A high throughput screening for TLR3-IRF3 signaling pathway modulators identifies several antipsychotic drugs as TLR inhibitors

Abstract

Toll-like Receptor 3 (TLR3) is one of the major innate immune sensors of double stranded RNA (dsRNA). The signal transduction pathway activated by TLR3, upon binding to dsRNA, leads to the activation of two major transcription factors: NF-κB and IRF3. In an effort to identify specific chemical modulators of TLR3-IRF3 signal transduction pathway we developed a cell-based read out system. Using the interferon stimulated gene 56 (ISG56) promoter driven firefly luciferase gene stably integrated in a TLR3 expressing HEK293 cell line, we were able to generate a cell line where treatment with dsRNA resulted in a dose dependent induction of luciferase activity. A screen of two pharmacologically active compound libraries using this system, identified a number of TLR3-IRF3 signaling pathway modulators. Among them we focused on a subset of inhibitors and characterized their mode of action. Several antipsychotic drugs, such as Sertraline, Trifluoperazine and Fluphenazine were found to be direct inhibitors of the innate immune signaling pathway. These inhibitors also showed the ability to inhibit ISG56 induction mediated by TLR4 and TLR7/8 pathways. Interestingly, they did not show significant effect on TLR3, TLR7 and TLR8 mediated NF-κB activation. Detailed analysis of the signaling pathway indicated that these drugs may be exerting their inhibitory effects on IRF3 via PI3K signaling pathway. The data presented here provides mechanistic explanation of possible anti-inflammatory roles of some antipsychotic drugs.

Introduction

Toll-like Receptors (TLR) have recently emerged as key components in sensing microbial infections and trigger antimicrobial host defense responses (1). TLRs are type I integral membrane glycoproteins, characterized by extracellular domains containing varying number of leucine-rich-repeat (LRR) motifs, and a cytoplasmic signaling domain, called the TIR (Toll/IL-1R homology) domain. They recognize conserved molecular patterns primarily found in invading microorganisms. Among the ten known human TLRs, TLR3 is responsible for sensing double stranded RNA (dsRNA) – a common byproduct or intermediate in viral genome replication (2). Besides TLR3, cytoplasmic dsRNA is also sensed by DExD/H box RNA helicases: RIG-I (retinoic acid-inducible gene I) and MDA5 (melanoma differentiation-associated gene-5) (3, 4). Binding of dsRNA by either TLR3 or RNA helicases causes transcriptional induction of a set of genes, mainly via the NF-κB and IRF (Interferon Regulatory Factor) family of transcription factors (5). Among them are the antiviral cytokine – type I interferons (IFN), which sensitize cells for detection of invading pathogens, inhibit protein synthesis and limit viral replication.

TLR3 is expressed by immune cells like conventional DCs, macrophages (6) and sometimes by cells of epithelial origins (7, 8). TRIF (TIR domain-containing adapter inducing IFNβ) is the unique adaptor protein used by TLR3 and TLR4 for downstream signaling (9). The TLR3 signaling pathway diverges into two branches from TRIF. One branch leads to the activation of NF-κB, c-Jun and ATF-2 via downstream mediator TRAF6 while the second branch causes activation of IRF3 via TRAF3 (10). TRIF, via TRAF6, helps activate the IκḄ kinases (IKK) and MAP kinases (JNK and p38) (11). IKKs activate NF-κB by phosphorylating its inhibitor IκB and causing its degradation. Released from IκB, NF-κB translocates to the nucleus and induces gene transcription.

IRF3 and IRF7 are the transcription factors mainly responsible for inducing IFN-β and other viral stress-inducible genes (5, 12, 13). IRF3 is mostly cytoplasmic and must be phosphorylated on specific Ser/Thr residues to form dimers, translocate to the nucleus, and bind to the specific cis-elements in gene promoters. IKK family protein kinases, TBK1 and IKKε, were found to phosphorylate IRF3 (14, 15). Induction of cytokine IFN-β is driven by a complex promoter requiring both NF-κB and IRF3, whereas, IFN-α subtypes, and several ISGs are directly induced by IRF family transcription factors – IRF3 or IRF7.

Despite their importance in protecting the host from invading pathogens, uncontrolled and sustained innate immune response via TLRs can result in chronic inflammatory diseases and cancer (16). Thus, modulation of TLR pathways offers an attractive method to fight diseases such as atherosclerosis, SLE, rheumatoid arthritis and many more (17-19). TLR3 has been shown to mediate inflammation and pathogenesis of viral infection. TLR3^–/– mice are more resistant to lethal infection by West Nile virus (WNV) than wild type mice (20). Similarly, TLR3 increases disease morbidity and mortality from Vaccinia and Phlebovirus infection (21, 22). Thus, in specific viral infection models, TLR3 may contribute not only to host defense but also to pathogenesis.

In order to search and identify small-molecule chemical modifiers of TLR3-IRF3 signaling pathway, we have developed a cell-based assay amenable to high throughput screening. Efficient use of small-molecule chemical libraries or small interfering RNA libraries has been successfully employed in drug discovery and/or pathway analysis research. However, only very few TLR signaling pathways have been subjected to high-throughput screening to identify modifiers (23). The advantages of applying these approaches to innate immune signaling pathways are many fold. i) A number of steps in these signaling pathways are dependent on protein-protein interactions, which make them amenable to small-molecule mediated disruptions. ii) It has the potential to identify novel reagents which can effectively modify innate immune signaling pathways. This may provide enormous medical benefits in treating a large number of inflammatory diseases including some forms of cancer. iii) In the process of determining the specificity of the modifiers and possible targets, we may discover new proteins or new roles of known proteins that play crucial roles in the signaling process. iv) Some of these modifiers may have potential to be used as high affinity agonists or antagonists for the biochemical purifications and characterization of signaling pathway components.

Here we report the identification and mechanism of action of a few novel inhibitors of TLR3-IRF3 signaling pathway identified by screening small-molecule chemical libraries. We screened two commercially available small-molecule chemical libraries. Out of total 31 modulators found from our secondary screening, we focused on a group of 5 compounds which have known biological functions and some of them are antipsychotic drugs. Detail characterization of two such molecules, Sertraline (SRT) and Trifluoperazine (TFP), showed that they specifically inhibit TLR mediated IRF activation by affecting PI3K signaling pathway while showing no effect on the NF-κB activation by TLRs. Our study reveals that these compounds, in addition to their well known anti-depressant and antipsychotic activities, may provide novel anti-TLR function.

Materials and Methods

Cell culture and Reagents

HEK293 cells and HEK293 derived stable cell lines, Wt11, C1 have been described before (24). HEK293 cells stably expressing TLR7 and TLR8 were generous gifts from Kate Fitzgerald (University of Massachusetts). All HEK293 based cell lines and HT1080 cells were cultured in DMEM containing 10% fetal bovine serum and Penicillin/Streptomycin. Human dendritic cells (DC) were generated from peripheral blood mononuclear cells (PBMC) as previously described (25). Briefly, leukocyte-enriched buffy coats from healthy donors were obtained from the Pittsburgh Central Blood Bank through an IRB-exempt protocol. PBMC were separated by density-gradient centrifugation through Ficoll-Paque Plus (Amersham Biosciences). PBMC were resuspended in serum-free AIM-V medium and plated at a density of 5×10⁶ cells/ml. After 2 h at 37°C, the non-adherent cells were washed away. To generate immature DCs, the adherent monocytes were cultured in serum-free AIM-V medium supplemented with recombinant human GM-CSF (1000 U/ml) and recombinant human IL-4 (1000 U/ml) for 6 days. Fresh cytokines were added on day 3 of culture. Synthetic dsRNA, poly(I):poly(C) was obtained from GE Healthcare and dissolved PBS (1 mg/ml). LPS was purchased from Sigma and resiquimod (R848) and LY294002 were from Alexis Biochemicals. Dual-luciferase reporter assay system and dual-Glo luciferase assay system were both purchased from Promega. HEK293 derived cells were transfected with Fugene6 (Roche Applied Sciences). Anti-actin, anti-IκBα, anti-tubulin, anti-AKT, and anti-Ser⁴⁷³-phospho-AKT antibodies were from Cell Signaling Technology Inc. Anti-ISG56, anti-ISG60, anti-DRBP76 anti-IRF3 antibodies have been described before (24).

Screening Methods

RL24 cells at a density of 3×10⁴ cells/well were plated in white wall, flat-bottomed, clear 96-well plates (Corning Inc.) in 100 μl media. The cells were allowed to grow for 24 h followed by the addition of compounds and dsRNA. We used plastic ‘replicators’ to manually apply the library directly in the media at a volume of 0.2 μl/well (final concentration of the compounds 20 μM). After 1 h incubation with the library compounds, poly(I):poly(C) was added to the cells at 1 μg/ml final concentration. A sub-saturating concentration of dsRNA was used to identify both positive and negative modulators. The firefly luciferase activities were measured after 6 h of dsRNA addition by adding 15 μl/well Dual-Glo reagent followed by luminometry in a Wallac Plate Reader. Following the firefly luciferase activity measurement, another 15 μl/well of Stop-and-Glo regent was added to the samples and the Renilla luciferase activities were measured after 15 min incubation at room temp. Luciferase activity from each well was normalized by converting to a ‘z’ score (26). Briefly, the raw mean and standard deviation were calculated from all test wells of a single plate. We then excluded those data points, which were more than three standard deviations away from the raw mean. The corrected mean (x_c) and standard deviation (s) of these filtered data were calculated and used for final z score calculation of all raw data using the formula z = (x_n – x_c)/s, where x_n is the raw measurement on the n^th compound (Supplementary Fig. 1). The compound library was formatted such that it had 80 compounds in a plate with two edge columns (1^st and 12^th) for controls. We added dsRNA in all 80 test wells and every alternate wells of 1st and 12th column (Shaded Boxes Supplementary Fig. 1) so that we could use these positive and negative control values for Z’ factor calculation (27). The above protocol regularly gave us Z’ -factor values of 0.7 and above for each plate. Luciferase activity of each test well was normalized by converting to a z score (Supplementary Fig. 1). Compounds with a z score greater than +2 or less than -2 were designated as significant hits; meanwhile, compounds which resulted in Renilla values lower than 90% of the average were discarded as overly toxic. Primary hits were selected for secondary and tertiary screening (Supplementary Fig. 2).

Transfection and reporter assays

Wt11 Cells were seeded into flat-bottomed 96-well cell culture plates at a density of 4×10⁴ cells/well in 100 μl media. The cells were transfected next day by Fugene6 with ISG56-luciferase reporter (10ng/well) and β-actin Renilla luciferase reporter (0.3 ng/well), together with RIG-I, IRF-3-5D mutant or vector control (5-40 ng/well). The total DNA per well was normalized to 50 ng by adding empty pcDNA3.1 vector. Twenty-four hours post-transfection, the cells were treated for 8 h with inhibitors, and luciferase activities were measured. The results were expressed as fold induction of ISG56-luciferase relative to that of vector control. For TLR7/8 reporter assays, TLR7 or TLR8 expressing cells were transfected similarly with NF-κB-luciferase and Renilla reporters, treated with inhibitors, and stimulated with R848 for 8 h followed by luciferase activity measurements.

Nuclear fractionation and Western-blotting

Wt11 cells cultured in 100 mm pates were treated with inhibitors and stimuli, washed in ice cold PBS and harvested by scraping. Nuclear fraction was prepared as described before (24). The pellet nuclei were washed and lysed in lysis buffer (20 mM HEPES pH 7.4, 0.5 % Triton-X 100, 150 mM NaCl, 1.5 mM MgCl₂, 12.5 mM β-glycerophosphate, 2 mM EGTA, 10 mM NaF, 2 mM DTT, 1 mM Na₃VO₄, 1 mM PMSF and 1X protease inhibitors). Nuclear extracts and whole cell lysates in lysis buffer were electrophoresed on 8% SDS-PAGE followed by western blotting with appropriate antibodies.

Quantitative PCR analysis of gene expression

RNA were isolated from cells treated with inhibitors and/or stimuli by TRIzol (Invitrogen), and cDNA was synthesized using iScript cDNA Synthesis Kit (Bio-Rad) and subjected to real-time PCR using a CFX96 Real Time System (Bio-Rad) according to manufacturer's instructions. The primers used were: human ISG56 (5’-CGCTATAGAATGGAGTGTCCA-3’; 5’-TTTCCTCCACACTTCAGCA-3’), human ISG60 (5’-AGTCTAGTCACTTGGGGAAAC-3’; 5’-ATAAATCTGAGCATCTGAGAGTC-3’), IFNα (5’-GTGAGGAAATACTTCCAAAGAATCAC-3’; 5’-TCTCATGATTTCTGCTCTGACAA-3’), IFNβ (5’-TGGGAGGATTCTGCATTACC-3’; 5’-CAGCATCTGCTGGTTGAAGA-3’) and IL-8 (5’-GTTTTTGAAGAGGGCTGAGAATTC-3’; 5’-CATGAAGTGTTGAAGTAGATTTGCTTG-3’). Each PCR amplification was normalized to RPL32 (5’-CAACATTGGTTATGGAAGCAACA-3’; 5’-TGACGTTGTGGACCAGGAACT-3’).

IL-8 ELISA

Wt11, HT1080 or human dendritic cells seeded in 24-well plate at a cell density of 4×10⁵ cells/well in 1 ml media. Cells were treated after 24 h with inhibitors for 1 h followed by TLR ligand stimulations for 18 h. The supernatants were harvested and assayed for IL-8 protein amount by ELISA using the human IL-8 ELISA set reagent from BD Biosciences following manufacturer's protocol.

Results

Development of cell-based assay system for high-throughput screening of TLR3-IRF3 signaling pathway modulators

Due to the absence of endogenous TLR3 expression, HEK293 cells do not respond to dsRNA applied to the culture medium, and provide an ideal cell system for developing cell-based screening for TLR3 signaling. We stably expressed human TLR3 in these cells and characterized the dsRNA response of the derived cell line (Fig. 1A, Wt11 cell line) (24). As a reporter for TLR3 mediated IRF3 activation, we used the promoter of human IFN stimulated gene 56 (ISG56) promoter driven firefly luciferase (ISG56-luciferase). ISG56 is the most well characterized member of the ISG56-family of proteins with translation inhibitory activity (28). In the uninduced condition ISG56 has undetectable level of expression, but it is one of the most highly induced genes after dsRNA treatment (29), thereby providing a very large dynamic range of induction. We co-transfected ISG56-luciferase plasmid along with Puromycin resistance plasmid (pBabePuro) into HEK293-TLR3 cells and selected for Puromycin resistance. After screening individual clones for luciferase activity following 6 h of dsRNA treatment, we picked one clone (C1), which showed highest inducibility with dsRNA treatment. One of the major obstacles to successful high-throughput screening for inhibitors in reporter based formats is the high number of ‘false positive hits’ due to the cytotoxicity of compounds in the library. To overcome the toxicity issue we integrated a constitutively active reporter (Renilla luciferase driven by HSV Thymidine Kinase promoter: pRLTK, Promega Corp.) in the C1 cells (Fig. 1A). The resulting clonal cell lines obtained were treated with dsRNA as well as dsRNA and LY294002 (a known inhibitor of ISG56 induction by dsRNA (24)) followed by dual-luciferase assay (Fig. 1B). As shown in Fig. 1B, all the clones showed dsRNA dependent induction of firefly luciferase activity, albeit at varying degrees, and inhibition of the same, in presence of LY294002. As expected, the Renilla luciferase activities remained same in all three conditions for each sample indicating that only the inducible ISG56 promoter driven transcription was specifically inhibited by LY294002 without affecting the constitutive TK promoter. We picked the clone RL24 for further characterization and screening purpose. The dose dependent induction of ISG56-luciferase activity was tested with increasing concentrations of dsRNA (Fig. 1C, solid line), while the Renilla luciferase activity remained same (Fig. 1C, broken line). We tested the effectiveness of using the Renilla luciferase activity as an indirect measure of cytotoxicity, by treating RL24 cells with cytotoxic doses of Staurosporine. This resulted in drastic reduction of Renilla luciferase activity (data not shown) confirming the validity of using Renilla luciferase activity as an indicator of cytotoxicity. Additionally, normalizing and expressing the firefly luciferase signal as a ratio to the Renilla luciferase signal effectively reduces the well to well variations in a plate while providing the means to flag cytotoxic false positives (Supplementary Fig. 1).

Fig. 1. Establishment of the cell-based assay system for high-throughput screening of TLR3-IRF3 signaling pathway.

(A) Schematic representation of sequential generation of RL24 stable reporter cell line. (B) Screening of individual stable clones containing dual reporters for dsRNA response in presence or absence of PI3K inhibitor LY294002. Individual clones were seeded in 96 well plates followed by 100 μg/ml dsRNA treatment in presence or absence of 5 μM LY294002. Firefly and Renilla luciferase activities were measured after 6 h with Dual-Glo luciferase kit. For each clone, luciferase values from both reporters were normalized against untreated firefly reporter values. Every sample was measured in triplicates and plotted as bar graph representing the mean and standard deviation. The clone RL24 was chosen for further characterization for its robust induction of ISG56-luciferase (firefly luciferase), characteristic inhibition of induction by LY294002, and steady constitutive expression of Renilla luciferase in all three experimental conditions. (C) Characterization of the selected cell clone RL24. RL24 cells showed dose-dependent firefly luciferase induction with increasing concentrations of dsRNA and constant levels of Renilla luciferase activities. Both luciferase activities were expressed as the fold inductions relative to the firefly luciferase values of non-stimulated cell controls.

Screening of small-molecule chemical libraries to identify TLR3-IRF3 pathway inhibitors and their Characterization

We used two commercially available small-molecule chemical libraries to screen for TLR3-IRF3 pathway modulators – the LOPAC¹²⁸⁰ library (1280 compounds) from Sigma-Aldrich, Inc. (St. Louis, MO) and the Spectrum Collection (2000 compounds) from Microsource Discovery Systems, Inc. (Gaylordsville, CT). Both of these libraries are composed of pharmacologically active compounds and natural products, with known biological profiles. Several screening parameters, such as cell seeding density, treatment duration, DMSO tolerance etc., were empirically optimized using Z’-factor calculations. Supplementary Fig. 1 shows a sample plate from the screening.

After the primary screen, the ‘positive hits’ were selected based on two criteria. The z scores – any compound that had z score either higher than + 2 or lower than – 2; and the Renilla luciferase activity that did not change beyond 90 % of the control. As shown in Supplementary Fig. 2A, based on these criteria we selected total 91 modifiers from our primary screen. Each of these 91 ‘primary hits’ were ‘cherry picked’, and used at three different concentrations in the same dual luciferase assay for secondary screening. Examples of positive and negative regulators in the secondary screen are shown in Supplementary Fig. 2B. After secondary screening, we looked for compounds with strong response, consistent dose dependence, and Renilla luciferase activities that remained consistently close to the control. Among the resulting modifiers, there were several known inhibitors of TLR3-IRF3 pathway including LY294002 and several non-specific protein tyrosine kinase inhibitors. These provided us with important ‘positive hits’ and demonstrated the robustness of the screening protocols. The resulting 31 compounds (Supplementary Table 1) were further analyzed according to their known functions (Tertiary screen) and we focused on a list of 5 inhibitors for further experiments (Table 1).

Table 1.

Final five compounds identified from our screen, corresponding z score and their known functions.

Compounds (Abbr. used)	z Score	IC50 (μM)	Known Function
Amlodipine Besylate (AB)	– 2.03	18.75	Calcium Channel Blocker, Antihypertensive
Fluphenazine (FLU)	– 2.35	15.25	H1 Antihistamine, Antipsychotic
SB224289 (SB)	– 4.77	3.5	Selective 5HT1B Serotonin Receptor Antagonist
Sertraline (SRT)	– 4.27	14.75	Selective Serotonin Reuptake Inhibitor, Antidepressant
Trifluoperazine (TFP)	– 4.24	16.5	Blocks Adrenergic, Dopamine, and Anticholinergic Receptors, Antipsychotic

The five inhibitors were further examined for their dose dependent inhibition of ISG56-luciferase activity at five different doses in order to obtain their corresponding IC₅₀ values (Fig. 2A). As expected, all five inhibitors showed reproducible inhibition of ISG56 induction by dsRNA with IC₅₀ values in the micromolar range. The inhibition of endogenous ISG56 protein induction was confirmed at two different concentrations of dsRNA treatments (1 and 10 μg/ml) in presence of 20 μM inhibitors by western blotting with ISG56 antibody (Fig. 2B and 2C).

Fig. 2. Specific Inhibition of TLR3-IRF3 signaling pathway by a selective panel of inhibitors identified from screening.

(A) IC₅₀ determination of tertiary hit compounds. RL24 cells were treated with the selected inhibitors Amlodipine Besylate (AB), Trifluoperazine (TFP), Fluphenazine (FLU), Sertraline (SRT) and SB224289 (SB) at the indicated concentrations for 1 h and then stimulated with 1 μg/ml dsRNA for another 6 h. Firefly luciferase activity was measured and expressed as fraction of no-inhibitor control. (B) and (C), Wt11 cells were either treated with 20 μM of above inhibitors, or with equivalent quantities of DMSO for 1 h, followed by stimulation with 1 or 10 μg/ml dsRNA for another 6 h. The cells were harvested and subjected to Western-blotting using anti-ISG56 and anti-β-actin antibodies respectively. The results were representative of two similar experiments.

SRT and TFP specifically inhibit IRF3 mediated signaling

Having established the inhibition of TLR3 mediated IRF3 dependent gene induction by all five inhibitors, we picked two specific inhibitors: Sertraline (SRT) and Trifluoperazine (TFP) for further characterization of their specificity and mechanism/s of inhibition. Wt11 cells were treated with dsRNA in presence of 10 μM of each inhibitor, and the effects of these inhibitors on endogenous mRNA induction by dsRNA, were tested by real-time qPCR (Fig. 3A). IRF3/IRF7 dependent genes: ISG56, ISG60, IFNβ and IFNα were strongly downregulated, whereas, IL-8 transcription, a representative NF-κB regulated endogenous gene, was largely unaffected. The inhibitor-dose dependence of endogenous ISG56 protein induction by dsRNA was tested in Wt11 cells (Fig. 3B and C top panels), showing expected reduction of ISG56 levels with increasing inhibitor concentrations. Besides IRF3, NF-κB is strongly activated by TLR3 activation with dsRNA. We tested one of the early steps of NF-κB activation by monitoring IκBα levels after dsRNA stimulation. As shown in Fig. 3B and 3C bottom panels, in the resting condition substantial quantities of IκBα was present in the whole cell lysates of Wt11cells. Following 50 min stimulation with dsRNA, IκBα is phosphorylated and degraded (second lanes Fig. 3B and 3C bottom panels) signifying the release of NF-κB for nuclear translocation and transcriptional activation. However, upon treatment with increasing concentrations of SRT and TFP, there was no significant inhibition of IκB degradation (Fig. 3B and 3C bottom panels, lanes 3 to 5). This reconfirms our findings from qPCR experiments that these inhibitors do not affect NF-kB mediated gene induction by dsRNA. In fact, when tested IL-8 protein induction by dsRNA, there was no significant inhibition of IL-8 induction in Wt11 (Fig. 3D) and HT1080 cells (data not shown) after SRT or TFP treatment.

Fig. 3. Effects of SRT and TFP on TLR3 signaling pathways.

Total RNA was isolated from Wt11 cells pretreated for 1 h with SRT (10 μM) and TFP (10 μM) followed by 10 μg/ml dsRNA for 6h. Real-time qPCR were performed on reverse transcribed cDNA obtained from the total RNA with specific primer for ISG56, ISG60, IFNβ, IFNα and IL-8 (A). Wt11 cells were pretreated for 1 h with SRT (B) and TFP (C) at the indicated concentrations, then stimulated with 10 μg/ml dsRNA for another 6h or 50 min. The 6h dsRNA treated cell lysates were western blotted with anti-ISG56 antibody to measure IRF3 mediated gene induction (upper panel, NS, non-specific) and anti-IκBα antibody to measure NF-κB activation (lower panel). Each blot was reprobed with β-actin antibody to establish equal loading. (D) Wt11 cells were pretreated with SRT or TFP at the indicated concentrations for 1 h, and then stimulated with 10 μg/ml dsRNA for another 24 h. Cell culture supernatants were harvested and assayed for IL-8 production by ELISA. The bars represent mean ± SD of triplicate samples.

SRT and TFP inhibit IRF3 mediated signaling in primary dendritic cells

In order to validate our findings in a physiologically relevant system, we tested the effects of SRT and TFP on dsRNA mediated gene induction in human primary dendritic cells (DC). Monocyte derived dendritic cells were treated with dsRNA in presence of different doses of SRT and TFP followed by detection of ISG56 and ISG60 (another member of ISG56 family proteins (28)) protein induction by western blotting (Fig. 4A). As expected, dsRNA mediated ISG56 and ISG60 protein induction in DC was strongly inhibited by SRT and TFP. Fig. 4B shows similar results where the transcriptional induction of IFNβ mRNA was monitored by real-time qPCR in presence of SRT and TFP.

Fig. 4. Effects of SRT and TFP on TLR3 and TLR4 signaling pathways in primary dendritic cells.

Human dendritic cells were pretreated with SRT or TFP for 1 h , then stimulated with either 50 μg/ml dsRNA for 6 h (A) or 1μg/ml LPS for 20 h (C). Cell lysates were western blotted for ISG56 and ISG60 induction. Total RNA isolated from dendritic cells treated with 10 μM inhibitors and 50 μg/ml dsRNA were reverse transcribed and subjected to real-time qPCR using IFNβ specific primers (B). The results were representative of two similar experiments.

Besides TLR3, IRF3 is also activated by other TLR signaling pathways. The most well characterized one among them is the activation of IRF3 by LPS via TLR4 signaling pathway. This pathway is TRIF dependent but MyD88 independent. Upon activation of TLR4 by LPS, TLR4 activates IRF3 kinases TBK1/IKKε via TRIF and TRAM dependent signaling pathway (1). We examined the effects of SRT and TFP on LPS mediated IRF3 activation by measuring ISG56 and ISG60 protein induction by LPS in human primary dendritic cells. As shown in Fig. 4C, both ISG56 and ISG60 were strongly induced in human DC after LPS treatment (Fig. 4C, second lanes). However, their inductions were inhibited after treatment with SRT and TFP in a dose dependent manner.

SRT and TFP selectively dampen the TLR7/8 medicated type I interferon induction

Similar to TLR3, TLR7 and TLR8 are also involved in ribonucleic acid recognition. Upon induction by their cognate ligands such as ssRNA or nucleotide analogue R848, they mediate NF-κB activation via adaptor MyD88 (1). We investigated whether SRT and TFP can specifically affect TLR7 and TLR8 mediated NF-κB signaling (Fig. 5A, 5B and 5C). HEK293 cells stably expressing TLR7 or TLR8 were co-transfected with NF-κB-luciferase reporter plasmid along with Renilla luciferase for 24 h. Transfected cells were pre-treated with increasing concentrations of SRT and TFP followed by 8 h treatment with 2 or 20 μg/ml of R848. Results from the dual luciferase assays performed at the end of the R848 treatment showed that there were no inhibition of either TLR8 (Fig. 5A) or TLR7 (Fig. 5B) mediated NF-κB activation by SRT or TFP. In a similar experiment we found that SRT and TFP do not affect NF-κB activation when activated through TLR5 using flagellin (Supplementary Fig. 3). In order to validate these results for endogenous NF-κB mediated gene induction by TLR7/8, we performed similar experiments using human dendritic cells. Human dendritic cells were treated with 20 μg/ml R848 for 16 h in presence of indicated amounts of SRT and TFP followed by IL-8 ELISA on culture supernatants. Fig. 5C shows that there was no inhibition of IL-8 induction by R848 in dendritic cells. In addition to NF-κB, activation of TLR7 and TLR8 in human macrophage and dendritic cells lead to type I IFN production via a poorly understood IRF mediated pathway. Upon stimulation of TLR7/8 ligand, R848, the DCs were able to induce, albeit low levels, ISG56 and ISG60 (Fig. 5D, lane 2). Pre-treatment with SRT and TFP showed dose dependent inhibition of both ISG56 and ISG60 induction by R848 (Fig. 5D). Together, these results suggest that SRT and TFP inhibit TLR7/8 mediated IRF response and type I IFN induction without affecting NF-κB signaling.

Fig. 5. Effects of SRT and TFP on TLR7/8 signaling.

TLR8 (A) and TLR7 (B) expressing cells were co-transfected with NF-κB firefly luciferase reporter along with β-actin Renilla luciferase reporter for 24 h. Transfected cells were pre-treated with SRT and TFP at the indicated concentrations for 1h, followed by 20 μg/ml (TLR8) and 2 (TLR7) R848 stimulation for another 8 h. NF-κB activation was expressed as fold induction relative to non-stimulated cells after normalizing for transfection efficiencies with Renilla luciferase. The bars represent mean ± SD of triplicate samples. (C) Human dendritic cells were pretreated with SRT and TFP at the indicated concentrations for 1 h , then stimulated with 20 μg/ml R848 for another 18 h. Cell supernatants were collected and assayed for IL-8 by ELISA. The bars represent mean ± SD of duplicate samples. (D) Human dendritic cells were treated and stimulated as above and cell lysates were analyzed for ISG56 and ISG60 induction by western blotting. The results were representative of two similar experiments.

SRT and TFP do not inhibit RIG-I mediated signaling pathway

In addition to TLR3, which senses dsRNA in endosomal vesicles, cytoplasmic dsRNA is sensed by DExD/H box RNA helicases: RIG-I (retinoic acid-inducible gene I) and MDA5 (1). Upon binding to dsRNA, these receptor recruits the adaptor protein IPS-1 (also known as MAVS, CARDIF and VISA) and signal downstream to IRF3 and NF-κB. A number of RNA virus activate the IFN induction via RIG-I mediated signaling pathway (30). We wanted to examine the effects of SRT and TFP on cytoplasmic dsRNA signaling pathways using Sendai virus (SeV) infection mediated induction of ISG56 as a model system. However, SeV infected cells underwent rapid apoptosis mediated cell death in presence of SRT and TFP precluding us from studying the effects of these inhibitors on RIG-I mediated signaling (data not shown). A similar observation has been made before, where pretreatment of cells with PI3K inhibitor LY294002 followed by SeV infection led to rapid apoptosis of cells via IRF3 dependent mechanism (31). This observation is consistent with the phenotypes observed with SRT and TFP and as we shall see below the inhibitory effects of SRT and TFP are at least partially mediated via PI3K signaling pathway.

Thus, we used an alternative approach to examine the RIG-I pathway. It has been observed that heterologous expression of constitutively active RIG-I, RIG-I-N-terminal fragment can activate the RIG-I signaling pathway leading to IFN induction (4). We used HEK293 cells to transiently express RIG-I-N-terminal and tested ISG56-luciferase induction in presence of SRT and TFP. As shown in Fig. 6A, SRT and TFP did not inhibit ISG56-luciferase activation, tested at two different RIG-I expression levels, indicating that these inhibitors specifically affect TLR3 mediated activation of IRF3.

Fig. 6. The RIG-I signaling pathway and downstream components of IRF3 activation are not affected by SRT and TFP.

HEK293 cells were co-transfected with ISG56-luciferase and β-actin Renilla luciferase, together with RIG-I (A) and IRF-3-5D mutant (B) for 24 h. Following transfection, cells were treated with SRT or TFP at the indicated concentrations for another 8 h. Cell lyasates were analyzed for firefly and Renila luciferese activities using the Dual Luciferase Reporter Assay System. The results were corrected for transfection efficiencies with Renilla luciferase and expressed as fold induction of ISG56 luciferase for RIG-I or IRF-3-5D transfected cells relative to those of vector transfected controls. The results were representative of two similar experiments.

Although there has been evidence of multiple post translational modifications of IRF3, it is primarily activated by Ser/Thr phosphorylation as evidenced by generating constitutively active IRF3 where all five important phosphorylation sites have been mutated to aspartate residues (IRF3-5D) (32). We used the IRF3-5D mutant in transient transfection assay to test if the inhibitors affect transcriptional activity of IRF3 downstream of its phosphorylation. As shown in Fig. 6B, co-transfection of IRF3-5D, at two different DNA concentrations, with ISG56-luciferase in HEK293 cells induced ISG56 promoter activity. However, SRT or TFP did not inhibit the induction indicating that these inhibitors do not affect steps downstream of IRF3 phosphorylation.

SRT and TFP inhibit PI3K signaling pathway

In the resting condition the transcription factor IRF3 shuttles between nucleus and cytoplasm, predominantly being present in the cytoplasm. Upon activation it is phosphorylated and translocated to the nucleus to induce transcription of ISRE containing promoters. In order to decipher the mechanism of action of SRT and TFP, we assayed this downstream signaling event by monitoring the IRF3 levels in nuclear fractions of dsRNA treated and untreated cells. Upon 2 h treatment with 10 μg/ml dsRNA, the amount of IRF3 increased substantially in the nuclear fraction (Fig. 7A, lane 2 and lane 3). However, in presence of inhibitory doses of SRT and TFP, there was no inhibition of IRF3 translocation. This indicates that in presence of the inhibitors IRF3 is partially activated, and it is able to translocate to the nucleus, but unable to support gene transcription. Previously, we have reported similar phenotype for IRF3, where TLR3 mediated PI3K activation was shown to be essential for complete activation of IRF3 (24). We found that dsRNA-TLR3 mediated activation of IRF3 involved a two step process. In the first step, upon activation by dsRNA, IRF3 is phosphorylated and translocates to the nucleus. However, in the absence of PI3K activation via specific tyrosine phosphorylation of TLR3, the nuclear IRF3 is unable to form a stable complex at ISG56 promoter and does not support transcriptional induction. Additional modifications of IRF3, via a second step, mediated by PI3K are needed to achieve complete transcriptionally active IRF3. Therefore we tested TLR3 mediated PI3K activation status in the presence of SRT and TFP. PI3K activation was measured by monitoring its target AKT phosphorylation by immunoblotting dsRNA treated cell lysates with anti-phospho-AKT antibody. As expected, upon dsRNA treatment AKT is phosphorylated via PI3K pathway (Fig. 7B and 7C). However, in the presence of 20 μM of either SRT or TFP, dsRNA mediated AKT phosphorylation is completely inhibited. This indicates that SRT and TFP inhibit PI3K signaling pathway to exert their inhibitory effects on TLR3-IRF3 signaling.

Fig. 7. Both SRT and TFP blocks dsRNA mediated AKT activation.

(A) Wt11 cells were treated for 1 h with 20 μM SRT and TFP respectively, and then stimulated with 10 μg/ml dsRNA for another 2 h as indicated. Nuclear fractions were isolated form treated and untreated control cells and western blotted with anti-IRF3 antibody (top panel). A control lane with whole cell extract (WCE) from untreated cells was included in the same blot. Following IRF3 detection, the same membrane was reprobed with DRBP76 (nuclear protein) and tubulin (cytoplasmic protein) antibodies to demonstrate integrity of the nuclear preparation. (B) and (C) HT1080 cells were treated or left untreated with 20 μM SRT or 20 μM TFP for 1 h, followed by stimulation with 10 μg/ml dsRNA for 1, 2 and 3 h, respectively. Cell lysates were probed with anti-Ser⁴⁷³-phospho-AKT (pAKT) and anti-AKT antibodies respectively by Western-blotting. The results were representative of two similar experiments. (D) SRT and TFP do not block TRIF mediated activation of IRF3. HA tagged TRIF expression construct was co-transfected with ISG56-luciferase and β-actin Renilla luciferase to HEK293 cells for 24 h. Following transfection, cells were treated with SRT or TFP at the indicated concentrations for another 8 h. Cell lyasates were analyzed for firefly and Renila luciferese activities as described in Fig. 6.

To confirm the above findings that SRT and TFP exert their inhibitory effects on IRF3 solely via PI3K pathway, we used TRIF mediated activation of IRF3. The overexpression of adaptor protein TRIF itself can activate IRF3 mediated gene induction without requiring TLR3 or dsRNA and independent of PI3K activity (24). As it is shown in Fig. 7D, the TRIF mediated activation of ISG56-luciferase is not inhibited by SRT or TFP. This confirms that the inhibition described here may be PI3K mediated, and is only seen when IRF3 is activated using dsRNA as the ligand for TLR3.

Discussion

Using a cell-based screening approach we have identified novel anti-innate immune properties of several well established antipsychotic drugs. We have shown that these compounds exhibit specific inhibitory properties against IRF3 dependent gene induction mediated by several TLRs without affecting the NF-κB activation. Mechanistic studies revealed a possible role of PI3K in mediating the inhibition. Furthermore, we have established a robust and highly specific cell-based screening system which can be used for further broad screening and identification of TLR signaling pathway modulators.

Double stranded RNA via TLR3 or RIG-I triggers at least four signaling pathways leading to the activation of the transcription factors N-κB, IRF3, c-Jun and ATF-2 (5). All of these factors together drive transcription of several complex cytokine gene promoters like the IFNβ gene (33). In contrast, the promoter of the ISG56 gene contains only IFN-stimulated response elements (ISRE) to which activated IRF3 or IRF7 binds. Activation of neither NF-κB nor ATF-2 is necessary for ISG56 gene induction by dsRNA (13). Although ISG56 is an IFN stimulated gene (IFN-β/α mediated induction of ISG56 is independent of IRF3, and driven by ISGF3 complex composed of STATs and IRF9) and dsRNA treatment induce synthesis of IFNs, HEK293 cells do not respond to type I IFNs due to the lack of receptors. This prevents the secondary autocrine induction of ISG56 by IFNs, and eliminates the likelihood of picking up IFN pathway inhibitors in the screen. Moreover, because ISG56 activation is only dependent on IRF3 or IRF7, it further limits the possibility of identifying generalized inhibitors of NF-κB or c-Jun in the screen. Therefore, as we have demonstrated here, the ISG56 promoter driven luciferase can be used as a reporter to successfully identify specific modulators of IRF3 activation in high throughput screens. Additionally, our strategy of integrating a constitutively active promoter-driven Renilla luciferase in the primary screen was also very successful in effectively reducing the number of cytotoxic agents and luciferase activity inhibitors as false positive hits. We used a sub-maximal dose of dsRNA, (1 μg/ml poly(I): poly(C)) in order to identify both enhancers and inhibitors of TLR3-IRF3 signaling pathway (Supplementary Table 1). However, as shown in Fig. 1C and Supplementary Fig. 1, using Renilla luciferase normalization we were consistently able to achieve more than 20 fold induction of ISG56-luciferase activity with minimum sample-to-sample variability (Z’≥0.7).

The findings reported here that several commonly used antipsychotic drugs specifically affect TLR signaling has major implication in therapy and treatments. An important extrapolation of these findings would be that patients who are treated with any of these antidepressant or antipsychotic drugs might have changed susceptibility for virus infection. As discussed before, depending on the specific virus infection, TLR3 signaling can be protective or deleterious for the host. Similarly, IRF3 is primarily associated with anti-viral interferon induction, but recent evidences suggest its pro-inflammatory function via TNFα induction (34). Therefore, depending on the particular virus infection, selective serotonin reuptake inhibitors (SSRI), such as Sertraline, may have either beneficial or harmful effects on the host. Some of the anti-inflammatory and immunomodulatory effects of SSRIs have been found empirically (35-37). Our studies provide confirmation of these results and a link to the mechanism of such immunomodulatory activities of SSRIs.

The involvement of PI3K in TLR signaling has been well established (38). Depending on the particular TLR ligand and the cell systems that have been used, PI3Ks have been shown to exhibit either negative or positive regulatory effects on TLR signaling. We show here that the SRT and TFP mediated inhibition of TLR3-IRF3 signaling pathway may be mediated by inhibition of PI3K-AKT signaling pathway (Fig. 7). It is in accord with our previous observation that PI3K plays an essential role in complete activation of IRF3 via TLR3. Inhibition of PI3K, either directly by LY294002 or in this case, indirectly by SRT/TFP, results in abolition of IRF3 signaling without affecting NF-κB signaling. It is likely that similar mechanism may be involved in case of IRF inhibitory action of these inhibitors on TLR7/8 signaling pathways (39, 40). Additionally, it may involve signaling molecules upstream of PI3K pathways such as Bruton's tyrosine kinase (41). The nature of IRF3 inhibition in TLR4 signaling pathway by SRT and TFP may be more complicated. Several reports have suggested either positive or negative effects of PI3K activation on TLR4 signaling pathway (42-45). However, in epithelial cells the LPS mediated IFN-β induction via IRF3 is dependent on PI3K activity (46). Thus, it is likely that SRT and TFP mediated inhibition of IRF3 activation in case of TLR4 signaling pathway, follows a very similar mechanism as that of TLR3.

One question that remains to be elucidated is the biochemical basis of PI3K inhibition by SSRIs. In spite of being highly effective in treating depression and other psychotic disorders, the exact mechanism of action for SSRI and antipsychotic drugs remains elusive. SSRIs bind directly to the serotonin-transporter protein, a member of the neurotransmitter:sodium symporter family proteins, and inhibit neurotransmitter recycling. However, they also bind to the homologous norepinephrine and dopamine transporters with lower affinity. Only recently we have started to understand their binding specificities from crystallographic studies on the SSRI and homologous bacterial symporter complexes (47, 48). In order to confirm that these inhibitors are working through their target proteins in our experimental system, we determined the expression levels of serotonin transporter and serotonin receptors in HEK293 and human dendritic cells by RT-PCR. Supplementary Fig. 4 shows that both the experimental cell types express serotonin receptor and serotonin transporter mRNAs indicating that the inhibitors are most likely working through their cognate receptors. The recognition of dsRNA by TLR3 occurs in the endosomal vesicles. Inhibition of clathrin mediated endocytosis or lysosomal acidification shows marked inhibition in TLR3 mediated gene induction (49). Therefore, it remains a possibility that these inhibitors affect the TLR3 mediated PI3K activation by interfering with any of the above steps.

In summary, in this study, we established a high throughput cell-based screening assay and applied it to screen TLR3-IRF3 specific inhibitors. Several antipsychotic drugs emerged as specific TLR3-IRF3 inhibitors, and also selectively inhibited TLR4 and TLR7/8 mediated type I IFN induction by inhibiting PI3K activation. Our study provides mechanistic insight on the anti-inflammatory effects of antipsychotic drugs and serves as a model for screening relevant inhibitors for certain TLR signaling pathway.

Abbreviations

Poly(I):poly(C)

Polyinosinic acid:polycytidylic acid

dsRNA

double stranded RNA

ssRNA

single stranded RNA

IFN

Interferon

NF-κB

Nuclear Factor κB

IRF

Interferon Regulatory Factor

TLR

Toll-like Receptor

ISG

Interferon Stimulated Gene

ISRE

Interferon Stimulated Response Element

PI3K

phosphatidylinositol 3 kinase

TRIF

TIR-domain-containing adapter-inducing interferon-β

RIG-I

Retinoic acid-inducible gene I

RPL32

Ribosomal Protein L32

SeV

Sendai virus

SSRI

Selective Serotonin Reuptake Inhibitor

SRT

Sertraline

TFP

Trifluoperazine