Microglia in Glia–Neuron Co-cultures Exhibit Robust Phagocytic Activity Without Concomitant Inflammation or Cytotoxicity

Abstract

A simple method to co-culture granule neurons and glia from a single brain region is described, and microglia activation profiles are assessed in response to naturally occurring neuronal apoptosis, excitotoxin-induced neuronal death, and lipopolysaccharide (LPS) addition. Using neonatal rat cerebellar cortex as a tissue source, glial proliferation is regulated by omission or addition of the mitotic inhibitor cytosine arabinoside (AraC). After 7–8 days in vitro, microglia in AraC⁻ cultures are abundant and activated based on their amoeboid morphology, expressions of ED1 and Iba1, and ability to phagocytose polystyrene beads and the majority of neurons undergoing spontaneous apoptosis. Microglia and phagocytic activities are sparse in AraC⁺ cultures. Following exposure to excitotoxic kainate concentrations, microglia in AraC⁻ cultures phagocytose most dead neurons within 24 h without exacerbating neuronal loss or mounting a strong or sustained inflammatory response. LPS addition induces a robust inflammatory response, based on microglial expressions of TNF-α, COX-2 and iNOS proteins, and mRNAs, whereas these markers are essentially undetectable in control cultures. Thus, the functional effector state of microglia is primed for phagocytosis but not inflammation or cytotoxicity even after kainate exposure that triggers death in the majority of neurons. This model should prove useful in studying the progressive activation states of microglia and factors that promote their conversion to inflammatory and cytotoxic phenotypes.

Introduction

Microglia residing in the brain parenchyma support development and maintain homeostasis by remodeling synapses, removing cellular debris, and engulfing apoptotic cells and foreign pathogens. The early literature focused on resting and active functional states, corresponding to ramified and amoeboid morphologies, respectively, as well as different secretory profiles and patterns of gene expression (Raivich et al. 1999). Resting microglia maintain compact somas and ramified branches, and were previously considered to be inactive guardians that were mobilized to an active phagocytic state in response to injury or infection. Subsequent in vivo studies revealed that microglia in the basal state are not idle; rather, they perform important homeostatic functions by actively surveying their environment for minor infractions, and clearing cellular and tissue debris without undergoing profound changes in morphology, cytokine secretion, or patterns of gene expression (Nimmerjahn et al. 2005). When exposed to foreign pathogens or tissue injury, microglia become ‘reactive’ or activated, undergoing pronounced morphological, chemotactic, and proliferative changes in conjunction with coordinated transcription of numerous genes encoding cytokines, cell surface antigens, inflammatory mediators, cell adhesion molecules, and associated proteins that subserve their transition to motile and phagocytic phenotypes. Emerging studies indicate that the number and characteristics of these functional effector states are more complex than previously appreciated, and likely represent a continuum that depends on the duration and context of injury (Kettenmann et al. 2011; Salter and Beggs 2014). Importantly, activated microglia can also become cytotoxic and exacerbate tissue damage (Araki et al. 2001; Bal-Price and Brown 2001; Choi et al. 2003; Yang et al. 2007; van Eldik et al. 2007; von Bernhardi et al. 2007; Cho et al. 2008; Blaylock 2013; Xing et al. 2015; Perry and Holmes 2014; but see Streit et al. 2014), prompting investigation of the therapeutic efficacy of microglia inhibitors in a variety of injury models, including multiple sclerosis, ischemic and hemorrhagic stroke, neurodegenerative diseases, and various forms of cerebral edema and brain trauma (Bachstetter et al. 2013; Sloka et al. 2013; Neher et al. 2013; Xie et al. 2014; Chen et al. 2014; Huang et al. 2014; Kabadi et al. 2014). From a clinical perspective, it is critical to distinguish between protective versus cytotoxic activation states of microglia because effective therapy could include agents that further enhance protection or attenuate toxicity, depending on the nature of the insult.

In vitro models using microglia alone, or microglia co-cultured with neurons or astrocytes, have provided important insights about conditions that influence microglia phenotypes and effector functions, as well as potentially useful pharmacological agents (Duan et al. 2013; Claycomb et al. 2014; Dinkins et al. 2014; Dambach et al. 2014). There are, however, important considerations when applying in vitro culture models to enhance our understanding of microglia function in vivo: (1) microglia in primary cultures are typically in a state of activation that is further along the continuum than the basal state, as exemplified by their predominant amoeboid morphologies and expressions of immunoregulatory proteins (Ransohoff and Perry 2009; Ousman and Kubes 2012). Thus, use of in vitro cultures are predicted to provide meaningful insights regarding the continuum of activation beyond the basal state, for example, conditions regulating the conversion from protective to cytotoxic phenotypes, but not the transition from highly ramified to active amoeboid states; (2) microglia are influenced by and, in turn, influence both astrocytes and neurons (Sudo et al. 1998; Rosenstiel et al. 2001; Faustmann et al. 2003; Min et al. 2006; Yang et al. 2007; Kettenmann et al. 2011; Ousman and Kubes 2012; Streit et al. 2014); and (3) in the CNS there appears to be regional heterogeneity in morphology, expressions of immunoregulatory proteins, and sensitivity to injury in microglia (Lawson et al. 1990; Elkabes et al. 1996; Ren et al. 1999; Kim et al. 2000; de Haas et al. 2008). Thus, it should prove advantageous to study co-cultures of neurons, astrocytes, and microglia derived from the same brain region.

Objective

In an effort to preserve the important interactions between neurons and glia, we describe a model using cultures of microglia, astrocytes, and cerebellar granule neurons (CGNs), derived from the same brain region, the postnatal cerebellar cortex. We use this simple and convenient model to characterize the activation states of the microglia, and to assess their phagocytic versus inflammatory and cytotoxic activities under conditions of naturally occurring apoptotic neuronal death, in response to the excitatory amino acid kainate where cytotoxic effects have been reported in other models (Tikka et al. 2001; Lee et al. 2003; Cho et al. 2008; Zheng et al. 2010; Zhu et al. 2010; Hong et al. 2010; Neher et al. 2011) and, for comparison, after addition of the potent endotoxin lipopolysaccharide (LPS).

Materials and Methods

Sprague–Dawley neonatal rats were purchased from Taconic Farms; the institutional review committee, in accordance with governmental guidelines, approved all procedures involving use of animals for experimentation. Antibodies against ED1 were purchased from Serotec Inc, GFAP from Sigma, COX-2 and iNOS from Santa Cruz and Iba1 from WAKO. β-tubulin antibodies were purchased from Promega. Kainic acid, fluorescein diacetate, propidium iodide, and fluorescent polystyrene beads (2 μm) were purchased from Sigma. Larger fluorescent polystyrene beads (5.5 μm) were obtained from Bangs Laboratories. Other reagents for RT-PCR and tissue culture were molecular biology and tissue culture grade, respectively, and were obtained from commercial sources.

Tissue Culture

Granule cell-enriched cultures were prepared as previously reported by our laboratory (Vallano et al. 1996) except that the mitotic inhibitor cytosine arabinoside (AraC) was omitted from the growth medium, as indicated, so that astrocytes and microglia would thrive and proliferate (Beaman-Hall et al. 1998; note that microglia were not examined in this study). Briefly, cerebella from 8-day-old Sprague–Dawley rats were minced, trypsinized, and triturated to dissociate cells. The cells were plated at a density of 1.25 × 10⁶ cells/ml of medium onto 12-well Corning dishes (for FDA/PI, live-dead assays, and RT-PCR), or 24-well Corning dishes containing glass coverslips precoated with poly-l-lysine (for immunocytochemistry) and incubated at 37 °C in a humidified atmosphere containing 5 % CO₂/95 % air (pH 7.4). Microglia-rich neuronal cultures were grown in basal Eagle’s medium with Earle’s salts, supplemented with heat-inactivated fetal bovine serum (10 %), gentamicin sulfate (100 μg/ml), and l-glutamine (2 mM). Growth medium was typically supplemented with 20 mM KCl (25 mM f.c.) to enhance long-term survival of CGNs (Thangnipon et al. 1983; Kingsbury et al. 1985; Gallo et al. 1987). In some experiments, medium was not supplemented with 20 mM KCl so that the phagocytic function of microglia could be assessed in the context of a substantial amount of naturally occurring neuronal death (~75 % in serum-containing growth medium). Microglial and astrocyte-poor cultures containing greater than 90 % CGNs were obtained by adding AraC (10 μM) to the culture medium 24 h after plating (Thangnipon et al. 1983; Kingsbury et al. 1985). Note that at this concentration, AraC is a mitotic inhibitor used for prevention of glial proliferation in neuronal preparations, and is not cytotoxic (Kolodny et al. 1985; Martin et al. 1990; Dessi et al. 1995). Where indicated, CGNs lacking AraC were switched to a chemically defined medium lacking serum components after 1 day in vitro (DIV; Leahy et al. 1994) and directly compared to sister cultures grown in serum-containing medium. Unless otherwise stated, cells were used for experiments at 7–8 DIV, at which time the CGNs had developed sensitivities to addition of excitototoxic agents (Leahy et al. 1994). Medium was not exchanged during this time, but 5.5 mM glucose was added to each well at 6 DIV to enhance long-term survival (Schramm et al. 1990). Cell viability was assessed routinely throughout the culture period using phase-contrast microscopy.

Excitotoxicity

Neurotoxicity was measured by fluorescent staining for live and dead cells (Jones and Senft 1985). The growth medium was collected, and the cells were washed twice with Lockes solution containing 154 mM NaCl, 5 mM KCl, 2.3 mM CaCl₂, and 8.6 mM HEPES, pH 7.4. Locke’s solution containing vehicle, 250 μM kainic acid (KA) or 500 μM KA was added and dishes were placed in a sterile hood at room temperature (22–23 °C) for 30 min, as originally described by Costa and associates (Favaron et al. 1988). After incubation with vehicle or drug, cultures were washed twice with Locke’s solution, the original preconditioned growth medium was replaced, and the cultures were returned to the incubator. After 24 h, cultures were washed with Locke’s solution before and after incubation with fluorescein diacetate (FDA; 10 μg/ml) and propidium iodide (PI; 4.6 μg/ml) for 10 min, and then examined using fluorescent microscopy, as previously described (Leahy et al. 1994). Where indicated, cultures were harvested at earlier times for RT-PCR and immunocytochemical analyses. In some cases, the relative proportions of living and dead cells were quantified using a live/dead assay kit where calcein was used to stain living cells and PI to stain dead cells (Invitrogen). Three or four fields per coverslip containing at least a hundred cells were photographed. The total number of living CGNs, which have a small rounded morphology compared to microglia and astrocytes, versus the number of dead CGNs were quantified by cell counting as previously described by our laboratory (Leahy et al. 1994).

Immunocytochemistry

Production of microglia-rich neuronal cultures (AraC⁻) was confirmed by observation using a phase-contrast microscope and immunocytochemical staining with the microglial markers ED1 and Iba1 (estimated as ~11 % of cells at 7–8 DIV), and the abundant neuronal protein β-tubulin. An antibody against GFAP was used to label astrocytes within the glial population (estimated as ~10 % of cells at 7–8 DIV; Beaman-Hall et al. 1998). Cultures were fixed with 4 % paraformaldehyde 24 h after toxicity (7–8 DIV), permeabilized with 0.3 % Triton X-100, and 4 % normal donkey serum in phosphate-buffered saline (PBS) was added for 20 min at room temperature (RT). Cultures were then incubated with the following antibodies: ED1(1:300) mouse monoclonal antibody, Iba1(1:1000) rabbit polyclonal, GFAP (1:1000) rabbit polyclonal antibody, COX-2 (1:300) rabbit polyclonal, iNOS (1:1000) rabbit polyclonal, and β-tubulin (1:1000) mouse monoclonal. Incubation was according to manufacturer’s protocol for 24 h at 4 °C, except for β-tubulin, which was incubated for 1 h at RT. Cultures were then washed 3× with PBS and incubated in Alexa fluor 488 donkey anti-mouse secondary antibody (1:400) or Alexa fluor 488 donkey anti-rabbit (1:400), and Alexa fluor 594 donkey anti-rabbit (1:400) for 1 h at RT. Coverslips were mounted using Vectashield Antifade (Vector Laboratories) or Vectashield Antifade with PI if ED1 was used alone. The coverslips were then photographed under a fluorescent microscope. The proportions of microglia with amoeboid (round with no processes) or rod-shaped (bipolar) or process-bearing morphologies were estimated by counting Iba1-positive cells in two randomly photographed fields per coverslip in 3 different cell culture preparations.

Phagocytosis Assay

A phagocytosis assay using two sizes of polystyrene beads was performed as previously described (Bocchini et al. 1988a, 1988b). Briefly, cultures were prepared as above and plated on 12 mm glass coverslips at a density of 1.25 × 10⁶ cells/ml and cultured for 7–8 DIV. At the end of the culturing period, 1 μl/ml of the 2 μm fluorescent microspheres or 1.5 μl/ml of 5.5 μm fluorescent microspheres were added directly to the cultures and incubated for 1 h at 37 °C. Cells were washed three times with PBS to remove residual microspheres and fixed with 4 % formalin in PBS. Cells were then labeled with Iba1 or GFAP antibody as described above. Four random fields were photographed and counted.

RT-PCR

Real-time quantitative RT-PCR (qRT-PCR) analysis was done to assess the induction of mRNAs encoding inflammatory molecules using a Light-cycler 480 System (Roche Applied Sci) using SYBR Green, a fluorescent double-stranded DNA dye as previously described by our laboratory (Gerber et al. 2010). For quantitative analysis, each sample was analyzed in triplicate PCR reactions on a single 384-well plate. Amplification in the absence of template was evaluated to ensure lack of signal due to primer dimerization and extension. End point melt-curve analysis was also done to confirm the presence of a single amplicon in each reaction well. Analysis of data was performed using the ΔΔCT method (with 18S as an internal reference), and a repeated measures ANOVA model using the ΔCT data. Validation of the ΔΔCT method was done by demonstrating equivalent efficiencies of amplification of the 18S RNA normalization control versus the target mRNAs over a range of serial dilutions. A conventional RT-PCR strategy with visualization of amplicons using ethidium bromide staining of agarose gels was also used for analysis of iNOS mRNA (Vallano et al. 1996). The primer sets are shown in Table 1.

Table 1.

Primers used for PCR

Name	Sense	Antisense
TNFα	5′-GTAGCCCACGTCGTAGCAAA-3′	5′-CCCTTCTCCAGCTGGGAGAC-3′
COX2	5′-CCATGTCAAAACCCGTGGTGAATG-3′	5′-ATGGGAGTTGGGCAGTCATCAG-3′
iNOS	5′-GCAGAATGTGACCATCATGG-3′	5′-ACAACCTTGGTGTTGAAGGC-3′
18S	5′-CGCCGCTAGAGGTGAAATTC-3′	5′-TTGGCAAATGCTTTCGCTC-3′

Statistical Analysis

Values are presented as mean ± standard error of the mean. Statistical comparison between groups is performed using analysis of variance (ANOVA) followed by Fisher’s protected least significant difference (PLSD) or Tukey/Kramer post hoc test. A probability of p < 0.05 is considered statistically significant.

Results

Culture of CGNs Without AraC Promotes Proliferation of Microglia and Astrocytes

AraC is typically added to the growth medium of CGN cultures prepared from postnatal cerebellum to suppress the proliferation of non-neuronal cells, primarily glia (Thangnipon et al. 1983; Kingsbury et al. 1985). Figure 1 shows a scarcity of Iba1-stained microglia (A) and GFAP-stained astrocytes (B) in AraC⁺ cultures grown for 7–8 DIV. Alternatively, AraC⁻ cultures contain numerous microglia (D) and astrocytes (E). Both culture conditions exhibit a highly interconnected network of CGNs, demonstrated by staining with antibodies against β-tubulin (C and F). In AraC⁻ cultures grown for 7–8 DIV, it is estimated that microglia constitute ~ 11 % of the cells, and those with a rounded amoeboid morphology predominate (63 %), with significantly lesser amounts of rod-shaped (31 %) and process-bearing (6 %) microglia (p < 0.004 when comparing amoeboid versus rod-shaped or process-bearing microglia, n = 3). Microglia with highly ramified processes and compact soma are not apparent, which is consistent with reports that microglia grown in vitro typically exhibit a more active phenotype compared to so-called ‘resting’ microglia in the brain parenchyma. When grown in a chemically defined medium lacking both AraC and serum, the number of microglia is significantly decreased (serum AraC⁻: 30.0 ± 4.2; defined AraC⁻: 7.0 ± 1.2, p < 0.01, n = 4).

Fig. 1.

AraC⁻ cultures contain numerous microglia and astrocytes compared to AraC⁺ cultures. Co-cultures of CGNs were grown in the presence or absence of the mitotic inhibitor AraC to regulate proliferation of microglia and astrocytes. Upper panels a–c are cultures grown in the presence of AraC to suppress microglia and astrocyte proliferation. Lower panels d–f are cultures grown in the absence of AraC to permit microglia and astrocyte proliferation. At 7–8 DIV cultures were fixed, incubated with specific antibodies, and photographed (×20 magnification). a AraC⁺ cultures incubated with antibodies against the microglial marker Iba1. b AraC⁺ cultures incubated with antibodies against the astrocyte marker GFAP. c AraC⁺ cultures incubated with antibodies against the CGN marker β-tubulin. d AraC⁻ cultures incubated with antibodies against the microglial marker Iba1. e AraC⁻ cultures incubated with antibodies against the astrocyte marker GFAP. f AraC⁻ cultures incubated with antibodies against the CGN marker β-tubulin

Microglia in Co-cultures Express Markers Associated with an Activated but Not an Inflammatory Phenotype

The amoeboid morphology of the majority of microglia in the cultures indicates that they are somewhere along the continuum of activation beyond the ramified/resting state. To further examine this, expression of immunoreactive ED1 (aka CD68), a single chain glycoprotein of 90–100 kDa localized on the lysosomal membrane of phagocytes, was assessed in cultures grown for 7–8 DIV. Expressions of the inflammatory molecules COX-2, iNOS, and TNF-α were also assessed using immunochemistry and qRT-PCR. For comparison, sister cultures were exposed to the potent endotoxin LPS. Figure 2a is a western immunoblot of whole culture homogenates, equalized for protein content, after overnight exposure of cultures to increasing concentrations of vehicle or LPS (0, 12, 25, and 50 ng/ml). As shown, COX-2 protein is undetectable in the basal state, but increases in a concentration-dependent manner in response to incubation with LPS. Immunocytochemical staining demonstrates that microglia in the co-cultures express the activation marker ED1, and are also the source of LPS-mediated COX-2 generation (Fig. 2b). Qualitatively, similar results are obtained when co-cultures are exposed to vehicle or LPS, and subsequently stained with antibodies against ED1 and iNOS (Fig. 2c). Cell nuclei are labeled with Hoechst stain.

Fig. 2.

Immunoreactive COX-2 and iNOS are undetectable in microglia in control cultures, and are induced after addition of LPS. a Upper panel Western immunoblot of COX-2 immunoreactivity in whole culture homogenates of AraC⁻ cultures grown for 8 DIV, and equalized for protein content after overnight addition of LPS (12, 25, 50 ng/ml) or vehicle (0 ng/ml). Lower panel Western immunoblot of β-tubulin in corresponding samples, used as a loading control. b AraC⁻ cultures were grown for 7–8 DIV, vehicle (no LPS) or 25 ng/ml LPS, (+LPS) were added overnight, then cultures were fixed and incubated with antibodies against the microglial marker ED1 (green) and COX-2 (red). Hoechst stain was used to label cell nuclei (blue). c AraC⁻ cultures were processed as described in b except that they were incubated with antibodies against ED1 (green) and iNOS (red). Hoechst stain was used to label cell nuclei (blue). All photographs were taken at ×20 magnification (Color figure online)

Next, the highly sensitive method of qRT-PCR was used to determine the temporal and concentration-dependent effects of LPS on COX-2 mRNA expression. Figure 3a shows a time-dependent increase in COX-2 mRNA with significant induction at all concentrations after exposure to LPS for 24 h, relative to control (fold-increases: 12 ng/ml = 83.8; 25 ng/ml = 116.0; 50 ng/ml = 60.7). LPS-mediated induction of TNF-α mRNA was also examined (Fig. 3b). Increases in expression of TNF-α mRNA are significant and greatest 1 h after addition of all concentrations of LPS (fold-increases: 12 ng/ml = 209.6; 25 ng/ml = 248.3; 50 ng/ml = 221.6), and rapidly decrease to non-significant values at 6 and 24 h. Induction of iNOS in response to LPS appears to be especially robust with the greatest increases 6 and 24 h after LPS addition, but the extremely low amounts of iNOS transcripts in control cultures interfered with reliable quantification of fold-changes by qRT-PCR and yielded exaggerated increases in response to LPS (data not shown). Thus, conventional RT-PCR (35 cycles) followed by visualization of amplicons using ethidium bromide-stained agarose gels was performed. Figure 3c shows that a single 557 bp amplicon corresponding to iNOS mRNA is detected in each of 4 separate culture preparations treated for 6 h with LPS (25 ng/ml). Vehicle-treated control samples do not contain detectable amounts of the iNOS amplicon. Altogether these data suggest that the microglia in the cultures are activated, based on the predominant amoeboid morphology and expression of immunoreactive ED1. However, microglia do not express detectable amounts of the inflammatory markers COX-2, iNOS, and TNF-α protein or mRNA unless they are treated with the potent endotoxin LPS. When comparing serum-AraC⁻ cultures to those grown in a chemically defined medium lacking both AraC and serum, on western immunoblots there is little or no detectable COX-2 induction in response to 12 or 25 ng/ml LPS, and COX-2 is detectable in response to 50 ng/ml LPS, but not in the amounts observed in serum-AraC⁻ cultures treated with 50 ng/ml LPS (data not shown). Nevertheless, the ED1-stained microglia in the defined-AraC⁻ cultures are the source of any detectable LPS-mediated increases in COX-2 and iNOS proteins (data not shown). Real-time RT-PCR analysis reveals significant LPS-mediated induction of TNF-α at 1 h but not 6 and 24 h time points in defined-AraC⁻ cultures compared to vehicle controls (1 h fold-increases: 12 ng/ml = 33.6; 25 ng/ml = 47.4; 50 ng/ml = 56.6; p < 0.02 for each concentration, compared to control, n = 3). However, these fold increases are significantly lower than those observed in the corresponding serum-AraC⁻ cultures (p < 0.0001 for each concentration, compared to respective concentration in serum). Similarly, significant increases in COX-2 mRNA are observed 24 h after 25 ng/ml and 50 ng/ml LPS, compared to control cultures (24 h fold-increases: 25 ng/ml = 29.8; 50 ng/ml = 35.0; p < 0.04, n = 3), and the fold-increase in response to 25 ng/ml LPS is significantly less than that observed in the corresponding serum-AraC⁻ cultures (p < 0.0001, compared to same concentration in serum). In summary, the data comparing LPS-mediated changes in TNF-α and COX-2 mRNAs in cultures reared in chemically defined versus serum-containing media are consistent with the reduced numbers of microglia in chemically defined cultures (~23 % compared to serum).

Fig. 3.

mRNAs encoding COX-2, iNOS, and TNF-α are induced in a time- and concentration-dependent manner after addition of LPS. AraC⁻ cultures were grown for 7–8 DIV, and vehicle or LPS (open bar 12 ng/ml; black bar 25 ng/ml; striped bar 50 ng/ml) were added 1, 6, or 24 h before harvesting. Whole cultures homogenates were processed for qRT-PCR (COX-2, TNF-α) or conventional RT-PCR (iNOS). a Fold-induction of COX-2 mRNA compared to vehicle control following addition of LPS (*p < 0.004 compared to control, **p < 0.04 compared to control, n = 3). b Fold-induction of TNF-α mRNA compared to vehicle control following addition of LPS (*p < 0.0003 compared to control, n = 3). c Amplicon corresponding to iNOS mRNA (557 bp, stained with ethidium bromide) in vehicle control (−) or 6 h after addition of 25 ng/ml LPS (+) in 4 different culture preparations (1–4)

Microglia in Co-cultures are Phagocytically Active

Phagocytosis by both astrocytes and microglia in culture has been reported, although most evidence points to activated microglia as being the primary means by which cell debris, dead neurons, and pathogens are cleared (Bechmann and Nitsch 1997; Ito et al. 2007; Koizumi et al. 2007; Neher et al. 2013). To assess whether microglia are the principal phagocytic cell type, AraC⁻ cultures grown for 7–8 DIV were incubated with fluorescently labeled polystyrene beads for 1 h, then cultures were fixed and stained with antibodies against Iba1 or GFAP, to stain microglia or astrocytes, respectively. Beads of two sizes (2 μm and 5.5 μm) were used to ensure that phagocytosis, and not pinocytosis, was the principal mode of engulfment (Pratten and Lloyd 1986; Koval et al. 1998; Hewlett et al. 1994). As exemplified in Fig. 4a, AraC⁻ cultures include numerous Iba1-labeled microglia (red) with the majority of them containing one to several fluorescent beads (2 μm, green). Figure 4b shows a lack of fluorescent beads within GFAP-labeled astrocytes (red). Similar results are obtained using 5.5 μm beads which are unequivocally too large to be pinocytosed. Figure 4c shows that several Iba1-labeled microglia (green) in AraC⁻ cultures contain fluorescent beads (note that in this case the larger beads are red). Figure 4d shows that no beads are contained within GFAP-positive astrocytes (green). Similar results were observed in 3 different co-cultures with each size beads. Also, the microglia present in cultures grown in chemically defined media lacking AraC are able to engulf fluorescent beads (data not shown). Altogether, these data indicate that microglia, not astrocytes, are the principal phagocytic cell type in AraC⁻ co-cultures of glia and neurons.

Fig. 4.

Microglia, but not astrocytes, in AraC⁻ cultures phagocytose polystyrene beads. AraC⁻ cultures grown for 7–8 DIV were incubated for 1 h with small (2 μm, upper panels) or large (5.5 μm, lower panels) fluorescent beads, cultures were fixed and incubated with antibodies against Iba1 or GFAP, and photographed (×20 magnification). a, b Numerous small fluorescent beads (green) are contained within Iba-labeled microglia (left, a) but not GFAP-labeled astrocytes (right, b). c, d Numerous large fluorescent beads (red) are contained within Iba1-labeled microglia (left, c) but not GFAP-labeled astrocytes (right, d). Similar results were observed in 3 different culture preparations incubated with 2 or 5.5 μm beads (not shown) (Color figure online)

The next series of experiments were designed to determine if microglia can phagocytose a natural cargo, CGNs that have undergone apoptosis. It is well established that the majority of CGNs originally plated undergo spontaneous apoptosis, beginning at ~3 DIV, if the standard growth medium is not supplemented with elevated amounts of extracellular KCl (f.c. 25 mM KCl, Gallo et al. 1987). To determine if microglia effectively phagocytose CGNs under these conditions, live-dead assays were performed in AraC⁻ and AraC⁺ cultures grown for 8 DIV and also for 2–3 DIV in medium containing 5 mM KCl. Figure 5b is a representative photograph demonstrating numerous calcein-containing/living microglia, astrocytes, and CGNs, but very few PI-stained/dead CGNs in AraC⁻ cultures grown for 8 DIV. In contrast, Fig. 5d shows numerous PI-stained/dead CGNs (red) in AraC⁺ cultures grown for 8 DIV, as well as some living CGNs and some glia. Note in these experiments, the live-dead assay in which the strongly fluorescent calcein accumulates in living cells (green) underscores the dramatic differences in numbers of microglia and astrocytes in AraC⁻ versus AraC⁺ cultures at this stage. Figure 5f is a histogram summarizing data obtained from multiple preparations, showing significant differences in the numbers of dead CGNs in AraC⁺ versus AraC⁻ cultures. Counting the numbers of surviving neurons at 8 DIV was challenging due to the presence of numerous calcein-containing microglia and astrocytes as well as CGNs in AraC⁻ cultures. However, to verify that a similar degree of CGN apoptosis occurs in AraC⁻ cultures, compared to AraC⁺ cultures, estimates of living CGNs were made in two different fields that afforded good visualization of CGNs from 4 culture preparations. These estimates indicate that the numbers of surviving CGNs are not significantly different from each other (AraC⁻: 113.0 ± 11.3; AraC⁺: 112.2 ± 4.7, n.s.). However, there is little evidence of PI-labeled/dead CGNs or CGN fragments within microglia in AraC⁻ cultures grown for 8 DIV, suggesting that the phagocytic process is virtually complete by this time. To further assess apoptosis and phagocytosis in this context, live-dead assays were performed in AraC⁻ and AraC⁺ cultures grown for 2–3 DIV. At this earlier stage of culture, the number of microglia in AraC⁻ cultures is less than observed at 8 DIV, possibly affording opportunities to visualize apoptotic CGNs as well as microglia engulfing apoptotic CGNs/fragments. Indeed, Fig. 5a, b are representative photographs showing numerous apoptotic CGNs in both culture conditions, confirming that glia in AraC⁻ cultures do not provide protection against naturally occurring cell death. Further, Fig. 5e shows several PI-stained apoptotic bodies within Iba1-immunolabeled microglia in AraC⁻ cultures grown for 2–3 DIV, providing direct evidence for engulfment of dead CGNs by phagocytic microglia. Altogether, these results indicate that microglia in the AraC⁻ cultures effectively phagocytose CGNs undergoing spontaneous cell death, even when most of the originally plated CGNs have undergone apoptosis.

Fig. 5.

Microglia phagocytose the majority of CGNs undergoing spontaneous apoptosis in AraC⁻ but not AraC⁺ cultures. In this series of experiments, AraC⁻ and AraC⁺ cultures were grown without medium supplementation with 20 mM KCl to induce apoptosis. Live-dead assays were performed at 2–3 DIV (a, c) and 8 DIV (b, d). Three random fields of living (green/calcein) and dead (red/PI) cells were photographed (×20 magnification). a Representative field in 2–3 DIV AraC⁻ cultures. b Representative field in 2–3 DIV AraC⁺ cultures. c Representative field in 8 DIV AraC⁻ cultures. d Representative field in 8 DIV AraC⁺ cultures. Note that the calcein stain for living cells reveals the presence of numerous microglia, astrocytes, and CGNs in AraC⁻ cultures, compared to AraC⁺ cultures, especially after 8 DIV. e AraC⁻ cultures grown for 2–3 DIV were incubated with PI to label apoptotic CGNs/fragments (red), followed by fixation and labeling with Iba1 antibodies to stain microglia (green). f Histogram showing significant differences at 8 DIV in numbers of dead CGNs in AraC⁺ cultures compared to AraC⁻ cultures (*p < 0.001; n = 4) (Color figure online)

Exposure to the Excitotoxin Kainate Triggers CGN Death and Phagocytosis but Not Conversion of Microglia to an Inflammatory or Cytotoxic Phenotype

Experimental paradigms that link inflammatory challenges, including kainic acid exposure, with phagocytic activation of microglia have reported exacerbation of injury and neuronal death due to release of neurotoxins by microglia (Tikka et al. 2001; Lee et al. 2003; Cho et al. 2008; Zheng et al. 2010; Zhu et al. 2010; Hong et al. 2010; Neher et al. 2011). To determine if challenge with excitotoxic concentrations of excitatory amino acids exacerbates neuronal death in microglia-rich neuronal cultures, CGN survival was assessed by FDA/PI staining 24 h following an acute 30 min exposure to kainic acid (250 or 500 μM). Figure 6a–d are representative photomicrographs showing FDA-labeled/living CGNs (green) and PI-labeled/dead CGNs in control and kainate-treated (250 μM) cultures grown in the presence or absence of AraC. Figure 6e, f are histograms summarizing data from multiple experiments (E: 250 μM kainate; F: 500 μM kainate). Compared to their respective control cultures, kainate triggers significant and equivalent loss of CGNs in both culture conditions. However, few if any dead CGNs are detectable in AraC⁻ cultures, whereas numerous dead neurons are visible in AraC⁺ cultures, and these differences are statistically significant. Both concentrations of kainate triggered massive death of CGNs with no significant differences between them. These results strongly suggest that microglia in AraC⁻ cultures, but not AraC⁺ cultures, effectively phagocytose the majority of dead neurons by 24 h after kainate exposure. However, there is little evidence of PI-labeled/dead CGNs in microglia after 24 h (data not shown), suggesting that the phagocytic process is essentially complete at this time. To directly determine whether microglia in AraC⁻ cultures phagocytose dead CGNs, cultures were examined for an evidence of PI-stained apoptotic bodies within microglia at earlier times after exposure to kainate. In these experiments, PI was added to cultures at 30 min, 1, 2, 3, and 24 h after acute exposure to kainate (250 μM, 30 min), cultures were subsequently fixed, and microglia were labeled with antibodies against Iba1. A minimum of three randomly selected fields per culture preparation were photographed. As exemplified in Fig. 6g, h, PI-labeled inclusions are visible within Iba1-labeled microglia 30 min and especially 3 h after kainate exposure. Examples of microglia that had engulfed dead-PI-stained neurons are also observed 1 and 2 h after exposure to kainate (data not shown). Qualitatively similar data were obtained in 5 different culture preparations. In agreement with Christensen et al. (2006), microglia exposed to kainate exhibited pronounced morphological alterations associated with phagocytosis; they became broader and flatter. Consistent with the significant reduction in microglia numbers in cultures grown in a chemically defined medium compared to sister cultures reared in serum-containing medium, there is a significant difference in the number of dead CGNs 24 h after exposure to 250 μM kainate (defined-AraC⁻: 69.5 ± 18.2; serum-AraC⁻: 3.4 ± 1.1, n = 4, p < 0.04). However, there is no significant difference in the number of living CGNs after kainate exposure (defined-AraC⁻: 69.5 ± 6.0; serum-AraC⁻: 76.1 ± 14.7, n = 4).

Fig. 6.

Challenge with the excitotoxin kainate triggers widespread CGN death, but preserves the phagocytic phenotype in microglia without exacerbating neurotoxicity. AraC⁻ and AraC⁺ cultures were grown for 8 DIV in standard medium (supplemented with 20 mM KCl). At 8 DIV, cultures were exposed to kainate (250 or 500 μM, as indicated) for 30 min, and then after overnight incubation in conditioned medium, cultures were assayed for living (FDA, green) and dead (PI, red) CGNs and photographed (×20 magnification). In one series of experiments, the presence of PI-labeled apoptotic fragments or dead CGNs that had been engulfed by microglia was assessed at different times after exposure to kainate (×40 magnification). a–d Representative photomicrographs of living and dead CGNs in AraC⁻ cultures (upper panels, a control, b kainate) versus AraC⁺ cultures (lower panels, c control, d kainate). Note that the FDA stain for living cells is not as sensitive as calcein, and green fluorescence is limited to living CGNs in this assay. (Experimental evidence indicates that calcein is better retained in viable cells than fluorescein and tends to have brighter fluorescence in several mammalian cell types (Lifetechnologies.com: Viability and Cytotoxicity Assay Reagents—Sect. 15.2).) e, f Histograms summarizing data from different culture preparations showing that kainate-treated AraC⁺ cultures have a significantly greater number of PI-stained/dead CGNs compared to kainate-treated AraC⁻ cultures (open bars; **p < 0.001), and that both culture conditions have significantly reduced but comparable numbers of FDA-stained/living CGNs after kainate treatment (filled bars, *p < 0.001, compared to respective untreated control culture). Shown are control cultures and cultures treated with 250 μM kainate (e, n = 3) and 500 μM kainate (f, n = 6). g, h Cultures grown in AraC⁻ medium were exposed to kainate (250 μM) for 30 min, and then incubated with PI to label dead neurons and apoptotic CGNs/fragments (red), followed by fixation and labeling with Iba1 to stain microglia (green). Time points ranging from 30 min, 1, 2, 3, and 24 h were sampled and photographed. Shown are 30 min (g) and 3 h (h) time points (Color figure online)

As indicated, other model systems using neuron–microglia co-cultures or slices report that microglia become cytotoxic and exacerbate neuronal death after exposure to toxic concentrations of kainate. Moreover, conversion of microglia to a cytotoxic phenotype is associated with a robust inflammatory response. The next series of experiments was designed to assess the inflammatory status of microglia in AraC⁻ cultures after exposure to kainate, for comparison with their response to the potent inflammatory mediator LPS. To test this, AraC⁻ cultures were exposed to kainate (500 μM) for 1 or 6 h (24 h time points were also examined but did not show alterations, possibly due to extensive CGN death by this time), then RNA was extracted and processed for qRT-PCR to examine alterations in mRNAs encoding COX-2, iNOS, and TNF-α (as described for LPS, see Fig. 3). Relative to the vehicle-treated control cultures, kainate treatment generates a significant increase in COX-2 mRNA at the 1 h time point (2.5-fold increase, p < 0.02, n = 3). A modest ~ twofold induction of iNOS mRNA in response to kainate is observed at the 6 h time point, compared to vehicle-treated control cultures, but this difference did not reach statistical significance (p > 0.09, n = 3). Altogether, these data indicate that exposure of AraC⁻ cultures to excitotoxic concentrations of kainate triggers widespread CGN death and activates phagocytosis in microglia. However, kainate treatment does not induce a sustained or robust inflammatory response in microglia, and does not convert microglia to a cytotoxic phenotype based on the observation that CGN death is not exacerbated in AraC⁻ cultures, compared to AraC⁺ cultures. When compared to LPS treatment, the RT-PCR data underscore the distinct effects of different stimuli on microglial activation to a phagocytic phenotype as opposed to a pro-inflammatory phenotype, supporting the notion of a continuum of activation that depends on the context of injury.

Discussion

There is ample evidence that microglia both influence and are influenced by astrocytes and neurons (Sudo et al. 1998; Rosenstiel et al. 2001; Faustmann et al. 2003; Min et al. 2006; Yang et al. 2007; Kettenmann et al. 2011; Ousman and Kubes 2012; Streit et al. 2014). There is also an evidence supporting regional heterogeneity in microglia (Lawson et al. 1990; Elkabes et al. 1996; Ren et al. 1999; Kim et al. 2000; de Haas et al. 2008). To examine the functional effector states of microglia with greater fidelity using an in vitro model, a primary goal in the present study was to characterize microglia in co-culture with astrocytes and neurons, derived from the same brain region. In rodents, CGNs migrate from the external granule layer to the internal granule layer after birth, so P7–8 cerebellum is often used as a tissue source for primary neuronal cultures. Typically, the mitotic inhibitor AraC is added to the growth medium 24 h after plating to inhibit replication of proliferating cell populations, primarily astrocytes and microglia, thereby producing cultures that are more than 90 % enriched in CGNs (Thangnipon et al. 1983; Kingsbury et al. 1985; Gallo et al. 1987). Notably, AraC concentrations used for this purpose are well below those that have been widely reported to trigger death of CGNs and other neuronal types (Kolodny et al. 1985; Martin et al. 1990; Dessi et al. 1995). Based on immunocytochemical staining at 7–8 DIV, microglia and astrocytes represent ~11 and ~10 %, respectively, of the mixed cell population in AraC⁻ cultures, and they are sparse in AraC⁺ cultures (and also significantly reduced in cultures grown in a chemically defined medium lacking AraC). The numbers of CGNs are comparable in AraC⁺ and AraC⁻ cultures (Fig. 6). Thus, omission of AraC from the growth medium of CGN cultures grown in serum-containing medium is permissive for proliferation of microglia and astrocytes so that neuron–glia co-cultures from the same brain region can be studied.

The primarily amoeboid morphology and microglial expression of ED1, an antigen expressed on the membrane of phagolysosomes, indicate that they are further along in the continuum of activation than the least active state (Damoiseaux et al. 1994; Ito et al. 2001), previously referred to as resting microglia and prevailing in normal healthy brain tissue in vivo. Activated microglia have an important protective role as phagocytes, engulfing dead cells, tissue debris, or foreign particles in response to changes in their local environments (Ransohoff and Perry 2009; Ousman and Kubes 2012). By rapidly removing these materials, inflammation is mitigated and homeostasis is re-established. Microglia, but not astrocytes, in AraC⁻ cultures are highly effective phagocytes; they rapidly engulf polystyrene beads (Fig. 4). In addition, it is well established that the majority of CGNs originally plated undergo apoptosis, beginning at ~3 DIV, if the serum-containing growth medium is not supplemented with additional KCl, typically 20 mM (Thangnipon et al. 1983; Kingsbury et al. 1985; Gallo et al. 1987). When examined at 8 DIV, after apoptosis is complete under these growth conditions, there is little or no evidence of dead CGNs in AraC⁻ cultures. This lack of dead cells contrasts with the numerous dead CGNs visible in cultures reared in the presence of AraC to suppress microglial growth (Fig. 5). A lack of dead CGNs in AraC⁻ cultures, compared to AraC⁺ cultures, is also observed after acute exposure to the potent excitotoxin kainate, and numerous microglia vacuoles containing dead CGNs and fragments are visible within a few hours after kainate exposure (Fig. 6). Notably, the presence of numerous phagocytically active microglia in AraC⁻ cultures is not associated with exacerbation of neuronal death, based on quantitation of living CGNs in AraC⁻ and AraC⁺ cultures. Altogether, these results indicate that microglia in AraC⁻ co-cultures are in a functional effector state that is phagocytic and protective, not cytotoxic, under conditions triggering apoptotic (5 mM KCl growth conditions) or excitotoxic (acute kainate exposure) death in the majority of CGNs.

From a clinical perspective, it is important to understand the environments and types of injuries that result in conversion of microglia from protective to cytotoxic effector states because such information would guide the use of pharmacological agents that inhibit microglial function. Notably, various injuries that are pro-inflammatory, including generation of nitric oxide, TNF-α, and COX-2 in microglia, are associated with exacerbation of neurotoxicity (Araki et al. 2001; Bal-Price and Brown 2001; von Bernhardi et al. 2007; Cho et al. 2008; Blaylock 2013; Xing et al. 2015; Perry and Holmes 2014). In a model of chronic neurodegenerative disease, for example, microglia proliferate and exhibit an amoeboid morphology in response to changes in their microenvironment (Perry and Holmes 2014). In response to pro-inflammatory challenges, these ‘primed’ microglia mount exaggerated inflammatory responses that are neurotoxic, and thus represent viable targets for microglia inhibitor therapies. In comparison, microglia in AraC⁻ cultures are similar in their amoeboid morphologies, expression of ED1, and ability to phagocytose polystyrene beads, which are characteristics of activated but not resting microglia (Pratten and Lloyd 1986; Koval et al. 1998). In contrast, however, they do not mount a robust or exaggerated inflammatory response under conditions leading to death of the majority on CGNs, as demonstrated by comparing inflammatory markers in AraC⁻ cultures exposed to vehicle alone, acute kainate (30 min), and LPS. Note that LPS is the prototypical endotoxin and arguably the most potent inducer of inflammation triggering microglial secretion of pro-inflammatory cytokines, nitric oxide, and eicosanoids (Paludan 2000). Moreover, inflammation and microglial cytotoxicity have also been reported after kainate treatment (Tikka et al. 2001; Lee et al. 2003; Cho et al. 2008; Zheng et al. 2010; Zhu et al. 2010; Hong et al. 2010) and, in some cases, are consequences of phagocytosis (Neher et al. 2011). Based on Western immunoblotting and immunocytochemistry with antibodies against COX-2 and iNOS, microglia in untreated AraC⁻ cultures do not express detectable amounts of these inflammatory proteins, whereas addition of LPS generates a concentration-dependent increase that is localized to microglia (Fig. 2). Using RT-PCR to assess mRNAs corresponding to COX-2, iNOS, and TNF-α, we observed that incubation with LPS turns on iNOS expression and induces robust increases in TNF-α mRNA and COX-2 mRNA (up to ~250- and ~116-fold, respectively). In comparison, a modest ~2.5-fold increase in COX-2 mRNA is the only significant change observed following kainate treatment. Thus, an important distinction between the effects of kainate exposure reported herein and the cytotoxic effects observed in other contexts is the modest pro-inflammatory response in AraC⁻ cultures. Kainate triggers a mixed form of cell death in CGNs with both apoptotic and oncosis (often referred to as necrosis) components (Ankarcrona et al. 1995; Cebers et al. 1997; Giardina and Beart 2001). Conditions leading to cell death by apoptosis are associated with expression of anti-inflammatory molecules and immunosuppression, whereas oncosis is commonly associated with a pro-inflammatory state (Voll et al. 1997; Gregory and Devitt 2004; Griffiths et al. 2009). Possibly, expressions of anti-inflammatory molecules by CGNs undergoing apoptosis in the context of acute kainate exposure mitigate its effects on CGNs undergoing oncosis that would otherwise promote a pro-inflammatory state leading to cytotoxicity. In future studies using this model, it will be instructive to characterize these relationships with emphasis on the types of insults that convert microglia from a protective to a pro-inflammatory cytotoxic phenotype.