Astrocytes and microglia are primary responders to injury, perturbation and cellular stress to, as well as infections of the CNS and induce production and release of molecular signals that initiate glial responses leading to excitotoxicity, inflammation, neurodegeneration, and apoptosis. As part of the CNS innate immune response, both cell types express receptors capable of recognizing not only pathogens but also endogenous ligands released upon stress and injury. The best described of these receptors are the TLRs. Mice and humans express 10 and 12 TLRs, respectively, of which TLR2 and -4 have been primarily implicated in neurodegeneration (Kawai and Akira,2010; Lehnardt et al.,2008; Owens,2009).
Microglia and astrocytes originate from different germ layers in ontogeny and play very different roles in the CNS. Microglia share myeloid lineage with macrophages, dendritic cells, and granulocytes. Often referred to as the macrophages of the brain, microglia have been shown capable of phagocytosis, expression of chemokines and cytokines, and antigen presentation (Hanisch and Kettenmann,2007). Microglia express most TLRs and show strong responses to stimulation in both human and murine tissue (Bsibsi et al.,2002; Kielian,2006).
Astrocytes are of ectodermal origin and are the most predominant cell type in the CNS (Hansson and Ronnback,2003). Astrocytes are intimately associated with neurons, axons and myelin, and astrocyte foot processes surround CNS capillaries to form the glia limitans which delineates the brain parenchyma (Bechmann et al.,2007). Astrocytes play a vital role in maintaining CNS homeostasis by regulating extracellular pH, K+, and glutamate levels via ion channels and membrane transporters (Hansson and Ronnback,2003). Astrocytes modulate synaptic neurotransmission and the activity of neurons (Simmons and Murphy,1992). Astrocytes also play a central role in regulating leukocyte infiltration in experimental autoimmune encephalomyelitis (Toft-Hansen et al.,2011; Voskuhl et al.,2009) and CNS injury (Fuchtbauer et al.,2010; Fuchtbauer et al.,2011; Khorooshi et al.,2008). Adjacent astrocytes are joined by gap junctions and form a syncytium; The gap junctions facilitate the shuttling of molecules such as glucose and its metabolites to and from the glia limitans (Esen et al.,2007; Froes and de Carvalho,1998). Whereas microglia actively probe their local environment by extending cellular projections, astrocytes remain more stationary.
In contrast to microglia, resting astrocytes have been reported to express low levels of only a few TLRs (Kielian,2006). When activated, however, both glial cell types have been shown capable of expressing high levels of most TLRs, for instance in a mouse model of neurocysticercosis (Mishra et al.,2006). In vitro-cultured murine astrocytes have been reported to express and upregulate TLRs 1, 2, 3, 4, 5, 6, and 9 upon ligation inducing the release of proinflammatory cytokines and chemokines (Bowman et al.,2003; Carpentier et al.,2005; Esen et al.,2004). Other studies, however, have shown both presence (Bsibsi et al.,2002) and absence of TLR2 on human astrocytes (Farina et al.,2005). Furthermore, in situ hybridization analyses of mouse brains activated by LPS or cytokines showed TLR2 message expression only in microglia (Owens et al.,2005; Rivest,2003). Similar findings have also been made for lack of TLR2 protein expression by astrocytes in two murine lesion models (Babcock et al.,2006; Wainwright et al.,2010). The fact that TLR2 protein expression was detected on astrocytes in murine neurocysticercosis (Mishra et al.,2006) suggests that astrocyte TLR2 expression may depend on the strength of induction signal.
LPS is a potent stimulator of microglia, whereas resting murine astrocytes in culture have been shown to express no (Falsig et al.,2004a; Sola et al.,2002) or very low levels of TLR4 (Bowman et al.,2003). In spite of this, astrocytes in culture have been reported to respond to LPS (Bowman et al.,2003; Carpentier et al.,2005; Esen et al.,2004; Park and Murphy,1994; Simmons and Murphy,1992). Such findings, however, are complicated by the fact that even a few residual microglia can affect the response of almost pure astrocyte cultures (Saura,2007). It is, therefore, important to be certain of both depletion and any possible residual functionality of contaminating microglia in astrocyte cultures. We have addressed this question by using rigorous strategies to remove or disable microglia, and our findings show that astrocytes are quite dependent on microglia for response to LPS, and their response to other TLR agonists is suboptimal in the absence of co-cultured microglia.
MATERIALS AND METHODS
Adult wild-type C57BL/6 mice were purchased from Taconic (Taconic Europe, Ry, Denmark). Transgenic CD11b-HSV-tk mice (FVB background) were kindly provided by Dr Jeppe Falsig, University Hospital Zürich, and were bred in the Biomedical Laboratory, University of Southern Denmark. All mice were kept in a temperature and humidity controlled environment with 12 h light–dark cycle and were provided with food and water ad libitum. Experiments were conducted in the Biomedical Laboratory according to the guidelines of the National Danish Animal Research Committee.
Primary Mixed Glial and Macrophage Cultures
Primary mixed glial cultures (MGCs) were prepared from whole brains of mice at postnatal day 3–5. Brains were dissociated by trypsin digestion after which the glial cells were isolated by percoll density centrifugation. The cells were counted and plated in flasks (Nunc A/S, Roskilde Denmark) and maintained in DMEM Glutamax + 10% FBS (Gibco/Invitrogen, Taastrup, Denmark) 100 U/mL penicillin-streptomycin Gibco/Invitrogen).
Peritoneal macrophages were obtained by injecting 1 mL 3% thioglycollate (Sigma–Aldrich Denmark A/S, Copenhagen Denmark) intraperitoneally into adult C57BL/6 mice. Three days after injection, mice were killed by cervical dislocation, and 10 mL Ca2+/Mg2+-free HBSS was injected into the peritoneum. Subsequently, the peritoneal fluid was withdrawn. After two washes in HBSS, the cells were counted and plated in flasks (Nunc A/S). Cell cultures were stimulated for 24 h with the agonists shown in Table 1.
Table 1. TLR Agonists and Cytokines Used in this Study
LPS and ultrapure LPS were from E. coli 0111:B4.
HPLC purified ultrapure LPS (upLPS) was used to exclude signals from contaminants that might stimulate other TLRs such as TLR2. UpLPS was used in all experiments unless otherwise stated.
Cultures reaching confluence were shaken overnight in an orbital shaker at 200 rpm to remove microglia. The remaining monolayer was trypsinized and replated. In later experiments, shaking was initiated before the cultures having reached confluence (from day 3–4 and onwards until usage). The culture flasks were shaken overnight repeatedly until evaluated to be devoid of microglia by phase contrast microscopy. Typically, four rounds of overnight shaking were required to obtain astrocyte cultures that did not respond to LPS stimulation. Medium was exchanged every 4 days or after each shaking. We typically used 2-week-old cultures. Older cultures (3–4 weeks) were sometimes used, without noticeable difference.
Microglial Depletion by Suicide
MGCs were prepared from CD11b-HSVtk transgenic mice (Heppner et al.,2005) and their wild-type littermates at postnatal day 3–5. Cells were counted and plated in flasks (Nunc A/S) in DMEM Glutamax + 5% FBS (Gibco/Invitrogen) and 100 U/mL penicillin-streptomycin (Gibco/Invitrogen) and 5–10 μg/mL Ganciclovir (GCV, Sigma–Aldrich). Medium was exchanged every 3–4 days. Paralysis of residual microglia was indicated by a reduction in surface expression of TLR2 and CD45 analyzed by flow cytometry (Fig. 1F,G), plus lack of microglial responsiveness to TNFα stimulation (by upregulation of TLR2) (data not shown).
Overnight shaken cultures were washed three times in Ca2+/Mg2+-free HBSS and trypsinized using EDTA/trypsin. After washing in HBSS, the cells were resuspended in DMEM Glutamax + 5% FBS (Gibco/Invitrogen) and 100 U/mL penicillin-streptomycin (Gibco/Invitrogen) and plated in plastic Lab-Tek™ Chamber Slides (Nunc A/S). The cells were washed three times in HBSS after reaching confluency and subsequently fixed in 4% PFA, PBS for 15 min 4°C. The upper structure of the chamber slides was removed and after washing in PBS, astrocytes and microglia were stained using rabbit-anti-GFAP (DAKO, Glostrup, Denmark) and rat-anti-CD11b (Serotec, Düsseldorf, Germany) in PBS w 5% FBS 1% TritonX100 (Sigma–Aldrich) for 1 h at room temperature. After rinsing in PBS, the cells were incubated with secondary antibodies: Donkey-anti-rabbit conjugated to A488 (Gibco/Invitrogen) and donkey-anti-rat conjugated to A555 (Gibco/Invitrogen) in PBS w 5% FBS 1% TritonX100 for 1 h at room temperature. Cell nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI, Gibco/Invitrogen) before mounting with gelvatol, a polyvinyl alcohol and glycerol-based mounting media. Images were acquired using a digital camera model Olympus DP71 mounted on an Olympus BX51 microscope (Olympus Danmark A/S, Ballerup Denmark). Images were combined using Adobe Photoshop CS5 (Adobe Systems Denmark A/S, Copenhagen, Denmark), using RGB channels to visualize double- and triple-labeled cells.
Flow Cytometric Analysis of Astrocytes and Microglia
Cultures were washed three times in Ca2+/Mg2+-free HBSS before the cells were removed by scraping and dissociated using gentle pipetting. After pelleting, the cells were resuspended in flow cytometry blocking solution (HBSS with 1.5% FBS (Sigma–Aldrich Denmark A/S) containing 2 μg/mL Fc Block (BD Biosciences, Brondby Denmark), 50 μg/mL hamster IgG (Jackson ImmunoResearch, West Grove, PA USA), and 0.1% sodium azide. Table 2 lists the antibodies and staining reagents used.
Table 2. Antibodies and Staining Reagents for Flow Cytometry
Peridinin Chlorophyll Protein Complex – Cyanine5.5 (PerCP Cy5.5)
Santa Cruz Biotech., Inc, Heidelberg, Germany
eBioscience, San Diego, CA, USA
Alexa Fluor (A)488
Ultrapure LPS E. coli 0111:B4
The staining reagents were added and incubated for 15 min at 4°C. The cells were washed once in flow cytometry buffer (1.5% FBS in HBSS, 0.1% sodium azide). Bound LPS was visualized using APC-conjugated streptavidin. The cells were fixed and permeabilized using Cytofix/Cytoperm™ Fixation/Permeabilization Solution Kit (BD Biosciences). The cells were then incubated with anti-GFAP. Astrocytes were identified by GFAP expression and microglia by expression of CD45 or CD11b. Expression levels were measured as the geometric mean fluorescent intensity of gated cell populations. Flow cytometric data were collected using a FACSCalibur (BD Biosciences) and analyzed using Flowjo (Tree Star, Inc., Ashland OR, USA).
The CCL2 content of supernatants from glial cultures was determined using the Mouse CCL2 ELISA Ready-SET-Go!® (eBioscience) according to the manufacturer's instructions.
Statistical analysis was performed using Graphpad Prism 5.0 (Graphpad Software Inc. La Jolla, CA, USA). Comparison of medians in two groups was performed using the Mann–Whitney rank sum test. For multiple comparisons, Kruskal–Wallis one-way ANOVA was performed followed by Dunns multiple comparisons test. P-values are indicated as follows: *P < 0.05; **P < 0.01; ***P < 0.001.
Depletion of Microglia from MGCs by Differential Adhesion
Astrocytes in confluent MGCs adhere to the plate surface forming a syncytial layer of cells (Fischer and Kettenmann,1985) on top of which microglia sit as single rounded-up cells. Traditional methods for removing microglia rely on their differential adhesion by which the rounded-up microglia detach from the astrocyte bed layer when culture flasks are subjected to prolonged orbital shaking (Giulian and Baker,1986). This methodology is usually combined with trypsination and replating. In our hands, a single round of overnight shaking followed by trypsination and replating did not produce completely microglia-free astrocyte cultures, as determined by immunofluorescent labeling of astrocytes (GFAP, green) and microglia (CD11b, red) (Fig. 1A). The question then was whether the residual microglia contributed to astrocyte response. To address this, we had to remove them. We initially removed the microglia by multiple rounds of shaking. Phase contrast microscopy was used to estimate the number of residual microglia as a measure of depletion efficiency after each round of shaking. MGCs before and after depletion are shown as phase contrast images (Fig. 1B,C) and contained as few as <0.5% microglia when analyzed by flow cytometry (Fig. 1D,E). However, several rounds of shaking were required to obtain this level of microglial depletion, and this was both time-consuming and subject to contamination.
Astrocytes and Microglia in Co-culture Respond to LPS
TLR2 expression was used as a measure of glial activation because it has the advantage that it allows double-staining for lineage identification. TLR2 expression by astrocytes and microglia in MGCs was measured by flow cytometry (Fig. 2). Unstimulated astrocytes did not express TLR2 above background levels, whereas microglia did (Fig. 2 columns 1 and 2 versus 4 and 5, respectively). Stimulation for 24 h with 1 μg/mL LPS caused a significant induction of TLR2 on both astrocytes and microglia (Fig. 2 columns 3 and 6, respectively).
Astrocytes Do Not Express CD14 or TLR4, and Do Not Bind LPS
Responsiveness to LPS has been shown to depend on the expression of TLR4/MD2 as well as the co-receptor receptor, CD14 (da Silva Correia et al.,2001). Astrocyte and microglial TLR4 and CD14 expression was measured by flow cytometry (Fig. 3A,B). The lack of TLR4/MD2 and CD14 expression suggested that astrocytes should not be able to bind or respond to LPS and might also show limited responsiveness to TLR2 and TLR3 agonists (Bowdish et al.,2009; Finberg and Kurt-Jones,2006).
The ability of astrocytes and microglia to bind LPS was also tested by flow cytometry. Astrocytes in MGCs showed very low binding of LPS even when prestimulated by TNFα or LPS (Fig. 3C, columns 1 through 4). In contrast, microglia clearly bound LPS (Fig. 3C, columns 5 vs. 6). Interestingly, TNFα stimulation produced only an insignificant increase in LPS binding by microglia in contrast to LPS stimulation (Fig. 3C, columns 7 and 8, respectively).
TLR Response of Primary Cultured Astrocytes Depends on Functional Microglia
Astrocyte responsiveness was assayed by TLR2 expression. Astrocytes in co-culture with microglia responded strongly to the TLR agonists Pam3CSK4, Poly(I:C), and LPS (Fig. 4). Interestingly, neither astrocytes purified by differential adherence (A) or by GCV depletion (B) responded to LPS (Fig. 4A, columns 5 and 6 and 4B, columns 7 and 8) and both showed reduced responses to Poly(I:C) (Fig. 4A columns 3 and 4 and 4B columns 5 and 6). GCV-depleted astrocytes showed reduced response to Pam3CSK4 (Fig. 4B columns 3 and 4). We have confirmed that astrocytes remained viable after shaking or GCV-depletion by assaying their response to Poly(I:C) stimulation. Furthermore, cellular integrity was confirmed using propidium iodide and trypan blue, and we did not find any signs of astrocyte cell death such as granulated cytoplasm or nuclear condensation. The fact that non-HPLC-purified LPS was used on the astrocytes prepared by shaking emphasized the significance of lack of astrocyte response.
Astrocyte response to Pam3CSK4 (Fig. 4B) may reflect influence of residual microglia in GCV-depleted cultures, or that borderline TLR2 expression was sufficient for response.
Astrocyte LPS Response Can Be Facilitated by Factors Released by Microglia
As shown previously, astrocytes in presence of microglia responded strongly to LPS (Fig. 2, column 3). To determine whether astrocytes themselves were induced to respond to LPS in mixed glial cell cultures, and to assay whether LPS response of astrocytes could be induced independently of direct cell-to-cell contact with microglia, we added supernatants of MGCs with LPS to purified astrocytes and assayed TLR2 expression. Astrocytes showed a twofold induction of TLR2 (Fig. 5A, columns 3 and 5). A similar TLR2 induction could be achieved by adding supernatants from MGCs or from peritoneal macrophages that had been prestimulated with LPS (data not shown). The level of TLR2 expression on unstimulated astrocytes was not affected by the presence of microglia in MGCs (Fig. 5A, columns 1 and 3). The enhanced responses induced by addition of LPS to astrocytes treated with MGC supernatants suggest that microglia and macrophages secrete soluble factors that either synergize with or facilitate astrocyte response to LPS, although not to the same extent as in co-cultures. This indicates that cell–cell contact and soluble mediators synergize to induce astrocyte response.
CCL2 Release from Glial Cultures
To further investigate the dependency of astrocytes on microglia, we assayed the production of CCL2. This physiologically relevant mediator is known to be produced by both astrocytes and microglia (Babcock et al.,2003; Khorooshi et al.,2008). Thus, the CCL2 titers detected from LPS-stimulated MGCs in Fig. 5B column 2 could not be assigned to either cell type. Determination whether MGC supernatant could induce a LPS response (as in 5A) was compromised by the fact that MGC supernatants already contained measurable CCL2. However, a significant increase in CCL2 was detected in presence of both MGC supernatant and LPS. As before, the cell source of this additional CCL2 could not be identified. Nevertheless, in contrast to the very high levels of CCL2 in MGCs, purified astrocytes generated background levels of CCL2 when treated with LPS (Fig. 5B columns 1–3). However, when treated with LPS in the presence of supernatant from MGCs, the CCL2 levels rose significantly, indicating that astrocytes responded to the combination of MGC supernatant and LPS (Fig. 5B column 4).
TNFα and IL-1β Directly Induce Activation of Astrocytes
Microglia are known to release cytokines, including TNFα and IL-1β, which can act on glial cells in either an autocrine or paracrine fashion (Phulwani et al.,2008). These have been shown to induce TLR2 specific responses measured as NFκB activation in macrophages and endothelial cells (Faure et al.,2001; Wang et al.,2000). We tested the ability of these cytokines to induce TLR2 expression by GCV-purified astrocytes. Pure astrocytes responded to both TNFα and IL-1β by upregulation of TLR2 after 24 h exposure (Fig. 6). Combining both cytokines caused even higher expression of TLR2. LPS did not induce additional response by astrocytes (not shown). Importantly, purified astrocytes responded to cytokines as well as astrocytes in MGC.
Using two different methods to obtain pure astrocytes, we have shown that astrocytes depend on microglia for their ability to respond to LPS. Furthermore, astrocytes depend on the presence of functional microglia for their ability to mount a full response to TLR2 and -3 ligation. Soluble factors released by microglia and macrophages facilitated a low but detectable response of astrocytes to LPS stimulation and the cytokines IL-1β and TNFα could directly induce response.
The traditional method used to achieve pure cultures of astrocytes involves shaking. In our hands, this method was not reliable and required several repeated treatments to ensure high purity, with each round of shaking increasing the risk of contaminating the cultures. We, therefore, used GCV treatment of CD11b-HSVtk transgenic cultures for effective ablation of microglial contribution. Microglia from transgenic CD11b-HSVtk mice become paralyzed or die in the presence of GCV (Heppner et al.,2005). LPS has been shown to contain impurities capable of signaling through other TLRs. We, therefore, used upLPS and a specific TLR2/1 agonist, Pam3CSK4 to bypass concerns about specificity.
In contrast to astrocytes in cultures from wild-type littermates that were treated with GCV, the astrocyte population from transgenic mice was slightly affected by GCV and did not always reach confluence. This may reflect inadvertent effects of microglial thymidine kinase metabolized GCV on proliferative capacity of neighboring cells. Importantly, however this did not compromise our ability to reproduce results obtained from astrocytes purified by shaking. Further optimization of GCV concentrations may reduce this effect. We also chose to scrape off the astrocytes rather than trypsinizing them to preserve surface expression of markers used in their evaluation. Upregulation of TLR2 was used to measure glial response and was assayed by flow cytometry. The data show a clear dependency of astrocytes on microglia for TLR stimulation.
Published reports draw different conclusions regarding the TLR4 responsiveness of astrocytes (Bsibsi et al.,2002; Carpentier et al.,2005; Falsig et al.,2004b). Differences in the preparation of astrocyte cultures may account for this. As an example, we have noted that prior coating of culture flasks with charged compounds such as polyornithine facilitates tighter binding of astrocytes as well as microglia and thus counteracts their subsequent depletion. This has been confirmed by others (Falsig et al.,2004a; Falsig et al.,2006). Also, protocols for removal of microglia by shaking often require that shaking not be initiated until cultures have reached confluence. However, as confluency approaches 100%, removing microglia by mechanical shaking becomes increasingly difficult, as microglia are also lodged within or situated below the astrocyte monolayer/syncytium as ramified microglia, and so shielded from the forces exerted on them during the shaking (Saura,2007). Trapped microglia have been reported to be a source of continued renewal for the microglia top layer (Crocker et al.,2008). The problem of microglia being shielded by astrocytes only becomes apparent once the cultures approach confluency, around 10–14 days after plating. The astrocytes adhere within 24 h. This leaves open a window of around 10 days in which the microglia directly attached to the bottom of the flask can be removed by shaking. Our strategy, therefore, was to initiate shaking from day 2 and monitor the efficiency of depletion using the morphological differences between astrocytes and microglia. Effective depletion typically required 3–4 rounds of shaking. Microglia also can be removed by adding reagents such as L-leucine methyl ester, a lysosomotrophic agent destroying phagocytic cells (Esen et al.,2004; Thiele et al.,1983). This is sometimes combined with antimitotic drugs such as cytosine arabinoside (Ara-C), preventing further proliferation of microglia (Hamby et al.,2006). Others have specifically depleted microglia using antibody-conjugated saporin toxin (Montero et al.,2009). The use of reagents to kill or inhibit proliferation of microglia may limit applicability. Despite the high reproducibility of the GCV depletion/paralysis approach, the requirement for transgenic microglial expression of thymidine kinase also limits applicability, and the shaking method remains a versatile option.
Our results show that medium conditioned by microglia or peritoneal macrophages enable astrocytes to respond to LPS. It is possible that astrocytes were induced by soluble factors to express receptors required to respond to LPS. This has so far been difficult to verify, in part due to the low level of response. Response to LPS depends on cooperation between the LPS binding protein, CD14, MD-2, and TLR4. Given that LBP and CD14 exist in soluble form and that astrocytes remained unresponsive to conditioned medium in the absence of LPS, these two proteins are likely candidates. This could explain our observations that astrocytes in the presence of microglia responded more strongly to TLR2 and -3 ligation because CD14 has been shown to act as a co-receptor for both. The fact that both IL-1β and TNFα were capable of inducing an astrocyte response makes it very likely that specific cytokines play a central role in relaying TLR activation signals. Finally, given the low level of response induced by supernatants, it is probable that direct cell-to-cell contact plays a major role in the activation of astrocytes.
Our data have implications for interpretation of the role of TLRs in controlling astrogliosis in neuroinflammatory diseases. Observations of expression of TLR2 and TLR4 by human astrocytes in multiple sclerosis (MS) may reflect stimulus from adjacent microglia (Bsibsi et al.,2002). Our data do not exclude the possibility of induction of such responsiveness by astrocytes in the highly inflammatory milieu of MS, where soluble mediator responses such as the one we have described may be more effective. The action of peptidoglycans and related molecules that have been demonstrated in the CNS in MS (Schrijver et al.,2001) may contribute to such a microglial-dependent astrogliosis.
Although LPS is known to downregulate TLR4 surface expression (Nomura et al.,2000; Ziegler-Heitbrock,1995), we have shown that LPS stimulated microglia in fact increase their surface binding of LPS (Fig. 3C). These two observations may be explained by the fact that LPS initially is bound by LPS binding protein and CD14 before reaching the TLR4/MD2 complex. A reduction in TLR4/MD2 surface expression does, therefore, not necessarily predict reduced ability to scavenge LPS from the extracellular environment. Rather, LPS binding depends on the expression of CD14, which we show to be highly upregulated in response to LPS stimulation (Fig. 3B).
In this study, we have shown that astrocytes in the presence of microglia are capable of responding to TLR2, -3, and -4 ligation. However in the absence of functional microglia, the astrocytes no longer responded to TLR4 ligation and responded weakly to TLR2 and -3 stimulation. Our results, therefore, suggest that the response of astrocytes to TLR agonists is in large part due to bystander activation by microglia, or the release of soluble factors that permit autonomous astrocyte responses. Consequently, cells within the CNS may become activated in the presence of TLR agonists even though they do not express the TLR receptor themselves. This may have relevance for CNS-protective glial responses.
The authors thank Dr Frank Heppner, (Charité Universitätsmedizin Berlin) and Dr Jeppe Falsig (University Hospital Zürich) for the transgenic CD11b-HSVtk mice. Also, huge thanks go to Pia Nyborg Nielsen and Dr. Peter H. Larsen.