Neuron-astrocyte co-cultures were done using rat astrocytes, because of tissue culture advantages of rat over mouse astrocytes, and mouse neurons to make it possible for future experiments to use neurons of genetically modified mice. Importantly, in these co-cultures, rat astrocyte provide the required support for optimal synaptic functioning of co-cultured mouse neurons (Wierda et al., 2007), an important requirement for our screen of neuron-induced expression changes in astrocytes. Primary rat astrocytes were obtained from Wistar rat pups (P1). Briefly, dissected cortices were treated with trypsin (Gibco) followed by addition of trypsin inhibitor and gentle trituration, after which cells were resuspended in Dulbecco’s modified Eagle’s medium (DMEM) + Glutamax (Gibco) containing 10% fetal bovine serum and cultured in non-coated plastic flasks. Medium was changed every 2 days. The purity of astrocyte cultures was determined by immunostaining for glial fibrillary acidic protein (GFAP) and S100β, while counterstained for all cell nuclei using Hoechst, and was determined to be higher than 98%. For neuron-astrocyte co-cultures, astrocytes were plated at a density of 30K/well in 12-wells plates in DMEM + 10% medium, which resulted in an about 80% confluent monolayer at day in vitro (DIV) 3, after which neurons were plated. For co-plating of neurons, hippocampi of E18 mouse embryos were collected in Hanks buffered salts solution (HBSS; Sigma), buffered with 7 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES). After removal of the meninges, hippocampi were minced and incubated for 20 min in 0.25% trypsin in HBSS at 37°C. After washing, the neurons were triturated with fire-polished Pasteur pipettes, counted, and plated in culture medium [Neurobasal medium (Invitrogen, Carlsbad, USA) supplemented with 2% B-27 (Invitrogen), 1.8% HEPES, 1% glutamax (Invitrogen), 1% Pen/Strep (Invitrogen), and 0.2% beta-mercaptoethanol] at a density of 60K/well onto the astrocyte monolayers in 12-well plates (4 cm × 4 cm; Greiner bio-one; Wierda et al., 2007). Wells were coated with 20% d-poly-lysine (Sigma) and 20% collagen diluted in a 17 μM acetic acid H₂O solution. Cultures were maintained in 5% CO₂ at 37°C, and half of the medium was changed at 1 day and 7 days after plating. Control conditions (astrocytes alone) were kept under similar tissue culture conditions as neuron-astrocyte co-cultures.

For removal of neurons from the co-cultured astrocytes, we modified a method previously used for removal of Schwann cells from co-cultured fibroblasts (Jirsova et al., 1997). 12-well tissue culture plates were placed on ice and the medium was gently aspirated. The cells were carefully rinsed once with ice-cold phosphate buffered saline (PBS; pH 7.4). Next, using a 1 mL pipette tip, 1 mL of ice-cold PBS was pipetted onto the co-cultures using a concentric circular motion, ensuring the force of the stream contacted the entire well surface area, which was repeated three times. This resulted in specific detachment of neurons, which were aspirated, leaving purified astrocytes, which were harvested for molecular analysis. All together, the isolation of astrocytes from co-cultures for molecular isolation per well was performed within a timeframe of 30 s. When co-culture cold jet (CC-CJ) cultures were compared to astrocyte alone cultures, the later were also treated by cold jet, therefore called astrocyte-alone cold jet (AA-CJ) samples.

Astrocyte monolayers or astrocyte-neuron co-cultures, before or after cold jet, were fixed by 4% paraformaldehyde in 0.1 M PBS for 15 min at room temperature (RT). Cells were incubated 30 min in 0.1 M PBS containing 0.5% TritonX-100 and 10% normal goat serum, and subsequently incubated with appropriate primary antibody concentrations in incubating medium (0.1 M PBS, 0.1% TritonX-100, 2% normal goat serum) at 4°C overnight. Cells were washed three times for 10 min in 0.1 M PBS and incubated in incubating medium containing the appropriated secondary antibody for 2 h. Sections were washed two times with 0.1 M PBS, one time with water, and subsequently DNA counterstained in Hoechst reagent 33258 solution (Sigma). The primary antibodies were rabbit anti-GFAP (1:500; Dako) and chicken anti-microtubule-associated protein 2 (MAP2; 1:5000; Abcam). Secondary antibodies were donkey anti-rabbit Cy3 (1:400; Jackson laboratories), goat anti-chick Alexa 488 (1:400; Molecular probes).

Total RNA was extracted from 12-well plate and mRNA was isolated with Trizol Reagent (Invitrogen). RNA samples for microarray analysis were derived from two to four tissue culture wells, and microarray replicates were derived from different pools of animals, with five to eight animals per pool. An additional chloroform extraction was performed to remove traces of phenol and samples were suspended in H₂O. The quality and quantity of RNA samples for microarray analysis were assessed on Agilent RNA chips with the Agilent 2100 Bioanalyzer (Agilent Technologies). For quantitative PCR, four samples were used that were different than those used for microarray analysis, and each sample was derived from different animals and four to six tissue culture wells. RNA integrity was determined using electrophoresis with RNA loading dye on a 1% agarose gel, RNA quantity was determined by photospectrometry (ND-1000; Thermo Scientific). Quantitative PCR reactions were done as described (Camargo et al., 2012).

Agilent whole rat genome 4x 44K chips were used. cDNA synthesis and labeling for microarray hybridization were performed according to the manufacturers’ instructions. Briefly, oligo dT primers were used and Cy3 and Cy5 labeling was done using the Agilent two-color RNA Spike-In Kit. 0.5 μg of Cy5 and Cy3-labeled cRNA targets were hydrolyzed and mixed for 30 min. Arrays were hybridized with Cy5 and Cy3 solution for 18 h at 60°C in a rotating hybridization chamber, using 1× hybridization solution (Agilent technologies). Subsequently, arrays were washed, spun dry and scanned using an Agilent scanner.

Analysis of the microarrays served two purposes: (1) detection of neuron-induced gene expression in astrocytes and (2) determination of the cold jet efficiency on the mRNA level. In total, four arrays were used. For the first purpose, two CC-CJ samples were labeled with Cy3 and two AA-CJ samples were labeled with Cy5. Each CC-CJ sample was hybridized together with an AA-CJ sample to an array. For checking the cold jet procedure on the mRNA level, two co-culture (CC) samples were labeled with Cy5 fluorescence and hybridized on an array together with a CC-CJ sample (Cy3). The microarray dataset can be found at National Center for Biotechnology Information Gene Expression Omnibus (NCBI GEO; accession number GSE52481).

Normalization of the raw data was performed, after feature extraction by Agilent software, by optimized local intensity-dependent normalization (OLIN) and optimized scaled local intensity-dependent normalization (OSLIN; Futschik and Crompton, 2005) using the R project for statistical computing (version 6.0; R project). Next, M values were scaled so that upregulated probes were associated with positive values and downregulated probes with negative values. The following criteria were used for selection of regulated transcripts: (1) for each array, the mean and standard deviation (SD) of the M values of all probes was calculated, and only mRNAs with M values equal or higher to the mean + 2× SD, or equal or lower to the mean - 2× SD were selected, (2) M values had to fulfill our criteria on both replicate arrays, (3) for mRNAs that were represented on the microarray with multiple probes, data were checked for consistency, and probes were only included in the analysis when replicate sequences gave comparable results, (4) mRNAs upregulated in the CC-CJ with AA-CJ comparison could represent left-over neuronal RNA when also strongly downregulated in the CC with CC-CJ comparison. Therefore, mRNAs upregulated in the CC-CJ with AA-CJ comparison where only included when downregulation in the CC with CC-CJ comparison was less than the mean - 2× SD. To assess consistency between technical and biological replicate samples, as well as between replicate probes, non-parametric correlations (using Spearman’s rho) within non-normalized red or green fluorescence samples were calculated. For technical/biological replicate samples, correlations were only performed on regulated probes as indicated by the cold jet-astrocyte alone arrays.

Thereafter, data were analyzed by TopGo (Alexa et al., 2006) using the R project for statistical computing. Overrepresented biological processes were identified by scoring statistical significance of predefined annotated groups based on the gene ontology (GO) database. Statistical significance of overrepresented predefined GO biological processes was scored using the Fisher exact test (Alexa et al., 2006).

Neuronal development in hippocampal mixed neuron-glia cultures is a temporal process involving neurite outgrowth (1 DIV onward), and synapse formation from 4 DIV onward (Basarsky et al., 1994). Since attachment of neurons to the extracellular culture environment might vary with time, cold jet removal of neurons from neuron-astrocyte co-cultures was tested at 10 and 14 DIV. Quantification of the number of neuronal somas and neurites, indicated that cold jet removed over 98% of neuronal somas and 80% of neurites at 10 DIV, and was only slightly less efficient at 14 DIV (Figure 1). Cold jet removal of neuronal mRNAs was tested by microarray RNA profiling, which confirmed robust removal of neuronal mRNAs, e.g., Mapt, Gap43, Snap25, and Snca (Figure 2A). Neuron-enriched mRNAs are expected downregulated in coculture by cold jet, while being absent or low expressed in astrocytes. To determine the efficiency of the cold jet in removing these neuron-enriched mRNAs, their fluorescence levels in CC and CC-CJ was taken and the level in astrocytes substracted. The data indicated that the cold jet effectively removed approximately 90% of neuronal RNA (Figure 2B).

Microarray analysis of CC-CJ versus AA-CJ samples, yielded 394 neuron-induced regulations of astrocyte mRNAs, of which 270 were upregulated and 124 downregulated. The 36 strongest up- and down-regulated mRNAs are shown (Figure 3). The group of 270 mRNAs included many which are involved in astrocyte-neuron interactions, such as cell adhesion (Itgb, Pcdh10, Cdh6, Ctnnal1, Glycam, L1cam, and TSP precursor 4), extracellular matrix (Cspg5 and Reck) and cytokine signaling (Megf6 and Lcn2). Several mRNAs involved in lipid synthesis were found down-regulated, including mRNAs involved in cholesterol synthesis (Dhcr7, Hmgcs, Sc4mol, Cyp26b, Tm7sf2, FPS, Sql, and Insig1).

To gain more insight into biological processes represented by the data set, we scored statistical significance of predefined functional genes groups, based on GO biological processes. From the complete list of regulated mRNAs (394), 10 significant overrepresented GO terms were identified (with p < 0.005, see Table 1). Overrepresented GO terms included mainly metabolic processes, e.g., sterol and nutrient related processes. GO analysis of only up-regulated (Table 2) or only down-regulated (Table 3) gene groups showed most strikingly that mRNAs associated with “sterol metabolism” were mainly down-regulated, whereas “response to nutrient” processes were up-regulated.

To validate the microarray data, qPCR was performed on a selection of mRNAs that were regulated between astrocytes alone (AA-CJ) and astrocytes in co-culture (CC-CJ). qPCR confirmed the down-regulation of three mRNAs involved in lipid metabolism (Insig1, Hmgcs, and Dhcr7) and confirmed the up-regulation of the transcription factor Egr1 in astrocytes when co-cultured with neurons (Figure 4).

Neuron-glia co-cultures are an established in vitro model to investigate the reciprocal interactions of these cell types that underlie neurite outgrowth, synapse formation, and synaptic transmission. Investigating the molecular changes in astrocytes that take place during interaction requires the efficient isolation of astrocytes from the co-culture in a manner that ensures cellular integrity. Here, we present cold jet as a fast and efficient isolation method of astrocytes from differentiating neuron-astrocyte co-cultures. We showed that at 10 and 14 DIV, when neurites and mature synapses have been formed (Basarsky et al., 1994), cold jet removed nearly all neuronal RNA in these co-cultures. This permits selective transcriptional analysis of only astrocytes in these co-cultures. Cold jet isolation of astrocytes from neuron-astrocyte co-cultures may also be suitable for analysis of proteins and lipids, since more than 90% of neuronal cell bodies and nearly as much neurites are removed. We conclude that cold jet offers important advantages over other techniques that are used for cell separation, such as FACS or immuno-panning. Unlike those other techniques, cold jet is fast (within 1 min), easy to perform, done without addition of proteases, and conducted under ice-cold conditions to ensure no detriment to molecular integrity.

Through a developmental in vitro model, in which neurons in co-culture with astrocytes differentiate and form synapses, we set out to identify neuron-responsive astrocyte gene expression using microarray analysis. We found regulation of many genes involved in astrocyte biological processes that have been reported previously to affect neuronal differentiation and function, such as lipid metabolism (Camargo et al., 2009, 2012; Pfrieger and Ungerer, 2011), cell-ECM adhesion (Christopherson et al., 2005; Tournell et al., 2006), complement factor signaling (Stevens et al., 2007), cytokine signaling (Acarin et al., 2001), and synaptic transmission (Halassa and Haydon, 2010). It should be noted that most of the astrocyte mRNAs involved in cholesterol metabolism were down-regulated by neuronal presence, which is surprising considering previous findings illustrating the importance of astrocyte cholesterol for synapse maturation and neurite outgrowth (Mauch et al., 2001; Fan et al., 2002; Hayashi et al., 2004; Camargo et al., 2009). However, we did find down-regulation of Insig1, an important negative regulator of the lipogenic sterol regulatory element binding protein (SREBP) transcription factors, which therefore would be expected to lead to increased SREBP-mediated transcription of lipogenic enzymes, as found in other cell systems (Engelking et al., 2005). Our recent observation that SREBPs are also required for expression of lipogenic enzymes in astrocytes (Camargo et al., 2012), warrants further investigation of the control of Insig, SREBP, and downstream transcription in astrocytes due to contact with neurons. Taken together, our observations suggest that astrocyte-factors involved in these processes do not only influence neuronal differentiation but are furthermore induced in astrocytes by neuronal presence, underscoring the reciprocal nature of the neuron-glia interactions.