Lipopolysaccharide-induced interleukin-6 production is controlled by glycogen synthase kinase-3 and STAT3 in the brain

Protein-free E. coli (K235) LPS was prepared as described [12]. Sources of substances used and solvents were as follows: IFNγ (R&D Systems), SB216763 (in DMSO; Tocris), kenpaullone, indirubin-3'-monoxime, BIO (6-bromoindirubin-3'-oxime), GSK3 inhibitor II, AG490 (each in DMSO; Calbiochem), LiCl (in water; Sigma), and JSI-124 (cucurbitacin, in DMSO; National Cancer Institute Developmental Therapeutic Program). Mouse IL-1β, TNFα, IL-6, IL-10 and IFNγ were measured with ELISA kits according to the manufacturer's instructions (eBioscience). The concentrations of 62 inflammatory proteins (listed in Additional files 1 and 2) were measured with inflammatory molecule antibody arrays according to the manufacturer's instructions (Raybiotech).

For murine primary neurons, the hippocampus from 1 day old C57Bl/6 mice was isolated and incubated in 0.1% trypsin (Invitrogen) and the cells mechanically dissociated using a fire-polished pipette. Cells were plated in DMEM/F12 medium supplemented with 10% FBS, 0.3% glucose, 2 mM glutamine, 10 U/mL penicillin and 10 μg/mL streptomycin in poly-D-lysine-coated six-well plates. Twelve hours after plating, the medium was replaced with Neurobasal medium supplemented with B27 and 0.5 mM glutamine (Invitrogen) to promote neuronal survival and to inhibit the growth of non-neuronal cells. Neurons were used for experiments after seven days in culture. Primary glia were prepared from the cerebral cortex of 1 day old C57Bl/6 mice as described [21], cultured in DMEM/F12 medium supplemented with 10% FBS, 0.3% glucose, 2 mM L-glutamine, 10 U/mL penicillin and 10 μg/mL streptomycin. For separation of astrocytes and microglia, after 10 days of culture the cells were shaken (30 hr; 250 rpm), resulting in >99% pure astrocytes as determined by immunostaining with the astrocyte marker glial fibrillary acidic protein (GFAP). After the first hour of shaking, the medium containing microglia cells was collected and microglia were cultured in the same medium as astrocytes. For expression experiments, cells were rinsed twice with DMEM/F12 medium without supplements, infected with 50 moi of the designated adenovirus for 30 min, supplemented medium was added for incubation for 36–48 h, and infected cultures were examined for adequate infection efficiency (80%) as assessed by GFP fluorescence after infection with GFP adenovirus. For knockdown experiments, cells were transfected using liposome-mediated transfection reagent LipofectAMINE RNAiMAX (Invitrogen) with 50 nM siRNA according to the manufacturer's instructions with STAT3 prevalidated siRNA [22], silencer negative control (Ambion), or GSK3α and GSK3β siRNA (Smart pool; Dharmacon). For experimental treatments, cells were pretreated with the indicated inhibitors for 30 min followed by treatment with 100 ng/mL LPS, 1 ng/mL IFNγ, or both, for the indicated times. Following treatments, cells were washed twice with phosphate-buffered saline (PBS) and were lysed with lysis buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, 10% glycerol, 1 μg/mL of leupeptin, aprotinin, and pepstatinA, 1 mM orthovanadate, 50 mM NaF, 0.1 μM okadaic acid, 1 mM phenylmethanesulfonyl fluoride). The lysates were centrifuged at 14 000 rpm for 10 min. Protein concentrations were determined in duplicate using the Bradford protein assay. Immunoblotting was carried out as described before [20] using antibodies to phospho-Tyr705-STAT3, total STAT3 (Cell Signaling Technology), GFAP, GSK3α/β (Millipore), and β-actin (Sigma). Immunoblots were developed using horseradish peroxidase-conjugated goat anti-mouse, or goat anti-rabbit IgG, followed by detection with enhanced chemiluminescence, and the protein bands were quantitated with a densitometer. Cell viability was assessed by a colorimetric 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay. A multi-well scanner was used to measure the absorbance at 570–630 nm dual wavelengths. Statistical significance between groups was evaluated by Student's t-test.

IL-6 in the cerebral cortex, cerebellum, and hippocampus was markedly increased 4 hr after peripheral LPS administration, which reflected the increased serum IL-6 level, whereas LPS administration slightly increased cortical and serum levels of TNFα, but not of IL-1β, at this time (Figure 1A–C). IL-6 present in brain regions was likely partly produced in the brain because after icv administration of LPS, IL-6 accumulated both in the ventricular infusion zone and in the distant occipital cortex, indicating centrally-induced spreading inflammation (Figure 2A). Central LPS infusion also moderately increased brain levels of TNFα and IFNγ in both brain regions but with a persistence in the ventricular infusion zone, while IL-1β and the anti-inflammatory cytokine IL-10 levels slightly increased at late time points in the ventricular infusion zone only (Figure 2B–E). Administration of saline vehicle had more modest effects than LPS on the levels of cytokines in the brain. Thus, in response to sepsis, IL-6 robustly accumulates in the brain and is at least partly produced by the central inflammatory system, although some contribution from the periphery cannot be ruled out.

IL-6 and other cytokines, such as interferons, induce the activation of STAT3 by receptor-induced Janus kinase (JAK)-mediated tyrosine-705 phosphorylation [23]. Phospho-Tyr705-STAT3 in the brain was increased 4 hr after either ip (Figure 2F) or icv LPS administration (Figure 2G), confirming that STAT3 is activated in the brain after sepsis [24, 25]. Examined 6 hr after treatment with LPS (100 ng/mL), STAT3 was activated in primary astrocytes and microglia (Figure 3A). Treatment with LPS alone did not stimulate STAT3 in primary neurons, but neuronal STAT3 was activated after treatment with conditioned medium from treated astrocytes, indicating that a factor released from glia, such as IL-6, can stimulate neuronal STAT3. Thus, STAT3 can be activated in all three types of cells during in vivo inflammation.

The production of IL-6 was also evaluated in primary glia after stimulation with LPS or with IFNγ co-administration with LPS, which amplifies LPS-induced cytokine production. There was a time-dependent increase in IL-6 production in response to LPS (100 ng/mL) that was amplified by co-stimulation with 1 ng/mL IFNγ (Figure 3B). Inflammatory molecule array analysis showed a large increase in the production of IL-6 by primary glia in response to stimulation with LPS plus IFNγ (Figure 3C). Several other products implicated in the pathogenesis of sepsis were also induced by this treatment, including IL-12p40, CCL5/RANTES, CXCL1/KC, CCL9/MIP-1γ, CXCL2/MIP-2, P-Selectin, TIMP-1 and CCL2/MCP-1 (Figure 3C).

To examine the mechanisms regulating the production of IL-6, we tested if STAT3 not only was activated by IL-6 but also participated in a feed-forward loop promoting IL-6 production in primary glia as was recently identified in B lymphocytes [26]. IL-6 production in glia induced by treatment with LPS plus IFNγ was greatly reduced by downregulation of STAT3 expression by siRNA (Figure 4A), by pharmacological inhibition with the STAT3-specific inhibitor JSI-124 (cucurbitacin) [27] (Figure 4B), or with the JAK2 inhibitor AG490 (Figure 4C), neither of which impaired cell viability (Figure 4D). These findings indicate that STAT3 may not only propagate the IL-6 signal after sepsis but also contributes to inflammation-induced IL-6 production in glia.

Since STAT3 activation requires active GSK3 [20], systemic IL-6 production is controlled by GSK3 [12], and GSK3 is abundantly expressed in the brain, we tested if GSK3 controls IL-6 production by astrocytes and microglia, first with the selective inhibitor lithium [28] using a cytokine array. Among the 12 molecules detected to be induced by treatment with LPS plus IFNγ in glia, the production of TIMP-1 and VCAM-1 were independent of GSK3, whereas active GSK3 decreased the expression of CXCL2/MIP-2 and increased the expression of CXCL1/KC, IL-12p40, CCL9/MIP-1γ, CCL2/MCP-1, P-Selectin and CCL5/RANTES (Figure 3C and 5A). However, most prominently, active GSK3 promoted IL-6 production by 16-fold, showing that IL-6 production by glia is strongly dependent on GSK3. Confirmation by ELISA showed that inhibition of GSK3 with lithium treatment reduced the time-dependent production of IL-6 induced by LPS and its potentiation by IFNγ in primary astrocytes (Figure 5B). The dependency of IL-6 production in primary astrocytes on GSK3 was confirmed by the strong inhibition exerted by four other structurally diverse GSK3 inhibitors on IL-6 produced in response to LPS or to a combination of LPS plus IFNγ (Figure 5C), which was not due to changes in cell viability (Figure 5D). Inhibition of GSK3 also strongly reduced IL-6 production in primary microglia (Figure 5E). siRNA-mediated knockdown by 75% of GSK3β, but not of GSK3α, greatly reduced the production of IL-6 by primary enriched astrocytes induced by LPS and its potentiation by IFNγ (Figure 5F), indicating that GSK3β predominantly regulates IL-6 production.

Since both active STAT3 and active GSK3 were necessary for IL-6 production, the expression of active forms of each was introduced using adenoviruses in primary enriched astrocytes to examine their effects on IL-6 production. These included STAT3C, in which covalent bonds between cysteines establishes the active dimer conformation [29], and constitutively active S9A-GSK3β and S21A-GSK3α, where serines subject to inhibitory phosphorylation were mutated to alanines to prohibit inhibitory serine phosphorylation. Expression of active S21A-GSK3α and S9A-GSK3β together increased IL-6 production and this was increased a further 4.5-fold with co-expression of STAT3C (Figure 5G). This demonstrates that GSK3 and STAT3 were both necessary, and together sufficient, to stimulate IL-6 production in glia cells.

The capacity of GSK3 inhibitors to dampen IL-6 production in glia raised the possibility that GSK3 inhibitors may be beneficial for reducing neuroinflammation. To test this hypothesis, mice were treated with lithium because it is the only GSK3 inhibitor that is well-established to be effective in the CNS after peripheral administration [26]. In mice treated with lithium, the serum IL-6 level 18 hr after LPS administration was 67% lower than in mice not given lithium (Figure 6A). This matches a previous report that the GSK3 inhibitor SB216763 greatly reduced serum IL-6 after LPS challenge [12]. IL-6 levels were elevated in the cerebral cortex and cerebellum (Figure 6A), but not hippocampus (not shown) 18 hr after LPS administration, and lithium pretreatment reduced these increases in IL-6. In mice pretreated with lithium, there was also a significant reduction in the activating tyrosine-phosphorylation of STAT3 following LPS administration in the cerebral cortex, hippocampus, and cerebellum (Figure 6B). Furthermore, inhibition of GSK3 with lithium also reduced the activation of astrocytes, as measured by the astrocyte marker GFAP immunoreactivity, following LPS treatment (Figure 6C). Since STAT3 promotes GFAP expression [30], this indicates that lithium's inhibition of IL-6 production and STAT3 activation reduces expression of STAT3-dependent astrocyte-specific genes such as GFAP, resulting in reduction of astrogliosis induced by LPS.