Luteolin triggers global changes in the microglial transcriptome leading to a unique anti-inflammatory and neuroprotective phenotype

C57BL/6 mice were purchased from Charles River Laboratories. Mice were kept in an air-conditioned barrier environment at constant temperature of 20-22°C on a 12-h light-dark schedule, and had free access to food and water. The health of the animals was regularly monitored, and all procedures were approved by the University of Regensburg animal rights committee and complied with the German Law on Animal Protection and the Institute for Laboratory Animal Research Guide for the Care and Use of Laboratory Animals, 1999.

Brain microglia were isolated and cultured as described earlier [36]. BV-2 microglia-like cells were provided by Professor Ralph Lucius (Clinic of Neurology, Christian Albrechts University, Kiel, Germany). BV-2 cells were cultured in RPMI/5% FCS supplemented with 2 mM L-Glutamine and 195 nM β-mercaptoethanol. Primary brain microglia or BV-2 cells were stimulated with 10 ng/ml or 50 ng/ml LPS and various concentrations of luteolin for 24 h. 661W photoreceptor-like cells were a gift from Prof. Muayyad Al-Ubaidi (University of Illinois, Chicago, IL) and the culture conditions have been described elsewhere [36].

BV-2 cells were plated overnight on coverslips, fixed with 3.7% paraformaldehyde for 10 min at 37°C, permeabilized with 0.2% Triton X-100 for 5 min, blocked with 5% non-fat milk, 0.2% Triton X-100, and stained with DAPI for 10 min at room temperature (0.1 μg/ml in PBS, 4',6-diamidino-2-phenylindol, Molecular Probes). Filamentous actin was stained by addition of 1.5 μM TRITC-conjugated phalloidin (Sigma). The coverslips were mounted on microscopic glass slides and viewed with a Axioskop 2 fluorescence microscope equipped with an Eclipse digital analyzer (Carl Zeiss).

NO concentrations were determined by measuring the amount of nitrite secreted by BV-2 cells into the culture medium using the Griess reagent system (Promega). 50 μl cell culture supernatant was collected and an equal volume of Griess reagent was added to each well. After incubation for 15 min at room temperature, the absorbance was read at 540 nm on a BMG FluoStar Optima plate reader (Labtech, Offenburg, Germany). The concentration of nitrite for each sample was calculated from a sodium nitrite standard curve.

Apoptotic cell death of 661W cells was determined with the Caspase-Glo^® 3/7 Assay (Promega). Cells were lysed and incubated with a luminogenic caspase-3/7 substrate, which contains the tetrapeptide sequence DEVD. Luminescence was then generated by addition of recombinant luciferase and was proportional to the amount of caspase activity present. The luminescent signal was read on a BMG FluoStar Optima plate reader (Labtech, Offenburg, Germany). A blank reaction was used to measure background luminescence associated with the cell culture system and Caspase-Glo^® 3/7 Reagent. The value for the blank reaction was subtracted from all experimental values. Negative control reactions were performed to determine the basal caspase activity of 661W cells.

Total RNA was extracted from cultured microglia according to the manufacturer's instructions using the RNeasy Protect Mini Kit (Qiagen, Hilden, Germany). Purity and integrity of the RNA was assessed on the Agilent 2100 bioanalyzer with the RNA 6000 Nano LabChip^® reagent set (Agilent Technologies, Büblingen, Germany). The RNA was quantified spectrophotometrically and then stored at -80°C. First-strand cDNA synthesis was performed with RevertAid™ H Minus First Strand cDNA Synthesis Kit (Fermentas, St. Leon-Rot, Germany).

Triplicate microarrays were carried out with three independent RNAs from non-stimulated BV-2 microglia or cultures treated for 24 h with 50 μM luteolin, 50 ng/ml LPS, or 50 μM LPS + 50 ng/ml LPS, respectively. Generation of double-stranded cDNA, preparation and labeling of cRNA, hybridization to Affymetrix 430 2.0 mouse genome arrays, washing, and scanning were performed according to the Affymetrix standard protocol. Minimum information about a microarray experiment (MIAME) criteria were met [37]. The microarray datasets of this study are publicly available at the National Center for Biotechnology Information Gene Expression Omnibus http://​www.​ncbi.​nlm.​nih.​gov/​geo/​ as series record GSE18740.

The Affymetrix Expression Console Software Version 1.0 was used to create summarized expression values (CHP-files) from 3' expression array feature intensities (CEL-files) using the Robust Multichip Analysis (RMA) algorithm. Integrative analysis of genome-wide expression activities from BV-2 cells was performed with the Gene Expression Dynamics Inspector (GEDI), a Matlab (Mathworks, Natick, MA) freeware program which uses self-organizing maps (SOMs) to translate high-dimensional data into a 2D mosaic [38]. Each tile of the mosaic represents an individual SOM cluster and is color-coded to represent high or low expression of the cluster's genes, thus identifying the underlying pattern.

Differentially regulated transcrips in 24 h luteolin stimulated versus non-treated and luteolin + LPS versus LPS-treated BV-2 cells, respectively, were retrieved with the Genomatix ChipInspector program (Genomatix Software GmbH, Munich, Germany), applying the Significance Analysis of Microarray (SAM) algorithm using a false-discovery rate of 0.1% and a minimum coverage of 3 independent probes.

Functional annotation of transcripts was performed using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) [39] and the Bibliosphere pathway edition (Genomatix).

Amplifications of 50 ng cDNA were performed with an ABI7900HT machine (Applied Biosystems) in triplicates in 10 μl reaction mixtures containing 1 × TaqMan Universal PCR Master Mix (Applied Biosystems), 200 nM of primers and 0.25 μl dual-labeled probe (Roche ProbeLibrary). The reaction parameters were as follows: 2-min 50°C hold, 30-min 60°C hold, and 5-min 95°C hold, followed by 45 cycles of 20-s 94°C melt and 1-min 60°C anneal/extension. Measurements were performed in triplicate. Results were analyzed with an ABI sequence detector software version 2.3 using the ΔΔCt method for relative quantitation. Primer sequences and Roche Library Probe numbers are listed in Table 1.

Statistical analysis were performed on ΔΔCt data using analysis of variance with a two-sample Student's t test (P < 0.05) unless otherwise indicated. Quantitative data are expressed as mean ± SEM. The levels of gene expression in treated BV-2 cells are shown relative to control cells.

As a basis to study the genome-wide transcriptional effects of luteolin on activated microglia and to validate our cell culture system, we initially performed a dose-response curve for luteolin. Four pro-inflammatory microglia markers with different expression levels and ranges of induction were selected as positive controls for qRT-PCR analyses. Interleukin 6 (Il6) is a well known pro-inflammatory cytokine target of luteolin [34], chemokine (C-X-C motif) ligand 10 (Cxcl10), interferon-regulatory factor 7 (Irf7), and interferon-inducible protein 44 (Ifi44) are LPS-sensitive genes in microglia [36]. BV-2 microglia were pre-treated with different concentrations of luteolin (0, 5, 10, 25, and 50 μM) for 1 h and then stimulated with LPS (10 ng/ml and 50 ng/ml) for a further 24 h period. Neither LPS nor luteolin changed the proliferation rate or cell survival at the concentration levels applied to the cells (data not shown). mRNA levels of Il6 (Fig. 1A), Cxcl10 (Fig. 1B), Irf7 (Fig. 1C), and Ifi44 (Fig. 1D), which were all induced by 10 ng/ml LPS and further increased by 50 ng/ml LPS, were dose-dependently reduced by luteolin (Fig. 1A-D). Luteolin's effects on the four genes were most prominent at 50 μM (Fig. 1A-D) and therefore this concentration was used in all further experiments.

Our next goal was to capture and compare the transcriptional profile of non-activated and LPS-activated BV-2 microgli treated each with 50 μM luteolin for 24 h using Affymetrix Mouse Genome 430 2.0 GeneChips. Twelve microarray analyses from three independent stimulation experiments were performed and high stringency criteria with a minimal signal intensity of 50 fluorescence units were used. The complete microarray RMA datasets and all raw data (Affymetrix CEL-files) were stored in the NCBI GEO repository as record GSE18740. We used the Gene Expression Dynamics Inspector (GEDI) to determine the global patterns of gene expression in the four different conditions, untreated, luteolin-treated, LPS-treated, and luteolin + LPS-treated microglia. GEDI is based on self-organizing maps to identify genome-wide transcriptome activity via 'gestalt' recognition [38]. GEDI is sample-oriented rather than gene-oriented, which facilitates the identification of genome-wide patterns. Each mosaic tile in the GEDI map represents a gene group or cluster that is expressed at similar levels, with blue color indicating a low level and red corresponding to high expression. The four GEDI maps clearly show a dynamic regulation of gene expression in stimulated versus non-stimulated microglia (Fig. 2). Especially the upper right positions in the control dataset and more pronounced in the LPS-treated dataset display an inverse regulation of the gene clusters following luteolin stimulation (Fig. 2A, white circles). These results demonstrate that stimulation with luteolin has a major impact on the global pattern of gene expression notably in activated microglia and to a lower extend in resting cells.

We have previously shown that a single probe analysis of Affymetrix microarrays using the Genomatix ChipInspector tool with the Significance Analysis of Microarray (SAM) algorithm improves detection of regulated transcripts [40]. We therefore used ChipInspector to analyze our dataset applying a false discovery rate (FDR) of 0.1%, a minimal probe coverage of 3, and a minimum log2 ratio of 1 (fold change of 2.0). Thereby, 18 significantly regulated genes were identified in luteolin-treated versus non-treated microglia (Table 2) and 54 differentially expressed genes were detected in luteolin + LPS versus LPS-stimulated cells (Table 3). Comparison of the total numer of differentially expressed genes indicated that the effects of luteolin were more pronounced in LPS-activated BV-2 cells than in resting microglia. Also, the overall fold change range was higher in the LPS-treated microglia than in the resting BV-2 cells when stimulated with luteolin. These finding correlate with the gene cluster patterns observed in GEDI analysis and implicate that luteolin is an effective counter-player of pro-inflammatory microglial activation.

Our next aim was to put the newly identified luteolin-regulated expression patterns into a specific biological context. We used all differentially expressed genes to perform a classification into functional categories with the Database for Annotation, Visualization, and Integrated Discovery (DAVID). The major enriched functional categories in luteolin-treated resting microglia were anti-oxidants, phagocytosis, and anti-inflammation for up-regulated genes and pro-inflammation and apoptosis for down-regulated genes (Fig. 2B, upper part). These pathways indicate that luteolin stimulated anti-oxidant and anti-inflammatory transcriptional programs which could potentially reflect M2 macrophage polarization. Moreover, basal pro-inflammatory and pro-apoptotic gene expression was blocked in non-activated BV-2 microglia. In luteolin + LPS co-treated versus LPS-treated cells, 11 induced transcripts represented pathways of lipid metabolism, antigen presentation, and anti-oxidants (Fig. 2B, bottom left). Interestingly, a large number of down-regulated genes covered the pathway categories pro-inflammation, toll-like receptor (TLR) signaling, acute phase response, apoptosis, and complement factors (Fig. 2B, bottom right). These results implicate that luteolin-induced transcriptomic changes promote a stable anti-inflammatory, anti-oxidant, and anti-apoptotic microglial phenotype reminescent of M2 macrophages. Our microarray data also identified a significant number of genes that could not be grouped into larger immune-related pathways and that have not been previously connected to microglial activation or flavonoid stimulation (Tables 2 and 3).

To validate a significant portion of the DNA microarray results, real-time qRT-PCR analyses were carried out with RNA samples from three independent BV-2 replicate experiments. We focused on representative candidates from different biological pathway which have not been previously described as flavonoid targets. 22 genes fulfilled these criteria and were analzyed with qRT-PCR. In the first set of experiments, mRNA levels of genes up-regulated by luteolin treatment were assessed (Fig. 3). Transcripts of Sulfirexodin 1 (Srxn1), Biliverdin reductase B (Blvrb), Heme oxigenase 1 (Hmox1), and Glutamate-cysteine ligase (Gclm) are components of the cellular anti-oxidant response [41‐43] and were all induced by luteolin treatment (Fig. 3A-D). We could further validate increased levels of Lysophosphatidylcholine acyltransferase (Lpcat1), Ring finger protein 145 (Rnf145), Cd36 antigen, Kinase insert domain receptor (Kdr, alias Vegfr2), and Cd83 antigen. Lpcat1 catalyzes the inactivation of inflammatory lipids [44], whereas Cd36 and Kdr are surface receptors required for phagocytic uptake and chemotactic migration, respectively [45, 46]. The function of Rnf145 is currently unknown.

In the next set of qRT-PCR experiments, down-regulated transcripts involved in pro-inflammatory activation and acute phase response were analyzed. Complement factor C3 (C3), Complement factor b (Cfb), Secreted leukocyte peptidase inhibitor (Slpi), and the pro-inflammatory Guanylate binding proteins 2 (Gbp2) and Gbp3 [47] showed diminished mRNA levels in luteolin-treated samples (Fig. 4A-D). The same tendency was seen in the newly identified microRNA miR-147, which is involved in toll-like receptor signaling [48], the acute phase gene Haptoglobin (Hp), the stress-related gene Nuclear protein 1 (Nupr1, alias p8) [49], and Cystatin 7 (Cst7) (Fig. 4F-I).

As third group of genes validated by qRT-PCR, four luteolin-repressed genes related to apoptosis and microglial shape formation were studied (Fig. 5A-D). The apoptotic mediators DNA-damage inducible transcript 3 (Ddit3, alias Chop, Gadd153) and Tribbles homolog 3 [50, 51] were down-regulated by luteolin in non-activated as well as LPS-activated microglia (Fig. 5A, B), indicating anti-apoptotic protection mechanisms elicited by luteolin. In line with these findings, we also observed reduced expression of the Macrophage receptor with collagenous structure (Marco) and Lipocalin 2 (Lcn2), two molecules with dual roles in apoptosis and de-ramification of activated microglia [52, 53].

To assess whether the particular gene expression profiles measured in luteolin-stimulated microglia translate into detectable functional properties, phenotypic characterization was performed. The general activation state and morphological phenotype of microglia is particularly reflected by their cell shape and cytoskeletal organization. To detected morphological changes and for visualization of filopodia we performed phalloidin staining. Conventional BV-2 microglia cultures were low level activated cells with a flat shape and some filopodia (Fig. 6A). Culture of BV-2 cells in the presence of luteolin lead to considerable ramification and formation of long protrusions (Fig. 6B), indicating induction of a surveillance state. LPS-activation of BV-2 cells caused formation of a round cell shape with retracted filopodia (Fig. 6C). In contrast, co-incubation of LPS-treated cells with luteolin sustained the ramified microglial morphology (Fig. 6D). A similar effect was seen in primary mouse microglia, where luteolin increased the length of filopodia in non-activated cells (Fig. 6E, F) and caused flatening and ramification of LPS-treated cells (Fig. 6G, H). To test the direct effect of luteolin on microglial secretion of toxic metabolites, NO levels were determined in the supernatant of BV-2 cells. Treatment of BV-2 cells with luteolin alone did not result in increased NO concentrations (Fig. 6I), whereas stimulation with LPS markedly increased secreted NO levels. Co-incubation of LPS-activated cells with luteolin nearly completely abolished NO secretion (Fig. 6I). These data indicate that luteolin favors the ramified surveillance state of microglia and effectively inhibits cytotoxic NO formation. Kim et al. previously reported that luteolin triggered a blockade of NO secretion in LPS-stimulated BV-2 microglia, which was mediated by inhibition of inducible NO synthase (iNos) protein expression [54]. To indenpendently verify these data by qRT-PCR, we analyzed iNos mRNA levels in BV-2 cells treated with luteolin, LPS, or both simultaneously. LPS caused a more than 40-fold increase in iNos transcripts and luteolin co-treatment significantly reduced iNos levels more than 2-fold. Thus, our expression data corroborate the findings by Kim et al. and provide a reasonable explanation for reduced NO secretion levels.

To test the hypothesis that luteolin leads to decreased microglial neurotoxicity, a culture system of 661W photoreceptor cells with conditioned medium (CM) from BV-2 cells or primary microglia was established. 661W, a retina-derived cell line [55], represents a valuable model to study microglial neurotoxicity in the special context of retinal degeneration [36]. 661W cells were incubated for 24 h with culture supernatants from unstimulated, luteolin-, LPS- or LPS + luteolin-treated cells and 661W photoreceptor cell morphology was assessed by phase contrast microscopy. 661W cells in their own medium grew confluent after 24 h and the presence of CM from control- or luteolin-treated BV-2 cells did not affect cell morphology (Fig. 7A, B). As described previously [56], confluent 661W cells flattened out and often multiple cells were connected to each other through their projected extensions (Fig. 7A, B). In contrast, 661W cells co-incubated with LPS-stimulated BV-2 supernatant appeared elongated and smaller, leading to prominent cell-free areas present in the culture (Fig. 7C). When adding CM from LPS + luteolin-stimulated BV-2 cells, nearly normal cell characteristics were retained (Fig. 7D). Similar morphological changes of 661W cells were observed when cultured in the presence of primary microglia CM (Fig. 7E-H). Direct incubation of photoreceptor cells with LPS, luteolin, or both had no effects on cell morphology (data not shown), indicating that the observed changes in 661W cell growth arise from secreted microglial compounds.

To correlate the microscopic findings with functional data, we next studied the influence of microglia-secreted products on caspase-related apoptotic signaling in the neuronal cell model. 661W cells cultured in the presence of CM from LPS-stimulated BV-2 cells displayed a strong induction of caspase 3/7 activity (Fig. 7I). When CM from microglia co-treated with LPS + luteolin was used, 661W apoptosis was still present but was significantly diminished (Fig. 7I). Culture media from BV-2 cells kept in the presence of luteolin alone had no effect on 661W apoptosis. These findings implicate that luteolin either stimulates microglia to produce less pro-apoptotic substances or actively promotes the release of survival signals.