Toll-Like Receptor Engagement Enhances the Immunosuppressive Properties of Human Bone Marrow-Derived Mesenchymal Stem Cells by Inducing Indoleamine-2,3-dioxygenase-1 via Interferon-β and Protein Kinase R†‡
Christiane A. Opitz,
Department of Neurooncology, University Hospital of Heidelberg and German Cancer Research Center, Heidelberg, Germany
Department of General Neurology and Hertie Institute for Clinical Brain Research, University of Tübingen, Tübingen, Germany
Author contributions: C.A.O.: designed and performed research, analyzed data, and wrote the article; U.M.L. and C.L.: performed research and analyzed data; T.V.L., I.T., and A.K.: performed research; E.T.: designed research and contributed vital new reagents; M.H., J.A., and W.K.A.: contributed vital new reagents and performed research; M.W.: designed research; W.W.: designed research and wrote the article; M.P.: designed research, analyzed data, and wrote the article.
First published online in STEM CELLSExpress January 15, 2009.
Mesenchymal stem cells (MSC) display unique suppressive properties on T-cell immunity, thus representing an attractive vehicle for the treatment of conditions associated with harmful T-cell responses such as organ-specific autoimmunity and graft-versus-host disease. Toll-like receptors (TLR) are primarily expressed on antigen-presenting cells and recognize conserved pathogen-derived components. Ligation of TLR activates multiple innate and adaptive immune response pathways to eliminate and protect against invading pathogens. In this work, we show that TLR expressed on human bone marrow-derived MSC enhanced the immunosuppressive phenotype of MSC. Immunosuppression mediated by TLR was dependent on the production of immunosuppressive kynurenines by the tryptophan-degrading enzyme indoleamine-2,3-dioxygenase-1 (IDO1). Induction of IDO1 by TLR involved an autocrine interferon (IFN)-β signaling loop, which was dependent on protein kinase R (PKR), but independent of IFN-γ. These data define a new role for TLR in MSC immunobiology, which is to augment the immunosuppressive properties of MSC in the absence of IFN-γ rather than inducing proinflammatory immune response pathways. PKR and IFN-β play a central, previously unidentified role in orchestrating the production of immunosuppressive kynurenines by MSC. STEM CELLS 2009;27:909–919
Mesenchymal stem cells (MSC) are multipotent cells that can differentiate along multiple lineages into osteoblasts, adipocytes, chondrocytes, tenocytes, myocardiocytes, myoblasts, hepatocytes, and cells with neuronal properties [1–4]. As stromal cells MSC represent integral cellular components of the bone marrow and support the ex vivo culture of hematopoietic stem cells (HSC) by supplying extracellular matrix components, growth factors, and cytokines . MSC support the survival, self-renewal, migration, expansion, and differentiation of HSC , contribute to the stem-cell niche , and may participate in the repair of damaged and inflamed tissue .
Cotransplantation of MSC with HSC may improve HSC function and reduces the incidence and severity of graft-versus-host disease . The immunomodulatory function of MSC is now well-recognized . MSC suppress inflammatory reactions by inhibiting T-cell responses [11, 12] and maturation of antigen-presenting cells (APCs) . The mechanisms of MSC-dependent immunosuppression remain controversial, and may include the secretion of transforming growth factor-β (TGF-β) , hepatocyte growth factor , prostaglandin E2 , nitric oxide , and the interferon (IFN)-γ-induced degradation of tryptophan [14, 18–20].
Toll-like receptors (TLR) are key molecules bridging innate and adaptive immune responses . TLR recognize multiple pathogen-associated molecular patterns (PAMPs), including bacterial lipoproteins and lipoteichoic acids (TLR1, TLR2, TLR6), lipopolysaccharide (LPS) (TLR4), flagellin (TLR5), the unmethylated CpG DNA of bacteria and viruses (TLR9), double-stranded RNA (dsRNA, TLR3) and single-stranded viral RNA (TLR7, TLR8) .
After sensing microbial infections by recognizing conserved PAMPs, TLR trigger multiple steps in the inflammatory cascades that help eliminate invading pathogens. Recent reports demonstrate that TLR ligation modulates the proliferation and differentiation of both human and mouse MSC, indicating that TLR are functional in MSC biology [22, 23]. In addition, TLR activation influences the immunological and migratory behavior of MSC [24, 25]. In the present study, we sought to investigate the effect of TLR activation on the immunosuppressive properties of MSC and the molecular pathways involved in TLR-mediated immune modulation.
MATERIALS AND METHODS
Cell Culture and Reagents
MSC were obtained from bone marrow from total hip replacement surgeries of nine different patients following informed consent. After density gradient centrifugation, MSC isolated by plastic adherence were grown in basal medium for human MSC with 10% stimulatory supplement (CellSystems, St. Katharinen, Germany, http://www.cellsystems.de). Passages 4-11 were used for experiments.
Peripheral blood mononuclear cells (PBMCs) were isolated from five healthy blood-donors by density-gradient centrifugation using lymphocyte separation medium 1077 (PAA Laboratories GmBH, Pasching, Austria, http://www.paa.at). For the generation of dendritic cells (DCs), CD14+ cells were separated using the magnetic activated cell sorting (MACS) technology (Miltenyi Biotec, Bergisch Gladbach, Germany, http://www.miltenyibiotec.com) and plated in Roswell Park Memorial Institute (RPMI) 1640 (Cambrex, Verviers, Belgium, http://www.cambrex.com) containing 10% fetal bovine serum (FBS), 200 IU/ml interleukin (IL)-4, and 1,000 IU/ml granulocyte macrophage-colony stimulating factor (Immunotools, Friesoythe, Germany, http://www.immunotools.de) for 7 days.
Twenty millimolar stock solutions of 1-methyl-D-tryptophan (1-D-MT) and 1-methyl-L-tryptophan (1-L-MT) (Sigma-Aldrich, Taufkirchen, Germany, http://www.sigmaaldrich.com) were prepared by dissolving the inhibitors in 0.1 N NaOH. The pH was adjusted to 7.5 using hydrochloric acid. To avoid contamination of the cell cultures, the stock solutions were filtered through 0.2-μm filters.
Resveratrol (Sigma-Aldrich) dissolved in dimethyl sulfoxide (DMSO) was used at a final concentration of 50 μM. MSC were stimulated with 50 μg/ml poly(cytidylic-inosinic) acid (pI:C; Sigma-Aldrich) and treated with either 50 μM resveratrol or the respective DMSO control.
To inhibit the binding of IFN-β, 5 × 105 MSC were incubated with 1 μg/ml anti-interferon-β receptor 1 (IFNAR1; abcam, Cambridge, U.K., http://www.abcam.com) or isotype control 30 minutes before stimulation with 50 μg/ml pI:C. After 24 hours supernatants were harvested and analyzed by high performance liquid chromatography (HPLC).
To inhibit protein kinase R (PKR) the classical inhibitor 2-aminopurine (2-AP; Sigma-Aldrich), and the RNA-dependent protein kinase inhibitor (PKRI; C13H8N4OS, Merck KGaA, Darmstadt, Germany, http://www.merck-chemicals.com) were used. Cells were pretreated with RPMI with a final concentration of 4 mM 2-AP or 300 nM PKRI 1 hour before stimulation with either 50 μg/ml pI:C or 5 μg/ml LPS (Sigma-Aldrich). The respective solvents, either 0.05% (v/v) acetic acid in phosphate-buffered saline for 2-AP or 1.1% (v/v) DMSO in PBS for PKRI, served as controls. All cultures were incubated at 37°C in a 5% CO2 atmosphere.
Osteogenic Differentiation of MSC
Osteogenic differentiation of the MSC was performed in D-MEM (PAA Laboratories) medium containing 10% FCS, 100 U/ml penicillin and streptomycin, 15 mM N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid, 10 mM glycerol-2-phosphate disodium salt hydrate, 0.1 mg/ml ascorbic acid and 0.1 μM dexamethasone (Sigma-Aldrich). Dexamethasone (0.1 μM) was supplemented twice a week. Cells were stained with Alizarin Red S solution (aqueous 2% Alizarin Red, pH 4.2; Sigma-Aldrich) after 4 weeks of culture.
Karyotyping of MSC
Karyotype analyses of each MSC donor were performed in AmnioMax Medium (Invitrogen, Carlsbad, http://www.invitrogen.com) as described .
Flow cytometry was performed using a CyAn ADP flow cytometer (Dako Cytomation, Glostrup, Denmark, http://www.dakocytomation.com). MSC were either left untreated or treated with 50 μg/ml pI:C or 5 μg/ml LPS 24 hours before flow cytometry. MSC were detached using 5 mM ethylenediaminetetraacetic acid (EDTA) buffer, washed, and resuspended at 105 per 100 μl in PBS + 1% BSA, followed by incubation with the specific antibody at 4°C. Cells were washed with PBS + 1% BSA and analyzed by flow cytometry. Antibodies against human CD29-FITC, CD44-FITC, CD105-PE, CD34-PE (eBioscience, San Diego, CA, http://www.ebioscience.com), CD14-FITC (Acris Antibodies; GmbH, Hiddenhausen, Germany, http://www.acris-antibodies.com), CD80-PE-Cy5, CD86-Pacific Blue, human leukocyte antigen (HLA1)-PE-Cy7 and HLA-DR-PE-Cy5 (BioLegend, San Diego, CA, http://www.biolegend.com) were used.
Mixed Leukocyte Reaction
MSC were seeded in flat-bottom 96-well plates in RPMI 1640 (Cambrex, Verviers, Belgium) containing 10% FBS, 100 U/ml penicillin, and 100 μg/ml streptomycin. Twenty-four hours after seeding MSC were pretreated with 50 μg/ml pI:C or 5 μg/ml LPS (Sigma-Aldrich). After another 24 hours MSC were washed three times with PBS (PAA Laboratories) and 2 × 105 irradiated (30 Gy) PBMC as stimulators and 2 × 105 PBMC from unrelated donors as responders were added. When using 1-L-MT, it was added to the cultures at 1 mM immediately after addition of the PBMC. Six-day mixed leukocyte reactions (MLR) were performed and cultures were pulsed with [3H]-methylthymidine (Amersham Radiochemical Centre, Buckinghamshire, U.K.) for the last 18 hours. The cells were then harvested, and radionuclide uptake was measured by scintillation counting. The counts of MSC without PBMC were subtracted from the counts of MSC/MLR cocultures to exclude the proliferation of MSC from the measurements. No differences in proliferation between untreated and TLR-activated MSC in the absence of PBMC were observed. For MLR with MSC supernatant indicated numbers of MSC were seeded in flat bottom 96-well plates, 24 hours after seeding MSC were treated with 50 μg/ml pI:C or 5 μg/ml LPS. After another 24 hours MSC were washed three times and fresh RPMI 1640 was added. After 7 days media was harvested. MLR as described earlier were then performed in this media.
Total RNA was isolated with the Qiagen RNAeasy RNA isolation kit (Qiagen, Hilden, Germany, http://www.qiagen.com) and cDNA was synthesized with the SuperscriptTM Choice System (Invitrogen) using random hexamers. Primers were designed across exon-boundaries and provided by Invitrogen. Quantitative reverse transcription polymerase chain reaction (qRT-PCR) was preformed in an ABI 7000 thermal cycler with SYBR Green PCR Mastermix (Applied Biosystems, CA, http://www.appliedbiosystems.com) according to standard protocols. PCR reactions were checked by including no-RT-controls, by omission of templates and by both melting curve and gel analysis. The size of the amplicons was analyzed by loading the samples and a 100 bp ladder (Invitrogen) on a 2% agarose gel, which was then stained with ethidium bromide and analyzed under UV light. To exclude amplification of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) pseudogenes, GAPDH primer sequences have been described . Standard curves were generated for each gene and the amplification was 90%–100% efficient. Relative quantification of gene expression was determined by comparison of threshold values. All results were normalized to GAPDH, which varied neither with pI:C and LPS treatment nor between MSC and DC. For TLR3 and TLR4 stimulation, cells were treated with 50 μg/ml pI:C and 5 μg/ml LPS for 24 hours, respectively.
HPLC analysis was performed as described  using a Beckman HPLC with photodiode array (PDA) detection and Lichrosorb RP-18 column (250 × 4 mm2 ID, 5 μm; Merck, Darmstadt, Germany). Kynurenine release and tryptophan degradation were measured in RPMI 1640 (Cambrex) containing 10% FBS (Perbio, http://www.perbio.com), 100 U/ml penicillin and 100 μg/ml streptomycin (PAA Laboratories) supplemented with 50 μg/ml tryptophan (Sigma-Aldrich). The medium was harvested from 24-well plates at the indicated time points, centrifuged and frozen until further analysis. After thawing, the samples were supplemented with trichloroacetic acid for protein precipitation, centrifuged and 100 μl of the supernatant was analyzed by HPLC. Standard curves were generated with L-kynurenine and L-tryptophan (Sigma-Aldrich) in the same medium. As FBS contains kynurenine, low kynurenine concentrations (∼1 μM) were detected in all samples and medium without cells, which was always measured for comparison.
Western Blot Analysis
Whole cell lysates were prepared in ice-cold tris(hydroxymethyl)aminomethane hydrochloride (TRIS-HCl, 50 mM, pH 8,0; Carl Roth, Karlsruhe, Germany, http://www.carl-roth.de) containing 150 mM NaCl (J.T. Baker, Deventer, Holland), 1% Triton X-100 (AppliChem, Darmstadt, Germany), 10 mM EDTA (Gerbu Biotechnik, Gaiberg, Germany, http://www.gerbu.de), 200 mM dithiothreitol (Carl Roth), 3% 2-mercaptoethanol (Sigma), 100 μM phenylmethylsulphonyl fluoride (PMSF), 10 μg/ml aprotinin and 5 μg/ml leupeptin (Carl Roth). After sonication, the lysates were centrifuged at 4°C (10 min, 13,000 rpm). The protein concentration of the supernatants was determined using the Bio-Rad protein assay (Bio-Rad, Hercules, CA, http://www.bio-rad.com) at 595 nm. The desired amount of protein (20 μg per lane) was separated by 10% SDS-PAGE and transferred to a 0.2-μm-pore nitrocellulose membrane (Whatman, Dassel, Germany). After 1 hour of blocking in PBS supplemented with 0.2% Tween 20 (Sigma-Aldrich) and 5% skim milk powder (Carl Roth), the membrane was incubated with rabbit anti-IDO1 antibody (1:2,000; Alexis, Lausen, Switzerland, http://www.alexis-biochemicals.com), anti-STAT1 antibody (α + β), anti-pSTAT1 (Y701, α + β) antibody (1:1,000; both Cell Signaling Technology, http://www.cellsignal.com), or goat anti-GAPDH (1:2,000; Abcam, Cambridge, U.K.) as loading control, over night at 4°C. After a 2-hour incubation at room temperature with secondary antibodies IRDye800 anti-rabbit (1:10,000) or IRDye800 anti-goat (1:10,000; both Rockland, Gilbertsville, PA, http://www.rockland-inc.com), protein detection was performed using an Odyssey LI-COR scanner (Lincoln, NE, http://www.lincolnlaser.com).
To knockdown IDO1 (INDO), STAT1, and IFN-β, ON-TARGETplus SMART-pool siRNA by Dharmacon RNA Technologies (Lafayette, CO, http://www.dharmacon.com) was used.
ON-TARGETplus siCONTROL Nontargeting Pool (D-001810-10-05, Dharmacon) and a transfection without siRNA were used as negative controls.
For transfection of siRNA, the Amaxa Human MSC Nucleofector Kit (Amaxa biosystems, Koeln, Germany, http://www. amaxa.com) was used. Briefly, 5 × 105 cells were resuspended in 100 μl of the human MSC nucleofector solution and mixed with 1.5 μg of siRNA, then electroporated using program U-23. Cells were immediately transferred into 37°C prewarmed culture medium (RPMI 1640) and plated into 6-well plates, containing 1.5 ml prewarmed RPMI. After 24 hours the cells were treated with 50 μg/ml pI:C or 5 μg/ml LPS for additional 24 hours. The supernatants and the cells were harvested, the knockdown was analyzed by qRT-PCR and the kynurenine and tryptophan content of the medium was measured by HPLC.
Enzyme-Linked Immunosorbent Assay
Human IFN-β, IFN-γ, and IL-6 in the supernatants of 5 × 105 MSC were measured by enzyme-linked immunosorbent assay (ELISA) according to manufacturer instructions. For IFN-β detection the human IFN-β ELISA Kit from PBL InterferonSource (Piscataway, NJ, http://www.interferonsource.com), for IFN-γ detection the BD OptEIA Set (BD Bioscience, San Diego, CA, http://www.bdbiosciences.com) and for IL-6 detection the ELISA Ready-SET-Go Kit (eBioscience, SanDiego) were used.
Data are expressed as mean ± SEM. Experiments were repeated at least three times with similar results. Analysis of significance was performed using the Student's t test (SigmaPlot, Systat Software Inc., San Jose, CA). p values < 0.05 were considered significant.
MSC differentiated along osteogenic lineages as verified by Alizarin Red staining (supporting information Fig. 1A). Karyotype analyses of the MSC showed normal karyotypes (supporting information Fig. 1B) in all but one of the nine MSC donors. The MSC with altered karyotype were passage 7 and displayed a trisomy 7; however, they did not differ from the other MSC regarding their biochemical and immunological properties. MSC constitutively expressed CD29, CD44, CD105, and HLA class I, and were negative by flow cytometry for CD14, CD34, CD80, and CD86 and HLA-DR (supporting information Fig. 1C). Stimulation of MSC with pI:C or LPS did not alter the expression of these surface markers (supporting information Fig. 1C).
TLR3 and TLR4 Ligands Enhance the Immunosuppressive Properties of MSC
TLR mRNA expression in MSC was analyzed using real-time RT-PCR and compared with DC known to express TLR. Both MSC and DC expressed TLR 1, 2, 3, 4, 5, and 6. Generally, TLR mRNA expression in MSC was lower than in DC except for TLR3 and TLR4, which showed comparable levels in MSC and DC (Fig. 1A). TLR3 and TLR4 in MSC were functional as shown by upregulation of IL-6 mRNA and protein in response to TLR ligands (Fig. 1B). Stimulation with the TLR3 and TLR4 ligands pI:C or LPS increased the expression of TLR1-3 mRNA (Fig. 1C).
To determine the effect of TLR activation on the immunomodulatory phenotype of MSC, MSC/MLR coculture experiments were performed. Inhibition of T-cell proliferation was dependent on the number of MSC (supporting information Fig. 2A). Surprisingly, pretreatment of MSC with pI:C and LPS significantly enhanced the immunosuppressive effects of MSC and induced immunosuppression at a MSC/T-cell ratio that is not suppressive under basal conditions (Fig. 1D, supporting information Fig. 2B), indicating that TLR3 and TLR4 mediate immunosuppressive signals in MSC. At a ratio of 1:30 TLR-stimulated MSC/PBMC immunosuppression compared to cocultures with the same ratio of untreated MSC was less than at higher ratios (1:100, 1:300) as 1:30 untreated MSC/PBMC already exerted strong suppressive effects (supporting information Fig. 2A, 2B).
TLR3 and TLR4 Activation Induces Degradation of Tryptophan into Kynurenine
Next, we investigated the immunosuppressive effects mediated by TLR3 and TLR4 activation in MSC. Real-time PCR analyses revealed that transcripts of the immunosuppressive cytokines IL-10 and TGF-β remained unaltered in TLR-stimulated MSC (data not shown). As degradation of tryptophan has been proposed to be responsible for the immunosuppression mediated by MSC, we analyzed tryptophan and kynurenine, the immediate product of tryptophan oxidation, in the media of MSC and DC after pI:C and LPS treatment. Although untreated MSC did not degrade tryptophan and release kynurenine, treatment with pI:C and LPS resulted in the depletion of tryptophan and accumulation of kynurenine in the cell culture media (Fig. 2A). MSC treated with pI:C for 96 hours released up to 20-50 μM kynurenine and MSC treated with LPS for 96 hours released 5-10 μM kynurenine. DC also released kynurenine after treatment with pI:C and LPS but released only 1-2 μM kynurenine in response to these TLR ligands (Fig. 2B). To mimic the situation in the MSC/MLR cocultures, we analyzed the kynurenine content of the media of 6,000 MSC stimulated with pI:C or LPS for 24 hours, then washed and harvested after 7 days. The LPS-activated and pI:C-activated MSC released 1.3 and 6.2 μM kynurenine, respectively (Fig. 2C). As IFN-γ is produced by the PBMC in the MSC/MLR cocultures, we also measured the kynurenine release of MSC treated with pI:C or LPS in combination with IFN-γ and found a synergistic effect on kynurenine production (Fig. 2D). MSC stimulated with the combination of TLR ligand and 125 IU IFN-γ released more kynurenine than MSC stimulated with 4,000 IU IFN-γ alone (Fig. 2D, supporting information Fig. 3A).
MSC Express IDO1, IDO2, TDO2, and Downstream Kynurenine-Metabolizing Enzymes
To determine the tryptophan-degrading enzyme responsible for TLR-mediated kynurenine production in MSC we analyzed the expression of IDO1, IDO2, and TDO2 upon stimulation with pI:C and LPS. Unstimulated MSC expressed IDO1, IDO2, and TDO2 mRNA (Fig. 3A). After stimulation with LPS, pI:C, or IFN-γ TDO2 mRNA was significantly downregulated (Fig. 3B). Although both IDO1 and IDO2 mRNA increased after treatment of MSC with pI:C or IFN-γ, only IDO1 mRNA was upregulated by LPS (Fig. 3C, 3D). In accordance, we observed a strong induction of IDO1 protein in response to TLR3 and TLR4 activation in MSC (Fig. 3E). In addition, TLR3 and TLR4 activation led to an upregulation of the downstream KYNU (Fig. 3F) and KMO (Fig. 3G).
IDO1 is Responsible for Tryptophan Degradation and Immunosuppression in Response to TLR3 and TLR4 Activation
To further test whether the immunosuppressive tryptophan degradation is due to IDO1 or IDO2, the two stereoisomers of 1-methyl-tryptophan were used. 1-Methyl-D-tryptophan (1-D-MT) was recently reported to inhibit IDO2 , whereas 1-methyl-L-tryptophan (1-L-MT) is a well-known inhibitor of IDO1 . Kynurenine release from pI:C-stimulated MSC (Fig. 4A) and LPS-stimulated MSC (data not shown) was not affected by 1-D-MT, but was significantly inhibited by 1-L-MT, suggesting that IDO1 is the enzyme responsible for tryptophan degradation in response to TLR3 and TLR4 signaling. By using siRNA targeting IDO1 we achieved a knockdown of 98% in pI:C-stimulated MSC and 92% in LPS-stimulated MSC as measured by qRT-PCR (Fig. 4B). Kynurenine production in these MSC in response to pI:C and LPS was inhibited by 88 and 85%, respectively, in comparison to the mean kynurenine production of MSC transfected without siRNA or with a nontargeting siRNA control (Fig. 4C). This result confirms that IDO1 indeed is responsible for the TLR-induced tryptophan degradation.
IDO1-mediated immunosuppression due to tryptophan depletion and generation of tryptophan metabolites should not require cell–cell contact. Indeed, the conditioned media of 6,000, 2,000, and 600 MSC stimulated with LPS or pI:C for 24 hours, washed, and then incubated for 6 days significantly suppressed T-cell proliferation in MLR in the absence of MSC (Fig. 4D).
To study the effect of tryptophan metabolites on T-cell proliferation, kynurenine and the downstream metabolite hydroxy-anthranilic acid were added to MLR. Kynurenine showed a significant inhibition of T-cell proliferation at 50 μM and hydroxy-anthranilic acid at 0.5 μM (supporting information Fig. 2C). To corroborate the involvement of IDO1-mediated tryptophan degradation in the immunosuppressive properties of TLR-activated MSC, the IDO1 inhibitor 1-methyl-L-tryptophan (1-L-MT) was used in MSC/MLR cocultures to inhibit tryptophan degradation. 1-L-MT not only restored, but even increased T-cell proliferation in pI:C and LPS pretreated cocultures (Fig. 4E). This indicates that TLR3 and TLR4 activation induces a proinflammatory phenotype in MSC when immunosuppressive IDO1 activity is bypassed.
An Autocrine IFN-β Loop Activating STAT1 is Involved in the TLR-Induced Upregulation of IDO1 in MSC
As STAT1 is involved in the induction of IDO1 in many different cells and tissues , we investigated whether pI:C and LPS induce phosphorylation of STAT1 in MSC. STAT1 was phosphorylated after 2 hours of pI:C treatment and 6 hours of LPS treatment in MSC (Fig. 5A). In addition, we observed an increase in IRF1 mRNA 24 hours after stimulation of MSC with pI:C or LPS (supporting information Fig. 4). To analyze, whether the STAT1/IRF1 pathway is necessary for TLR-mediated induction of IDO1 activity, we transfected MSC with STAT1 siRNA for 24 hours, then stimulated with pI:C or LPS and measured kynurenine production in the medium by HPLC after another 24 hours. We achieved a knockdown of 94% in pI:C-stimulated MSC and 83% in LPS-stimulated MSC as measured by qRT-PCR (data not shown). The knockdown of STAT1 significantly reduced IDO1-mediated kynurenine production in response to both pI:C (92% reduction) and LPS (62% reduction) (Fig. 5B), thus indicating that signaling via STAT1 is involved in TLR-induced IDO1 upregulation.
The delay in TLR3/TLR4-induced tyrosine phosphorylation of STAT1 raised the possibility that an autocrine mediator of STAT1 activation may be induced by pI:C and LPS. In line with this hypothesis, the supernatants harvested after 4 hours from MSC that were pretreated with pI:C for 2 hours and then washed thoroughly, induced IDO1 mRNA in untreated MSC after 6 hours of incubation (Fig. 5C). Likewise supernatants harvested after 24 hours from MSC that were pretreated with pI:C for 2 hours and then washed thoroughly, induced 43 μM kynurenine in the supernatant of untreated MSC after 48 hours of incubation (Fig. 5D). As STAT1 is phoshorylated by interferons, we tested the hypothesis that interferons mediate the phosphorylation of STAT1 in TLR-activated MSC. Both IFN-γ and IFN-β induced kynurenine production in MSC (Fig. 5E). Although constitutive expression was not detected, both IFN-β and to a lesser extent IFN-γ mRNA was induced after stimulation of MSC with pI:C (Fig. 5F) and LPS (data not shown). Accordingly, IFN-β concentrations in the supernatants of pI:C-treated MSC were considerably higher than IFN-γ as measured by ELISA (Fig. 5G). Treatment of MSC with resveratrol, an inhibitor of the Toll-IL-1 receptor domain-containing adapter inducing IFN-β (TRIF) pathway , suppressed the pI:C-mediated induction of IDO1 mRNA and kynurenine in MSC (Fig. 5H, 5I), further supporting the notion that TRIF-mediated IFN-β signaling is involved in the upregulation of IDO1 following TLR activation. Knockdown of IFN-β by siRNA lead to a reduction of pI:C-induced IFN-β and IDO1 mRNA expression by 88 and 78%, respectively (data not shown) and kynurenine formation by 76% (Fig. 5J). Finally, blocking of the IFNAR by a neutralizing antibody significantly inhibited pI:C-induced kynurenine formation, further supporting the notion that autocrine IFN-β signaling is involved in the induction of IDO1 activity by TLR activation in MSC (Fig. 5K).
The dsRNA Activated PKR Mediates Induction of IDO1 Activity by TLR3 and TLR4
As PKR is involved in both TLR and STAT1 signaling [33–35], we investigated whether it also plays a role in TLR-induced IDO1 activity in MSC. PKR mRNA expression was induced by both LPS and pI:C (Fig. 6A). The inhibition of PKR significantly reduced pI:C-induced IFN-β mRNA and protein production (Fig. 6B, 6C), but had no effect on IFN-γ mRNA (data not shown). Further, the inhibition of PKR significantly suppressed the pI:C-induced and LPS-induced phosphorylation of STAT1 (Fig. 6D). In agreement with the finding that IFN-β production and subsequent STAT1 signaling is necessary for IDO1 induction, PKR inhibition also significantly reduced the pI:C-mediated (Fig. 6E) and LPS-mediated (data not shown) induction of IDO1 mRNA. IDO1 protein expression was significantly reduced by 95% (LPS) and 88% (pI:C) by the PKR inhibitor 2-AP and 73% (LPS) and 65% (pI:C) by the PKR inhibitor PKRI, whereas IFN-γ induced IDO1 was unaffected by PKR inhibition (Fig. 6F). The decrease in IDO1 expression resulted in a reduction of kynurenine formation by 98 and 52% in MSC pretreated with 2-AP and PKRI, respectively for 1 hour and then stimulated with pI:C (Fig. 6G). A significant reduction in kynurenine formation by 2-AP and PKRI was also observed in LPS-stimulated MSC (data not shown). In summary, these data suggest that immunosuppressive tryptophan degradation is mediated by TLR3 and TLR4 via PKR activated autocrine IFN-β signaling.
MSC not only represent a fascinating tool for regenerative medicine but also emerged as an attractive vehicle for the treatment of conditions associated with harmful T-cell responses such as organ-specific autoimmunity and graft-versus-host disease. The fact that MSC can be obtained from adult tissues and cultured and expanded ex vivo has sparked a rapid translation into clinical trials . The elucidation of immunosuppressive mechanisms used by MSC is not only important for clinical trials but also expands our understanding of the physiological function of MSC in maintaining the homeostasis of HSC in the bone marrow. TLR are primarily expressed on APC and recognize conserved pathogen-derived components. Ligation of TLR activates multiple innate and adaptive immune response pathways to eliminate and protect against invading pathogens. Although engagement of TLR on professional APC can break tolerance or induce adaptive and innate immune suppression depending on the cellular context and timing of TLR engagement , we show that TLR expressed on human bone marrow-derived MSC enhance the immunosuppressive phenotype of MSC (Fig. 1D). This novel finding contrasts with a previous report by Liotta et al. , where TLR ligation mitigated the immunosuppressive phenotype of human MSC. Liotta et al. used CD4+ T-cells as responder cells stimulated with irradiated T-cell depleted PBMC, whereas we used total PBMC from two unrelated donors, one of which was irradiated, raising the possibility that CD4− T cells are important in mediating immunosuppression induced by TLR-stimulated MSC. In addition, while Liotta et al. either left TLR ligands in the cocultures or treated the MSC with TLR ligands for 5 days before initiating the coculture experiments, we pretreated MSC with TLR ligands for 24 hours, extensively washed and then cocultured them with the MLR. The presence of TLR ligands in the cocultures may lead to different results as also CD4+ T-cells express TLR. On the other hand, incubation of MSC with TLR ligands for 5 days followed by 4 days of coculture could already lead to a downregulation of factors induced shortly after the initial ligation of TLR.
We found that TLR3-activated and TLR4-activated MSC displayed a significant increase in tryptophan degradation and kynurenine production (Fig. 2). Recent studies point to a synergistic role of tryptophan depletion  and downstream tryptophan metabolites [38–42] in suppressing T-cell proliferation. Initiation of tryptophan degradation by TLR activation has been described in other cell types such as DC and epithelial cells [43, 44]. The immunosuppression of adaptive T-cell responses mediated by TLR ligation in mice has been attributed to the induction of tryptophan degradation . Although in human monocyte-derived DC TLR ligation alone is not sufficient to induce tryptophan degradation , we observed accumulation of high micromolar levels of kynurenine in the supernatant of TLR3-activated MSC indicating that these cells may be constitutively geared towards an immunosuppressive phenotype (Fig. 2). At a concentration of 50 μM kynurenine suppressed proliferation of MLR, whereas the downstream metabolite hydroxy-anthranilic acid was suppressive at 0.5 μM (supporting information Fig. 2C). The conditioned media of MSC stimulated with LPS or pI:C significantly suppressed proliferation of MLR (Fig. 4D). As kynurenine concentrations of 50 μM are not reached in these conditioned media, downstream tryptophan metabolites, that are more immunosuppressive than kynurenine such as hydroxy-anthranilic acid (supporting information Fig. 2C; [38, 40, 42]) are likely to be the key mediators of the observed immunosuppression. LPS may have led to more suppression than pI:C (Fig. 1D, supporting information Fig. 2B) due to stronger induction of kynureninase (Fig. 3F) resulting in increased production of downstream kynurenine metabolites such as hydroxy-anthranilic acid. However, as they are more difficult to measure than the surrogate of IDO1 activity, kynurenine, future studies are necessary to analyze the expression and regulation of the downstream kynurenine metabolites. Consistent with previous data obtained from DC , our data show that kynurenine production by TLR-activated MSC can be attributed to IDO1 activity as kynurenine release was suppressed by the IDO1 inhibitor 1-L-MT and IDO1-specific siRNA (Fig. 4A–4C). IDO2 mRNA was induced by pI:C (Fig. 3C), but it did not contribute to tryptophan degradation, neither did the liver-specific TDO2. The functional relevance of these findings was corroborated by the fact that addition of 1-L-MT to the cocultures restored T-cell proliferation with both LPS-pretreated and pI:C-pretreated MSC (Fig. 4E). We even observed an increase in T-cell proliferation after inhibition of IDO1 with 1-L-MT, suggesting that in the absence of IDO1 activity the net-effects of pI:C or LPS may be proinflammatory as previously shown .
We further explored the signaling pathways involved in the upregulation of IDO1 in response to TLR3 and TLR4 signaling. Active STAT1 is directly involved in the induction of IDO1 by binding to GAS elements in the IDO1 regulatory region and indirectly by inducing the production of IRF-1, which binds to ISRE elements in the IDO1 regulatory region . In previous studies, STAT1 has been shown to be phosphorylated in response to both pI:C and LPS in murine cells [33–35]. We show that STAT1 is phosphorylated by TLR3 and TLR4 activation in human MSC (Fig. 5A). STAT1 phosphorylation, however, was not directly induced by pI:C or LPS, but mediated by autocrine IFN-β signaling. Both TLR3 and TLR4 are known to induce IFN-β gene expression as they share the possibility to activate the TRIF-dependent pathway, which induces the phosphorylation-dependent dimerization of IFN response factor 3 (IRF-3) . In addition, induction of IFN-β by TLR activation was mediated by PKR (Fig. 6B, 6C), which has previously not been implicated in the regulation of tryptophan metabolism in MSC or other cell types.
The dsRNA-activated serine/threonine kinase PKR is an important component of host responses to infection and other conditions of cellular stress. PKR is activated by trans-autophosphorylation, induced by viral dsRNA or signals mediated by cytokines, growth factors, or cellular stress . dsRNA is thought to activate PKR either directly if it has access to the cytosol at very low micromolar concentrations or indirectly via an upstream membrane receptor such as TLR3 if applied extracellularly . pI:C, which can not permeate the plasma membrane to directly access the cytosolic PKR , and LPS have been shown to activate PKR via TLR3 and TLR4, respectively [33–35]. In addition, PKR modulates STAT1 activation , and is known to physically interact with STAT1 . Our data imply that in MSC PKR is upstream of IFN-β in the signaling cascade leading from TLR3 and TLR4 activation to the induction of IDO1 activity as the inhibition of PKR significantly reduced the production of IFN-β. In contrast to IFN-β, IFN-γ production by MSC was not altered by PKR inhibition. In line with the finding that the effects of IFN-γ are not inhibited by 2-AP in murine macrophages , the induction of IDO1 by exogenous IFN-γ in MSC was also not influenced by PKR inhibitors. In MSC stimulated with pI:C or LPS, IFN-β rather than IFN-γ was produced and the knockdown of IFN-β and blocking of the IFNAR significantly inhibited kynurenine formation (Fig. 5G, 5J, 5K). Whereas, previous studies demonstrated that IFN-γ is necessary to induce tryptophan catabolism in human MSC [14, 18–20], our data provide evidence that TLR3 and TLR4 activation in the absence of IFN-γ is sufficient to induce functional IDO1 and an immunosuppressive phenotype in MSC.
In summary, we have established TLR3 and TLR4 signaling as a novel mechanism that enhances the immunosuppressive properties of MSC. The signaling pathway leading from TLR3 and TLR4 activation to the induction of functional IDO1 involves the activation of PKR, autocrine IFN-β signaling, and the activation of STAT1/IRF-1 (Fig. 7).
As inflammation can severely harm HSC , the physiological function of TLR signaling in MSC may represent a support mechanism by which MSC protect HSC from infectious injury. Furthermore, this mechanism could be utilized therapeutically to augment the immunosuppressive properties of MSC in cell therapy to treat autoimmune disease or transplant rejection .
We thank Sabrina Koch and Andreas Mlitzko for expert technical assistance and Dr. H.D. Hager for help with the karyotype analyses. This work was supported by the Hertie Foundation and grants from the Center for Interdisciplinary Research Tübingen (1496-0-0 to M.P. and 1546-0-0 to W.W. and M.P.), by the Helmholtz Foundation (VH-NG-306 to M.P.), and by the Landesstiftung Baden-Württemberg, State of Baden-Württemberg, Germany (P-LS-AS/HSPA7-12 to W.W).
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.