Inhibition of tumour necrosis factor-α (TNFα)-induced NF-κB p52 converts the metabolic effects of microglial-derived TNFα on mouse cerebellar neurones to neurotoxicity

Cell culture: microglia

Strain 129 mice were prepared as previously described (McCarthy and de Vellis 1980). Cultures were incubated in Dulbecco's modified Eagle's medium (DMEM, Gibco, Paisley, UK) + 10% fetal calf serum (Gibco) + 1% penicillin and streptomycin (Gibco) (DFP) at 37°C in 7% CO₂. The medium was changed after 24 h. Microglia were prepared from 6 day mixed glial cultures after 2-h shaking at 200 r.p.m. After centrifugation, the cell pellet was resuspended in DFP, counted and plated onto sterile glass coverslips or plastic for 15 min. The nonadherent cells were then removed by washing twice in Hank’s balanced salt solution (HBSS). The remaining microglia were > 95% CD11b positive (Ox42, Serotec, Oxford, UK). These microglia were cultured in N2B3 medium for 20 h alone or with 100 units/mL interferon-γ (Serotec). The media collected from the non-activated microglia was used as conditioned medium from resting microglia and the media from interferon-γ treated microglia was used as conditioned medium from activated microglia. N2B3 consisted of DMEM ± 0.5% fetal calf serum, 1 mm glutamine, 50 U/mL penicillin and streptomycin and 1 mL Sato mix per 100 mL basal medium leading to a final concentration of 100 µg/mL bovine serum albumin, 60 ng/mL progesterone water soluble, 16 µg/mL putrescine, 40 ng/mL sodium selenite, 40 ng/mL thyroxine, 40 ng/mL triiodothyronine, 5 µg/mL insulin and 10 µg/mL transferrin (all Sigma). Conditioned media were collected from microglial cultures, filtered (0.2 µm) and incubated with neurones at a 1 : 2 dilution unless otherwise stated, or tested using the L929 bioassay for TNFα as described previously (Zajicek et al. 1992). The TNFα (Santa Cruz, Santa Cruz, CA, USA) antibody was a goat polyclonal purified IgG raised against mouse TNFα with no cross-reactivity to TNFβ. TNFα was from Serotec.

Cell culture: neurones

Preparation of cerebellar cells from 6-day-old mice (P6) was as previously described (Brown et al. 1996). Briefly, the cerebella were dissociated in Hank's medium (Gibco) containing 0.5% trypsin (Sigma) and plated at 1–2 × 10⁶ cells/cm² in 24 well trays (Falcon, Becton-Dickinson, Franklin Lanes, NJ, USA) coated with poly-d-lysine (50 µg/mL, Sigma). Cultures were maintained in DMEM (Gibco) supplemented with 10% fetal calf serum, 2 mm glutamine and 1% antibiotics (penicillin, streptomycin, fungizone) (Gibco). Cultures were maintained at 37°C with 6% CO₂. After 24-h culture, the cerebellar cells were transferred to serum-free medium consisting of DMEM with 1% penicillin/streptomycin and addition of the TCM® supplement (ICN, Irvine, CA, USA). Cells were maintained in this medium for 1–2 days before being used for experiments.

MTT, LDH and Hoechst assays

3-[4,5-Dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide (MTT; Sigma, Poole, UK) was diluted to 200 µm in Hanks' solution (Gibco) and added to cultures for 2 h at 37°C. After incubation the substrate was removed. The cells were then washed briefly with Hanks' solution. The MTT formazan product was released from cells by addition of 700 µL of dimethylsulfoxide (Sigma) and measured at 570 nm in a Unicam UV2 spectrophotometer (ATI Unicam, Cambridge, UK). Survival of treated cerebellar cells in comparison to untreated controls could then be determined. The relative values were expressed as a percentage of control values with untreated control values in the MTT assay being equate to 100%. The assay for total lactate dehydrogenase (LDH) activity was the Sigma commercial assay. This was performed according the manufacturer's instructions. Briefly, cerebellar neurones were extracted in a buffer containing phosphate-buffered saline (PBS) and 0.1% NP-40 (Sigma). The extracted cells (in 100 µL) were centrifuged and the supernatant added to 2 mL of the LDH assay reagent. After 20 min incubation at room temperature the samples were measured spectrophotometrically at 340 nm. Numbers of apoptotic cells were determined using the Hoechst staining method. The Hoechst reagent (Sigma) was applied to live cells at 50 µm for 10 min. Stained cells were examined using a Leitz DMRB fluorescence microscope. Fields at 20 times magnification were examined and the number of cells with fragmented or condensed nuclei counted.

ATPase and cytochrome oxidase assays

The assay for total ATPase activity was according to the method of Lowry (1957). Briefly, cells were extracted in 0.03 m Tris pH.8.4 and then diluted in buffer (0.03 m Tris, 0.03 m HCl, 2 mm MgCl₂, 0.03 m 2-amino-2-methyl-1,3-propandiol, pH 8.4) until the protein concentration was 20 µg/mL. Samples are incubated on ice for 10 min after which time 3% trichloroacetic acid was added. The reactants were then centrifuged and the supernatant collected. After further incubation on ice for 30 min 100 µL of sample was added to 1 mL of 2.5% ammonium molybdate in 0.1 m acetic acid buffer. After 15 min the colour was measured spectrophotometrically at 870 nm. Values were compared with a standard curve based on bovine ATPase (Sigma) and expressed as units per/mg protein. The assay for cytochrome oxidase was carried out as previously described (Wharton and Tzagoloff 1967). Briefly, the assay was based on whole cell extracts prepared from homogenized cells in PBS and 1% NP-40. 0.8 mL of 1% ferrocytochrome c prepared in 0.01 m potassium phosphate buffer pH 7. The protein sample was diluted to 100 µg/mL in 0.01 m potassium phosphate buffer and 100 µL added to the reaction. A blank with 0.1 m potassium ferricyanide instead of ferrocytochrome C was subtracted from the values. Increased absorbance was measured over 15 min at 550 nm. Protein content of extracts was determined using the BCA assay (Sigma).

Analysis of NF-κB and STAT transcription factors

Nuclear extracts were prepared and incubated with 32P-αATP radio-labelled double stranded oligonucleotides containing consensus binding sequences for STAT or NF-κB as previously described (Girdlestone and Wing 1996). Cells were plated at 1–2 × 106 cells/well for up to 24 h after washing twice in serum-free plating media, test media was added. Nuclear extracts were made after washing the cells twice in PBS before transferring them to a microfuge tube where they were counted and pelleted (4500 r.p.m., 4 min). On ice 30 µL of buffer A (10 mm HEPES, pH 7.9, 1.5 mm MgCl₂, 10 mm KCl, 0.5 mm DTT, 0.5 mm PMSF, 0.5% NP40, Aproteinin 10 µg/mL in distilled water) was added to the cell pellet for 1 min. After centrifugation (14 000 r.p.m., 2 min) 15 µL of buffer C (20 mm HEPES, pH 7.9, 25% glycerol, 0.42 mm NaCl, 1.5 mm MgCl₂, 0.2 mm EDTA, 0.5 mm DTT, 0.5 mm PMSF, aproteinin 10 µg/mL in distilled water) was added to the pellet for a total of 10 min and vortexed regularly. After a further centrifugation the DNA pellet was removed and the nuclear extract was frozen at − 70°C. At room temperature 2 µL nuclear extract, 5 µL distilled water and 1 µL nonspecific DNA (poly-dI-dC) were incubated for 10 min in 96-well plate. Where appropriate 1 µL STAT-1 antibody (a rabbit anti-STAT-1 human antibody (Santa Cruz) without cross-reactivity to other STATs), was added for a further 10 min 2 µL ³²P-labelled consensus sequence was added for 20 min, then 2 µL loading buffer (50% glycerol, 0.25% bromophenol blue in sterile water) before loading onto a nondenaturing 6% acrylamide gel in 0.5 × TBE and running at 200 V, 12 mA for 90 min (5 mL acryl stock [50% 37.5 : 1 acrylamide, 37.5% 19 : 1 acrylamide made up with sterile water], 10 × TBE 0.65 mL [0.045 m Tris-borate, 0.001 m EDTA], TEMED 50 µL, 25% AMPS 50 µL). The gel was dried and autoradiographed for up to 10 days. Optical density was calculated using NIH image software (NIH, Washington, WA, USA) and statistical analysis carried out using the paired t-test.

Soluble factors secreted by interferon-γ treated microglia impair neuronal reduction of MTT

Cultured cerebellar neurones from 6-day-old mice were grown first in serum-free medium and then treated with conditioned medium either from resting or interferon-γ treated microglia. As further controls neurones were treated with unconditioned medium with and without interferon-γ. Cerebellar neurones were incubated for 2 h in increasing concentrations of conditioned media (Fig. 1a) and an MTT assay was performed directly. Unconditioned media, interferon-γ alone and media conditioned by resting microglia did not alter measured values of the MTT assay at 2 h. However, a 2-h treatment with medium from interferon-γ treated microglia caused a reduction (∼30%) in MTT values measured for cerebellar neurones (Fig. 1a). MTT assays directly measure the reducing activity of cells. The majority of this activity is contained in cells. Therefore MTT assays are a measure of mitochondrial activity. As reduced mitochondrial activity can also correspond to a decreased number of cells further assays were carried out. Parallel cultures were assayed for survival with a standard LDH assay to identify cell loss at 24 h (Fig. 1b). Neurones were treated for 2 h with conditioned media and then restored to unconditioned medium before assaying survival. LDH assays carried out at 24 h following a 2-h treatment with medium conditioned from interferon-γ treated microglia showed no significant difference (t-test, p > 0.05) implying that the medium was not toxic (Fig. 1b). MTT assays carried out at 24 h in parallel with the LDH assay also showed no significant change (p > 0.05) indicating that the change in MTT assay value for neurones treated with conditioned medium from interferon-γ treated microglia was transitory. Additionally we also examined neurones treated with conditioned medium from interferon-γ treated microglia using the Hoechst reagent. Directly after the 2 h treatment there was no significant increase in cells with condensed or fragment nuclei (treated = 4.2 ± 0.5 cells per field; control, 4.7 ± 0.6 cells per field, 10 fields, n = 3). Similarly after 24 h (cells returned to unconditioned medium) there was also no difference in the number of apoptotic nuclei (treated = 5.6 ± 0.8 cells per field; control, 6.7 ± 0.6 cells per field, 10 fields, n = 3). These results suggests that the measured reduction of MTT assay value after 2 h treatment of neurones with conditioned medium from interferon-γ treated microglia is not a result of cell death but of a transitory effect on mitochondrial ability to reduce MTT.

Prolonged culture with media conditioned by interferon-γ treated microglia is toxic for mouse cerebellar neurones

Cerebellar neurones were given a longer exposure to medium consisting of microglial conditioned medium diluted 1 in 2 with unconditioned medium. At each time point (2, 4, 24 and 48 h), cells were either assayed with the MTT assay directly (Fig. 1c) or transferred to unconditioned medium and maintained in culture before carrying out an MTT assay at 48 h. The MTT assay values measured directly were reduced for cells treated with medium from interferon-γ treated microglia even after brief exposure (2 h, Student's t-test, p < 0.05). Further exposure led to a progressive reduction in MTT at 24 or 48 h (Student's t-test, p < 0.05). Medium from resting microglia, or interferon-γ added with medium from resting microglia had no effect on MTT assay values. The changes in MTT assay values measured directly after 2 and 4 h treatment were transitory and could not be detected at 24 (Fig. 1b) or 48 h (2 h, 97 ± 3%; 4 h, 96 ± 2% compared with control value 100 ± 2% n = 4) suggesting that removing the medium activated microglial medium allowed the neurones to recover from the effect causing the reduction in MTT assay value. However, the value from 24 h treated neurones did not return to control values when measured 24 h after returning the cells to the unconditioned medium, i.e. at 48 h (24 h, 88 ± 3% compared with control value 100 ± 2%). This suggests that the effect after 24 h of continuous treatment is no longer transitory. To determine if these changes in MTT reflect cell loss, LDH assay was carried out at 48 h after initial exposure for 2, 4, 24 and 48 h with test media and then maintained with unconditioned medium up to 48 h (Fig. 3d). Only the 48-h incubation with medium conditioned by interferon-γ treated microglia led to a reduction in LDH implying that the onset of cell death was at ≥ 24 h. This result was confirmed using Hoechst staining of neurones treated with medium conditioned from interferon-γ treated microglia for 48 h. There was a significant difference (p < 0.05) in the number of apoptotic nuclei (treated = 3.6 ± 0.8 cells per field; control, 27.7 ± 0.6 cells per field, 10 fields, n = 3).

Soluble factors secreted by interferon-γ treated microglia impair neuronal mitochondrial activity

To confirm the impairment of mitochondrial metabolism, extracts from cells treated for 2 h with microglial conditioned medium were assayed for total ATPase and cytochrome oxidase activity (Fig. 2). The activity of these enzymes was significantly (p < 0.05, n = 7 both assays) reduced in cerebellar neurones treated with medium conditioned by interferon-γ treated but not resting microglia. This suggests that cells treated with this extract have reduced metabolic capacity due to impaired mitochondrial activity.

Interferon-γ treated microglia stimulate NF-κB expression in mouse cerebellar neurones

We wished to determine if NF-κB activation was associated with the changes in the cerebellar neurones as it has been suggested that such activation might be protective against neuronal death. We cultured cerebellar neurones from 6-day-old mice for up to 24 h with media conditioned by interferon-γ treated and resting microglia and with unconditioned medium with and without interferon-γ added. Nuclear extracts were made from these cultures and run in an electrophoretic mobility shift assay (EMSA) with NF-κB and STAT consensus sequences. In the presence of interferon-γ treated but not resting microglia conditioned media or controls, an additional band bound to the NF-κB consensus sequence after 2 h (Fig. 3a) and this was still present after 24 h incubation (Fig. 3c). Controls were also carried out using cold NF-κB consensus sequence peptides. These abolished the bands in the extracts induced by conditioned medium from interferon-γ microglia (data not shown).

A single band was observed in the presence of the STAT consensus sequence after a 2-h incubation with interferon-γ treated microglia conditioned medium (Fig. 3b). The STAT band was also present on incubation with interferon-γ alone (data not shown) and could be super-shifted with anti-STAT-1 antibody (data not shown), compatible with a direct effect of interferon-γ ligation on its receptor within neuronal membranes.

Anti-TNFα antibody inhibits cerebellar responses to interferon-γ treated microglial medium

Using the L929 TNFα bioassay, we demonstrated a rise in TNFα secretion from 1.05 ± 0.5 in resting to 5.46 ± 1.2* U/mL in interferon-γ treated microglia conditioned medium (four experiments, mean ± SEM, *p < 0.05) confirming microglial activation. TNFα is known to have deleterious effects on neurones. Therefore, cerebellar cells were treated with conditioned medium from interferon-γ activated microglia for 2 h in the presence of increasing concentrations of an antibody to TNFα (Fig. 4a). This abolished the transient decrease in MTT suggesting that either TNFα or TNFα acting synergistically with other factors is responsible for the observed reduction in MTT caused by medium from interferon-γ treated microglia. NF-κB induction seen with medium conditioned by interferon-γ treated microglia was blocked in the presence of anti-TNFα antibody (5 ng/mL) confirming that the effect is TNFα dependent (Fig. 3d, lanes 1 and 2).

Interferon-γ treated microglia stimulated NF-κB expression is inhibited by SN50

We attempted to block the NF-κB activation seen in cerebellar cells on treatment with interferon-γ treated microglia conditioned medium. Experiments were carried out using SN50 (Maggirwar et al. 1998; a peptide which inhibits entry of the p52 subunit of NF-κB into the nucleus) and a mutated control peptide (SN50M; which has a similar sequence but no effect on NF-κB transport). The peptides were added to cerebellar neurone cultures together with conditioned media from interferon-γ treated microglia for 2 h before making nuclear extracts. These extracts were run on an EMSA gel run and analysed in coded assays (i.e. blind to treatment) on three separate occasions. The SN50 peptide but not the SN50M peptide inhibited entry of NF-κB subunit p52 into the nucleus (Fig. 3d, lanes 3 and 4).

SN50 causes medium conditioned by interferon-γ treated microglia to be neurotoxic

Cerebellar neurones treated with medium conditioned by interferon-γ treated microglia for 2 h were cotreated with SN50 and SN50M at increasing concentration. After 2 h the neurones were returned to unconditioned medium but were maintained in the same concentration of the peptides until MTT assays carried out at 24 h. Treatment of cerebellar cells with SN50 or SN50M alone had no effect on MTT assay values. MTT assay values of conditioned media treated cerebellar neurones were reduced by SN50 but not by SN50M (Fig. 4b).

The effects of SN50 on cerebellar was also assessed using LDH assay (Fig. 4c). SN50 that was not toxic to cerebellar cells at 50 µg/mL. SN50 was significantly toxic to cerebellar neurones treated for 2 h with medium from interferon-γ treated microglia (50% of total medium) (p < 0.05). Parallel cultures were treated with SN50 and anti-TNFα antibody. The toxic effect of SN50 was blocked by the anti-TNFα antibody.

In another set of experiments cerebellar neurones were treated for 24 h with TNFα at 10 units/mL. Neurones were also treated with TNFα with addition of from 100 units/mL interferon-γ or 50 µg/mL SN50 peptide. At 24 h an assay showed that TNFα reduced the assay values of MTT but not LDH. Interferon-γ did not reduce the effect further (Fig. 4d). Addition of the SN50 peptide enhanced the effect of TNFα on the MTT value and caused a reduction in the LDH assay value indicating a toxic effect. These results indicate that NF-κB protects against the toxic effects of TNFα in medium conditioned by interferon-γ treated microglia and SN50 blocks this effect of NF-κB by preventing its entry into the nucleus.