mRNA Is an Endogenous Ligand for Toll-like Receptor 3*

Toll-like receptors (TLRs) are the basic signaling receptors of the innate immune system. They are activated by molecules associated with pathogens or injured host cells and tissue. TLR3 has been shown to respond to double stranded (ds) RNA, a replication intermediary for many viruses. Here we present evidence that heterologous RNA released from or associated with necrotic cells or generated by in vitro transcription also stimulates TLR3 and induces immune activation. To assess RNA-mediated TLR3 activation, human embryonic kidney 293 cells stably expressing TLR3 and containing a nuclear factor-κB-dependent luciferase reporter were generated. Exposing these cells to in vitro transcribed RNA resulted in a TLR3-dependent induction of luciferase activity and interleukin-8 secretion. Treatment with in vitro transcribed mRNA activated nuclear factor-κB via TLR3 through a process that was dose-dependent and involved tyrosine phosphorylation. Furthermore, in vitro transcribed natural or 2′-fluoro-substituted mRNA induced the expression of TLR3, interferon regulatory factor-1, tumor necrosis factor-α, and interleukin-1 receptor-associated kinase-M mRNA in human dendritic cells (DCs). DCs responded to mRNA treatment by expressing activation markers, and this maturation was inhibited by antagonistic TLR3-specific antibody. Endogenous RNA released from or associated with necrotic cells also stimulated DCs, leading to interferon-α secretion, which could be abolished by pretreatment of necrotic cells with RNase. These results demonstrate that RNA, likely through secondary structure, is a potent host-derived activator of TLR3. This finding has potential physiologic relevance because RNA escaping from damaged tissue or contained within endocytosed cells could serve as an endogenous ligand for TLR3 that induces or otherwise modulates immune responses.

Cell Culture-Human embryonic kidney (HEK) 293 cells (ATCC, Manassas, VA) were propagated in Dulbecco's modified Eagle's medium supplemented with glutamine (Invitrogen), and 10% fetal calf serum (Hyclone, Ogden, UT) (complete medium). A cell line, TLR3-293, derived from HEK293 cells, which was first stably transformed with pELAM-luc and then with pCMV6-XL5 containing human TLR3 cDNA (Origene Technologies, Rockville, MD) was generated. TLR3-293 cells were grown in complete medium with 125 g/ml Zeocin and 400 g/ml G418 (Invitrogen). Leukophoresis samples were obtained from HIVuninfected volunteers through an IRB-approved protocol. Peripheral blood mononuclear cells were purified by Ficoll-Hypaque density gradient purification. DCs were produced as described previously (13). The resulting immature DCs were used between 6 and 9 days after the initial culture of monocytes.
Treatment of TLR3-293 Cells-TLR3-293 cells were seeded into 96well plates (5 ϫ 10 4 cells/well) 1 day prior to stimulation and cultured without antibiotics. The following day, the cells were exposed to RNA without or with prior complexing to Lipofectin (Invitrogen) as described previously (15). When the treatment was performed in the absence of Lipofectin, cells were exposed to 50 g/ml mRNA (F-mRNA1 and F-mRNA2) containing 2Ј-deoxy-2Ј-fluoro-substituted nucleotides or mRNA1, or poly(I)⅐poly(C) that was added to complete medium and incubated for 8 or 16 h. When Lipofectin-complexed RNA was used, cells were exposed to 5 g/ml poly(C), poly(A), poly(U), poly(G), poly(I)⅐poly(C), or in vitro transcribed mRNA with or without a cap structure or poly(A) tail, for 1 h. The RNA was then removed, and the cells were incubated further in complete medium for 7 or 15 h. Where noted, cells were pretreated with 3 nM staurosporine, 0.5 M genistein, 5 mM 2-aminopurine (Sigma), or 40 M TIRAP peptide (Novagen, Madison, WI) for 30 min prior to stimulation with RNA. Supernatants were collected for IL-8 measurement, and cells were lysed with luciferase lysis buffer (Promega, Madison, WI) and analyzed using luciferase assay system (Promega) in a Dynatech MLX luminometer (Chantilly, VA).
In Vitro RNA Transcription-Messenger RNAs encoding gag, firefly luciferase, Renilla luciferase, and the nonstructural polyprotein of Semliki Forest virus were obtained from corresponding plasmids (NdeI, NdeI, SspI, and SpeI-linearized pTEVluc, pTEVgag, pSVren, and pSFV1, respectively) using in vitro transcription assays (MessageMachine and MegaScript kits; Ambion, Austin, TX) as described previously (15). The transcripts (TEVluc, TEVgag, Renilla) were designated mRNA1, mRNA2, and mRNA3, respectively. In certain instances, the Dura-Script kit (Epicentre, Madison, WI) was used to generate nuclease-resistant, 2Ј-F-deoxycytidine and 2Ј-F-deoxyuridine nucleotide-containing mRNAs (TEVluc and SFV1, designated as F-mRNA1 and F-mRNA2) as described by the manufacturer. All mRNA contained a cap structure and poly(A) tail unless otherwise noted. Assays for LPS in RNA preparations using the Limulus Amebocyte Lysate gel clot assay were negative with a sensitivity of 3 pg/ml (Department of Genetics, Cell Center Service Facility, University of Pennsylvania). Transcribed mRNA was treated with DNase I and then precipitated with LiCl to remove any trace amount of DNA. The transcripts were also analyzed by agarose gel electrophoresis to confirm that the mRNAs were intact, migrate with the expected size, and were free of DNA contaminants.
Secondary Structure Analysis of the RNA-RNA secondary structures for mRNA1, mRNA2, and mRNA3 were predicted using mFOLD 3.1 (www.bioinfo.rpi.edu/applications/mfold), which utilizes a free energy minimization algorithm established by Zuker and Stiegler (16,17). The longest uninterrupted self-complementary sequences in F-mRNA2 were determined by DotPlot analysis using MegAlign 5.5 (DNASTAR Inc., Madison, WI). 20 g of in vitro transcript was digested with S1 nuclease (Promega) according to the manufacturer's recommendation. S1 nuclease-resistant RNA products were recovered by ethanol (2.5 v/v) precipitation performed in the presence of glycogen carrier and 0.75 M LiCl at Ϫ20°C for 20 h. Precipitate was washed with 75% ethanol and reconstituted in water. Using agarose gel electrophoresis only very low molecular weight cleavage products were detected as a diffuse band capable of retaining ethidium bromide.
Analysis of DC Maturation-DCs were treated with the indicated ligands for 24 h. Supernatants were collected for cytokine quantitation, and cells were harvested for flow cytometry. Isotype control or anti-TLR3 antagonistic mAbs (40 g/ml) (3.7) (eBioscience, Inc., San Diego, CA) were added 30 min prior to stimulation, when indicated. Necrotic cells were prepared by freeze-thawing HEK293 cells four times. Then, necrotic cells were incubated with or without 1-10 units/10 6 cells of Benzonase (Novagen) at 4°C for 6 -24 h and added to DCs at a ratio of 2:1. DCs were stained with CD83-phycoerythrin mAb (Research Diagnostics Inc., Flanders, NJ) and HLA-DR-fluorescein isothiocyanate mAb and analyzed on a FACScalibur flow cytometer using CellQuest software (BD Biosciences).
Enzyme-linked Immunosorbent Assays for Cytokines-Supernatants of activated DCs, and HEK293 and TLR3-293 cells were collected at the indicated times for cytokine measurement. The levels of IL-12 (p70) (Pharmingen), IFN-␣, IFN-␤, and IL-8 (BIOSOURCE International, Camarillo, CA) were measured in supernatants by sandwich enzymelinked immunosorbent assay. Cultures were performed in duplicate to quadruplicate and measured in duplicate.
Northern Blot Analysis-DCs plated in 48-well plates (3.3 ϫ 10 5 cells/well) were treated for 8 or 16 h, as indicated, with the exception of samples containing Lipofectin, which either alone or within an RNA complex was left on the cells for 1 h followed by replacement with fresh medium. Where noted, cells were treated with 2.5 g/ml cycloheximide (Sigma) 30 min prior to the stimulation and throughout the entire length of incubation. At the end of the 8-or 16-h incubation, cells were lysed in 100 l of guanidinium thiocyanate (Master Blaster, Bio-Rad), and RNA was extracted according to the manufacturer. To enhance the RNA yield, 70 g of glycogen (Roche Diagnostics) was added as carrier, and the precipitation was performed in siliconized tubes at Ϫ20°C overnight. The RNA pellet was reconstituted in 4 l of nuclease-free water (Promega) and stored at Ϫ20°C. 2-l RNA samples (ϳ1 g) were denatured then separated in denaturing, 1.4% agarose, 0.22 M formaldehyde gel submerged into MESA buffer (Sigma) supplemented with 0.22 M formaldehyde. RNA was transferred to NYTRAN SuperCharge filters (Schleicher & Schuell) and UV cross-linked. The filters were prehybridized at 68°C for 1 h in MiracleHyb (Stratagene). To probe the Northern blots, 50 ng of DNA was labeled using Redivue [␣-32 P]dCTP (Amersham Biosciences) with a random prime labeling kit (Roche Applied Science). The filters were hybridized at 68°C for 20 h with MiracleHyb containing the labeled and denatured probe. The filters were washed and exposed to Kodak MS film using an MS intensifier screen at Ϫ70°C for 2-72 h.
Probes-Templates for human TNF-␣ and GAPDH probes were excised and gel purified from plasmids (pE4 and pHcGAP, respectively) obtained from ATCC. Coding sequences for human IRF-1 were excised from a plasmid (accession number BC009483) purchased from Open Biosystems (Huntsville, AL). Probes for human TLR3 and IL-1 receptor-associated kinase (IRAK)-M were purified cDNAs derived from plasmids generated by TOPO TA cloning (Invitrogen). The TLR3-and IRAK-M-specific PCR products were generated from RNA of phorbol 12-myristate 13-acetate-treated U937 cells. For TLR3, the 5Ј primer (5Ј-AAGGAAAGGCTAGCAGTC-3Ј) and 3Ј primer (5Ј-TCGGAGCAT-CAGTCGT-3Ј) of the coding sequence of human TLR3 were used (Accession: U88879). This probe detected two transcripts (ϳ3 and 5 kb) in RNA samples. The IRAK-M-specific 5Ј-primer (5Ј-TTTCAACCCAAAC-TAACTGAT-3Ј) and 3Ј-primer (5Ј-CAAGAATGCCCTGGAGCTTC-3Ј) of human IRAK-M CDS were used (accession number AF113136). This probe detected a 2.5-kb RNA species and two additional spurious transcripts (8 and 9 kb) in the RNA samples. The specificity of all probes was confirmed by sequencing (DNA Sequencing Facility, University of Pennsylvania).

In Vitro Transcribed mRNA Activates the NF-B Signaling
Pathway through TLR3 in TLR3-293 Cells-We demonstrated previously that mRNA containing a poly(A) tail signaled and activated DCs. Two mechanisms were described. The first acted through a P2Y nucleotide receptor and was mediated by the poly(A) tail. The second induced TNF-␣ and was independent of the poly(A) tail (18). TLR3 has been demonstrated to signal in response to poly(I)⅐poly(C) dsRNA but not ssRNA (poly(C) homopolymer) (5). Although mRNA is a single strand, it differs from synthetic poly(C) homopolymer in that it contains considerable secondary structure, including double stranded regions. This led us to test whether mRNA signaled through TLR3 and might be the TNF-␣-inducing activity observed in our earlier experiments (18). mRNA is intrinsically unstable in tissue culture because of the presence of RNases, so different formulations of ssRNA and dsRNA were complexed to Lipofectin, which provides protection against RNase-mediated degradation. These Lipofectin-complexed RNAs were added to HEK293 cells stably transfected with human TLR3 and the ELAM-luc reporter plasmid, which responds to NF-B activation by production of luciferase. Neither Lipofectin alone nor poly(C) nor poly(U) homopolymer complexed to Lipofectin activated TLR3. However, HIV gag-encoding mRNA complexed to Lipofectin induced luciferase activity, demonstrating signaling through TLR3. The presence of a poly(A) tail or cap structure on the mRNA was not required for TLR3 signaling (Fig. 1A). The addition of mRNA to HEK293 cells that contained only the ELAM-luc reporter plasmid demonstrated no signaling, but the addition of poly(I)⅐poly(C) dsRNA yielded low level luciferase production (data not shown). A subsequent Northern blot analysis confirmed the presence of a small amount of endogenous TLR3 mRNA in the parental HEK293 cells (data not shown).
The instability of mRNA in tissue culture systems required the use of Lipofectin as a stabilization reagent to be able to measure mRNA signaling. Lipofectin did not seem to affect TLR3 signaling because similar induction was seen with both Lipofectin-complexed and free poly(I)⅐poly(C) dsRNA ( Fig. 1, A and B). However, it is conceivable that Lipofectin complexing might lead to structural alteration of mRNA, causing it to signal through TLR3. To exclude this possibility, we generated RNase-resistant mRNA containing 2Ј-deoxy-2Ј-fluoro-substituted nucleotides (2Ј-F-deoxycytidine and 2Ј-F-deoxyuridine). Two mRNAs, encoding luciferase and the Semliki Forest virus nonstructural genes, were made with fluoro-substituted nucleosides. We found that neither of these mRNAs was translated in HEK293 cells. 2  HEK293 cells stably expressing TLR3 and the reporter plasmid ELAM-luc were exposed to different RNAs complexed with Lipofectin (A, C, and D) or without (B). A, luciferase activity in RNA-stimulated cells. Cells were incubated with Lipofectin complexed to poly(C), poly(U), poly(I)⅐poly(C), and to in vitro transcribed gag-encoding mRNA with or without a cap structure or poly(A) tail (5 g/ml), or Lipofectin alone (control) for 1 h. Lipofectin-complexed RNA was removed, the medium was replaced, and luciferase activity was assayed 7 h later. B, luciferase activity in cells treated with mRNA containing 2Ј-deoxy-2Јfluoro-substituted nucleotides. F-mRNA1, F-mRNA2, or poly(I)⅐poly(C) (50 g/ml) was added to medium containing 10% serum without Lipofectin complexing for 8 h. C, IL-8 secretion by cells treated with Lipofectin-complexed RNA homopolymers, poly(I)⅐poly(C), or in vitro transcribed mRNA2 and mRNA3 as described for A. IL-8 was measured in supernatants collected 8 h later. D, dose-dependent IL-8 secretion in cells exposed to different concentrations of mRNA2 or poly(I)⅐poly(C). Cell treatment with the Lipofectin-complexed RNA was performed as described for A. IL-8 was measured in supernatants collected 8 h after RNA stimulation. Values for luciferase and IL-8 in quadruplicate were averaged and S.E. calculated. Experiments were performed at least three times for each data point.
We were also concerned about the use of a synthetic promoter reporter system. Transcriptional regulation, even of the same promoter, can differ considerably between a stably integrated promoter versus the endogenous locus because of differences in chromatin structure and other variables in the promoter environment (19,20). Additionally, the luciferase transcription unit has been reported to confer ambiguous promoter-independent responsiveness in certain cell lines (21). Therefore, it was important to demonstrate that stimulation of TLR3 by mRNA could induce transcription of an endogenous locus. It was reported previously that HEK293 cells transfected with TLR4 produced IL-8 in response to specific ligand (22). Therefore, we analyzed supernatants from TLR3-expressing HEK293 cells for IL-8 production after stimulation with RNA. Interestingly, in the absence of Lipofectin, these cells exhibited a high base-line secretion of IL-8 (typically 100 -400 pg/ml) with a significant increase in IL-8 after poly(I)⅐poly(C) or mRNA stimulation. In the presence of Lipofectin, endogenous IL-8 secretion was greatly reduced (less than 10 pg/ml), although ligand-induced secretion remained intact. We found that mRNAs encoding HIV gag or Renilla luciferase induced the secretion of IL-8 at levels comparable with that observed for poly(I)⅐poly(C) (Fig. 1C). Using IL-8 as a readout of TLR3 signaling allowed us to observe more easily concentration-dependent induction by mRNA (Fig. 1D). These data demonstrate that mRNAs can efficiently signal through TLR3 and are functionally different from synthetic ssRNA homopolymers, which cannot signal through TLR3.
mRNA Treatment Induces IRF-1, IRAK-M, TNF-␣, and TLR3 mRNAs in Human DCs-To confirm the role of extracellular mRNA as a TLR3 ligand in primary cells, we treated DCs with various RNA constructs and used Northern blot analysis to monitor changes in the expression of genes known to be involved in dsRNA-mediated signaling. First, we examined IRF-1, a transcription factor whose mRNA accumulates in response to stimulation by various cytokines, including type I IFN and dsRNA (for review, see Refs. [23][24][25]. Fibroblasts from IRF-1 knock-out mice exhibit defective production of type I IFN after poly(I)⅐poly(C) treatment, suggesting the involvement of IRF-1 in type I IFN induction by dsRNA (26). To test whether TLR3 stimulation with different RNA constructs would lead to the transcriptional up-regulation of IRF-1, Northern blot analysis was performed on DCs that were treated for 8 h with RNA. These RNAs included in vitro transcribed conventional and 2Ј-deoxy-2Ј-fluoro-substituted mRNA, poly(C) homopolymer, and poly(I)⅐poly(C). We observed that IRF-1 mRNA was induced in response to both mRNA and poly(I)⅐poly(C), but not to RNA homopolymer ( Fig. 2A). IFN-␣ is known to induce IRF-1, but we found that addition of cycloheximide did not block IRF-1 induction ( Fig. 2A), suggesting that IRF-1 induction by mRNA does not require new protein synthesis and is therefore a direct result of primary signal transduction.
Although the above results suggest that the exogenous mRNAs may utilize TLR3 to signal DCs, we cannot exclude the possibility that other mediators such as PKR (27,28) or nucleotide receptors may be involved in mRNA-induced signaling (18). The IRAK-M kinase was originally identified in cells of monomyeloic origin and shown to be induced upon TLR stimulation with LPS and negatively regulate TLR signaling to subsequent LPS challenge (29). We analyzed monocyte-derived DCs and found that they also produced and up-regulated IRAK-M mRNA subsequent to treatment with various TLR ligands, including lipoteichoic acid (TLR2), poly(I)⅐poly(C) (TLR3), LPS (TLR4), and R-848 (TLR7) (Fig. 2B). However, IRAK-M was not up-regulated by other non-TLR stimulators of DCs including CD40L and ATP (data not shown). Interestingly,  (F-mRNA1), poly(C), or poly(I)⅐poly(C); or Lipofectin alone (none). The experiment was performed in the presence or absence of 2.5 g/ml cycloheximide (CHX) as described under "Experimental Procedures." RNA was harvested after an 8-h incubation and analyzed for IRF-1 expression. B, DCs were exposed to several TLR ligands: 0.1 g/ml LPS, 50 g/ml poly(I)⅐poly(C), 1 g/ml R-848, 5 M CpG oligodeoxynucleotide (ODN), and 2 g/ml lipoteichoic acid (LTA) or to 2 ng/ml TNF-␣ for 16 h. Expression of IRAK-M, TNF-␣, IRF-1, and GAPDH mRNAs was analyzed by Northern blotting. C, cells were exposed to 50 g/ml conventional RNA (mRNA1) or fluoro-substituted RNA (F-mRNA1), or poly(I)⅐poly(C) without Lipofectin complexing in the presence or absence of 2.5 g/ml cycloheximide for 8 h. Cellular RNA was analyzed for expression of IRAK-M, TNF-␣, and GAPDH genes. D, DCs were exposed to Lipofectin alone (none), RNA complexed to Lipofectin (5 g/ml mRNA1 or poly(C)), or to uncomplexed RNA (50 g/ml poly(I)⅐poly(C), poly(C), poly(U), or poly(A)) added directly to the medium as described under "Experimental Procedures." In addition, cells were unstimulated or exposed to 2 ng/ml TNF-␣ or 1 g/ml CD40L, 10% serum with 1 g/ml LPS, or serum alone. Cellular RNA, harvested 16 h after initiation of the treatments, was tested for expression of TLR3 and GAPDH mRNAs. Experiments were performed in triplicate. monocyte-derived DCs did not respond to nonmethylated CpG DNA (TLR9). Both human type A and B CpG oligonucleotides and supercoiled and linear plasmids were added to DCs, but no induction of IRAK-M mRNA or increase in DC activation markers was detected (data not shown), in agreement with previous findings (30). Next, we tested whether in vitro transcribed RNA could induce IRAK-M. We used mRNA containing 2Ј-deoxy-2Јfluoro-substituted nucleotides because conventional mRNA is unstable in cell culture conditions in the absence of the protective Lipofectin carrier, but, as we showed in earlier experiments, mRNA stabilized with fluoro-derivatized C and U nucleotides was able to induce luciferase (Fig. 1B) and IL-8 (data not shown) in TLR3-293 cells. Fig. 2C shows that mRNA containing 2Ј-deoxy-2Ј-fluoro-substituted nucleotides (F-mRNA) induced IRAK-M and TNF-␣ mRNA to levels similar to poly(I)⅐poly(C) induction. This induction did not require new protein synthesis because cycloheximide was not inhibitory (Fig. 2C). Conventional mRNA complexed with Lipofectin delivered to DCs also induced IRAK-M mRNA accumulation (data not shown).
mRNA-triggered IL-8 Secretion Is Tyrosine Phosphorylationdependent and PKR-independent-Signaling from the TLR receptors is modulated by the differential use of adaptor proteins and may be MyD88-dependent and/or -independent (8,32,33). TLR3 requires tyrosines in its cytoplasmic tail for signaling (34). TLR3 can also signal in an IRAK-independent manner through PKR (33). The use of these various signaling pathways by TLR3 has not been fully elucidated. In the simple NF-B readout system, all of these activities, TRIF (8,32), IRAK, and PKR, can lead to reporter activation. To discern better the signaling pathways used by mRNA, several inhibitors of TLR signaling were employed. The tyrosine phosphorylation inhibitors, genistein and staurosporine, completely blocked NF-Bdriven luciferase and IL-8 production by poly(I)⅐poly(C) and mRNA (Fig. 3, A and B, respectively). The inhibitory TIRAP peptide (35,36), which potently blocks TLR4 but not TLR2 or TLR9 signaling (36), did not block TLR3 signaling by poly(I)⅐poly(C) or mRNA (Fig. 3, A and B, respectively). Signaling by dsRNA, as well as by TLR3, has been demonstrated to activate PKR which leads to NF-B induction (33). To rule out the possibility that mRNA caused increased reporter and IL-8 production by acting directly on PKR instead of signaling through TLR3, the PKR inhibitor 2-aminopurine was employed. Both poly(I)⅐poly(C)-and mRNA-mediated induction of ELAM-luc activity and IL-8 production was unaffected by 2-aminopurine treatment (Fig. 3, A and B). Similar data were obtained whether mRNA with Lipofectin or F-mRNA without Lipofectin was used. These data support the idea that mRNA signals through TLR3 in a manner similar to dsRNA and is dependent on tyrosine phosphorylation and independent of PKR.
Necrotic Cell-associated RNA Induces IFN-␣ Secretion by DCs-We reported earlier (13) that DCs treated with in vitro transcribed mRNA up-regulate major histocompatibility complex and costimulatory molecules, secrete IL-12, IL-8, and IFN-␣, and become potent T cell stimulators. These data were obtained using in vitro transcribed mRNAs produced from linearized plasmids by RNA polymerase. In the next series of studies, we tested whether endogenous RNA, either released from or associated with necrotic cells, can signal and activate DCs under physiologic conditions. It is well established that extracts of necrotic cells can activate DCs (for review, see Ref. 7). This activation is mediated at least in part by TLR2 and TLR4 and may involve heat shock proteins (6). DCs can mature into several types of cell with different effects on T cell development. The phenotype of activated DCs, broadly defined as Th0-, Th1-, Th2-, or Treg-inducing, are determined, in part, by the cytokines secreted in response to TLR stimulation (3,4,13,30,37,38). For example poly(I)⅐poly(C) induces high levels of IL-12 and IFN-␣, whereas LPS induces little IFN-␣ (13). We utilized HEK293 cells freeze-thawed four times as a model for necrotic cells and tested whether RNA from these cells stimulates TLR3 and modulates the DC phenotype. To confirm that the effector molecule in these experiments was RNA, necrotic cell lysates were also pretreated with Benzonase, a potent and nonspecific nuclease that degrades all RNA into oligomers of 2-5 nucleotides in length. Analysis of RNA isolated from necrotic cells that were incubated with or without Benzonase verified the effectiveness of nuclease digestion. Using ethidium bromide staining, intact ribosomal RNAs were still detectable in samples isolated from necrotic cells incubated without Benzonase, but not in those incubated with the nuclease (data not shown). When necrotic cells, with or without prior Benzonase treatment, were added to immature DCs, both resulted in

FIG. 3. TLR3 signaling is inhibited by the tyrosine kinase inhibitors genistein and staurosporine but not by inhibitors of TIRAP and PKR.
A, TLR3-293 cells were pretreated with medium (Control), 40 M TIRAP peptide, 5 mM 2-aminopurine (2-AP), 0.5 M genistein, or 3 nM staurosporine for 30 min and then exposed to medium (gray bars) or 50 g/ml poly(I)⅐poly(C) (f) for 8 h. B, TLR3-293 cells were pretreated as in A, then exposed to Lipofectin alone (gray bars) or 5 g/ml gag-encoding mRNA (mRNA2) complexed to Lipofectin (f) for 1 h, at which time medium was replaced, and the cells were incubated further for 7 h with inhibitors. Levels of luciferase and IL-8 were measured in the cell lysates and supernatants, respectively. Values obtained in quadruplicate were averaged and S.E. calculated. Experiments were performed three times for each data point. activation of DCs as measured by an increase in CD83, CD80, CD86, and HLA-DR expression, suggesting that the removal of RNA did not alter the ability of necrotic cells to activate DCs ( Fig. 4A and data not shown). However, measurement of IL-12 (p70), IL-8, and IFN-␣ from necrotic cell activated-DCs revealed that the presence or absence of RNA affected the pattern of cytokine production. Necrotic cells induced high levels of IL-12 and IFN-␣ secretion by DCs. The removal of RNA from the necrotic cells resulted in a decrease in IL-12 and a return to unstimulated levels of IFN-␣ (Fig. 4B). The level of secreted IL-8, which is induced by most activated DCs, was similar regardless of Benzonase treatment (Fig. 4B), suggesting that the decrease in IL-12 and IFN-␣ secretion by DC was selective and not a nonspecific cytotoxic effect. As we stated, we selected Benzonase because it can degrade all forms of RNA (ssRNA, dsRNA, and RNA bound to DNA) to small oligomers, but it also degrades all forms of DNA as well. Although mammalian DNA has been implicated in TLR9 activation (39), such activation is unlikely to occur because no form of DNA (CpG containing oligonucleotides or plasmid) was found to activate the DCs used in these studies, and this type of DC has been demonstrated not to express TLR9 (30). Therefore, we conclude that the loss of RNA and not the loss of DNA was responsible for the changes in DC activation. As an additional test, necrotic cells were treated with DNase, and no loss of IL-12 or IFN-␣ secretion was observed (data not shown). Thus, necrotic cells have been shown to activate DCs through TLR2 by some undefined ligand, to release heat shock proteins that signal through TLR2 and 4, and now have been shown also to signal through TLR3 by RNA. Although RNA does not increase the overall levels of DC activation in our in vitro system, it was observed to alter the phenotype of the activated DCs.
Antagonistic TLR3-specific Antibody Inhibits mRNAinduced DC Activation-To confirm that mRNA acts as a TLR3 activator of DCs, an antagonistic antibody that has been demonstrated to block poly(I)⅐poly(C)-induced IFN-␤ secretion in human fibroblasts (40) was used. Pretreatment of DCs with the anti-TLR3 mAb 3.7 inhibited up-regulation of CD83, CD80, or CD86 (Fig. 5A) and IFN-␣ secretion (data not shown) when stimulated with poly(I)⅐poly(C) or F-mRNA without Lipofectin complexing. Isotype control antibodies had no such effect (Fig.  5A). Thus, poly(I)⅐poly(C) and mRNA activate DCs leading to costimulatory molecule expression and IFN-␣ production, by signaling through TLR3.
Dominant Negative TRIF Inhibits mRNA-induced IFN-␤ Secretion-TLR3 uses the adaptor protein MyD88 to activate NF-B signaling (5) and TRIF to activate type I IFN (8,32) signaling pathways. To confirm further that mRNA signaled through TLR3, a dominant negative form of TRIF was used. HEK293 cells were cotransfected with a TLR3 expression plasmid and either TRIF dominant negative expression plasmid or a control vector. Cells were then treated with poly(I)⅐poly(C) or mRNA complexed to Lipofectin, and supernatant associated IFN-␤ was measured. The dominant negative TRIF significantly reduced mRNA-induced IFN-␤ secretion (Fig. 5B). These data further demonstrate that mRNA signals through TLR3. DISCUSSION TLRs play an important role in innate immunity by recognizing microbial infection and orchestrating an appropriate immune response. Mammalian TLRs respond not only to pathogen-associated structures but also to host-derived molecules that are released from injured tissues and cells. The growing list of endogenous ligands that activate TLR include heat shock proteins (6,7,41,42), surfactant protein A (43), fibrinogen (44), hyaluronan (45), heparan sulfate (46), defensin (47), and chromatin-IgG complexes (39). In this report, we identify mRNA as a new endogenous ligand by demonstrating that RNA, associated or released from necrotic cells, can modulate DC activation and that RNA activation of DCs requires TLR3 signaling. We provide supporting evidence for this finding by demonstrating in a model system that in vitro transcribed mRNA can specifically activate TLR3 on DCs and on HEK293 cells overexpressing TLR3.
Alexopoulou et al. (5) were the first to report that dsRNA is the specific ligand for TLR3 by demonstrating that TLR3 recognizes and responds to viral or synthetic dsRNA but not to homopolymer ssRNA (5). Here, we present evidence that TLR3 also responds to stimulation by cellular or in vitro transcribed RNAs. Although these are typically regarded as ssRNA, cellular and in vitro transcribed RNAs do have dynamic secondary structures (e.g. hairpins) that contain double stranded sequences (27,28). It is conceivable that the first component of the TLR3 signaling cascade, a putative dsRNA-binding protein, might interact efficiently with one helical turn of dsRNA (ϳ11 bp) (27,28), as has been suggested for other dsRNA-binding proteins (28), thus allowing short segments of dsRNA in mRNA to signal. This possibility is supported by secondary structure analysis, which predicted formation of multiple double stranded regions that were up to 14 bp long in each of the mRNA used in the present study (data not shown). Additionally, when mRNA1 was digested with S1 nuclease, which cleaves ssRNA but not dsRNA, the S1-resistant RNA products (most likely short dsRNA) signaled DCs with efficiency similar to that of the intact, undigested mRNA (data not shown).
Induction of IRAK-M mRNA by TLR ligand has only been demonstrated for LPS in mouse macrophages (29). Here we demonstrate in human DCs and monocyte-derived macrophages (data not shown) that the expression of IRAK-M mRNA FIG. 5. Blocking of signaling through TLR3 inhibits mRNA-induced DC activation or IFN-␤ production by HEK293 cells transiently overexpressing TLR3. A, DCs were preincubated with 40 g/ml antagonistic anti-TLR3 mAb or isotype-matched control for 30 min at 37°C, then stimulated with 1 g/ml LPS, 10 g/ml poly(I)⅐poly(C) or 10 g/ml nucleaseresistant fluoro-substituted RNA (F-mRNA1) for 24 h. Cells were analyzed for expression of CD83, CD80, and CD86 by flow cytometry. B, HEK293 cells were transfected with TLR3 and either a TRIF dominant negative (gray bars) expression vector or a control vector (f). The following day, cells were stimulated with Lipofectin alone (control) or poly(I)⅐poly(C) or mRNA1 complexed with Lipofectin. 24 h later supernatants were analyzed for IFN-␤. Values obtained in triplicate were averaged and S.E. calculated. Experiments were performed three times.
was increased in response to other TLR ligands, including R-848, lipoteichoic acid, mRNA, fibrinogen, and poly(I)⅐poly(C) as well as LPS (Fig. 2B and data not shown). Induction of IRAK-M mRNA by these ligands did not require new protein synthesis ( Fig. 2C and data not shown). This demonstrates that IRAK-M is directly induced by multiple TLR ligands in human DCs and macrophages.
What could be the physiological relevance for TLR3 activation by cellular RNA? Examples certainly exist for nucleic acids being involved in induction of immune responses. For instance, it has been demonstrated recently that mammalian chromatin can induce the secretion of anti-DNA immunoglobulin by B cells through dual interaction with TLR9 and surface IgM (39). The autoimmune disease systemic lupus erythematosus is marked by autoantibodies against DNA and RNA (48). It therefore comes as no surprise that deficiencies in the removal of chromatin by DNase-I has been linked to systemic lupus erythematosus (49). Despite the ubiquitous distribution of RNases, additional pathophysiological conditions are associated with the appearance of extracellular RNA. An example of this is the demonstration that RNA is a constituent of the plaques found in patients with Alzheimer's, a disease that is increasingly being viewed as a chronic inflammatory process (50 -52). Given these precedents, it is reasonable to suppose that RNA might have a role similar to that already known to be played by chromatin. Such RNA could come in contact with the DCs or other immune system cells through release by necrotic cells or the phagocytosis of necrotic cells or their components. Such liberated cellular RNA could conceivably stimulate TLR3 signaling pathways, as we describe in this paper, which would lead to specific immune responses in vivo. The potential of RNA to induce immune reactions also suggests that physiologic mechanisms for the regulation of RNA-induced immune activation must exist. Indeed, RNases such as the eosin-derived neurotoxin, which is released from the granules of eosinophils attracted to sites of inflammation (53), may play such a role in delimiting the amount of immunologic activation that can occur at sites of injury through the rapid degradation of released RNA.
Once antigen is acquired, DCs have the peculiar capacity to shunt the exogenously captured material into the major histocompatibility complex class I pathway via a mechanism that is dependent on cytosolic antigen processing machinery. This process is known as antigen cross-presentation (for review, see Ref. 54). In a recent study, it was observed that stimulation with TLR3 and TLR9 ligands, but not other TLR ligands, increased the ability of DCs to cross-present antigen (55,56). Multiple mechanisms for how a DC presents exogenously acquired antigen on major histocompatibility complex class I molecules have been suggested (for review, see Refs. [57][58][59]. Considering that TLR3 and TLR9 are unique among the TLRs in their capacity to respond to nucleic acids with antigenencoding potentials, and considering the data presented here and previously, which demonstrated efficient uptake and translation of antigen-encoding mRNA by DCs with subsequent T cell activation directed against the encoded antigen (13, 60 -64), these results suggest that mRNA could be a novel source of antigen for presentation and cross-presentation by professional antigen-presenting cells.
In summary, we describe a new host-derived ligand for TLR3, mRNA, and demonstrate that RNA contained within necrotic cells activates DCs and alters their phenotype. mRNA therefore joins the list of endogenous ligands for TLRs and is the first endogenous ligand described for TLR3. The presence of these host ligands during inflammation activates the immune system. Pathogens continuously mutate and evolve to escape immune surveillance, including detection by TLRs. Therefore, activation of TLRs by host-derived ligands released during injury and infection is very important to alleviate the sole need of pathogen for activation of the immune system. Whether the same cellular mRNA that activates DCs through TLR3 can also serve as a source of encoded antigen is currently under study.