Glutathione S-transferases promote proinflammatory astrocyte-microglia communication during brain inflammation

See allHide authors and affiliations

Science Signaling  19 Feb 2019:
Vol. 12, Issue 569, eaar2124
DOI: 10.1126/scisignal.aar2124

Glutathione S-transferases enhance astrocyte-microglia communication

Astrocytes and microglia contribute to inflammatory responses in the brain. Kano et al. found that the glutathione S-transferases GSTM1 and GSTT2 contributed to astrocyte-mediated enhancement of microglia activation during brain inflammation. Knocking down either enzyme specifically in astrocytes reduced microglia activation in the brains of mice responding to systemic challenge with lipopolysaccharide (LPS). Coculture experiments with primary astrocytes and a microglial cell line showed that GSTM1 was required in astrocytes for nuclear factor κB (NF-κB)–mediated production of the proinflammatory factors GM-CSF and CCL2, which stimulate microglia activation. These findings identify a role for GSTM1 in enabling astrocytes to promote microglia activation during brain inflammation.


Astrocytes and microglia play critical roles in brain inflammation. Here, we report that glutathione S-transferases (GSTs), particularly GSTM1, promote proinflammatory signaling in astrocytes and contribute to astrocyte-mediated microglia activation during brain inflammation. In vivo, astrocyte-specific knockdown of GSTM1 in the prefrontal cortex attenuated microglia activation in brain inflammation induced by systemic injection of lipopolysaccharides (LPS). Knocking down GSTM1 in astrocytes also attenuated LPS-induced production of the proinflammatory cytokine tumor necrosis factor–α (TNF-α) by microglia when the two cell types were cocultured. In astrocytes, GSTM1 was required for the activation of nuclear factor κB (NF-κB) and the production of proinflammatory mediators, such as granulocyte-macrophage colony-stimulating factor (GM-CSF) and C-C motif chemokine ligand 2 (CCL2), both of which enhance microglia activation. Our study suggests that GSTs play a proinflammatory role in priming astrocytes and enhancing microglia activation in a microglia-astrocyte positive feedback loop during brain inflammation.


Astrocytes play a critical role in maintaining normal neuronal function by modulating synaptic activity, supporting neuronal survival, and providing metabolic support (14). In brain inflammation, astrocytes have been suggested to regulate the activity of microglia, neurons, oligodendrocytes, and immune cells infiltrating from the periphery (46). Because both astrocytes and microglia sense immune stimuli and produce inflammatory mediators, it is important to understand the mechanisms by which astrocytes and microglia influence each other’s proinflammatory activities.

Glutathione is a thiol-containing tripeptide and a major antioxidant within cells (7). Decreases in the reduced form (GSH) and increases in the oxidized form (GSSG) are associated with cellular susceptibility to oxidative stress. Glutathione also influences cellular functions through S-glutathionylation, which is the reversible conjugation of a GSH molecule to reactive cysteine residues in proteins (8, 9). Dysregulation of glutathione metabolism is associated with brain inflammation in various neurological and psychiatric disorders (1017). It is not clear, however, how the glutathione system influences inflammatory responses at the mechanistic level.

Glutathione S-transferases (GSTs) are the enzymes that conjugate GSH to target molecules in phase II of metabolic drug detoxification (18). GSTs are a diverse family of cytosolic, mitochondrial, and microsomal enzymes that prevent cellular damage from the noxious stimuli of xenobiotic metabolites (1820). GSTs are widely present throughout the body and are particularly abundant in the liver, kidney, and lung (1820). In addition, some of the alpha, mu, and pi classes of GSTs (GSTA4, GSTM1, and GSTP1, respectively) have been detected in both human and rodent brains (2123). A recent study suggested that GSTM1 is one of the most abundant proteins in astrocytes (24).

Accumulating evidence shows that GSTs also influence a wide range of biological mechanisms, such as redox homeostasis, signal transduction, cell proliferation, and cell death (2527). GSTs exert these regulatory functions by activating or inhibiting their target molecules, such as c-Jun N-terminal kinases (JNKs), apoptosis signal–regulating kinase 1 (ASK1), and nuclear factor κB (NF-κB), through either protein-protein interactions or S-glutathionylation (27, 28). The current knowledge about these non–phase II detoxification roles of GSTs, however, is still very limited. In the brain, neuronal GSTP1 has been shown to protect neurons from cell death in an animal model of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)–induced degeneration of dopaminergic neurons in the substantia nigra (29). In contrast, the roles of specific GSTs in glial cells have not been well characterized, particularly in vivo.

Genetic studies also suggest that variations in the genes encoding GSTs are involved in neurological and psychiatric disorders with immune dysregulation, such as Parkinson’s disease, Alzheimer’s disease, multiple sclerosis, schizophrenia, and autism (16, 3034). Altered expression of GST activity–related genes has been found to be one of the most drastically changed molecular signatures, together with immune- and microglia-related genes, in postmortem brains from patients with late-onset Alzheimer’s disease (35). Nevertheless, it has not been shown whether GSTs are involved in the regulation of astrocytes and microglia during brain inflammation.

Here, we investigated the role of GST enzymes in astrocytes in a mouse model of brain inflammation. We found that GSTs in astrocytes, particularly GSTM1, were required for microglial activation and tumor necrosis factor–α (TNF-α) production in brain inflammation induced by systemic lipopolysaccharide (LPS) administration. Mechanistically, GSTM1 activated NF-κB and induced proinflammatory mediators such as granulocyte-macrophage colony-stimulating factor (GM-CSF, also known as CSF2) and C-C motif chemokine ligand 2 (CCL2) in astrocytes in response to TNF-α and interleukin-1β (IL-1β). Thus, we propose that GSTs prime the inflammatory responses of astrocytes and enhance the activation of microglia through the secretion of proinflammatory mediators by astrocytes.


GSTM1 and GSTT2 are enriched in astrocytes in the mouse brain

Whereas previous studies showed that GSTM1 and GSTP1 proteins are present in the mouse brain (21, 29), region and cell type specificity of their distribution has not been fully addressed. It has been reported that GSTM1 is one of the most abundant proteins in astrocytes (24). In addition, it is not clear whether other GST enzymes, such as GST theta (GSTT), are also present in the brain. Thus, we examined GSTM1 and GSTT2 in various regions of the mouse brain. As expected, GSTM1 was abundant in the mouse brain, including the cortex, hippocampus, striatum, and cerebellum (Fig. 1, A and B, and fig. S2A). The amount of GSTM1 in the hippocampus, cerebellum, and the whole brain was nearly equivalent to that in the lung, where GSTM1 has been reported to be abundant (36, 37). GSTT2 also showed a similar pattern, although its abundance in the brain was much lower than that of GSTM1. Immunohistochemical analysis revealed that both GSTM1 and GSTT2 were enriched in astrocytes compared to neurons, oligodendrocytes, and microglia (Fig. 1, C and D, and fig. S1). There was no sex difference in the abundance or distribution of GSTM1 and GSTT2 (Fig. 1, B to D, and fig. S2B). These results suggest that GST enzymes, particularly GSTM1, may regulate astrocyte function in the mouse brain.

Fig. 1 Enriched expression of GSTM1 in astrocytes in the mouse brain.

(A) A representative Western blot for GSTM1 and GSTT2 in the cerebral cortex, hippocampus, striatum, cerebellum, whole brain, and lung of C57BL/6 wild-type (WT) mice (8 weeks of age, male). β-Actin is a loading control. (B) The abundances of GSTM1 and GSTT2 in each brain area relative to those in the lung were quantified by densitometry. Male (8 to 11 weeks of age), n = 3 mice; female (8 weeks of age), n = 3 mice. (C) Immunofluorescence showing GSTM1 in the medial prefrontal cortex (mPFC), specifically the prelimbic area, in 8-week-old C57BL/6J male mice. Cells were costained for cell type–specific markers to identify neurons (NeuN), astrocytes (S100β), oligodendrocytes (Olig2), or microglia (Iba1). (D) Quantification of the percentage of each indicated cell type that was positive for GSTM1 staining. Male (8 to 11 weeks of age), n = 3 mice; female (8 weeks of age), n = 3 mice. Scale bar, 25 μm. In (B) and (D), error bars represent means ± SEM. *P < 0.05 and **P < 0.01. n.d., not detected. Significance was determined by two-way analysis of variance (ANOVA) with Tukey’s post hoc test.

GSTM1 and GSTT2 are required in astrocytes for sufficient microglia activation in vivo

Because astrocytes play a critical role in brain inflammation, we examined the effect of GSTM1 and GSTT2 knockdown in astrocytes on inflammatory responses in the brain. Systemic administration of LPS is a well-established model to induce brain inflammation and is characterized by the activation of both microglia and astrocytes (38, 39). We used this model to examine the effect of astrocyte-specific knockdown of GSTs on the activation of microglia. We first knocked down GSTM1 by using an adeno-associated virus (AAV) vector encoding green fluorescent protein (GFP) and mir-30–based short hairpin RNA targeting Gstm1 transcripts (Gstm1 shRNAmir) downstream of a floxed stop codon [AAV–loxP-Stop-loxP (LSL)–GFP-Gstm1-shRNAmir] (fig. S3). The AAV was stereotactically injected into the mPFC of mice expressing Cre recombinase under the mouse Gfap promoter (mGfap-Cre mice) at 3 weeks of age. The expression of GFP was specific for astrocytes (S100β+ cells) in the mPFC of mGfap-Cre mice at 3 to 4 weeks after the AAV injection (Fig. 2A). We observed no GFP signals in neurons (NeuN+ cells), oligodendrocytes (Olig2+ cells), or microglia (Iba1+ cells). Using this system for knocking down GSTM1 specifically in astrocytes, we injected LPS intraperitoneally and examined the activation status of microglia in the area where GFP+ astrocytes were detected. An AAV virus carrying a nontargeting shRNAmir was used as a negative control. We found that GSTM1 knockdown in astrocytes attenuated the activation of nearby microglia, as judged by morphological changes, at 48 hours after LPS injection (Fig. 2, B to D). We observed no difference in microglia activation at 48 hours after saline injection (fig. S4). The percentage of microglia-producing TNF-α after LPS injection was also significantly decreased by GSTM1 knockdown in astrocytes, with substantially weaker TNF-α signals per microglia (Fig. 2, E and F). Microglia activation was similarly attenuated when GSTT2 was knocked down in astrocytes by injection of AAV-LSL-GFP-Gstt2-shRNAmir into mGfap-Cre mice (fig. S5). These data show that GSTM1 and GSTT2 in astrocytes were required for the sufficient activation of microglia during brain inflammation.

Fig. 2 Reduced activation of microglia in astrocyte-specific GSTM1 knockdown mice during brain inflammation induced by systemic injection of LPS.

(A) Experimental design. Mouse Gfap promoter–driven Cre transgenic (mGfap-Cre) mice (3 weeks of age) were stereotactically injected with floxed AAV vector encoding shRNAmir against Gstm1 (AAV-LSL-GFP-Gstm1 shRNAmir) into the mPFC and challenged with intraperitoneal (i.p.) injection of LPS 3 to 4 weeks later. After 48 hours, the brains were harvested and stained for the presence of virally encoded GFP together with cell type–specific markers (NeuN for neurons and S100β for astrocytes). DAPI, 4′,6-diamidino-2-phenylindole. (B) Slices from the mPFC of LPS-challenged mice injected with AAV encoding the control shRNA or Gstm1 shRNA were stained with the microglia marker Iba1, and their activation status was analyzed by morphological changes in the area of astrocyte-specific GSTM1 knockdown (GFP+) by confocal microscopy. (C) To quantify microglial activation, we morphologically classified each Iba1+ microglia as ramified, intermediate, amoeboid, or round. These morphologies correspond to surveying (ramified) or activated (intermediate, amoeboid, and round) microglia (58). (D) The microglia activation profiles were compared between the mice injected with control shRNA and those injected with Gstm1 shRNA (n = 1265 microglia from eight mice for control shRNA; n = 941 microglia from eight mice for Gstm1 shRNA). (E) Immunofluorescence showing TNF-α in microglia in the vicinity of astrocytes with GSTM1 knockdown in mice injected with AAV encoding the control shRNA or Gstm1 shRNA. (F) Quantification of the percentages of Iba1+ microglia positive for TNF-α in mice in (E). n = 560 microglia from seven mice for control shRNA; n = 616 microglia from eight mice for Gstm1 shRNA. Scale bars, 25 μm (A), 100 μm (B), 10 μm (C), and 25 μm (E). In (D) and (F), each dot represents one animal, and the bar represents mean ± SEM. Significance was determined by Mann-Whitney test. *P < 0.05 and **P < 0.01.

Astrocyte GSTM1 and GSTT2 promote microglia activation in vitro

Microglia and astrocytes can amplify each other’s activation by secreting proinflammatory mediators (Fig. 3A) (4043). To understand the mechanisms underlying the effects of GSTM1 knockdown in astrocytes on the activation of microglia, we used a coculture system of primary mouse astrocytes and immortalized microglial cell lines (BV2 microglia) (Fig. 3B). Purified primary mouse cortical astrocytes (fig. S6) infected with a lentivirus encoding shRNA targeting Gstm1 (Gstm1 shRNA) or control shRNA were mixed with BV2 microglia. Then, the mixed cultures and monocultures of astrocytes and BV2 cells were challenged with LPS for 6 hours. Under these conditions, LPS induced TNF-α production only from microglia (Fig. 3B). We then compared the effects of GSTM1 knockdown in astrocytes on microglial TNF-α production. Consistent with our in vivo findings, GSTM1 knockdown in astrocytes reduced the amount of TNF-α secretion and mRNA expression at 6 hours after LPS stimulation (Fig. 3, C and D). The induction of transcripts encoding IL-1β (Il1b) was also reduced (Fig. 3D). A previous study reported that microglia-derived TNF-α and IL-1β induced proinflammatory changes of astrocytes (43). Inhibition of TNF-α and IL-1β signaling by blocking antibodies attenuated the induction of Tnf, Csf2, and Ccl2 mRNAs in our cocultures (Fig. 3E). Previous studies showed that astrocytes produce GM-CSF (also called CSF2) and CCL2, both of which are potent activators of microglia (4043), during brain inflammation. Thus, these data support that GSTM1 in astrocytes is required for boosting microglial TNF-α production in a non–cell autonomous manner and indicate the requirement of microglia-derived signals for the induction of astrocyte inflammatory mediators. GSTM1 or GSTT2 overexpression in astrocytes, on the other hand, enhanced the induction of Tnf mRNA in coculture (fig. S7).

Fig. 3 Impaired production of microglial TNF-α by GSTM1 silencing in cocultured astrocytes.

(A) Amplification of inflammatory responses between astrocytes and microglia through soluble mediators. Previous studies suggest that microglia produce proinflammatory cytokines such as TNF-α and IL-1β, which in turn stimulate astrocytes to produce proinflammatory mediators such as GM-CSF and CCL2, which amplify inflammatory responses in microglia. (B) Experimental design. Primary mouse glial cultures were prepared from six to eight postnatal days 2 to 5 (P2-5) pups (mixed male and female) and infected with lentivirus encoding shRNA. Astrocytes were enriched and then cocultured with BV2 microglia overnight before LPS stimulation for 6 hours. Cells and culture supernatants were harvested for quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis and enzyme-linked immunosorbent assay (ELISA), respectively. The graph shows TNF-α production in response to LPS stimulation in astrocyte and microglia monocultures and in coculture as measured by ELISA. Representative data from two biologically independent cell culture experiments were shown. (C) Quantification of TNF-α production in cocultures of BV2 microglia with astrocytes expressing the control shRNA or Gstm1 shRNA. Quantification was performed by ELISA. (D) Quantification of Tnf and Il1b expression in cocultures of BV2 microglia with control or GSTM1 knockdown astrocytes. Quantification was performed by qRT-PCR of cell extracts. (E) Quantification of Tnf, Csf2, and Ccl2 expression in cocultures of BV2 microglia and WT astrocytes in the presence of TNF-α and IL-1β signaling-blocking antibodies. Quantification was performed by qRT-PCR of cell extracts. Each bar represents mean ± SD of triplicate measurements. For (C) to (E), representative data from three biologically independent cell culture experiments were shown. In (C) and (D), significance was determined by two-way ANOVA with Sidak’s post hoc test for control versus Gstm1 shRNA. In (E), significance was determined by one-way ANOVA with Sidak’s post hoc test for no antibodies versus anti–TNF-α/IL-1β. *P < 0.05 and **P < 0.01. h, hours; A.U., arbitrary units.

GSTM1 promotes the production of inflammatory mediators in astrocytes

To investigate the molecular basis by which GSTM1 regulates astrocyte function during inflammatory responses, we analyzed the effects of GSTM1 knockdown on purified primary mouse astrocytes. We examined the secretion of GM-CSF and CCL2 from astrocytes in response to the proinflammatory cytokines TNF-α and IL-1β. We found that the amount of GM-CSF and CCL2 in the culture supernatants from astrocytes was decreased by GSTM1 knockdown (Fig. 4A). We also performed qRT-PCR analysis to examine the expression of genes encoding proinflammatory and immunoregulatory mediators. The induction of Ccl2, Csf1, and Csf2 mRNAs, but not Nos2 mRNA, was impaired by GSTM1 knockdown in astrocytes (Fig. 4B). GSTM1 knockdown also affected the production of immunoregulatory cytokines. Tgfb1 expression decreased upon GSTM1 knockdown, irrespective of the presence of TNF-α and IL-1β. In contrast, the induction of Il33 mRNA by TNF-α and IL-1β was enhanced in the absence of GSTM1 (Fig. 4B). These findings showed that GSTM1 regulates the transcription of both proinflammatory and immunoregulatory genes in astrocytes.

Fig. 4 Altered induction of inflammatory mediators in cultured GSTM1 knockdown astrocytes.

(A) Quantification of GM-CSF and CCL2 in supernatants from control and GSTM1 knockdown primary mouse cortical astrocytes in response to TNF-α and IL-1β stimulation. Quantification was performed by ELISA. (B) Expression of Ccl2, Csf1, Csf2, Tgfb1, Nos2, and Il33 mRNAs in control and GSTM1 knockdown primary mouse cortical astrocytes stimulated with TNF-α or IL-1β. Transcripts were quantified by qRT-PCR. Each bar represents mean ± SD of triplicate measurements. Representative data from three biologically independent cell culture experiments are shown. Significance was determined by two-way ANOVA with Sidak’s post hoc test for control shRNA versus Gstm1 shRNA. **P < 0.01.

GSTM1 is required for NF-κB activation downstream of TNF-α and IL-1β signaling in astrocytes

To address the mechanisms underlying altered gene transcription downstream of TNF-α and IL-1β in astrocytes, we analyzed the activation of NF-κB, JNK, and extracellular signal–regulated kinase (ERK), all of which are major mediators of TNF-α and IL-1β signaling, in primary mouse astrocytes. GSTM1 knockdown in astrocytes reduced the phosphorylation of the p65 subunit of NF-κB in response to TNF-α stimulation for 6 hours (Fig. 5, A and B). In contrast, the phosphorylation of JNK, ERK, and the NF-κB inhibitor IκBα was not significantly affected by GSTM1 knockdown at the same time point. Because previous studies reported that pharmacological depletion of GSH dampens TNF-α–induced activation of the NF-κB pathway in primary mouse hepatocytes (44), we next examined the effects of GSH depletion by diethyl maleate (DEM) on TNF-α–induced phosphorylation of p65 in mouse astrocytes. Similar to GSTM1 knockdown, DEM treatment caused reduced p65 phosphorylation in response to TNF-α (Fig. 5C and fig. S8). Thus, these findings indicate that GSTM1 activates NF-κB signaling and induces the expression of proinflammatory mediators such as GM-CSF and CCL2, both of which enhance the activation of microglia (Fig. 5D).

Fig. 5 GSTM1 knockdown impairs NF-κB activation in primary astrocytes.

(A) Western blot for the p65 subunit of NF-κB (p65), JNK, ERK, and IκBα and the phosphorylated forms of these proteins (p-p65, p-JNK, p-ERK, and p-IκBα) in primary mouse cortical astrocytes expressing a control or Gstm1 shRNA. β-Actin is a loading control. (B) Quantification of p65, JNK, ERK, and IκBα phosphorylation relative to the total abundance of each protein in primary mouse cortical astrocytes expressing a control or Gstm1 shRNA. (C) Western blot and quantification of p65 phosphorylation in astrocytes treated with the GSH depletor DEM during TNF-α stimulation. (D) Schematic model of the role of GSTM1 in astrocytes and proinflammatory astrocyte-microglia interactions. Our findings support the role of GSTM1 in activating NF-κB and inducing the expression of Ccl2 and Csf2 in astrocytes. In the absence of GSTM1, microglia activation is attenuated by insufficient amounts of astrocyte-derived GM-CSF and CCL2. This also results in a decrease in TNF-α production by microglia due to reduced positive feedback mediated by GM-CSF and CCL2. For (A) and (C), representative blot data from three independent experiments were shown. For quantification in (B) and (C), bar graphs represent means ± SD of three biologically independent cell cultures. In (B), significance was determined by two-way ANOVA with Sidak’s post hoc test for control shRNA versus Gstm1 shRNA. In (C), significance was determined by two-way ANOVA with Tukey’s post hoc test. **P < 0.01.


In this study, we discovered that two GST enzymes, GSTM1 and GSTT2, were required in astrocytes for the enhancement of microglia activation during brain inflammation induced by systemic LPS administration. In addition, GSTM1 promoted the induction of proinflammatory mediators in astrocytes such as GM-CSF and CCL2. We further demonstrated that GSTM1 activated the NF-κB pathway in response to TNF-α stimulation in cultured astrocytes. Although further mechanistic studies are necessary, these data have revealed a critical role for GST enzymes in astrocytes in enhancing the activation of microglia in brain inflammation.

Astrocytes and microglia are widely involved in brain inflammatory responses related to infection, autoimmunity, neurodegeneration, injury, and other pathological conditions (13, 5, 6). Whereas astrocyte-neuron and microglia-neuron interactions have been extensively studied, less is known about the interactions between astrocytes and microglia in brain inflammation. Because both of these glial cells sense immune stimuli and produce inflammatory mediators, it is important to understand their mutual regulation. Our findings suggest that GSTM1-mediated astrocyte activation is required for boosting microglia activation. At the same time, our data were consistent with previous findings that microglia are the primary cells to sense inflammatory stimuli such as LPS (Fig. 3). Thus, we hypothesize that microglia undergo a two-step activation: first, by stimulation through cell-autonomous mechanisms of sensing inflammatory stimuli, such as LPS, and second, by astrocyte-derived “amplifier” signals such as GM-CSF and CCL2. Accordingly, the interaction of microglia and astrocytes may form a positive feedback loop to facilitate inflammatory responses in the brain. Recent studies have highlighted the role of activated microglia in the functional polarization of astrocytes, such as neurotoxic A1 and neuroprotective A2 astrocytes, in LPS-induced brain inflammation (39). Our study implies that such differentially activated astrocytes may cause differential microglial activation. Further studies to determine the significance of second-step activation of microglia would address this important question.

The mechanisms by which GSTM1 modulates NF-κB activation and gene expression patterns in astrocytes should be further investigated. Although DEM treatment resulted in attenuated NF-κB activation in astrocytes similar to GSTM1 knockdown, a more detailed study would be required to determine whether GSTM1-mediated S-glutathionylation enhances NF-κB activation. DEM treatment may cause NF-κB inactivation by mechanisms other than GSH depletion. Future experiments with a mutant GSTM1 lacking glutathione transferase activity may provide further insight into this important question. Thus, further mechanistic studies would determine the role of S-glutathionylation in astrocyte activation and address whether such GSTM1-mediated mechanisms are present in other brain inflammation models, such as infections, autoimmune diseases, and neurodegeneration.

Although GSTP was previously studied in an animal model of MPTP-induced degeneration of dopaminergic neurons in the substantia nigra (29), it was not clear whether GST enzymes are involved in additional physiological inflammatory conditions beyond xenobiotic-induced neurodegeneration. Our study demonstrated that GSTM1 and GSTT2 were required for LPS-induced brain inflammation unrelated to xenobiotics such as MPTP, adding to the increasing body of work that supports the role of GSTs as endogenous regulators of physiological processes distinct from phase II of drug detoxification (2527). Our findings are also consistent with recent reports on proinflammatory roles of glutathione and protein S-glutathionylation (4547). It is speculated that many proteins are S-glutathionylated by GSTs in various mouse and human cells (2527). The identification of those target proteins will further facilitate our understanding of immunoregulatory mechanisms by diverse GSTs in different tissues and cell types.

An increasing body of evidence suggests that sex difference considerably influences inflammation and microglia function (48, 49). In our study, we did not specifically analyze the impact of sex difference on the role of GSTM1 in astrocyte-microglia interaction, although some experiments, particularly the in vitro cell culture experiments, were conducted with mixed groups of cells from both male and female mice. In the brain expression data, we did not observe any significant difference between male and female for either GSTM1 or GSTT2 (Fig. 1 and fig. S2). Although additional experiments with a larger sample size are required, there was no clear evidence supporting a sex-related role of GSTM1 in astrocyte-microglia interaction during brain inflammation.

Although our in vitro and in vivo findings are consistent with each other, the time course of these experiments is not directly comparable. For the in vitro experiments, cells were challenged with LPS for 6 hours. We chose 48 hours after LPS injection as a time point for in vivo experiments because microglial morphological changes were not clear at earlier time points after LPS injection. Future studies would address whether microglia-astrocyte interaction occurs at earlier time points in the brain after LPS-induced brain inflammation.

On the basis of our findings, we propose that GST variations in human populations may dictate individual variations in inflammatory responses or susceptibility to immune pathology associated with various acute and chronic diseases. Despite the publication of numerous genome-wide association studies and exome sequencing studies, there has been no clear strong genetic evidence supporting a role for GST variations in specific diseases. This indicates that GST variations are not associated with specific disease(s) but rather modify individual responses to homeostatic imbalance, such as inflammation, in a wide range of disorders. Because GST enzymes belonging to different classes are distinct in structure (1820), small-molecule inhibitors or activators of specific GST enzymes may be useful as add-on therapeutic agents to modify the outcome of inflammation caused by various diseases.



mGfapCre mice (lines 73.12 and 77.6) and C57BL/6J mice were purchased from the Jackson Laboratory. C57BL/6 timed pregnant female mice for in vitro cell cultures were purchased from Charles River Laboratories or prepared in-house. Mice were housed in specific pathogen–free facilities at the Johns Hopkins University. All the procedures were approved by the Institutional Animal Care and Use Committee of the Johns Hopkins University.

Virus preparation

Lentiviruses (pLKO.1 and pGIPZ viruses) and AAVs were prepared following our established protocols (5052), and their titers were estimated by quantitative PCR (qPCR)–based methods (50, 51, 53). For lentiviruses, pLKO.1-Gstm1 shRNA lentiviral vectors (#1, TRCN0000103241; #2, TRCN0000103243; #3, TRCN0000103244; and #4, TRCN0000103240) were obtained from the RNAi Consortium (TRC) library via the HiT Center at the Johns Hopkins University School of Medicine; control pLKO.1-GFP shRNA lentiviral vector (no. 30323) was obtained from Addgene, and pGIPZ-Gstt2 shRNAmir lentiviral vectors (#1, V2LMM_67055; #2, V2LMM_218573; #3, V3LMM_449685; and #4, V3LMM_449688) and control nonsilencing (NS) shRNAmir lentiviral vector (RHS4346) were obtained from Open Biosystems. pHAGE-Gstm1 and pHAGE-Gstt2 were generated by subcloning mouse Gstm1 and Gstt2 complementary DNAs (cDNAs) into pHAGE vectors. For AAVs, AAV-LSL-GFP-Gstm1 shRNAmir was generated on the basis of the most efficient shRNA construct (#4) (fig. S2), following the established protocol (54), and AAV-LSL-GFP-Gstt2 shRNAmir and AAV-LSL-GFP-NS shRNAmir were generated by shuttling Gstt2-shRNAmir (#2) and NS-shRNAmir from pGIPZ lentiviral vectors into AAV-LSL-GFP vectors.

Cell culture

Primary mouse glial cell cultures were prepared from the cortices of P2-5 pups of C57BL/6 mice as described previously (5557). After careful removal of meninges, single-cell suspensions were obtained by serial trituration of cerebral cortices with 20-gauge (G) and 26G needles with a 10-ml syringe. Cells were suspended in Dulbecco’s modified Eagle’s medium (DMEM)/F12 supplemented with 15% fetal bovine serum (FBS) and penicillin/streptomycin (all from Thermo Fisher Scientific) and seeded onto T-75 flasks (Corning) precoated with poly-d-lysine (PDL; 25 μg/ml) at about two brains per flask. Medium was changed on day 3 and every 2 to 3 days thereafter. If necessary, then lentiviral infection was performed with glial cells on day 10 as described below. To enrich astrocytes, oligodendrocyte lineage cells and microglia on the surface of mixed glial cell culture were vigorously shaken off, and astrocytes were then collected as negative fractions after magnetic-activated cell sorting (MACS) with CD11b MicroBeads (Miltenyi Biotec). Collected astrocytes were >98% GFAP+ CD11b cells (fig. S6). BV2 microglia were maintained in DMEM/F12 supplemented with 15% FBS and penicillin/streptomycin (all from Thermo Fisher Scientific). Astrocyte-microglia cocultures were prepared by seeding 5 × 105 astrocytes onto PDL-coated six-well plates and adding 5 × 104 BV2 cells 2 days later. Purified astrocyte cultures were prepared by seeding 5 × 105 astrocytes immediately after MACS, onto PDL-coated (10 μg/ml) six-well plates.

Lentiviral infection

Mixed glia culture was passaged on days 9 to 12 and reseeded at 4 × 106 to 7 × 106 astrocytes onto PDL-coated T-75 flasks. On the next day, lentiviruses were added to the culture at 1:1 MOI (multiplicity of infection). For pLKO.1 lentiviruses, the virus-infected cells were enriched by antibiotic-based selection (puromycin; 2.5 μg/ml) for 72 hours beginning at 72 hours after infection. For pHAGE lentiviruses, infection efficiency was monitored by GFP signals under a fluorescence microscope. Astrocytes were purified as negative fraction with MACS with CD11b MicroBeads and reseeded onto PDL-coated six-well plates for further experiment.

In vitro cell stimulation

Astrocyte-microglia cocultures and purified astrocytes were stimulated with LPS (1 μg/ml) (O55B5, Sigma), TNF-α (50 ng/ml), or IL-1β (10 ng/ml) for 6 hours, respectively. For GSH depletion experiments using DEM (Santa Cruz Biotechnology), purified astrocytes were treated with serially diluted DEM for 6 hours. Cells and culture sups were immediately harvested at the end of the 6-hour incubation for the downstream assays.

Stereotactic surgery and LPS treatment

Mice at P21-28 were anesthetized and placed in the mouse stereotaxic frame [World Precision Instrument (WPI)] to secure the cranium. Then, the mice were injected with 250 to 500 nl of AAV (2.0 × 1010 genomic copies (GC)/μl) at the rate of 100 to 200 nl/min into the mPFC bilaterally using a NanoFil syringe (WPI) with a 35G beveled needle. The following stereotactic coordinate was used for injection: anteroposterior (AP), +1.5 mm; mediolateral (ML), ±0.2 mm; and dorsoventral (DV), −1.8 mm from the bregma. Three to four weeks later, LPS (5 mg/kg; Sigma, O55:B5) was injected intraperitoneally, and the brains were harvested 48 hours later.


Mice were anesthetized and transcardially perfused with ice-cold phosphate-buffered saline (PBS) followed by 4% paraformaldehyde. Free-floating sections (40 μm in thickness) were prepared by a Leica cryostat, and, if necessary, antigen retrieval was performed with 10 mM sodium citrate buffer (pH 8.5). The sections were then placed in blocking solution (PBS supplemented with 2% normal goat serum, 1% bovine serum albumin, 0.1% Triton X-100, 0.05% Tween 20, and 0.05% sodium azide) for 1 hour at room temperature and then incubated at 4°C overnight with the following primary antibodies: rabbit anti-Iba1 (1:400; 019-19741, Wako), goat anti-Iba1 (1:400; NB100-1028, Novus Biologicals), chicken anti-GFP (1:5000; ab13970, Abcam), mouse anti-NeuN (1:200 to 400; MAB377, EMD Millipore), rabbit anti-S100β (1:400; ab4066, Abcam), mouse anti-Olig2 (1:400 to 1000; MABN50, EMD Millipore), rabbit anti-GSTM1 (1:250; 12412-1-AP, Proteintech), rabbit anti-GSTT2 (1:250; 17622-1-AP, Proteintech), and mouse anti–TNF-α (1:200; 52B83, Abcam). After washing with PBS, the sections were further incubated with fluorophore-conjugated secondary antibodies at 1:400 dilution for 2 hours at room temperature, followed by DAPI staining (1:50,000) for 10 min at room temperature. The sections were mounted on glass slides with PermaFluor mounting medium or ProLong Diamond antifade mounting medium (Thermo Fisher Scientific). Images were acquired with Zeiss LSM 510 and LSM 700 confocal microscopes and a ZEN software (Zeiss).

Image analysis

For microglial analysis, z-stack images were analyzed with ImageJ [National Institutes of Health (NIH)]. Three images were taken from the coronal sections of mPFC of each mouse stained with anti-Iba1 and anti–TNF-α (n = 8 mice per group). Microglia activation status was categorized into ramified, intermediate, amoeboid, or round on the basis of cell morphology from Iba1 staining. Then, the activation status was categorized as either activated (intermediate, amoeboid, and round) or surveying (ramified) as previously described (58). The quantity and percentage of microglia in each category as well as “activated” microglia were reported per animal. Percentages of TNF-α–positive microglia per total microglia were also calculated with 20× z-stack images (n = 7 mice for control shRNA and n = 8 mice for Gstm1 shRNA).

Enzyme-linked immunosorbent assay

ELISA was performed using Ready-SET-Go! kits (eBioscience) for TNF-α, CCL2, and GM-CSF following a standard protocol (59).

qRT-PCR analysis

qRT-PCR analysis was performed as previously described (16, 59). Total RNA was isolated using RNeasy Plus Micro Kit (Qiagen), and cDNA was synthesized using a SuperScript III kit [Invitrogen with oligo(dT)20 primers]. qPCR was carried out with Maxima SYBR Green/ROX qPCR Master Mix (Thermo Fisher Scientific) on ABI 7900HT system (Applied Biosystems) and QuantStudio 5 (Thermo Fisher Scientific). Gene-specific primer sets were obtained from PrimerBank ( (60) or designed by using Primer-BLAST ( The primer sequences are available in table S1.

Western blotting

Cell lysates were prepared with radioimmunoprecipitation assay buffer and separated on NuPAGE Bis-Tris Mini Gels (Life Technologies), followed by the transfer to polyvinylidene difluoride membrane (Millipore) following a standard protocol. After blocking in 5% skim milk/phosphate-buffered saline supplemented with 0.1% Tween 20 (PBS-T), membranes were incubated with the primary antibody overnight at 4°C and then incubated with the secondary antibody for 1 hour at room temperature. Gel images were captured using ImageQuant LAS 4000 (GE Healthcare), and the intensities of bands were quantified using Quantity One imaging analysis software (Bio-Rad). For Fig. 1B, GSTM1 abundance in each brain area relative to that in the lung was calculated. For Fig. 5 (B and C), the amounts of phosphorylated proteins relative to those of total proteins were calculated. The following primary antibodies were used: p65 (1:1000; no. 8242, Cell Signaling Technology), phospho-p65 (1:1000; no. 3033, Cell Signaling Technology), JNK (1:1000; no. 9252, Cell Signaling Technology), phospho-JNK (1:1000; no. 4668, Cell Signaling Technology), Erk1/2 (1:1000; no. 4695, Cell Signaling Technology), phospho-Erk1/2 (1:1000; no. 4370, Cell Signaling Technology), GSTM1 (1:500; no. 12412-1-AP, Proteintech), GSTT2 (1:500; no. ab53942, Abcam), hemagglutinin (HA) tag (1:500; no. 3724, Cell Signaling Technology), and β-actin (1:2000; no. sc-47778, Santa Cruz Biotechnology).

Statistical analysis

Mann-Whitney test, Student t test, one-way ANOVA, and two-way ANOVA were used, and P < 0.05 was considered statistically significant. For multiple testing corrections, Tukey’s post hoc test, Dunnet’s post hoc test, or Sidak’s post hoc test was utilized as indicated in the figure legends and table S2. All the statistical analyses were performed with GraphPad Prism 7 (GraphPad Software Inc.). Detailed values for statistical analysis are provided in table S2.


Lentivirus vector encoding mouse Gstm1 shRNA (pLKO.1-Gstm1 shRNA-PuroR) was obtained from TRC libraries. Lentivirus vector encoding mouse Gstt2 shRNAmir (pGIPZ-turboGFP-Gstt2 shRNAmir) was obtained from Open Biosciences (currently Thermo Fisher Scientific). Mouse Gstm1 and Gstt2 cDNA was PCR-cloned from brain lysates and was inserted into the pRK5-HA vector. Lentiviral construct overexpressing HA-tagged cDNA was generated by inserting a blunt-ended Eco RI–Bam HI fragment from pRK5-HA-Gstm1 or Gstt2 vector into pHAGE–CMV–MCS-IZsGreen vector between a blunted site of Not I and a Bam HI site. AAV constructs encoding NS control shRNAmir and Gstt2 shRNAmir under an LSL cassette were generated by inserting an Xho I–blunt-ended Mlu I fragment from pGIPZ into pAAV-CMV-LSL-GFP-shRNAmir between an Xho I site and a blunt-ended site of Eco RI. AAV construct encoding Gstm1 shRNAmir was generated by modifying Gstm1 shRNA sequences into Gstm1 shRNAmir sequences with a published protocol (52, 54) and inserting the Xho I–Gstm1 shRNAmir–Eco RI fragment into pAAV-CMV-LSL-GFP shRNAmir between Xho I and Eco RI sites. The sequences of all the constructs were confirmed by conventional Sanger sequencing.

Lentivirus preparation

Human embryonic kidney 293FT cells (Invitrogen) were transfected with lentiviral vectors (pLKO-puro-shRNA, pGIPZ, and pHAGE) and packaging vectors (pMD.2G and psPAX2) by using Lipofectamine 3000 (Invitrogen) with a standard protocol. Briefly, culture supernatants were collected at 48 hours after transfection and ultracentrifuged at 25,000 rpm at 4°C for 2 hours to precipitate the virus. Viral pellets were dissolved in PBS and kept frozen at −80°C until use. Virus titers were estimated by using a qPCR-based method to detect long terminal repeat (LTR) sequences as described previously (53). The following primers and probe were used: LTR F, 5′-TGTGTGCCCGTCTGTTGTGT-3′; LTR R, 5′-GAGTCCTGCGTCGAGAGAGC-3′; and LTR probe, 5′-CAGTGGCGCCCGAACAGGGA-3′.

AAV preparation

AAV-293 cells (Agilent Technologies) were transfected with AAV vectors (pAAV-LSL-GFP-shRNAmirs) and helper vectors (pAd helper vector and pAAV2/5 packaging vector) following a standard CaCl2 method with some modifications (61). Transfected cells were harvested 48 hours later and lysed by three rounds of freeze-thaw cycles. The lysates were further sonicated, and virus-containing fractions were precipitated by adding ammonium sulfate. Then, the viruses were enriched by OptiPrep density gradient ultracentrifugation at 60,000 rpm for 1.5 hours, followed by concentration of viruses using Amicon Ultra Centrifugal filter unit (100,000 Da molecular weight cut-off) (Millipore). Virus titers were estimated using a qPCR-based method to detect inverted terminal repeat (ITR) sequence as described previously (50). The following primers and probe were used: AAV2 ITR F, 5′-GGAACCCCTAGTGATGGAGTT-3′; AAV2 ITR R, 5′-CGGCCTCAGTGAGCGA-3′; and AAV2 ITR probe, 5′-CACTCCCTCTCTGCGCGCTCG-3′.

Quantification of GSTM1 and GSTT2 knockdown efficiency in vivo

The abundance of GSTM1/GSTT2 in AAV-infected GFP+ cells was quantified as follows: First, the outline of a GFP-expressing cell was traced and GSTM1/GSTT2 signal intensities per pixel were averaged across the traced GFP+ area. Then, the averaged GSTM1/GSTT2 signal intensities were compared between cells expressing control and Gstm1/Gstt2 shRNA.

Measurement of intracellular total GSH

GSH was measured using Glutathione Assay Kit (703002, Cayman Chemicals) according to the manufacturer’s instructions. Briefly, astrocytes were harvested in MES buffer after 6-hour incubation with DEM and sonicated to disrupt cellular membrane. Supernatants were incubated with 10% meta-phosphoric acid (MPA) (Sigma) at room temperature for deproteination. Total GSH was detected by measuring the product of glutathionylated 5,5′-dithiobis(2-nitrobenzoic acid) (DTNB) at 415 nm with a kinetic method on an ultraviolet spectrophotometer. GSH amount was normalized by the protein amount per each sample.


Fig. S1. Enriched expression of GSTT2 in astrocytes in the mouse cerebral cortex.

Fig. S2. Enriched expression of GSTM1 in astrocytes in the female mouse brain.

Fig. S3. GSTM1 and GSTT2 knockdown efficiency.

Fig. S4. No difference in microglia activation in the vicinity of GSTM1 knockdown astrocytes in mice injected with saline.

Fig. S5. Reduced activation of microglia during neuroinflammation in mice with astrocyte-specific GSTT2 knockdown.

Fig. S6. Representative flow cytometry data of astrocytes used for in vitro cell culture experiments.

Fig. S7. Enhanced induction of Tnf expression in BV2 cells cocultured with astrocytes overexpressing GSTM1 or GSTT2.

Fig. S8. Total amounts of glutathione in astrocytes treated with different doses of DEM.

Table. S1. Primer sequences for qRT-PCR analysis.

Table. S2. Statistical analyses used in this manuscript.


Acknowledgments: We thank M. Niwa, S.-H. Kim, Y. Chen, L. Cortina, S. Chow, A. Murata, and J. See for technical help; A. Kamiya for BV2 microglia cells; Z. J. Huang (Cold Spring Harbor Laboratory) for AAV-LSL-GFP-shRNAmir constructs; and F. Wan for advice. Funding: This work was made possible by support from the NIH (MH093458 and MH113645 to S.-i.K.; MH094268, MH105660, MH092443, and DA040127 to A.S.), RUSK (to A.S.), BBRF (to A.S.), Stanley (to A.S.), Johns Hopkins Medicine Discovery Fund (to S.-i.K.), and the Department of Psychiatry Venture Discovery Fund (to S.-i.K.). Author contributions: S.-i.K. conceived and supervised the project with the guidance of A.S. S.-i.K., E.Y.C., and A.M.W. designed and constructed the virus vectors. E.Y.C., E.D., D.J.C., and A.M.W. prepared and titrated lentiviruses and AAVs. S.-i.K. and E.D. performed stereotactic virus injection experiments. E.Y.C., E.D., S.A., I.V.L.R., S.G., and D.J.C. conducted histology experiments. E.Y.C., E.D., S.A., B.D.L., T.I., and A.M.W. performed cell culture experiments, qRT-PCR analysis, Western blot experiments, and other related experiments. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials. AAV-Gstm1 and Gstt2 shRNAmir plasmids will be made available upon request and will require a material transfer agreement (MTA).

Stay Connected to Science Signaling

Navigate This Article