Roles of the Cleaved N-Terminal TLR3 Fragment and Cell Surface TLR3 in Double-Stranded RNA Sensing

Toll-like receptors recognize a variety of microbial products. Cell surface TLRs, including TLR4/MD-2, TLR1/TLR2, and TLR6/TLR2, recognize bacterial membrane lipids, whereas TLR3, TLR7, TLR8, and TLR9 recognize microbial nucleic acids (NAs) in intracellular organelles (1–3). The self/pathogen discrimination by NA-sensing TLRs is error prone and needs to be strengthened by compartmentalization of NA sensing in endolysosomes, not on the cell surface. Although self-derived NA is rapidly degraded by nucleases, microbial NAs are resistant to degradation because they are encased in bacterial cell walls or viral particles. Microbial NAs, therefore, more easily reach endolysosomes and stimulate NA-sensing TLRs than self-derived NAs (4, 5).

To compartmentalize NA sensing by TLRs in endolysosomes, two mechanisms have been proposed. The first mechanism depends on the proteolytic cleavage of TLRs in endolysosomes (6, 7). NA sensing is activated by proteolytic cleavage of TLR ectodomains after trafficking to endolysosomes. After proteolytic cleavage, requirements of N-terminal TLR fragments in NA sensing have been controversial. For example, the ectodomain of TLR9, a sensor for ssDNA, is cleaved by asparagine endopeptidase and/or cathepsins in endolysosomes (8, 9). Although TLR9C alone is reported to sense DNA in the previous study (7), another study showed that the N-terminal TLR9 fragment (TLR9N) remains associated with C-terminal TLR9 fragment (TLR9C) and that TLR9N association is essential for DNA sensing (10). TLR7, a ssRNA sensor, is also proteolytically cleaved (11). The TLR7N fragment is linked to the TLR7C fragment by a disulfide bond, which is indispensable for TLR7 proteolytic cleavage and responses (11). Furthermore, the structure of the complex consisting of TLR8 and a small chemical ligand has revealed that TLR8 is also cleaved, but TLR8N is associated with TLR8C (12). TLR8N and TLR8C both have ligand binding domains and dimerization interfaces (12), demonstrating an indispensable role of TLR8N in ligand binding and signaling. TLR3, a dsRNA sensor, is cleaved, but the precise cleavage site has been only suggested (13, 14). TLR3N is suggested to be associated with TLR3C and required for dsRNA sensing (14, 15). However, the direct evidence for TLR3N requirement in dsRNA sensing has not been shown.

The second mechanism for compartmentalization of NA sensing is based on TLR transportation. Trafficking of NA-sensing TLRs out of endoplasmic reticulum (ER) is regulated by Unc93B1, a multiple transmembrane protein (16, 17). In Unc93b1^3d/3d mice harboring loss-of-function mutation, NA-sensing TLRs fail to respond to NAs, because all the NA-sensing TLRs remain uncleaved in the ER (16, 17). Enforced TLR9 localization to the cell surface causes systemic lethal inflammation (18). Nonetheless, cell surface expression of TLR9 on splenic dendritic cells (DCs) is recently shown (10), suggesting a functional role of cell surface TLR9 in ssDNA sensing. Human TLR3 is also shown to be expressed on the surface of fibroblasts, but not of DCs (19, 20). Moreover, human TLR3 overexpressed in HEK293 cells is shown to be expressed on the cell surface (13, 21). Unc93B1 overexpression augments cell surface expression of TLR3 (21). Despite this progress, cell surface expressions of mouse TLR3 on primary immune cells remain unclarified. Furthermore, functional roles of cell surface TLR3 in responses to dsRNA have not been clear yet.

TLR3 is most highly expressed in CD8⁺ cross-presenting DCs (22). TLR3 induces rapid and massive production of type I IFN and proinflammatory cytokines. TLR3-deficient human patients show that TLR3 is crucial for the defense against HSV-1 encephalitis (23). TLR3 also has a role in sensing UV radiation-dependent cell damage (24). These results demonstrate an essential role of TLR3 in activating immune responses during virus infection and tissue damage. It is important to understand how TLR3 responses are controlled. To study mechanisms regulating TLR3 responses, the current study determined the cleavage site of the TLR3 ectodomain and revealed requirements for TLR3N in dsRNA sensing by TLR3. Furthermore, cell surface TLR3 on CD8⁺ DCs and marginal zone (MZ) B cells was detected by novel mAbs against mouse TLR3. TLR3 mAb enhanced dsRNA-dependent TLR3 responses without any activation with TLR3 mAb alone, suggesting that cell surface TLR3 can be a target for modulating TLR3 responses.

To establish anti-mouse TLR3 mAbs, BALB/c background Tlr3^−/− mice were immunized with Ba/F3 cells expressing Flag-6×His-conjugated mouse TLR3 (Ba/F3-mTLR3fH) in CFA/IFA used as adjuvants. Next, to boost the immunization, the mice were immunized three times with Ba/F3-mTLR3fH in PBS. Three days after the final immunization, splenic cells were fused with SP2/O myeloma cells. Hybridomas producing anti-TLR3 mAb were selected by flow cytometry staining of Ba/F3 cells expressing mTLR3fH. The novel mAbs, named CaT3 and PaT3, were used in this study.

C57BL/6N mice were purchased from Japan SLC, and Tlr3^−/− mice on C57BL/6N background were provided by S. Akira (Osaka University, Osaka, Japan). The Unc93b1^3d/3d mice were provided by B. Beutler (University of Texas, Dallas, TX) and the Scripps Research Institute. All animal experiments were conducted with the approval of the Animal Research Committee of the Institute of Medical Science, University of Tokyo.

Biotinylated anti-Flag Ab (clone M2) and anti-Flag M2 agarose affinity gel were purchased from Sigma-Aldrich. Rabbit GFP polyclonal Ab (pAb) and mAb (clone FM264) were purchased from Invitrogen and MBL, respectively. The pAb to the N-terminal fragment of TLR3 (TLR3N) was purchased from Millipore.

J774 cells were cultured in high-glucose DMEM supplemented with 10% FBS, penicillin-streptomycin-glutamine (Life Technologies), and 50 μM 2-ME. Ba/F3 cells were cultured in RPMI 1640 medium supplemented with IL-3, 10% FBS, penicillin-streptomycin-glutamine, and 50 μM 2-ME. Bone marrow (BM)–derived macrophage (BM-MC) was prepared. In brief, BM cells were plated at 1 × 10⁷ cells per 10 ml with 10% FCS-DMEM supplemented with 100 ng/ml mouse rM-CSF (PeproTech, Rocky Hill, NJ) in 10-cm cell culture dishes.

The C terminus of mouse TLR3, 7, 8, and 9 was tagged with Flag-6×His. They were amplified by PCR and cloned into retroviral pMX, pMXpuro, or pMXneo vectors (provided by Dr. T. Kitamura, Tokyo, Japan). We also constructed the short hairpin RNA-targeting Unc93B1 in retroviral vector pSSCH. For the constructs encoding TLR3 N fragments (Figs. 3 and 4), TLR3 fragments corresponding to aa 1–342 (TLR3N342), 1–346 (TLR3N346), or 1–356 (TLR3N356) were cloned into pMX vector. For the TLR3C constructs, TLR3 fragments corresponding to aa 1–35, the signal peptide, and following 10 aa, fused to aa 343–905 (TLR3C343), 347–905 (TLR3C347), or 357–905 (TLR3C357), were cloned into the pMX vector with tandem HA-tag. The In-Fusion HD cloning kit (Clontech Laboratories) and Rapid DNA Ligation kit (Roche Applied Science) were used for cloning.

pMX and pMXpuro vectors were transfected into Plat-E packaging cells with FuGene6 (Roche Applied Science). After 2 d of incubation, supernatants were obtained as virus suspensions. Ba/F3 and BaκB cells were infected by virus suspensions mixed with DOTAP (Roche Applied Science).

TLR3-GFP was stably expressed in the RAW264.7 cell line. Next, 5 × 10⁹ transfectants were collected and lysed in the buffer, 150 mM NaCl, 50 mM Tris-HCl (pH 7.4), 1.0% Triton X-100, and 1× complete EDTA-free mixture (Roche Applied Science). Anti-GFP agarose beads were added to the lysate and incubated overnight at 4°C. Beads were collected and washed twice by washing buffer, 150 mM NaCl, 30 mM Tris-HCl (pH 7.4), and 0.1% Triton X-100; and bound protein was eluted in elusion buffer, 150 mM NaCl, 30 mM glycine-HCl (pH 2.5), and 0.1% Triton X-100 and neutralization buffer, 1 M Tris-HCl (pH 8.0). Eluted sample was concentrated and dialyzed in buffer (10 mM NaCl, 0.1% Triton X-100) by Amicon Ultra (Millipore). Concentrated sample was dissolved in the sample buffer, 2% SDS, 10% glycerol, 62.5 mM Tris-HCl (pH 6.8), 0.025% bromophenol blue, and 5% 2-ME, and subjected to SDS-PAGE. Separated protein was transferred to polyvinylidene difluoride (PVDF) membranes. Transferred protein was visualized by Coomassie brilliant blue, and the membrane containing TLR3N and 3C was cut out. The membrane was subjected to N-terminal amino acid sequencing by Edman degradation (APRO Science).

Cells were collected and washed twice with 1× PBS. Washed cells were lysed with 1% Triton X-100 lysis buffer [20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 10% glycerol, 1 mM DTT, and 1× complete EDTA-free mixture (Roche Applied Science)]. After incubation for 30 min on ice, lysates were centrifuged and debris was removed. The cell lysates were rotated for 2 h at 4°C with beads coupled with anti-Flag or anti-TLR3 mAbs. The beads were then washed with 0.1% Triton X-100 washing buffer [20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 10% glycerol, 1 mM DTT] three times. Immunoprecipitates were subjected to SDS-PAGE after boiling in reduced sample buffer [125 mM Tris-HCl (pH 6.8), 20% glycerol, 4% SDS, 10% 2-ME, and 0.005% bromophenol blue] or nonreduced sample buffer [125 mM Tris-HCl (pH 6.8), 20% glycerol, 4% SDS, and 0.005% bromophenol blue]. After electrophoresis, samples were transferred to PVDF membranes and subjected to immunoblotting.

HEK293 cells (1 × 10⁶ cells/well) were transiently transfected in 6-well plates using Lipofectamine 2000 reagent (Invitrogen) with pMX, pMX-TLR3, or -TLR3 mutant vector (5 μg), together with a luciferase-linked IFN-β promoter reporter gene, p55C1B-Luc (100 ng). Twenty-four hours after transfection, cells were harvested and seeded into 96-well plates (5 × 10⁴ per well). After 24 h, cells were stimulated with medium alone or poly(I:C) for 6 h. Stimulated cells were lysed using lysis buffer (Promega), and luciferase activities were measured. Data are expressed as mean values with SD.

J774 cells and BM-MCs plated at 5 × 10⁴ per well on 96-well plates were stimulated with indicated TLR ligands for 24 h. The supernatant was subjected to ELISA for RANTES production (R&D Systems).

BM cells and splenocytes were stained with fluorescein-conjugated Abs specific for the following markers in flow cytometry analysis: CD4 (GK1.5), CD8α (53-6.7), CD11b (M1/70), CD11c (HL3), PDCA-1 (927), B220 (RA3-6B2), CD21 (7G6), CD23 (B3B4), Ly6C (HK1.4), Ly6G (1A8), and Gr-1 (RB6-8C5). Biotinylated anti-TLR3 (CaT3 and PaT3) mAbs were established in this study. For preparation of single-cell splenocyte suspensions, spleens were minced by scissors and incubated for 30 min at 37°C in RPMI 1640 with 0.09 U ml–1 Liberase TL (Roche) and 0.1 mg ml–1 DNase I (Roche). Finally, suspended splenocytes were teased through nylon mesh after pipetting and subjected to cell staining.

The cell staining was performed in staining buffer, 1× PBS with 2.5% FBS and 0.1% NaN₃. Single-cell suspensions were incubated on ice for 15 min with biotinylated mAbs diluted in staining buffer. Then cells were washed and incubated with PE-streptavidin (StAv; BioLegend) and fluorescein-conjugated mAbs for 15 min. Stained cells were analyzed by the FACSCalibur or FACSAria flow cytometers (BD Biosciences).

Wild-type, Tlr3^−/−, or Unc93b1^3d/3d BM-MCs were washed twice with HBSS and subjected to cell surface biotinylation with EZ-Link Sulfo-NHS-LC-LC Biotin (Thermo Scientific Protein Biology Products) for 1 h at 4°C. After washing with HBSS, BM-MCs were lysed in 1% Triton X-100 lysis buffer (20 mM Tris-HCl [pH 7.4], 150 mM NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 10% glycerol, 1 mM DTT, and 1× Complete EDTA-free mixture [Roche Applied Science]). Cell lysate was incubated on ice for 30 min and centrifuged to remove debris. Supernatant was collected and rotated at 4°C for 2 h with StAv-conjugated Dynabeads (Veritas). After incubation, beads were washed three times with 0.1% Triton X-100 washing buffer (20 mM Tris-HCl [pH 7.4], 150 mM NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 10% glycerol, and 1 mM DTT). Immunoprecipitates were subjected to SDS-PAGE after boiling in sample buffer (125 mM Tris-HCl [pH 6.8], 20% glycerol, 4% SDS, and 0.005% bromophenol blue). After electrophoresis, samples were transferred to PVDF membrane and subjected to immunoblotting with TLR3N pAb.

PaT3 or CaT3 was coated at 10 μg/ml on the 96-well plate and incubated at 37°C for 1 h. Then the plates were washed twice with 1× PBS.

J774 cells were cultured with Alexa488-labeled Abs at 4°C or 37°C for 24 h. Then these cells were washed twice with 1× PBS and stained with Anti-Alexa488 Ab (Life Technologies) at 4°C for 20 min. After washing the cells, mAb uptake was measured by the FACSCalibur flow cytometers (BD Biosciences).

J774 cells were cultured with CaT3, PaT3, or IgG1 control Ab at 37°C for 24 h. After washing with 1× PBS, these cells were lysed in 1% Triton X-100 lysis buffer [20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 10% glycerol, 1 mM DTT, and 1× Complete EDTA-free mixture (Roche Applied Science)]. Cell lysate was incubated on ice for 30 min and centrifuged to remove debris. Supernatant was collected and rotated at 4°C for 8 h with protein G Sepharose 4 Fast Flow (Amersham Biosciences). After incubation, beads were washed three times with 0.1% Triton X-100 washing buffer (20 mM Tris-HCl [pH 7.4], 150 mM NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 10% glycerol, and 1 mM DTT). Immunoprecipitates were subjected to SDS-PAGE after boiling in sample buffer (125 mM Tris-HCl [pH 6.8], 20% glycerol, 4% SDS, and 0.005% bromophenol blue). After electrophoresis, samples were transferred to PVDF membrane and subjected to immunoblotting with TLR3N pAb.

Data from triplicate samples were statistically analyzed by Student t test. The p value <0.01 was considered to be significant.

To study endogenous TLR3, novel two mAbs against mouse TLR3 were established. The specificity of the mAbs was confirmed by membrane-permeabilized staining (Fig. 1A). The mAbs, CaT3 and PaT3, reacted to TLR3, but not TLR7, TLR8, or TLR9. Neither PaT3 nor CaT3 showed cross-reactivity with human TLR3. Furthermore, the mAbs immunoprecipitated TLR3, but not TLR7, from a pro-B cell line Ba/F3 expressing TLR3 or TLR7 (Fig. 1B). Endogenous TLR3 was next precipitated from J774 macrophage cell line by TLR3 mAbs and immunoprobed with the pAb against TLR3N. Two signals were detected in J774 cells (Fig. 1C). We also studied Unc93B1-silenced J774, which showed ∼90% reduction in Unc93B1 mRNA (Supplemental Fig. 1). The ∼120-kDa signal was immunoprecipitated from Unc93B1-silenced J774 cells and is likely to be full-length, uncleaved TLR3 (TLR3F) in the ER. The ∼70-kDa signal disappeared in Unc93B1-silenced J774 cells (Fig. 1C), indicating that this signal is TLR3N. These results demonstrate that CaT3 and PaT3 are able to bind to both uncleaved and cleaved TLR3.

To study TLR3 proteolytic cleavage, the N-terminal sequence of the cleaved TLR3 fragments was determined. TLR3-GFP was expressed in RAW264.7 macrophage cell line and purified with anti-GFP mAb. Eluted fractions were analyzed by SDS-PAGE and Coomassie Brilliant Blue (CBB) staining (Fig. 2A). Immunoprobing with anti-GFP Ab identified TLR3F-GFP and TLR3C-GFP. The ∼70-kDa signal was predicted as TLR3N from its apparent molecular mass. The N-terminal amino acid sequences of TLR3C and TLR3N were determined. TLR3C and TLR3N started at 343S and 26T, respectively (Fig. 2B). The TLR3 ectodomain consists of tandem repeats of 22 leucine-rich repeats (LRRs). TLR3C contains LRR12-LRR22 (Fig. 2C). The result that TLR3N started at 26T experimentally demonstrated the signal peptide cleavage site at 25S.

Based on the determined amino acid sequence, TLR3N or TLR3C was expressed in Ba/F3 cells and subjected to membrane-permeabilized staining. PaT3 bound to TLR3N, but not TLR3C (Fig. 2D), whereas CaT3 bound to neither fragment. CaT3, however, reacted with the Ba/F3 cells coexpressing TLR3N and TLR3C (Fig. 2D). PaT3 and CaT3 did not interfere with the binding of the other mAb, demonstrating that these two mAbs react with completely different epitopes on the TLR3 ectodomain (Supplemental Fig. 2).

According to the previous reports (14, 15), dsRNA sensing by TLR3C alone is still controversial. One report shows that TLR3C responds to dsRNA (15), whereas another report shows that TLR3C fails to respond to dsRNA (14). After TLR3 cleavage, TLR3N is suggested to be still associated with TLR3C (14), but the requirement of TLR3N for TLR3 responses has not been clarified. We previously showed that the TLR9N+C complex, but not TLR9C alone, responds to ssDNA (10). Expression of TLR9N enables TLR9 responses in TLR9C-expressing cell lines. Furthermore, TLR9 responses in Tlr9^−/− BM-cDCs are complemented by TLR9N and TLR9C, but not by TLR9C alone. These previous results indicate that coexpression of TLR3N and TLR3C leads to the assembly of TLR3N+TLR3C complex. Notably, CaT3 reacted with the cells expressing both TLR3N and TLR3C, but not those expressing TLR3N or TLR3C alone (Fig. 2D). These results suggest that CaT3 binds to TLR3N+C complex. To confirm that separately expressed TLR3N and TLR3C are associated with each other, TLR3N and TLR3C-hemagglutinin (HA) were expressed in Ba/F3 cells and immunoprecipitated by PaT3 or anti-HA, and coprecipitation of the other TLR3 fragment was detected. TLR3N alone showed faster migration than TLR3N endogenously cleaved from TLR3 (Fig. 2E, lanes 4 and 5), probably due to distinct posttranslational modification such as glycosylation. TLR3C was coprecipitated with TLR3N by PaT3 (Fig. 2E, lane 4), whereas coprecipitation of TLR3N with TLR3C was difficult to detect (Fig. 2E, lane 9). PaT3 is likely to stabilize the association between TLR3N and TLR3C (see below).

To address a role of TLR3N in dsRNA sensing by TLR3, TLR3N and TLR3C were separately expressed in HEK293 cells together with the luciferase reporter plasmid p55C1B for IFN-β promoter activity. Comparable expression of TLR3 or its fragments was evaluated by membrane-permeabilized staining with PaT3 or anti-HA mAb (Supplemental Fig. 3). dsRNA-dependent IFN-β promoter activation was not found in HEK293 cells expressing TLR3N truncated at 342, 346, or 356 (N342, N346, or N356) or those expressing TLR3C starting at 343, 347, or 357 (C343, C347, or C357). TLR3 responses were observed by coexpression of N342 with C343, N346 with C347, or N356 with C343, but not by coexpression of N356 with C357 or N342 with C357 (Fig. 3A). TLR3-dependent NF-κB activation was next studied by using Ba/F3 cells expressing the reporter construct NF-κB-GFP. TLR3, TLR3N, and TLR3C were expressed in this cell line, and expression of each fragment was confirmed by membrane-permeabilized staining with PaT3 for TLR3N and anti-HA for TLR3C (Supplemental Fig. 4). These cells were stimulated with poly(I:C), and GFP induction was evaluated by flow cytometry (Fig. 3B). Poly(I:C)-dependent GFP induction was not seen in cells expressing any single fragment. TLR3-dependent GFP induction was detected by combined expression of N342+C343 or N346+C347, but not of N356+C357 (Fig. 3B). These results are consistent with those obtained by IFN-β induction assay and demonstrate that dsRNA is sensed by the TLR3N+C complex, not by TLR3C alone.

Human TLR3 is expressed on the surface of fibroblasts, but not of DCs, and TLR3 mAb inhibits TLR3 responses in fibroblasts (19, 20), suggesting a functional role of cell surface TLR3. In contrast, cell surface expression of mouse TLR3 on primary immune cells has never been reported. In a previous study, membrane-permeabilized staining with anti-TLR3 shows that CD8⁺ cDC subset expresses TLR3 more highly than other cDC subsets (22). We performed cell surface and membrane-permeabilized staining of splenocytes with PaT3. Membrane-permeabilized staining revealed that TLR3 was expressed in cDCs, but not pDCs (Fig. 4A). The specificity of staining was verified by staining Tlr3^−/− DCs. Consistent with the previous report (22), highest TLR3 expression was observed in CD8⁺ cDCs. Cell surface TLR3 was detected on CD8⁺ cDCs and very weakly on the other cDC subsets (Fig. 4A). Splenic B cell subsets were next studied. TLR3 was detected in both follicular and MZ B cells by membrane-permeabilized staining, whereas only MZ B cells expressed TLR3 on the cell surface (Fig. 4B). Cell surface TLR3 was also detected on BM-MCs (Fig. 4C). Although TLR3 was detectable by membrane-permeabilized staining of BM-derived cDCs (BM-cDCs), cell surface TLR3 was hard to detect.

To gain insight into a mechanism regulating cell surface expression of TLR3, Unc93b1^3d/3d mice were studied by using PaT3. In Unc93b1^3d/3d mice harboring loss-of-function mutation, NA-sensing TLRs fail to respond to NAs, because all the NA-sensing TLRs failed to exit the ER (16, 17). Cell surface TLR3 was not detected on splenic CD8α⁺ cDCs (Fig. 4D), demonstrating an essential role of Unc93B1 in cell surface expression of TLR3.

Next, we asked the origin of cell surface TLR3. If cell surface TLR3 comes from the ER, uncleaved TLR3 is likely to be enriched on the cell surface. BM-MCs were subjected to cell surface biotinylation, immunoprecipitation with StAv or PaT3, and immunoblotting with anti-TLR3 Ab. In Unc93b1^3d/3d BM-MCs, TLR3 was not detected in the precipitate with StAv (Fig. 4E, lane 2), demonstrating cell surface–specific biotinylation. Only cleaved TLR3 was detected by immunoprecipitation with StAv, and no increase in uncleaved TLR3 was observed when compared with immunoprecipitation with PaT3 (Fig. 4E, compare upper and middle in lane 1), indicating that cleaved TLR3 is a predominant form on the cell surface. These results demonstrate that cleaved TLR3 is predominantly expressed on the cell surface.

The importance of cell surface TLR3 in dsRNA sensing was addressed by studying effects of PaT3 and CaT3 on TLR3 responses. Cell surface TLR3 was detected on J774 cells. J774 cells were stimulated with poly(I:C) in the presence of TLR3 mAb, and upregulation of costimulatory molecules and RANTES production were analyzed. Interestingly, PaT3, but not CaT3, augmented poly(I:C)-dependent upregulation of CD40 and CD69 (Fig. 5A). RANTES production in response to poly(I:C) was also upregulated by PaT3 (Fig. 5B). RANTES production induced by lipid A was not altered by TLR3 mAbs. Next, the effect of TLR3 mAb on poly(I:C) responses of BM-MCs was studied. Although cell surface TLR3 was only weakly detected on BM-MCs (Fig. 4C), poly(I:C)-induced RANTES production was augmented by PaT3 (Fig. 5C). PaT3 did not enhance lipid A–dependent RANTES production by BM-MCs. Augmentation of IFN-β mRNA induction by PaT3 was also observed (Fig. 5C). In contrast to these strong effects, PaT3 alone did not induce any activation in J774 and BM-MCs.

To gain insight into a mechanism underlying the PaT3-dependent enhancement of TLR3 responses, PaT3 internalization was first studied. J774 cells were incubated with Alexa488-labeled PaT3 or CaT3 at 4 or 37°C and analyzed by flow cytometry after quenching the fluorescence of the mAb on the cell surface. All the mAbs, including control IgG1 mAb, were internalized at 37°C (Fig. 6A, lower). Considering that quenching did not alter fluorescence intensity, the mAbs were unlikely to remain on the cell surface. To confirm that internalized anti-TLR3 mAbs still bound to TLR3, internalized mAbs were precipitated by protein G and immunoblotting with anti-TLR3 Ab was performed. TLR3 was precipitated by CaT3 or PaT3, but not by IgG1 control mAb (Fig. 6B). These results suggest that internalized anti-TLR3 mAbs still bind to TLR3.

Requirement of PaT3 internalization in the enhancing effect of PaT3 was next asked. J774 cells were stimulated with immobilized PaT3. Immobilized PaT3 would ligate TLR3 on the cell surface without internalization. No amplification was observed by immobilized PaT3 (Fig. 6C), suggesting that PaT3 internalization is required for the enhancing effect of PaT3.

Considering that PaT3, but not CaT3, showed the enhancing effect, it is important to find the PaT3-specific effect on TLR3. TLR3N was required for TLR3 responses, and association between TLR3N and TLR3C was more clearly demonstrated by PaT3 than anti-HA Ab capturing TLR3C (Fig. 2E). We hypothesized that PaT3 may strengthen the association between TLR3N and TLR3C. To address this possibility, TLR3C-HA was immunoprecipitated by anti-HA Ab in the presence of PaT3 or CaT3 mAb from the Ba/F3 cells expressing TLR3C-HA and TLR3N. Coprecipitation of TLR3N with TLR3C was clearly enhanced in the presence of PaT3, but not CaT3 (Fig. 6D). Poly(I:C) stimulation did not alter TLR3N coprecipitation. These results indicate that PaT3 enhances TLR3 signaling in endosomes by stabilizing the association between TLR3N and TLR3C, indicating an important role of association between TLR3N and TLR3C in TLR3 responses.

TLR ectodomains consist of LRRs, the tandem repeats of a leucine-rich motif (LRM). LRM is typically 24-aa length and adopts a loop structure, beginning with a short β-strand followed by a β-turn (25). The extra residues in LRM form a loop that protrudes from the LRR solenoid. A tentative loop in TLR9 ectodomain is located between LRR14 and LRR15 and shown to be cleaved (10). A similar loop is located in the LRR11 of the TLR3 ectodomain, which is predicted to be cleaved (14). Consistent with this prediction, our present study experimentally determined the cleavage site at 343S, which resides in the protruding loop of LRR11.

Association of TLR3N and TLR3C was previously suggested by the result that cleaved TLR3 fragments are detectable only under denatured, but not nondenatured condition (14). However, proteolytic cleavage of the TLR3 ectodomain in endolysosomes leads to the presence of all the fragments, including uncleaved TLR3 (TLR3F), TLR3N, and TLR3C. It is difficult to exclude a possibility that TLR3N and TLR3C are indirectly associated with each other by forming a complex with uncleaved TLR3F. In the current study, the determination of the cleavage site enabled separate expression of TLR3N and TLR3C. Thereby, the association of TLR3N and TLR3C was directly shown. Interestingly, coprecipitation of TLR3N was hard to detect in immunoprecipitation of TLR3C by the Ab to C-terminal–tagged HA. Interaction between TLR3N and TLR3C may be unstable in the cell lysate. PaT3 was able to enhance TLR3N coprecipitation with TLR3C, indicating that PaT3 stabilizes the interaction between TLR3N and TLR3C. The persistent association of TLR3C and TLR3N was further supported by another TLR3 mAb Cat3, which is likely to react with the TLR3N+C complex, but not with either fragment. In the TLR ectodomains, tandem repeats of LRM form a solenoid structure with parallel β-sheets. TLR ectodomains are stabilized by their interior hydrophobic cores and hydrogen bonds between parallel β-sheets of LRRs (25). Proteolytic cleavage of the protruding loop is unlikely to alter the interaction between adjacent LRMs.

The interaction between TLR3N and TLR3C was found to be functionally important. TLR3N+C complex, not TLR3C alone, was able to sense dsRNA and activate both IFN-β and NF-κB promoters. Just 14-aa N-terminal deletion in TLR3C abolished TLR3N+C responses. Considering that the N-terminal portion of TLR3C locates in the LRRs, not the protruding loop, 14-aa deletion may prevent TLR3C association with TLR3N by altering the LRR structure. Similarly, the 14-aa extension of TLR3N impaired TLR3 responses. The extended portion may also impair TLR3N association with TLR3C by competing with the N-terminal portion of TLR3C. These results suggest that the cleavage site of TLR3 needs to be within the protruding loop, not in the LRRs.

Cell surface TLR3 expression on CD8⁺ DCs and MZ B cells was shown in the current study. Consistent with the results using in vitro cell lines (13, 21), cell surface expression of TLR3 on primary immune cells was dependent on Unc93B1. TLR3 was weakly detected on the J774 macrophage cell line and BM-MCs. TLR3 mAb PaT3 specifically enhanced TLR3 responses. This is not simply due to mAb-mediated TLR3 ligation, because PaT3 alone did not induce any activation. Moreover, only the mAb to TLR3N PaT3, but not CaT3, had the augmenting effect. Cell surface TLR3N may have a role in modulating TLR3 responses. Considering that PaT3 specifically augmented TLR3 responses, but not TLR4/MD-2 responses, PaT3 is likely to have a direct effect on TLR3 rather than inducing weak priming signal. PaT3 and CaT3 were both internalized, but still bound to TLR3. PaT3, but not CaT3, strengthened the association between TLR3N and TLR3C. PaT3-dependent augmentation of the association between TLR3N and TLR3C is likely to enhance TLR3 responses in endolysosomes, suggesting an important role of association between TLR3N and TLR3C in TLR3 responses. Taken together with the previous report showing that human TLR3 mAb is shown to inhibit TLR3 responses in fibroblasts (19), cell surface TLR3 may have a role in TLR3 responses, although downstream signaling is activated after internalization into endosomes.

Immunoprecipitation of endogenous TLR3 shows that a majority of TLR3 is cleaved in BM-MCs, probably due to rapid degradation of uncleaved TLR3 (13). Consistent with this, cell surface biotinylation shows that cell surface TLR3 is mostly cleaved on these cells. Previous study also shows cleavage of cell surface TLR3 (13). These results indicate that TLR3 traffics from endolysosomes to the cell surface. Considering that PaT3 mAb enhanced TLR3 responses, cell surface TLR3 is likely to be activated by dsRNA probably after internalization, suggesting the shuttling of the TLR3N+C complex between the endolysosomal systems and the plasma membrane. Cell surface TLR3 may be a target for enhancing or inhibiting TLR3 responses. Given that TLR3 promotes cross-priming in viral infection (26), targeting cell surface TLR3 by TLR3 mAb may deserve further study in terms of vaccine development.

We thank Profs. Shizuo Akira (Osaka University) and Bruce Beutler (University of Texas) for providing mice.

The authors have no financial conflicts of interest.