Inflammatory factors and amyloid β-induced microglial polarization promote inflammatory crosstalk with astrocytes

The immunological responses are a key pathological factor in Alzheimer’s disease (AD). We hypothesized that microglial polarization alters microglia-astrocyte immune interactions in AD. M1 and M2 microglia were isolated from primary rat microglia and were confirmed to secrete pro-inflammatory and anti-inflammatory factors, respectively. Primary rat astrocytes were co-cultured with M1 or M2 microglial medium. M1 microglial medium increased astrocyte production of pro-inflammatory factors (interleukin [IL]-1β, tumor necrosis factor α and IL-6), while M2 microglial medium enhanced astrocyte production of anti-inflammatory factors (IL-4 and IL-10). To analyze the crosstalk between microglia and astrocytes after microglial polarization specifically in AD, we co-cultured astrocytes with medium from microglia treated with amyloid-β (Aβ) alone or in combination with other inflammatory substances. Aβ alone and Aβ combined with lipopolysaccharide/interferon-γ induced pro-inflammatory activity in M1 microglia and astrocytes, whereas IL-4/IL-13 inhibited Aβ-induced pro-inflammatory activity. Nuclear factor κB p65 was upregulated in M1 microglia and pro-inflammatory astrocytes, while Stat6 was upregulated in M2 microglia and anti-inflammatory astrocytes. These results provide direct evidence that microglial polarization governs communication between microglia and astrocytes, and that AD debris alters this crosstalk.

AGING activation) and M2 (via alternative activation) [4]. However, it is unclear whether the polarization of microglia impacts their communication with astrocytes.
The activation of microglia and astrocytes is the primary immune response of the CNS, and is involved in the pathological process of Alzheimer's disease (AD). The reactivity of astrocytes and microglia was observed in APP/PS1 mice (an AD model). When these mice were 16 months old, they displayed abnormal microglia with enlarged cell bodies and shorter, thicker and fewer processes. Moreover, glial fibrillary acidic protein-positive cells (astrocytes) were clustered in the hippocampus [5]. AD products such as amyloid-beta peptide (Aβ) may induce microglial activation to the M1 phenotype. For instance, an in vitro study demonstrated that the microglial activity marker CD68 and the M1 phenotype markers tumor necrosis factor alpha (TNF-α), interleukin (IL)-1β and CD86 were upregulated in Aβ-stimulated primary microglia [6]. In another study, the levels of M1 microglia increased after Aβ was injected into the hippocampus in mice [7]. Additionally, Tau was found to induce complement 3 (C3) secretion through the nuclear factor (NF)-κB pathway in astrocytes. Astrocytic C3 was shown to combine with microglial C3 receptors, thus inhibiting debris removal and inducing Aβ and Tau accumulation [8,9]. However, it is not known whether AD debris alters the communication between astrocytes and microglia during microglial polarization.
Toll-like receptor 4 (TLR4) is a critical participant in microglial activation and polarization [10]. NF-κB signal transducer and activator of transcription 6 (Stat6) are downstream regulatory molecules of TLR4, and are also involved in glial inflammation [10,11]. While the TLR4/NF-κB pathway polarizes M1 microglia and promotes inflammatory cytokine secretion [11], Stat6 induces microglia toward a beneficial phenotype and improves phagocyte clearance as a downstream factor of IL-4 [12]. The balance between Stat6 and Stat1 seems to determine the microglial phenotype, as autophagy inhibitors promote anti-inflammatory microglial polarization by upregulating Stat6 and downregulating Stat1 [10]. The NF-κB and Stat6 pathways are also activated during astrocyte inflammation. Excretory/ secretory products activate the NF-κB pathway, which causes astrocytes to release IL-1β and IL-6 [13]. On the other hand, through the Stat6 pathway, astrocytes of the A2 phenotype produce antioxidant factors such as nuclear factor erythroid 2 and arginase-1 (Arg1) [14].
In this study, we used different inflammatory stimuli to promote M1/M2 microglial polarization, and studied the interactions of astrocytes with the polarized microglia. Furthermore, we used Aβ as simulate AD debris, and examined its effects on microglial polarization and microglia-astrocyte interactions. We measured NF-κB and Stat6 levels in microglia and astrocytes to examine the mechanism of inflammation.

Inflammatory factors activated microglia and directed their polarization
We obtained primary glial cells from newborn rats, and isolated primary microglia from the mixed cell cultures. The primary microglia were sometimes rounded and sometimes ovular in shape, and had spindle process. Then, we stimulated the microglia with lipopolysaccharide (LPS) and interferon (IFN)-γ to polarize them toward the M1 phenotype, or treated them with anti-inflammatory cytokines (IL-4 and IL-13) to obtain the M2 phenotype. Morphological changes were observed in the microglia 24 h after the administration of inflammatory substances: the soma was enlarged, the number of processes was reduced, some cells lost their dendrites and swelled to an ovular shape, and ionized calcium-binding adapter molecule 1 (Iba1) was detected in the soma and its projections ( Figure 1B-1F). These results support Tang and Le's claim that both proinflammatory and anti-inflammatory factors can activate microglia [4].
Aβ induced M1 microglial polarization and proinflammatory factor secretion, but was selectively inhibited by IL-4/IL-13 Next, we exposed microglia to Aβ with or without various inflammatory factors to examine the activation mechanism of microglia under AD conditions. The following groups were assessed: control, Aβ, LPS/IFN-γ+Aβ, and IL-4/IL-13+Aβ. The number of iNOS + Iba1 + cells was considerably greater in the Aβ and LPS/IFN-γ+Aβ groups than in the control group (p<0.01) (  Primary microglia were divided into six groups to be treated with different stimuli. Media from these microglia were then collected and co-cultured with primary astrocytes, and these astrocytes were divided into six groups according to the media with which they were treated. Microglial polarization was examined, and cytokine levels in both types of glia were measured. AGING markedly elevated in these groups (p<0.01) ( Figure  2O). In contrast, the number of iNOS + microglia and the protein levels of iNOS were reduced in the IL-4/IL-13+Aβ group (p<0.01) (Figure 2A, 2F, 2O). Thus, antiinflammatory factors prevented Aβ from inducing M1 microglial polarization. On the other hand, the number of Arg1 + microglia and the protein levels of Arg1 were significantly higher in the IL-4/IL-13+Aβ group than in the control group (p<0.01) ( Figure 2G, 2L). The number of Arg1 + microglia was lower in IL-4/IL-13+Aβ group than in the IL-4/IL-13 group, although there was no significant difference in Arg1 protein expression between these groups (p<0.05) ( Figure 2I, 2L). Thus, anti-inflammatory factors shifted Aβ-induced microglial polarization toward the M2 phenotype.

Pro-inflammatory factors and Aβ induced NF-κB expression, while anti-inflammatory factors enhanced Stat6 expression in microglia and microglial mediuminduced astrocytes
The NF-κB pathway component p65 is associated with inflammatory activity [6,18]; thus, we measured p65 expression in microglia and microglial medium-treated AGING astrocytes. The number of p65 + microglia and the mRNA levels of p65 were greater in LPS/IFN-γ, Aβ and LPS/IFN-γ+Aβ microglial groups than in the control group (p<0.01) ( Figure 5A1-5A3, 5I). Media originating from M1 microglia also activated the NF-κB pathway in astrocytes. The number of p65 + astrocytes and the mRNA levels of p65 were greater in the AS-LPS/IFN-γ, AS-Aβ and AS-LPS/IFN-γ+Aβ groups than in the AS-control group ( Figure 5C1-5C3, 5K). However, p65 expression was downregulated in microglia treated with IL-4/IL-13+Aβ and in astrocytes treated with medium from these microglia ( Figure 5A6, 5C6).
Given that Stat6 functions downstream of IL-4 [12], we also examined Stat6 expression in microglia and astrocytes. The number of Stat6 + cells and the mRNA levels of Stat6 were greater in the IL-4/IL-13 microglial group than in the control group (p<0.01) ( Figure  5B4, 5BJ). Interestingly, these measures were also AGING upregulated in the IL-4/IL-13+Aβ microglial group (p<0.05) ( Figure 5B6). In the astrocyte groups treated with microglial media, the number of Stat6 + cells and the mRNA levels of Stat6 were greater in the AS-IL-4/IL-13 and AS-IL-4/IL-13+Aβ groups than in the AScontrol group (p<0.01) ( Figure 5D4, 5L). However, Stat6 expression was lower in the AS-IL-4/IL-13+Aβ group than in the AS-IL-4/IL-13 group (p<0.01) ( Figure  5D6, 5L).

DISCUSSION
The crosstalk between astrocytes and microglia is crucial for many CNS functions, including development, neuronal development and immunological responses to diseases [19]. Proper bidirectional communication depends on a variety of molecules, including cytokines, neurotransmitters and metabolic substances. A review by Jha et al. indicated that proinflammatory substances (e.g., TNF-α, C1q) originating from microglia can influence astrocytes, for instance, by increasing TNF-α production and inducing astrocytosis. On the other hand, the astrocytic molecules LCN2 and ORM2 respectively enhance and inhibit microglial activity [3]. Thus, molecules produced by astrocytes can also alter microglial function.
It is well known that pro-inflammatory factors such as LPS and IFN-γ induce microglial polarization toward the M1 phenotype [20,21], while anti-inflammatory AGING factors such as IL-4 and IL-13 induce polarization toward the M2 phenotype [22]. However, it has been unclear how microglial polarization impacts astrocyte activity. Using different microglial media corresponding to these phenotypes, we found that M1 microglia stimulated pro-inflammatory astrocyte activity, while M2 microglia promoted anti-inflammatory astrocyte activity ( Figure 5).
Recent studies have indicated that active astrocytes also exhibit different subtypes, of which the A1 subtype can be induced by classically activated microglia [23]. Another study demonstrated that the M1 phenotype marker iNOS stimulated pro-inflammatory processes by enhancing nitric oxide production in the CNS [24]. We found that pro-inflammatory molecules such as IL-1β and TNF-α were upregulated in M1 microglia; thus, these substances may have induced pro-inflammatory activity in astrocytes. A1 astrocyte-induced inflammation can subsequently trigger neuronal and glial apoptosis [23]. Moreover, these astrocytes are able to diffuse through the blood-brain barrier, damaging it in the process. Then, as an immune response, leukocytes and macrophages move toward the brain and begin to accumulate in the CNS [25], causing a detrimental cycle. Thus, astrocyte polarization induced by microglial polarization could further damage the CNS.
In contrast to M1 microglia, M2 microglia mainly secrete anti-inflammatory factors such as IL-10 and IL-4. A previous study demonstrated that IL-10 triggered TGF-β secretion from astrocytes, ultimately attenuating IL-1β production [26]. We observed that both M2 microglia and M2 microglial medium-induced astrocytes expressed anti-inflammatory factors (IL-4 and IL-10) with neuroprotective effects in the CNS ( Figure 6).
The communication between different cells in the CNS is involved in the pathogenesis of AD. This is especially evident in the crosstalk between microglia and astrocytes. Astrocytic C3 activated through the microglial C3/C3a pathway increases Aβ pathology [8]. Aβ specifically activates neuroinflammation during the pathogenesis of AD by binding to certain receptors (CD14 and TLR4), thus stimulating pro-inflammatory activity in microglia [27,28]. Here, we observed that M1 microglia were induced by Aβ alone and Aβ combined with LPS/IFN-γ. However, the induction of the M1 phenotype was not greater in the LPS/IFN-γ+Aβ group than in the Aβ group. In a previous study, Aβ plaques surrounded by M2 microglia were observed in 6-month-old PS1 M146L /APP 751SL mice, but more abundant plaques were observed in 18-month-old mice when M2 microglia began to convert to the M1 phenotype [29]. Thus, the limited activation of M1 microglia in our study may have been due to regulatory mechanisms that maintain equilibrium between M2 and M1 activity, since we observed a small amount of M2 microglial polarization after administering Aβ and LPS/IFN-γ. Figure 6. Microglial polarization alters the inflammatory activity of astrocytes. Pro-inflammatory factors and Aβ induce microglia to polarize toward the M1 phenotype and to produce IL-1β and TNF-α, thus stimulating pro-inflammatory activity in astrocytes. Antiinflammatory factors (IL-4/IL-13) induce microglia to polarize toward the M2 phenotype and to produce IL-4 and IL-10, thus stimulating antiinflammatory activity in astrocytes. Furthermore, anti-inflammatory factors reduce pro-inflammatory activity in microglia and astrocytes.

AGING
We also observed that IL-4/IL-13 shifted Aβ-induced microglial polarization toward the M2 phenotype. By inducing the production of M2 microglia, IL-4 and IL-13 may stimulate Aβ degradation. Indeed, in previous studies, M2 microglia reduced the deposition of Aβ (Aβ38, 40 and 42) both in vitro and in vivo [30,31]. In the present study, the inflammatory crosstalk between astrocytes and microglia was also altered by Aβ treatment and microglial polarization. Aβ-induced M1 microglia enhanced the pro-inflammatory activity of astrocytes, whereas IL-4/IL-13 inhibited it. Thus, antiinflammatory factors improved the protective crosstalk between astrocytes and microglia in AD.
The NF-κB pathway is recognized as one of the most important mechanisms involved in M1 microglial polarization. In vitro, LPS induces M1 polarization by activating the TLR4/TLR2/NF-κB pathway, which upregulates iNOS and major histocompatibility complex II [15]. We found that p65 levels increased in LPS/IFN-γ-and Aβ-treated microglia, as well as in astrocytes treated with media from these microglia. Thus, both LPS/IFN-γ and Aβ induced M1 microglial polarization and pro-inflammatory molecule secretion by stimulating the NF-κB pathway. A significant inflammatory trigger of the NF-κB pathway is the binding of molecules to TLR4 [32]. Aβ can serve as an agonist of CD14, which can then bind to TLR4 and activate the NF-κB pathway [33]. Our finding that p65 expression also increased in astrocytes suggested that M1 microglia activate pro-inflammatory astrocytes in an NF-κB-dependent manner. On the other hand, during M2 microglial polarization, we observed the upregulation of Stat6, which is well known to participate in the primary pathways of IL-4 and IL-13 [34]. A previous report indicated that Arg1 reduced inflammatory signaling in microglia by activating Stat6 pathways [12]. Our results indicated that the antiinflammatory properties of M2 microglia and astrocytes were associated with Stat6.
In summary, this study demonstrated that M1/M2 microglial polarization alters the crosstalk between astrocytes and microglia. Moreover, Aβ promotes microglial polarization toward the M1 phenotype.

Primary culture
Primary glial cells were obtained from the brain cortexes of one-to three-day-old newborn Sprague Dawley rats. The cortexes were removed aseptically from the whole brain, and the meninges were stripped carefully under a dissecting microscope. Then, the cortexes were cut into 1-mm x 1-mm pieces and digested with pancreatic enzymes for 10 min. The resulting cell suspension was cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum in a 15-mL poly-L-lysine-coated flask and incubated at 37°C and 5% CO2. After two weeks, the mixed glial cells were shaken at 120 rpm on a gyratory shaker for 4 h. For microglia, the floating cells were collected and reseeded in DMEM/F12 supplemented with 10% fetal bovine serum in 2-mL poly-L-lysine-coated plates. Microglial growth reached the logarithmic phase after four weeks. For astrocytes, the adherent cultures were cultivated in DMEM supplemented with 10% fetal bovine serum. Astrocyte growth reached the logarithmic phase after three weeks.

Immunohistochemistry
Cells were fixed with 4% paraformaldehyde for 10 min and then washed in phosphate-buffered saline (PBS). After being blocked with 5% bovine serum albumin containing 0.5% Triton X-100 for 1 h at 37°C, the microglia were incubated with the following primary antibodies at 4°C overnight: mouse anti-Iba1 (1:200

Western blotting
For protein extraction, 1 mL of NP-40 lysis buffer (Biyuntian) was transferred to each well of a six-well plate of microglia on ice for 30 min, and the cells were then centrifuged at 12,000 rpm for 10 min. The soluble protein solutions were then mixed with 5x sample buffer (Biyuntian) and boiled at 90 o C for 10 min. Equal amounts of protein (60 μg) from each sample were separated via 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis (Sigma) and transferred to polyvinylidene fluoride membranes (Millipore, Billerica, MA, USA). The membranes were blocked in 5% milk with Tris-buffered saline for 1 h, and then were incubated with the following primary antibodies at 4°C overnight: rabbit anti-iNOS, anti-Arg1 (1:500, Bioss) and rabbit anti-actin (1:1000, Santa Cruz Biotechnology). The membranes were washed three times in Tris-buffered saline with Tween and then incubated with the secondary antibody (peroxidaseconjugated IgG, 1:5000, Santa) at room temperature for 2 h. Binding antibodies were visualized using enhanced chemiluminescence on an imaging system (Amersham Imager 600).

Statistical analysis
All experiments were performed at least three times independently. Data were analyzed with one-way analysis of variance (ANOVA) followed by Bonferroni's multiple comparison test (SPSS 20.0, IBM, USA). The results are expressed as the mean ± standard error of the mean (S.E.M.).