Tissue transglutaminase in astrocytes is enhanced by inflammatory mediators and is involved in the formation of fibronectin fibril-like structures

Primary rat glial cells were isolated from cerebral cortices of 2-day-old Wistar rats (Harlan CPB, Zeist, The Netherlands), as described previously [57], and approved by the Animal Experiment Committee of the VU University Medical Center (ID FGA 11-03). The meninges and blood vessels were removed from the cortices. Cortices were mechanically homogenized in Dulbecco’s Modified Eagle’s Medium-F10 (Gibco, Life Technologies, Breda, The Netherlands), supplemented with 10% v/v heat-inactivated fetal calf serum (Gibco), 2 mM L-glutamine (Sigma-Aldrich, St. Louis, MO, USA), 50 U/ml penicillin (Sigma-Aldrich), and 50 μg/ml streptomycin (Gibco) to make a cell suspension. Dissociated cells from 2 to 3 pups were plated in poly-L-lysine (PLL, 15 μg/ml (2 μg/cm²); Sigma-Aldrich) coated T75 culture flasks (Nunc, Hamstrop, Denmark) and incubated at 37 °C in humidified air containing 5% CO₂. The medium was changed at days 1, 6, and 8 after seeding. After 10 days in culture, microglia and astrocytes were separated by shaking the flasks on a rotary platform (Heidolph Unimax 2010) at 230 rpm for 16 h. Astrocytes were further purified by treatment with 5 mM leucine methyl ester (Sigma-Aldrich) in serum-free medium, for 24 h at 37 °C. Microglia were cultured in fresh medium mixed with conditioned medium (1:1 ratio) collected from the mixed cell culture before separation. The purity of astrocyte cultures was between 80 and 90%, as described previously [58], determined by immunofluorescent staining using glial fibrillary acidic protein (GFAP) antibody (1:6000; DAKO, Glostrup, Denmark). Purity of primary microglial cultures was between 90 and 94% as determined by immunofluorescent staining using Iba1 antibody (1:1000; WAKO Chemicals USA).

Astrocytes and microglia cultured on 2 μg/cm² PLL-coated 6-well plates (Nunc) were incubated for 48 h in medium alone (control) or with lipopolysaccharide (LPS, E. coli 055-B5, Difco, 100 ng/ml), human recombinant (rec) transforming growth factor (TGF)-β1 (BioLegend), human rec TGF-β2 (BioLegend), rat rec interleukin (IL)-1β (Glaxo), rat rec tumor necrosis factor (TNF)-α (Biolegend), rat rec TNF-α+IL-1β, rat rec interleukin (IL)-4 (BioVision), rat rec interleukin (IL)-6 (gift from Steve Poole), or rat rec interleukin (IL)-10 (Pharmingen) (all 20 ng/ml). In addition, astrocytes and ECM deposited by astrocytes were treated for 48 h with TNF+IL-1β in the absence or presence of 10 μM of the bona fide dihydroisoxazole TG2 activity inhibitor ERW1041E (Quinolin-3-ylmethyl(S)-2-((((S)-3-bromo-4,5-dihydro-isoxazol-5-yl)methyl)carbamoyl)pyrrolidine-1-carboxylate) diluted in 0.1% (v/v) dimethyl sulfoxide (DMSO); kind gift from C. Khoshla, Stanford University, USA) [59‐61]. Astrocytes were alternatively treated with exogenous recombinant guinea pig TG2 (Sigma-Aldrich) in pathophysiological concentrations (0.13 and 1.3 μM) according to previously published studies [62, 63] for 48 h. Additionally, deposited ECM, after removal of astrocytes, was treated with 0.64 μM exogenous recombinant guinea pig TG2 (Sigma-Aldrich) for 16 h combined with a pre-incubation with a selective inhibitor of TG2 (Z-DON, 1 μM, Zedira) diluted in 0.001% DMSO or only with DMSO for 30 min (min) at 37 °C. In some experiments, astrocytes were cultured in 8-well Lab-Tek Permanox chamber slides (Nunc) or 6-well plates (Nunc) coated with 2 μg/cm² laminin (mixture of laminin-1 and laminin-2, from Sigma-Aldrich) instead of PLL.

TG2 was downregulated in primary rat astrocytes by lentiviral transduction with TG2 specific shRNA. Therefore, astrocytes were plated on a laminin (2 μg/cm²; Sigma-Aldrich) coated 12-well plate (0.2 × 106 cells/well, Nunc). The following day, 0.2 × 106 infectious units of virus (IFU) of lentiviral particles (Santa Cruz, sc-270266-V for rat TG2 shRNA, or sc-108080 for control scrambled shRNA) were added to the cells. The next day, the same amount of lentiviral particles was added to the cells. The medium was changed the following day, and the cells were left for 5 days before treatment with TNF-α+IL-1β.

After sample collection, homogenates of whole cell lysate samples and TG activity samples were sonicated (Branson “Sonifier 250”; output 1, duty 30%, 8 pulses) and cleared by centrifugation (20,000 g for 10 min at 4 °C), and protein concentrations of supernatants were determined by the BCA method (Pierce Biotechnology, Perbio Science, Etten-Leur, The Netherlands). Samples were heated for 10 min at 95 °C. ECM samples were not subjected to sonication or heating to maintain the integrity of the ECM aggregates.

Equal amounts of protein (20 μg) were subjected to 8% SDS-polyacrylamide gel electrophoresis and transferred to a nitrocellulose membrane (Li-Cor Biosciences, Lincoln, NE, USA). Membranes were blocked at RT with Odyssey blocking buffer (Li-Cor Biosciences) diluted 1:1 with TBS. Membranes were incubated overnight at 4 °C with mouse anti-TG2 (ab3, Neomarkers 1:2000), rabbit anti-fibronectin (Neomarkers, 1:1000), or mouse anti-β-actin (Abcam, 1:5,0000) antibodies. After several washes with TBS containing 0.1% triton (TBS-T), blots were incubated for 1 h at RT with corresponding goat anti-rabbit or donkey anti-mouse IRDye 800CW IgG’s (1:10,000, Li-Cor Biosciences) for subsequent antigen detection. All antibody solutions were prepared in the Odyssey blocking buffer (Li-Cor Biosciences) diluted 1:1 with TBS-T. After washing with TBS-T and TBS, membranes were scanned for fluorescence emission at 800 nm, using an Odyssey infrared imaging system (Li-Cor Biosciences). Bands were visualized, and their signal intensities were measured using the Odyssey Sa Infrared scanning software (version 1.1, Li-Cor Biosciences).

TG activity was measured using the commercially available TG Covtest Transglutaminase Colorimetric Microassay (Covalab, Villeurbanne, France), which uses immobilized CBZ-Gln-Gly as the first substrate and biotinylated-cadaverine (biotin-Cd) as second substrate of the enzyme. To determine the amount of TG2 activity among the TG activity measured, samples were pre-incubated with a selective inhibitor of TG2 (Z-DON, 1 μM, Zedira) diluted in 0.001% DMSO or only with DMSO for 30 min at RT. The assay was performed following the manufacturer’s instructions [64]. In brief, 10 μg/well of each sample was incubated with 50 μl/well of biotin-Cd solution containing CaCl₂ for 1 h at 37°C on the CBZ-Gln-Gly coated plates. At the end of the incubation period, plates were washed three times with TBS (pH 7.5) containing 0.1% Tween 20. Then, 100 μl/well of streptavidin-labeled horseradish peroxidase (HRP) diluted to 1:2000 was added to the wells and incubated for 1 h at RT. After washing, peroxidase activity was revealed using 100 μl/well of 0.01% H₂O₂ as HRP substrate and (0.1 mg/ml) tetramethylbenzidine as electron acceptor (chromogen). The reaction was stopped by the addition of 2.5 N H₂SO₄, and TG activity was detected by absorbance measurement of streptavidin-labeled peroxidase activity in each well on a microplate reader (SpectraMax 250, Molecular Devices, Sunnyvale, CA, USA) at 450 nm. Guinea pig TG2 (T5398, Sigma-Aldrich) was used as a standard. One unit of guinea pig TG2 will catalyze the formation of 1.0 μmole of hydroxamate per min from Na-Z-Gln-Gly and hydroxylamine at pH 6.0 at 37 °C.

Astrocytes (4.0 × 104/well) were plated on laminin (2 μg/cm², Sigma-Aldrich) or PLL (2 μg/cm², Sigma-Aldrich) coated 8-well Lab-Tek Permanox chamber slides (Nunc) and treated with TGF-β1 (for ECM stainings only) or a combination of TNF-α + IL-1β cytokines (20 ng/ml each) for 48 h. Alternatively, exogenous recombinant guinea pig TG2 (0.13 and 1.3 μM, Sigma-Aldrich) was added for 48 h to untreated astrocytes. Astrocytes were fixed in ice-cold methanol for 10 mins and washed in TBS. Nonspecific binding was blocked with 3% bovine serum albumin in TBS containing 0.05% triton (TBS-T, pH 7.6) for 30 min at RT. Subsequently, astrocytes were double-labeled with mouse anti-TG2 (ab1, Neomarkers, 1:300) and the astrocytic marker rabbit anti-GFAP (DAKO, 1:4000) or with rabbit anti-fibronectin (Millipore, 1:200) to examine TG2 localization in astrocytes and any co-localization with fibronectin, respectively. Triton was alternatively omitted from all the incubation steps during double stainings to detect TG2 and fibronectin on the cell surface. Prior to ECM stainings, astrocytes were removed as described before. Triton was not omitted for the staining of the ECM. Secondary antibodies used were the appropriate Alexa Fluor© (488 or 594)-conjugated IgG’s (Molecular Probes, 1:400). After washing in TBS, astrocytes were cover-slipped in polyvinyl alcohol mounting medium with DABCO (Sigma-Aldrich) and examined using a confocal microscope (Leica TSC-SP2-AOBS; Leica Microsystems, Wetzlar, Germany).

ECM lysates were treated for 30 min at RT with 150 mM DTT to break down disulfide bonds and to obtain stable aggregates. Equal amounts of ECM samples (12 μg) were applied to a cellulose acetate membrane (GE water and process technologies) using a filter trap assay apparatus (FTA, Bio-Dot SF microfiltration apparatus, Bio-Rad). This apparatus was coupled to a vacuum-pump in order to wash through the monomers and dimers of ECM aggregates and trap aggregates larger than the pore-size of 5 μm of the membrane. Approximately 5 min after sample application, the membrane was washed in PBS and incubated overnight at 4 °C with rabbit anti-fibronectin (Neomarkers, 1:1000) antibodies. For subsequent antigen detection, blots were incubated for 2 h with corresponding goat anti-rabbit IRDye 800CW IgG’s (1:10,000, Li-Cor Biosciences). After washing with TBS-T and TBS, membranes were scanned for fluorescence emission at 800 nm, using the same Odyssey infrared imaging system (Li-Cor Biosciences) as used for western blotting.

Normal distribution of the data was tested using the Shapiro-Wilk procedure. One-way analysis of variance was performed followed by Dunnett’s post hoc test for multiple comparisons by using IBM SPSS software, version 20.0 (IBM Corp., NY, USA). P < 0.05 was considered to represent statistically significant differences, and a P value between 0.05 and 0.1 was regarded as a statistical trend.

We first examined whether primary rat astrocytes and microglia produced TG2 and fibronectin. Under control conditions and after treatment with LPS, astrocytes expressed more TG2 and fibronectin than microglia (Additional file 1). Upon subsequent treatment of astrocytes and microglia with a variety of inflammatory mediators, TG2 protein levels were significantly increased after LPS, TNF-α+IL-1β, or IL-4 treatment (p < 0.001 for all, Fig. 1a, b). Little protein was found in untreated microglia. Incubation of microglia with the various mediators resulted in significant changes in TG2 protein levels (Fig. 1d, e). LPS or TNF-α+IL-1β treatment resulted in a significant increase in TG2 protein level (p < 0.01 for LPS and p = 0.001 for TNF-α+IL-1β, Fig. 1d, e). Moreover, IL-4 treatment showed a trend (p = 0.080) toward an increase in TG2 protein level (Fig. 1e). Other conditions studied, i.e., TGF-β1, TGF-β2, IL-1β, TNF-α, IL-6, and IL-10, did not show significant effects on TG2 protein level in astrocytes or microglia (Fig. 1a–e). Double bands at approximately 78 kDa were visible and quantified for TG2 in most conditions studied. This has been shown by others as well [65, 66] and could represent the full-length TG2 protein and the short TG2 protein that is truncated at the 3′ end [67]. Fibronectin levels were significantly increased in astrocytes after treatment with TGF-β1 (P < 0.01, Fig. 1a, c), and TGF-β2 showed a trend (P = 0.062, Fig. 1c) toward an increased fibronectin level. Other conditions studied, i.e., LPS, IL-1β, TNF-α, TNF-α+IL-1β, IL-4, IL-6, and IL-10, did not result in significant effects on fibronectin protein levels in astrocytes (Fig. 1a, c). Moreover, none of the treatments affected the level of fibronectin in microglia (Fig. 1d, f). Considering the minimal amount of fibronectin detected in microglia (Fig. 1d), we decided to continue our study with a focus on astrocytes.

We subsequently visualized TG2 by immunohistochemistry in astrocytes. Under control conditions, little TG2 immunoreactivity was observed (Fig. 2a), but a clear increase was apparent in TG2 immunoreactivity in specific GFAP positive astrocytes after 48 h of TNF-α+IL-1β treatment (indicated by arrows in Fig. 2b). Inflammatory conditions not only affected expression of TG2 in astrocytes, but also increased its activity. In particular, LPS (P < 0.01) and TNF-α+IL-1β (P < 0.001) increased TG activity by astrocytes (Fig. 2c) which was reduced when adding the TG2 specific inhibitor Z-DON to the activity assay, indicating that the measured TG activity is mainly due to TG2. Of interest is that although IL-4 increased TG2 protein level, it showed a trend but did not significantly affect TG activity (Fig. 2c).

Astrocytes were subsequently plated onto the ECM protein laminin to mimic the situation of an MS lesion, where laminin expression is increased [22]. Moreover, laminin is of interest because it is expressed by cultured astrocytes from healthy subjects and MS patients [29], and laminin is a myelination permissive ECM molecule [68‐70]. Under those conditions, we again examined the effect of inflammatory conditions on TG2 and fibronectin levels. LPS or TNF-α+IL-1β treatment were as effective in enhancing TG2 and fibronectin protein levels (P < 0.05 for both treatments, Fig. 3a–c) as they were in astrocytes cultured on PLL (Fig. 1). However, IL-4 did not affect TG2 protein levels in astrocytes cultured on laminin (Fig. 3a, b) in contrast to astrocytes cultured on PLL (Fig. 1a, b), while fibronectin levels were not affected in astrocytes cultured on both ECM proteins (Fig. 1a, c and Fig. 3a, c). TGF-β1 treatment did not affect TG2 protein levels in astrocytes cultured on either PLL or laminin (Fig. 1a, b and Fig. 3a, b), whereas fibronectin protein levels in astrocytes cultured on laminin were slightly enhanced (Fig. 3a, c) but more significantly when astrocytes were cultured on PLL (Fig. 1a, c). We next examined the effect of different inflammatory conditions on the extracellular deposition of endogenously produced TG2 and fibronectin by astrocytes cultured on laminin. Astrocytes were therefore first cultured on laminin coated chamber slides and treated with TGF-β1 or a combination of TNF-α+IL-1β cytokines for 48 h. The astrocytes were then removed and the ECM, produced by these astrocytes during the 48 h incubation period, was immunohistochemically stained for TG2 and fibronectin. Without treatment, little extracellular TG2 and fibronectin immunoreactivity in the ECM could be detected (Fig. 3d). Similarly, after TGF-β1 treatment of the cells, TG2 in the ECM was almost undetectable (Fig. 3e), but fibronectin deposition was clearly visible (Fig. 3e) and showed fibril-like structures. Upon treatment with TNF-α+IL-1β, extracellular TG2 immunoreactivity was present in the ECM (Fig. 3f). This treatment did not visibly affect the amount of fibronectin deposited in the ECM, though co-localization between TG2 and fibronectin was apparent under this condition (indicated by arrows in Fig. 3f).

To follow-up on the observation that TG2 can be extracellular deposited by astrocytes, we determined if extracellular TG2 can have an effect on fibronectin deposition. As proof of principle, we applied exogenous TG2 to untreated astrocytes plated onto laminin and studied fibronectin deposition and aggregation. We observed that within a pathophysiological range of TG2 concentrations, the concentration of 0.13 μM exogenously added TG2 was visible in either the ECM or interacting with astrocytes, but no visible effect on fibronectin expression or deposition was observed (Fig. 4b). When the TG2 concentration of 1.3 μM was added to the cells, it resulted in a more general interaction with astrocytes and was possibly incorporated into the astrocytes (Fig. 4c). In addition, this concentration of TG2 resulted in morphologically altered fibronectin deposition which appeared with a more fibrillary structure (as indicated by arrows in Fig. 4c). Moreover, adding 0.64 μM TG2 to the deposited ECM, after removal of astrocytes, resulted in a significant increase in astrocyte-derived fibronectin aggregates in the ECM as detected by FTA (P < 0.001, Fig. 4d). Pre-incubation of the ECM with the specific TG2 inhibitor Z-DON partially, but significantly, reduced this increase in fibronectin aggregates (P < 0.05, Fig. 4d), indicating that TG2 activity contributes to an altered morphology of the fibronectin deposited and to fibronectin aggregation.

We next determined whether endogenously produced TG2, enhanced by MS relevant inflammatory stimuli, is involved in fibronectin deposition and aggregation. As already found that TNF-α+IL-1β did not affect fibronectin protein levels in astrocytes (Fig. 3c), we now observed that it neither significantly affected aggregation of extracellular fibronectin (Fig. 5a, b). Interestingly, the morphological appearance of extracellular fibril-like fibronectin deposits (Fig. 5c, see arrows) was clearly altered, i.e., shorter and thinner fibrils, upon co-incubation with the bona fide TG2 inhibitor ERW1041E (48 h) (Fig. 5d). Also, after lentiviral downregulation of TG2 (37% TG2 protein left of control scrambled shRNA treatment, data not shown), the morphological fibril-like fibronectin deposition, as found in astrocytes lentivirally transduced with control scrambled shRNA (Fig. 5e, see arrows), was altered (Fig. 5f).