTA (1x) (cat# 25300-054, Gibco, Carlsbad, CA, United States); GlutaMAX™-I supplement (cat# 35050-038); B-27 Serum-Free Supplement (50X) liquid (cat# 17504-001 or 17504-044), Neurobasal (cat# 21103-049) were from Gibco or Invitrogen, Carlsbad, CA, United States, and Primocin was from (cat# ant—pm-1, Invivogen, San Diego, CA, United States). The HEPES buffer, 1 M pH 7.3 (cat# 15630-080) and Nunclon Delta Surface, sterile 12 well cell cultures plastic dish (cat# 150628) from Thermo Scientific, Waltham, MA, United States. Tissue culture flask T75 (cat# 83.3911.002, Sarstedt Nürnbrecht, Germany). The 5-Fluoro-2′-Deoxyuridine (FUDR) (cat# F0503-100MG) and Bovine serum albumin (BSA), cat# A9647-50G from Sigma, Saint Louis, MO, United States. The Aβ42, Aβ42-TFA (cat# A-42-T-1), and Aβscr, Aβ-scrambled (cat# A-42-S-1) from GenicBio, Shanghai, China. Paraformaldehyde (PFA) 16% solution, EM Grade (cat# 15710) from Electron Microscopy Sciences, Hatfield, PA, United States. CellTiter 96 Non-radioactive Cell Proliferation Assay (cat# S G4000 and G4100, Promega, Madison, WI, United States). The 4′,6-Diamidino-2-phenylindole (DAPI) Fluoromount-G from (cat#0100-20, SouthernBiotech, Birmingham, AL, United States).

1° Antibodies: Astrocyte marker: anti-GFAP antibody (1:1,000) cat# MAB360, (clone GA5 Millipore, Burlington, MA, United States), anti-GLAST (1:1,000) cat# ab416, Abcam, Cambridge, United Kingdom. Neuronal and synaptic marker: anti-synaptophysin antibody (1:250), cat# ab32594 Abcam; anti-MAP2 antibody (1:500), cat# M9942 and anti-ß-III-Tubulin antibody (1:250), cat#T8328 were both from Sigma, St. Louis, MO, United States; anti-VGlut1 antibody (1:250), cat#135 302, Synaptic Systems, Göttingen, Germany. Amyloidβ antibody: Anti-ß-Amyloid (1:300), 1–16 antibody, cat#9320-02, Previously Covance catalog# SIG-39320, Biolegend, San Diego, CA, United States. Microglia marker: anti-Iba1 antibody (1:1,250), cat# 019-19741, Wako Pure Chemical Industries, Osaka, Japan. Oligodendrocyte marker: anti-Olig2 (1:100), cat# MABN50, Millipore, Burlington, MA, United States; anti-O4, (1:250), Cat# MAB345, Millipore. Fibroblast marker: anti-FSP antibody (1:1,000), cat# ABF32, Millipore; anti-smooth muscle actin antibody (1:1,000), cat# ab5694, Abcam.

2° Antibodies: Goat anti-mouse Alexa 488 (1:500), cat#150113; donkey anti-rabbit Alexa 594 (1:500), cat# 31572; Donkey anti-rabbit Alexa 488 (1:500), cat# A21206; donkey anti-mouse Alexa 594 (1:500), cat#15108 all from Thermo Scientific.

All research procedures involving animals were according to and approved by the animal care and ethics committee of the City of Hamburg (permit number: ORG739 Molecular Mechanisms of Dementia). Primary mouse astrocyte (dissection #1) and primary mouse neuronal cultures (dissection #2) (Figure 1) were derived from mixed CNS cultures from male or female p0-1 C57BL/6J mice (Charles Rivers, Wilmington, MA, United States), as described by Kaech and Banker (2006).

Proper sterile technique and handling were applied throughout the work with primary cell cultures. Cell culture media or soluble components were filter-sterilized before use. Dissection tools were regularly autoclaved and kept in 70% ethanol (EtOH) after unpacking. Immediately before their use, they were rinsed with dissection medium (DM) containing Hanks’ Balanced Salt Solution (HBSS) 1x, 1% penicillin-streptomycin, 10 mm HEPES, and 0.6% of 45% glucose solution. As in previous protocols, dissection (skin, skull, and brain) was performed with different dissection tools to minimize the contamination risk (Kaech and Banker, 2006; Beaudoin et al., 2012). After decapitation, the head was immersed for a few seconds in 70% EtOH by putting it into a cap of a 50 ml tube, then placed into another plastic falcon filled with chilled DM. Next, the skin flap of the skull was removed using forceps. Tissue preparation was done under a binocular microscope, as shown by Beaudoin et al. (2012). Briefly, separate curved forceps were used for meticulous removal of meninges and of visible blood vessels. It is well-known that insufficient removal of meninges and choroid plexus can lead to contamination with meningeal fibroblasts and endothelial cells (Kaech and Banker, 2006; Beaudoin et al., 2012). For each animal, two halves of the cortex (for astrocyte feeder cultures) and two hippocampi (for neuronal cultures) were obtained and placed in a 15 ml falcon with a chilled DM, then placed on ice. Tissue from four mice was pooled to obtain sufficient cell numbers for the primary cultures. Under the sterile bench, the freshly dissected tissue was transferred with a 5 ml Pasteur pipette into a 15 ml plastic tube on ice. Two milliliters of DM, including tissue pieces, were transferred into a 60 mm dish. Total volume of DM should reach 4.5 ml DM with 0.5 ml pre-warmed 2.5% Trypsin added (Kaech and Banker, 2006; Jones et al., 2012; Schildge et al., 2013). The resulting mixture in 60 mm dishes was covered with parafilm to avoid evaporation and then incubated at 37°C for 15 min in the cell incubator at horizontal agitation of 400 rpm (Schildge et al., 2013). After digestion, 100 μl of sterile 1 mg/ml DNase I was added and was gently swirled. After 1 min, 5 ml of glial growth medium (GGM) (DMEM, 10% FBS, 0.6% v/v from 45% glucose solution, and 1% penicillin-streptomycin) was added to quench the enzymatic reaction (Kaech and Banker, 2006; Jones et al., 2012). Subsequently, gentle trituration with a blue 1,000 μl pipet tip (10 times) was added and the mixture (10 ml total volume) was transferred into a new 15 ml vial. Cells were harvested by centrifugation (5 min at 1,000 rpm). Next, the supernatant was decanted, and 2 ml of GGM was added with a 5 ml plastic pipette to the small cell pellet, which was then gently triturated with a fire-polished glass pipette seven times. After residual tissue pieces were able to sediment, tissue was subjected to another trituration step (seven times). An additional 2 ml of neural maintenance medium (NMM) (Neurobasal medium, 1% 2 mM Glutamax, 2% B27 serum supplement, 100 μg/ml Primocin) was added to the cell suspension. Cells were then centrifuged for 5 min at 1,000 rpm and the resulting supernatant was decanted with the pellet being resolved in 2 ml GGM. A 70 μm cell strainer was placed over a new 50 ml vial and its membrane surface was pre-equilibrated with 500 μl of GGM. Subsequently, the cell suspension was resuspended by gently pipetting up and down 3 times. Cells were then filtered through the cell strainer. Residual cells in the 15 ml tube were washed with an additional volume of 500 ml of GGM and strained into the 50 ml vial. Another fresh 1,000 μl pipette tip was used to aspirate the remaining cell suspension from the bottom side of the cell strainer to harvest additional cells that still adhered to the surface of the cell strainer. Typically, the Hippocampi of the brains of four mice that were dissected and pooled together yield approximately 500,000 cells/ml in 4 ml of a total volume, whereas two cortices yield 1,200,000 cells/ml in 4 ml of total volume. Cell viability was semi-automatically evaluated using a cell counter based on the Trypan blue exclusion and should normally range between 85 and 90%. The cell suspension of the mixed CNS cells was then plated into a T75 cm² tissue culture flask, which was filled up to 12.5 ml with GGM. The next day, the cells were carefully washed with Dulbecco’s Phosphate Buffer Saline (DPBS) 1x, and the GGM medium was changed. Mixed CNS cultures were monitored every second day, with an additional wash step and medium change every 5–7 days.

One important modification introduced in this study, for the neuron-astrocyte co-cultures, consists of the incorporation of a second glass coverslip that is placed in the bottom well of a multi-well plastic dish for the lower portion of the co-cultures (astrocyte feeder layer). Glass coverslips of 12 mm in diameter for 24-well plates and 18 mm in diameter for 12-well format were used. Before placing each coverslip into a well of a multi-well dish, dots of heated paraffin were placed on the coverslips in a triangular fashion, which will later provide the physical separation between the lower coverslip (with the astrocyte feeder layer) and the upper coverslips (with the neuronal cell layer) (Figure 1; Kaech and Banker, 2006; Gottschling et al., 2016). The optimal temperature to produce the roundly-shaped paraffin wax dots adhering to the plate was found to be at 110°C. Temperatures below this led to an insufficient adhesion of paraffin wax dots and detachment during the cell culture. At the same time, paraffin should not be too hot to avoid the spreading of the paraffin over the surface of the coverslip (Kaech and Banker, 2006). Subsequently, the prepared multi-well dishes, each is well-equipped with glass coverslips with triangular wax dots, were placed under a sterile bench and irradiated with a UV light for 30 min to minimize the risk of contamination. Before plating the primary astrocytes for co-culture experiments onto the wax dot-coverslips in multi-well dishes, mixed CNS cell cultures were subjected to overnight shaking at 500 rpm at 37°C to reduce the amount of contaminating microglia, as published in previous reports (McCarthy and de Vellis, 1980; Kaech and Banker, 2006; Jones et al., 2012; Schildge et al., 2013). After shaking, the supernatant containing free-floating microglial cells was discarded. Contamination of the astrocyte feeder cells with other cell types was assessed by immunofluorescence analyses and by quantification of the different cell populations (see Figure 2). Of note, the astrocyte cultures with gross contaminations with fibroblasts should be discarded and not used for the co-culture studies. Plating of astrocyte feeder cells originating from the first dissection (dissection #1) was performed regularly for 48 h before the “neuronal” dissection (dissection #2) (see Figure 1). Astrocyte feeder cells were washed 1x with DPBS1x and were detached by trypsin for 10 min. Cells were harvested and the trypsin was removed by spinning down 1,000 rpm for 5 min, with the 80,000 astrocytes being plated per coverslip/well of a 12-well plate on top of an 18 mm wax dot-coverslip (lower coverslip) (see Figure 1).

The medium of the astrocytes was changed 24 h after plating (i.e., 24 h before neuronal dissection), from GGM to NMM (see above), with a total per-well volume of 2 ml, which marks the onset of pre-conditioning of the NMM medium. Astrocytes now start to produce paracrine factors that will subsequently facilitate the growth and maturation of co-cultured neurons (Kaech and Banker, 2006).

Importantly, the glass coverslips for neurons (upper coverslip) require a special pre-treatment before use, including the nitric acid-based treatment and banking at 200°C, whereas the coverslips used for astrocytes did not require any treatment (Kaech and Banker, 2006). Nitric acid is used to roughen the surface of glass coverslips which promotes neuron adhesion and differentiation (Kaech and Banker, 2006; Schildge et al., 2013).

Glass coverslips for neuronal cells (18 mm in diameter) were placed into ceramic racks, rinsed with ddH20, and placed in 65% nitric acid for at least 18–24 h (Kaech and Banker, 2006). Subsequently, glass coverslips were rinsed three times for 3 min with ddH2O, placed in 96% EtOH for 30 min, and again rinsed three times for 3 min and 5 min with ddH2O. It was then placed under the sterile bench to dry and to irradiate with UV for at least 30 min. Finally, the treated and dried coverslips were baked at 200°C for at least 4 h. Under sterile conditions, the pre-treated coverslips were then placed into separate wells of a 24-well plate, which was then packed into sterile plastic bags, sealed, and stored under sterile before use. In our experience, the advance production of pre-treated glass coverslips is crucial for the workflow of the assay.

For optimal adhesion, pretreated and baked coverslips were placed in wells of a separate 12-well multi-well dish and further treated with Poly-L-lysine (PLL; 100 μg/ml end concentration) for at least 24 h before cell plating. For this, PLL was dissolved in 1 mg/ml in 0.1 M borate buffer (Kaech and Banker, 2006; Beaudoin et al., 2012). The pH was set to 8.5 (1.9 g borax + 1.24 g boric acid in 400 mL H2O), then filter-sterilized (Beaudoin et al., 2012). Approximately 1 ml of PLL solution (working concentration 100 μg/ml) was added per well and coverslip. Shortly, the PLL was removed with three washes using DPBS1x before plating the neurons. The DPBS1x was left just until the plating (Kaech and Banker, 2006; Seibenhener and Wooten, 2012).

The cultivation of primary neurons, dissection of brain tissue, and generation of a single cell suspension were carried out as described for primary astrocytes with similar working steps (dissection #2) and with the following differences. Serum-free NMM was used for cultures of neurons. A total of 50,000 or 25,000 freshly dissected CNS cells were plated onto baked and poly-L-lysine-coated coverslips that were placed in wells of a 12-well dish. For plating cells on a 24-well, a format adaption of the number of plated cells was necessary.

Co-culturing primary neurons with feeder astrocytes aimed to improve neuronal survival and differentiation up to 14DIV with subsequent experimental manipulation. Briefly, freshly dissected mixed primary CNS cells (dissection #2) were plated onto coverslips that were placed into wells of a 12-well plastic dish. About 1 h after plating, coverslips were inverted and hung over previously prepared feeder astrocytes of another 12-well dish. Importantly, wax dots were placed on the lower coverslip before the astrocyte feeder cell plating. Hence, paraffin wax dots were functioning as spacers to ensure spatial separation between the upper (neurons) and lower (astrocyte) coverslips (Figure 1). Subsequently, 1,000 μl of fresh NMM was added to the remaining volume of the medium. The addition of the mitotic inhibitor 5-Fluoro-2′-Deoxyuridine (FUDR) with a working concentration of 10 μM to deplete other mitotic cell types, such as microglia, oligodendrocytes, and fibroblasts, and to dampen glial proliferation in the upper coverslips, was done only in the very first medium change of the co-culture at day 4 (4 DIV), whereas the changes in the subsequent medium performed every 3 days did not include a mitotic inhibitor (Kaech and Banker, 2006). Co-cultures were monitored daily, and medium changes were performed every 4 days. Importantly, medium changes included 50% removal of an old medium, while the addition of the same volume of fresh medium, using a 1,000 μl pipette to not fully deplete the medium of co-cultured neurons as media, contain paracrine astrocyte-derived factors being critical for proper growth and differentiation of the co-cultures.

For induction of reactive astrocytes and assessment of paracrine, non-cell-autonomous neurotoxic, and/or synaptotoxic effects, astrocyte feeders were pre-treated with oligomeric Aβ species; either Aβ42 or Aβ scrambled (Aβscr; negative control). Two millimolar stock solutions of Aβ42 or Aβscr were purchased (GenicBio, Shanghai, China) and prepared according to Falker et al. (2016). Briefly, stock solutions were dissolved in 110 μl ultra-filtered Dimethyl sulfoxide (DMSO) with a subsequent centrifugation step (5,000 × g 1 min) to obtain 2 mm stock solutions. Corresponding aliquots were immediately stored in an −80°C deep freezer and thawed directly before the next use. Using this preparation, the starting solution of Aβ mostly contained monomers and small oligomers, but rapidly aggregated into bigger oligomers during the handling (see Falker et al., 2016 for details such as Western Blot or Thioflavin aggregation assay). Of note, other preparations of Aβ, such as mixtures of Aβ40/Aβ42, oligomer preparations, or fibrillary Aβ, might also be used to assess differential effects of diverse Aβ subtypes and aggregation states of astrocytes, as well as their subsequent impact on the neurons. The working concentration for Astrocyte treatment was 2 μM for both Aβscr and Aβ42 (Lau et al., 2014). The LPS was prepared according to Liddelow et al. (2017), and different concentrations were tested with 100 ng/ml as the final LPS concentration for the experiments. After the treatment of the astrocyte feeder layer for 24 h, treatment media was changed to fresh media without LPS or Aβ42 (conditioning). Primary neurons at 14DIV were then inverted over and co-cultured with pre-treated feeder cells for 24 h.

Immunofluorescence staining was performed to stain fixed cells either for astrocyte and/or neuronal markers (Kaech and Banker, 2006; Beaudoin et al., 2012; Nwabuisi-Heath et al., 2012; Lian et al., 2015). Briefly, after media removal and a washing step with pre-warmed DPBS1x, coverslips were incubated for 10 min with pre-warmed 4% p-formaldehyde (w/v) and 4% sucrose (w/v) fixation solution. After subsequent washing steps, cells were permeabilized by incubating in DPBS1x containing 0.3% Triton X-100 for 10 min, then incubated in a blocking solution (DPBS1x) containing 1% bovine serum albumin (w/v) for 1 h. Primary antibody incubation (see section “Materials” for dilution) was performed for 18 h overnight at 4°C. Coverslips were intensively washed, followed by 2 h incubation with secondary antibodies conjugated to either Alexa Fluor 488 or 594, with subsequent final washing steps in DPBS1x. Changes in astrocyte morphology were visualized by the staining of actin cytoskeleton using the Alexa Fluor™647 Phalloidin (#A22287, Invitrogen by Thermo Fisher Scientific) according to the protocol of the manufacturers after the staining procedure was concluded. After repeated washing steps, mounting was performed with nuclear counterstaining involving 4’,6-Diamidino-2-phenylindole (DAPI) Fluoromount-G. Slides were dried and stored at room temperature in the dark. Images were stored as TIFF files measuring 1,024 × 1,024 pixels with a 63x immersion oil lens objective and a numerical aperture of 1.2 (N.A. = 1.2) using a Leica TCS SP5 confocal laser scanning microscope (Leica Microsystems, Mannheim, Germany). In the case of imaging astrocytes, images were stored as TIFF files measuring 512 × 512 pixels with a 20x objective microscope (Axiovert S100, Carl-Zeiss, Jena, Germany).

All experiments were independently performed at least three times. For imaging of primary neurons, 15 neurons were imaged per well and coverslip experimental condition (technical duplicates). Three random fields of neuronal dendrites per neuron (zoom = 1.5) were chosen with equal settings maintained in all neurons for each of the three experiments. Astrocytes were imaged as follows: five fields of a given coverslip were imaged (upper, lower, center, right, and left) with three coverslips of the analyzed separate wells (three technical replicates). In total, for experiments with astrocytes, three independent experiments, each with triplicates for each condition, were conducted.

Fluorescence (signal) intensity or integrated density (IntDen) of a given cell marker of interest can reflect protein content and can be used to compare markers in both untreated vs. treated groups of cells. Color threshold and exposure times of a selected marker of interest (e.g., MAP2 or GFAP) were set and maintained constant for each neuron and astrocyte analyzed for each of the three independent experiments. The IntDen values per image were calculated via ImageJ according to previous reports (Kaech and Banker, 2006; Marchetto et al., 2010; Gottschling et al., 2016; Odawara et al., 2016). To obtain IntDen values of markers of interest that are specific to neurons or astrocytes, raw confocal and fluorescence images stored as TIFF files had to be converted into and stored as 8-bit images. Images were processed via ImageJ (bundled with Java 1.8.0_172) and split into single channels and images. For instance, in the case of neurons, images were split into a red channel (MAP2) and a green channel (GFAP; Synaptophysin—Syn). In the case of astrocytes, the GFAP was used for IntDen quantification and served as a readout for GFAP signal intensity. Sholl analysis served as a readout for the neurotoxicity based on a common cell-fill marker, with MAP2 chosen in this study. As a cell-fill marker, MAP2 depicts the branching pattern of dendrites by plotting the number of branches as a function of the distance from the center of a cell (soma) (Ferreira et al., 2014). Here, we conducted an analysis of linear Sholl plots of our primary neurons. According to Ferreira et al. (2014), we extracted the key indices from linear Sholl plots, including the following: (a) Critical value (Nm), the maximum of the polynomial function, an indicator of maximal branching; (b) Critical radius (rc), the distance at which the Critical value occurs; and (c) Mean value of the polynomial function (Nav), which implies the average number of intersections. We used the Sholl analysis by accessing it as a plug-in from ImageJ14 programmed in Java and bundled in Fiji15. Here, the geometric center (i.e., soma) was manually set and marked using the point tool in Fiji, and the image was analyzed with the Fiji plugins Bitmap Sholl Analysis (Ferreira et al., 2014). Sholl analysis is a technique enabling the quantification of indices of axon length and neurite branching complexity and was carried out with a starting radius of 12 μm progressing with 10 μm intervals to a maximum radius of 120 μm. In addition, the lower threshold was set to 0, while the upper threshold to 128. All neurons of each group were subjected to the same settings.

Synaptic Puncta Analysis (SynPAnal) served for synaptic puncta analysis. Three randomly chosen non-overlapping dendritic segments per neuron were analyzed for quantification of synaptic puncta (Danielson and Lee, 2014). Synaptic puncta were recognized as puncta positive for the presynaptic marker Syn showing specific signal enrichment, and proximity to the dendritic marker MAP2, similarly to previous reports (Kaech and Banker, 2006; Beaudoin et al., 2012; Lian et al., 2015; Gottschling et al., 2016). Spine number and intensity were normalized with an in-built calibration tool provided by the SynPAnal software to the measured length of a given dendritic segment that was tracked with the free-hand selection tool to provide results of the number of synaptic puncta per 100 μm, the density of puncta per μm², and intensity of synaptic puncta per length of 100 μm or per area in μm² (Danielson and Lee, 2014). Soma size was manually examined according to previous studies in the field using confocal 1,024 × 1,024-pixel images (zoom = 1.2x) by outlining the soma in the red MAP2+ channel, omitting dendrites (Daub et al., 2009; Marchetto et al., 2010).

Astrocytes were cultured as described above. After treatment with different AD-related stimuli or respective controls for 24 h, the astrocytes were carefully washed in pre-warmed phosphate buffered saline (PBS) and directly lysed in RNA isolation kit lysis buffer. Total RNA was extracted using the mirVana™ miRNA isolation kit (Invitrogen by Thermo Fisher Scientific; Vilnius, Lithuania) according to the protocol of the manufacturer. For the conventional quantitative reverse transcription polymerase chain reaction (qRT-PCR), the total RNA (30 ng) with specific mRNA probes (Applied Biosystems) were used after reverse transcription reaction according to the manufacturer (high-capacity cDNA Reverse Transcription Kit; Applied Biosystems). All mRNA amplifications were performed using the TaqMan Gene Expression Master Mix 5 ml (#4369016) with commercially available FAM-labeled Taqman probes (Applied Biosystems by Thermo Fisher Scientific). Specifically, these are Cxcl10 (Mm00445235_m1), Tnf (Mm00443258_m1), Ccl5 (Mm01302427_m1), Ccl3 (Mm00441259_g1), Cxcl1 (Mm04207460_m1), Ccl2 (Mm00441242_m1), Timp1 (Mm01341361_m1), C3 (Mm00437858_m1), and GAPDH (Mm99999915_g1). The levels of the mRNA were normalized relative to the GAPDH. Real-time PCR reaction was performed using Quantstudio5 (Applied Biosystems by Thermo Fisher Scientific). All qRT-PCRs were performed in duplicates and the data are presented as a relative expression compared to GAPDH as mean ± s.e.m.

Astrocytes were cultured as described above. After treatment with different AD-related stimuli or respective controls for 24 h, the cell culture supernatant was harvested and cleared from the cellular debris by centrifugation. Cytokine levels in the astrocyte culture supernatants were measured with the Proteome Profiler Mouse Cytokine Array Panel A (ARY006, R&D Systems, Bio-Techne) according to the instructions of the manufacturer. Briefly, the membranes coated in duplicates, with capturing antibodies against 40 cytokines, were blocked and then incubated with the cleared astrocyte culture supernatant in the presence of a cocktail of the biotinylated detection antibodies overnight at 4°C. After washing, membranes were incubated with horse redish peroxidase (HRP)-conjugated antibodies, followed by chemiluminescent detection. The array data, duplicate spots representing the abundance of each cytokine or chemokine, were quantitated using the Licor Image Studio Light software to generate a protein profile, and results are presented as integrated densities.

Effects of Aβ42 or LPS treatment on astrocyte survival were probed by performing the CellTiter 96^® Non-radioactive Cell Proliferation Assay (Promega, Madison, WI, United States), according to the instructions of the manufacturer (Wang et al., 2012). Briefly, astrocytes were treated for 24 h with different treatment conditions (DMSO 1:1,000, Aβscr 2 μM, Aβ42 2 μM or water, and LPS 10 ng/ml or LPS 100 ng/ml). Experiments were terminated upon the addition of 15 μl of MTT dye solution per well of a 96-well plate, according to the protocol of the manufacturer. Next, after 4 h incubation at 37°C with MTT solution, 100 μl of solubilization/stop solution were added per well. After 1 h, absorbance at 550 and 650 nm were measured using a microplate reader (Molecular Devices, CA, United States).

Data analysis and visualization were performed with the GraphPad Prism 8 (GraphPad Software Inc., La Jolla, CA, United States). Data were represented as mean ± SD, if not otherwise specified. Ordinary one-way ANOVA for multiple comparisons was used followed by a post hoc Tukey’s test. A probability of p < 0.05 was considered statistically significant, whereas p values higher than 0.05 were considered as non-significant. Non-significant results were not indicated within the bar blots for the sake of readability and clarity of the data. All experiments were repeated at least three independent times (N ≥ 3). We performed at least technical duplicates for each experiment and used hippocampal neurons originating from different animals.

To set-up neuron-astrocyte co-cultures, we first assessed the amount and the viability of our single-celled suspension obtained from a freshly dissected and digested CNS tissue using a semi-automatic cell counter (Countess II Automated Cell Counter, Thermo Scientific, Waltham, MA, United States). The cell isolation protocol is essentially the same for both cell types (astrocytes and neurons). However, astrocytes were derived from the tissue dissection of cortices (cortical astrocytes), while neurons were derived from the tissue dissection of hippocampi (hippocampal neurons). The first dissection served for the generation of a layer of feeder astrocytes, whereas the second dissection is necessary to produce neurons (Figure 1). Minor differences exist with respect to the preparation of both cell types (e.g., nitric acid pre-treatment, baking, and poly-L-lysine coating of glass coverslips for neurons, see section “Materials and Methods”). The mean number of primary cortical CNS cells dissected per brain was 1.135.500 ± 49.797, whereas in the case of hippocampal tissue the number was 279.125 ± 16.543 per mouse per brain (i.e., per 2 hippocampi). The mean life cell fraction in the case of cortical tissue was 90.09 ± 0.8622 %, and in the case of hippocampal tissue, the mean life cell fraction was 93.2 ± 0.5925%. Coverslips with feeder cells were monitored over a time of 2–3 weeks to show the presence of GFAP+ astrocytes that are crucial in providing paracrine (trophic) support to neurons (Kaech and Banker, 2006; Jones et al., 2012; Gottschling et al., 2016). To assess the astrocyte differentiation state of feeder cells, immunofluorescence (IF) staining was performed (Figure 1).

Importantly, feeder cells showed positivity for the pan-astrocyte marker GFAP without the signs of gross contamination with other cells, degeneration, cell death, or overt reactivity during their use for “sandwich” co-cultures. The degree of contamination in the astrocyte feeder layer with other cell types was assessed using the double immunofluorescence staining (Figure 2A). Contaminations with fibroblasts (marker: αSMA and FSP) or oligodendrocytes (marker: Olig2 and O4) were very low or undetectable (Figures 2B,C). Of note, astrocyte cultures with gross contaminations of fibroblasts should be discarded. Since microglia were vigorously shaken off from the astrocyte cells during cultivation, contamination with microglia was also very low or undetectable (Figure 2A). The overall purity of the astrocyte layer in our experiments was ≥ 90%. Over time, we observed a steady growth of feeder cells reaching a sub-confluent to confluent levels typically at 14 days (Figure 1). Neurons pre-treated (nitric acid treated, baked, PLL-treated) coverslip were inverted over the feeder layer of astrocytes that was plated 14 days ago, where wax dots were working as spacers and physical barriers between both cell types (Kaech and Banker, 2006; Jones et al., 2012). Thereby, neurons were directly facing the lower coverslip with the astrocyte feeder layer, but without being in direct contact. Of note, high purity of neuronal cells (on the upper coverslip) was mandatory to assess the consequences of experimental manipulations on the neuronal health and morphology of the neurites and dendritic spines. An overgrowth of the upper layer with glia can impede a reliable analysis of neuronal morphology and downstream neuronal readouts (Kaech and Banker, 2006; Jones et al., 2012). To curb glial overgrowth of the neuronal (upper) coverslip, we implemented a glial depletion step by adding the thymidylate synthase inhibitor and mitotic inhibitor FUDR (working concentration 10 μM) at day 4 (4DIV) (Jones et al., 2012; Figure 2D). Later on, medium changes of NMM did not involve the addition of FUDR in order to avoid adverse effects on neuronal metabolism. Upper coverslips were subjected to IF staining for neuron-specific markers, such as MAP2 or β-III-Tubulin, which enables evaluation of the relative purity of neurons (percentage of β-III-Tubulin+ cells) among the total cell population seeded on top of the upper coverslip (Figures 2E,F). The mean percentage of ß-III-Tubulin+ neurons per field of a given coverslip at 4 DIV before glial depletion was 63.2389 ± 21.0176%, whereas, after glial depletion at 14 DIV, it was 92.0187 ± 8.8579%. In contrast, GFAP+ DAPI+ astrocytes accounted for 36.7611 ± 21.0176% and 7.9812 ± 8.8579% of all DAPI+ cells at 4 DIV and 14 DIV, respectively (Figure 2F).

Monitoring by phase-contrast microscopy and IF staining at 7, 14, 19, and 21 DIV of upper and lower coverslips, respectively, demonstrated continuous growth and differentiation with increasing age of the co-cultures. As early as 1 h of plating, before the inversion of the upper coverslip over feeder cells, phase-contrast imaging consistently showed the neurons attaching to glass-forming small minor neurites (Figure 3). After 7 days in culture, neurites have grown steadily with some of them resembling axons (Figure 3). Neurons showed signs of maturity in vitro at 14 DIV with increased soma size, neurite thickness, branching, and an increase in MAP2+ neurites and Syn+ synaptic puncta (Figures 1, 3). Further information, with respect to technical considerations and potential pitfalls of the protocol, are summarized in Supplementary Table 1.

To determine basic readouts and to quantify potential advantages of our neuron-astrocyte co-culture setup, we compared neurons in co-cultures vs. neurons in monocultures at 9 DIV for morphological parameters. Since neurons in the monoculture were plated in the density that was used throughout our study showed a reduced live span ex vivo, and already started to die and disintegrate at around 12 DIV, we specifically chose day 9 for the comparison experiment between the mono and co-cultures. Of note, the experiments assessing astrocyte-neuron interaction in the co-cultures should be performed at 14 DIV, when neuronal differentiation is optimal (see Figure 3). Readouts were as follows: soma size, IntDen of the fluorescence signal of the neuronal marker MAP2, and the synaptic marker Synaptophysin by immunofluorescence staining and dendritic complexity by Sholl metrics (Figure 4). Moreover, we investigated different seeding densities of neurons in monocultures or co-cultures to optimize our assay (Figure 4). Soma size of neurons was chosen as a readout for neuronal health and differentiation status. Soma size in μm² in neurons in the monocultures, seeded at a density of 25 × 10³ neurons/well (1 × N), showed a mean soma size of 180.70 ± 101.53 μm²; while neurons plated at the double density (2 × N) showed a soma size of 190.19 ± 97.74 μm², whereas co-cultured neurons showed significantly higher mean values of soma size of 363.90 ± 191.70 for 1x N and 2x A 234.40 ± 80.47 2x N and 1x A and 297.50 ± 138.80 for 2x N and 2x A, respectively (Figure 4). Between the groups of co-cultured neurons [1x N (25 × 10³ neurons) co-cultured with either 80 × 10³ astrocytes (2x A), 2x N co-cultured with 1x A and 2x N co-cultured with 2x A], there was no significant difference in the mean neuronal soma size (Figure 4). For Sholl analysis, we assessed neuronal dendritic complexity parameters, including mean intersections, mean value, critical radius, and intersecting radii. All these parameters from Sholl analysis are indicators and measures of neuronal health and differentiation and built the basis for further studies. Mean intersections were analyzed via the Sholl plugin and showed, again, a significantly increased mean number of intersections of co-cultured neurons as compared to monocultures. The 1x N had 3.185 ± 1.3190 mean intersections, 2x N 2.533 ± 6470, whereas 1x N/1x A co-cultures showed 5.015 ± 1.3440 mean intersections, 2x N/1x A 4.350 ± 1.0492, and 2x N/2x A co-cultures showing with 6.1667 ± 1.1034, the highest number of mean intersections. Likewise, the sum of intersections with 593.1820 ± 328.8418 and 481.6670 ± 138.9239 for 1x N and 2x N monocultures was significantly lower as compared to the values reached in co-cultured neurons. Particularly, 2x N/1x A co-cultures showing 1092.3800 ± 296.4193 sum of intersections and 1x N/1x A 938.167 ± 236.8746 and 2x N/2x A co-cultures showing 1342.22 ± 238.6329 sum of intersections (Figure 4). Key metrics were analyzed as they derive from a polynomial function that can be plotted after running the Sholl analysis plugin, and using a built-in heuristic algorithm in the Sholl analysis plugin, approximating the polynomial function from a linear Sholl plot and, thus, improving local variations of sampled data (das Neves et al., 2021). Interestingly, results of the key metrics were similar to the results above. Significant changes in the mean value were observed between neurons, either subjected to monocultures or co-cultures. For 1x N and 2x N monocultures, it was significantly lower as compared to the values reached in co-cultured neurons: 1x N showed a mean value of 3.0000 ± 1.3505. 2x Na value of 2.3667 ± 0.6345. Conversely, 1x N/2x A showed a Nav of 5.0154 ± 1.3440, 1x N/1x A a mean value of 4.3500 ± 1.0715, and 2x N/2x A showed a mean value of 6.1666 +/-1.1034 (Figure 4). Critical radius and intersecting radii were in line with the results above. In contrast, the so-called critical value showed no significant differences between the mono-cultured or co-cultured neurons (Figure 4). In addition, the signal intensity of the dendritic marker, MAP2, and the synaptic marker protein Syn was assessed by IntDen values. Here, co-cultured neurons generally showed higher MAP2 signal intensity values as compared to neuronal monocultures. However, the mean IntDen values for Syn were higher in monocultured neurons as compared to co-cultured neurons (Figure 4).

Astrocyte co-culture set-ups have been developed to provide trophic support and increase neuronal health and differentiation in vitro. However, paracrine non-cell-autonomous effects and mechanisms of neurotoxicity and synaptotoxicity coming from the astrocyte feeder layer are less well defined. Given the increasing appreciation of the impact of glial cells on neurodegeneration, including the neurotoxicity and synaptotoxicity, we aimed at assessing the effect of toxin-exposed astrocytes on neurons using our assay by pre-treating astrocytes with either Aβ42 or LPS, two known stimulants in the context of AD (Liddelow et al., 2017; Zhang et al., 2017). The water or Aβbscr served as controls, respectively (Figure 5; García-Matas et al., 2010; Li et al., 2016; Pekny et al., 2016; Liddelow et al., 2017). Astrocytes without any neuronal co-culture were exposed to 24 h lasting treatments with either Aβ42 or LPS. Staining with astrocytic markers, GFAP and GLAST, revealed a highly increased intensity for GFAP (Figure 5A). However, the cell number was unchanged by treatment (Figure 5B). Analyses showed that the GFAP IntDen was significantly elevated in Aβ42-treated astrocytes with 5.0685 × 10⁷ ± 110,45,240.84 arbitrary units (a.u.), as compared to Aβscr control with 3.4956 × 10⁷ ± 119,60,517.06 a.u. In contrast, there was a non-significant elevation of GFAP signal intensity upon treatment with LPS with 3.5846 × 10⁷ ± 14517849.33 a.u. as compared to control 2.6356 × 10⁷ ± 8720571.991 a.u. (Figure 5C). In line with that observation, MTT assay of primary astrocytes led to a significant decrease of the optical density (OD) value upon treatment with 2.5 μM Aβ42 compared to Aβscr control, whereas, again, no significant differences in the OD values were detected between water and LPS treatment of astrocytes (Figure 5D).

Microglia, the innate immune cells of the brain, are considered to be the professional phagocytes of the brain. Upon phagocytosis of apoptotic cells from the brain, their phenotype becomes highly dysregulated (Krasemann et al., 2017). Lesser known is that the astrocytes could also significantly contribute to the removal of misfolded proteins or cell debris (Prasad and Rao, 2018). Thus, it was possible that the Aβ-pretreated astrocytes would bind or phagocyte Aβ in our experiments. To assess this, we performed IF staining for Aβ42 (6E10 antibody) and astrocytes (GLAST) (Fu et al., 2015). Astrocytes pre-treated with Aβ42 showed robust signs of positivity for human Aβ42 (Figure 5E). Thus, we speculate that this might also contribute to the astrocyte phenotype.

To elucidate whether reactive astrocytes directly impact primary neurons, we placed 14-day-old primary neurons, that were co-cultivated with untreated astrocytes before, over the pre-treated astrocytes for 24 h. To determine the effect of reactive astrocytes ratio on the neuronal phenotype, we used high-density (Figure 6A) and low-density culture neurons (Figure 6B). The Syn+ synaptic puncta (clusters) were used as a readout to analyze the impact of the pre-treatment of astrocytes on neuronal dendritic spines of co-cultured neurons (Figures 6C–H). Analysis of Syn+ synaptic puncta of 3 dendritic segments per neuron, with at least five neurons per experiment, was done in a semi-automatic fashion with SynPAnal with its built-in dendritic area-drawing and thresholding tool. Comparing Syn+ synaptic puncta of co-cultured neurons, either with pre-treated or with naive astrocytes, demonstrated a significant difference in the mean number of synaptic puncta per 100 μm length in neurons after contact with the reactive astrocytes seeded at high density (50 × 10³), but not at low density (25 × 10³) (Figure 6C). Likewise, the neuron-to-astrocyte ratio (N/A ratio) was significantly associated with the number of synaptic puncta of the co-cultured neurons. Similar results were observed measuring puncta density/μm² based on SynPAnal analysis (Figure 6D). Moreover, the ratio of intensity of Syn+ synaptic puncta and dendritic length or area was assessed. A similar effect on Syn+ synaptic puncta intensity was seen after treatment of neurons with Aβ42 or LPS-pre-treated astrocytes, leading to a significant reduction compared to the control astrocytes (Figures 6F–H).

Since we could show that astrocytes significantly contribute to the neuronal phenotype in an astrocyte-neuron co-culture, we wanted to know more about the activation profile of the reactive astrocytes. Given that Amyloidβ deposits in the brain may comprise different Aβ species. We also included Aβ40 and its respective control, Aβ40scr. Astrocytes were exposed to LPS (100 ng/ml), Aβ40 (2.5 μM), or Aβ42 (2.5 μM), or the respective controls of water, Aβ40scr (2.5 μM), or Aβ42scr (2.5 μM) for 24 h. Astrocytes were then stained for GFAP and actin to display changes in cell morphology. As seen with Aβ42, the Aβ40 treatment and LPS have induced considerable upregulation in the abundance of GFAP (Figure 7A). In contrast, the treatment led to severe changes in the appearance of actin cytoskeleton and reduction of signal intensity (Figure 7B). To assess which factors might contribute to the effect of astrocytes on the neuronal phenotype that we detected (see Figure 6), we performed a Mouse Cytokine Array by measuring the abundance of 40 different mouse cytokines in astrocyte cell culture supernatant of astrocytes that were treated with LPS (100 ng/ml), Aβ40 (2.5 μM), or Aβ42 (2.5 μM), or Aβ42scr (2.5 μM) as control for 24 h. Interestingly, we could detect significant upregulation of several cytokines upon treatment including IP-10/CXCL10, JE/CCL2, KC/CXCL1, and RANTES/CCL5 (Figure 7C and Supplementary Figure 1). Other cytokines like TIMP-1 or TNFα were also elevated but did not reach significance. To confirm our findings, we performed a set of independent experiments and performed qPCR of the pre-selected set of the elevated cytokines from the protein assay in astrocyte cell lysates and included the reactive astrocyte A1 marker complement C3 (Figure 7D). Using this alternative method, we confirmed significant upregulation of IP-10/CXCL10, JE/CCL2, KC/CXCL1, and RANTES/CCL5 in treated astrocytes. Moreover, we could detect significant elevation of MIP1-α and TNFα on the RNA level. Although C3 expression was also increased in treated astrocytes, it did not reach significance.

In summary, we could show that astrocytes secrete a couple of cytokines upon treatment that might directly or indirectly impact neuronal health. We were able to show that astrocytes contribute considerably to the neuronal phenotype in an astrocyte-neuron co-culture. However, this effect is dependent on a specific astrocyte to neuron ratio. Thus, highly controlled experimental conditions are necessary on an in vitro co-culture set-up to be able to get reproducible and representative results.

Technical considerations and troubleshooting concerning our protocol are summarized in the following table, which can access via the Supplementary Table 1.