A Role for the Adaptor Proteins TRAM and TRIF in Toll-like Receptor 2 Signaling*

Background: Toll-like receptor 2 (TLR2) mediates innate immune responses by recognizing microbial components. Results: TLR2-mediated induction of the chemokines Ccl4 and Ccl5 and interferon-β is impaired in macrophages lacking the signaling molecules TRAM, TRIF, TBK-1, IRF1, and IRF3. Conclusion: The TRAM/TRIF pathway is involved in TLR2 signaling. Significance: TLR signaling pathways determine the immune response mounted against infectious organisms. Toll-like receptors (TLRs) are involved in sensing invading microbes by host innate immunity. TLR2 recognizes bacterial lipoproteins/lipopeptides, and lipopolysaccharide activates TLR4. TLR2 and TLR4 signal via the Toll/interleukin-1 receptor adaptors MyD88 and MAL, leading to NF-κB activation. TLR4 also utilizes the adaptors TRAM and TRIF, resulting in activation of interferon regulatory factor (IRF) 3. Here, we report a new role for TRAM and TRIF in TLR2 regulation and signaling. Interestingly, we observed that TLR2-mediated induction of the chemokine Ccl5 was impaired in TRAM or TRIF deficient macrophages. Inhibition of endocytosis reduced Ccl5 release, and the data also suggested that TRAM and TLR2 co-localize in early endosomes, supporting the hypothesis that signaling may occur from an intracellular compartment. Ccl5 release following lipoprotein challenge additionally involved the kinase Tbk-1 and Irf3, as well as MyD88 and Irf1. Induction of Interferon-β and Ccl4 by lipoproteins was also partially impaired in cells lacking TRIF cells. Our results show a novel function of TRAM and TRIF in TLR2-mediated signal transduction, and the findings broaden our understanding of how Toll/interleukin-1 receptor adaptor proteins may participate in signaling downstream from TLR2.


Toll-like receptors (TLRs) are involved in sensing invading microbes by host innate immunity. TLR2 recognizes bacterial lipoproteins/lipopeptides, and lipopolysaccharide activates TLR4. TLR2 and TLR4 signal via the Toll/interleukin-1 receptor adaptors MyD88 and MAL, leading to NF-B activation. TLR4 also utilizes the adaptors TRAM and TRIF, resulting in activation of interferon regulatory factor (IRF) 3.
Here, we report a new role for TRAM and TRIF in TLR2 regulation and signaling. Interestingly, we observed that TLR2-mediated induction of the chemokine Ccl5 was impaired in TRAM or TRIF deficient macrophages. Inhibition of endocytosis reduced Ccl5 release, and the data also suggested that TRAM and TLR2 co-localize in early endosomes, supporting the hypothesis that signaling may occur from an intracellular compartment. Ccl5 release following lipoprotein challenge additionally involved the kinase Tbk-1 and Irf3, as well as MyD88 and Irf1. Induction of Interferon-␤ and Ccl4 by lipoproteins was also partially impaired in cells lacking TRIF cells. Our results show a novel function of TRAM and TRIF in TLR2-mediated signal transduction, and the findings broaden our understanding of how Toll/interleukin-1 recep-tor adaptor proteins may participate in signaling downstream from TLR2.
Toll-like receptors (TLR) 4 are a family of 13 (TLR1-13) transmembrane pattern-recognition receptors (1). TLRs recognize conserved molecular motifs found in microorganisms and induce the production of pro-inflammatory cytokines, type I interferons, and up-regulation of co-stimulatory molecules. TLR4 recognizes bacterial lipopolysaccharide (LPS), whereas TLR2 recognizes Gram-positive bacteria and bacterial lipoproteins, but other ligands are also suggested (2)(3)(4)(5). TLR2 discriminates between diacylated and triacylated lipoproteins by heterodimerization with TLR6 and TLR1, respectively. Viral single-stranded RNA and the base analog resiquimod (R848) are ligands for TLR7 and TLR8, whereas the dsRNA poly(I:C) is recognized by TLR3. Viral RNA is also recognized by TLRindependent pathways and can activate the cytoplasmic RNA helicases RIG-I and Mda-5 (6 -8).
Here, we report a novel role for TRAM and TRIF in TLR2 signaling, as lipoprotein-triggered TLR2-mediated Ccl5 induction was abrogated in TRAM-and TRIF-deficient macrophages. Ccl5 release following lipoprotein challenge additionally involves the kinase Tbk-1 and Irf3, in addition to MyD88 and Irf1. Induction of the cytokines Ccl4 and Interferon-␤ (IFN-␤) in response to TLR2 ligands was also partially dependent on TRIF. Combined, our results suggest involvement of the TRAM and TRIF in TLR2 signaling.

EXPERIMENTAL PROCEDURES
Ethics Statement-Experiments involving animals were conducted in accordance with the recommendations made the American Association for Laboratory Animal Science. The animal studies, covered under the protocol A2332, have been approved by the University of Massachusetts Medical School Institutional Animal Care and Use Committee (Animal Welfare Assurance number A3306-01).
Antibodies-Mouse anti-TLR2 mAb 6C2 was produced and purified as described (14). Mouse anti-CD86 PE-Cy5 and control rat IgG PE-Cy5 antibody were from eBioscience (San Diego). TLR2 mAb 6C2 and rat IgG from Sigma or from Jackson ImmunoResearch (West Grove, PA) were labeled with an Alexa 488 kit from Molecular Probes (Eugene, OR).
Flow Cytometry Analysis and Cytokine Measurements-Cells were stimulated 16 -20 h at 37°C, 8% CO 2 before supernatant was harvested and assessed for Tnf or Ccl5 content using ELISA kits from R&D Systems (Oxon, UK) or Pharmingen. Stimulated cells were labeled as described (14) and analyzed by flow cytometry analysis on a BD-LSR II instrument.
In Vivo Experiments-WT or TRIF Lps2/Lps2 mice were injected intraperitoneally with PBS or Pam 3 Cys (25 g). Whole blood was collected after 4 h by terminal cardiac puncture, after isoflurane anesthesia. Serum was separated from whole blood using pedi-serum separator tubes (BD Biosciences), which were placed on ice for 30 min, prior to centrifugation at 900 ϫ g for 60 s. Ccl5 content in the serum was analyzed by ELISA.
Confocal Microscopy-HEK293 cells, seeded on 35-mm glass bottom ␥-irradiated tissue cell dishes (MatTek Corp., Ashland, MA), were transiently transfected with TLR2 Cherry (1.25 g/dish), TRAM YFP (0.25 g/dish), and early endosome marker EEA-1 CFP or TRIF CFP (0.25 g/dish) for 48 h. Cells were treated with medium or FSL-1 (200 ng/ml) for 1 h before cells were fixed and observed by confocal microscopy using an Axiovert 100-M inverted microscope, equipped with an LSM 510 laser scanning unit and a ϫ63 1.4-NA plan Apochromat oilimmersion objective (Zeiss, Jena, Germany). Co-localization maps showing co-localization events of TLR2 and/or TRAM and Eea-1 were created in Imaris 5.0.2, 64-bit version (Bitplane AG, Zurich, Switzerland). In these images, white denotes co-localization events between two channels, and pixels above the threshold that failed to co-localize were set to zero (black). User-defined thresholds were set conservatively in a rectangular selection mode chosen above the apparent noise level for each channel.
Gene Expression Analysis-Nanostring nCounter gene expression analysis (NanoString Technologies) of selected inflammatory genes was performed on peritoneal macrophages from wild-type (WT) or TRIF Ϫ/Ϫ mice treated with medium or FSL-1 lipopeptide for 4 h. Total RNA was isolated using RNeasy RNA isolation kit (Qiagen, Germany) and subjected to nCounter gene expression analysis of selected genes. Values are given as arbitrary units, normalized to internal GAPDH and hypoxanthine-guanine phosphoribosyltransferase standards. In other experiments, immortalized bone marrow macrophages (BMDM) from wild-type or from TRAM Ϫ/Ϫ -, TRIF Ϫ/Ϫ -, and TLR2 Ϫ/Ϫ -deficient C57Bl/6 mice were seeded in 24-well plates in triplicate and treated with medium or FSL-1 (200 ng/ml) for 3 h before cells were lysed. Total RNA was isolated with the Nucleo-Spin 96 RNA kit (Macherey-Nagel, Germany) and processed by vacuum. RNA quantity and purity were checked using NanoDrop (Thermo Scientific). cDNA was made using Maxima first-strand cDNA synthesis kit (Thermo Scientific). qPCR for murine ifnb1 (IFN-␤) and tnf was done with GAPDH as endogenous control using TaqMan Probes (Mm0043552_s1, Mm00443258_m1, and Mm99999915_g1; Applied Biosystems). Real time thermal cycling was done with StepOnePlus (Applied Biosystems) using Perfecta qPCR FastMix (Quanta Biosciences) with uracil N-glycosylase and ROX reference dye. Analysis was done with the StepOne software version 2.1.
Statistics-Analysis was performed with GraphPad Prism 6.0. The difference between the two groups was determined by the two-tailed t test or multiple t tests. Two or more groups were compared with one-way ANOVA with Tukey post-test, or two-way ANOVA with Bonferroni or Fischer's LSD post-test or multiple t-tests, as indicated.

TRIF Pathway Regulates LPS-induced TLR2 Expression and TLR2-mediated Ccl5
Induction-The level of surface expression of TLR2 may determine how efficiently cells respond to bacterial lipoproteins, or other pathogen-derived components, during an infection. We have previously shown that surface TLR2 is markedly up-regulated in response to a range of TLR ligands by an MyD88-dependent mechanism, except in response to LPS and poly(I:C), which induce TLR2 up-regulation independent of MyD88 (14). Because TLR4 utilizes the TIR adaptors TRAM and TRIF, we hypothesized that this pathway may play an important role in LPS-induced up-regulation of TLR2. We therefore assessed TLR2 expression on TRIF Ϫ/Ϫ peritoneal macrophages in response to LPS, as well as in response to Pam 3 Cys, R848, and Sendai virus. We found that LPS-induced up-regulation of surface TLR2 was only partially affected by the absence of TRIF, whereas TLR2 up-regulation was normal in response to Pam 3 Cys and R848 (Fig.  1A).
We proceeded to investigate whether LPS-induced TLR2 surface expression may be regulated by a mechanism similar to the regulation of the co-stimulatory molecule CD86 or the release of the cytokine CCL5, because TRIF is shown to play a role in these responses. We observed that surface CD86 was TRAM and TRIF Mediate TLR2 Signaling FEBRUARY 6, 2015 • VOLUME 290 • NUMBER 6 up-regulated in response to LPS and Sendai virus ( Fig. 1B) but not in response to Pam 3 Cys or R848 (Fig. 1B). The up-regulation of TLR2 was induced by all these ligands (Fig. 1A), suggesting differences in the regulation of surface TLR2 and CD86 molecules in response to different TLR ligands.
Upon assaying Ccl5, we found that all tested ligands induced Ccl5 release in wild-type peritoneal macrophages ( Fig. 1C) (14), including TLR2 ligand Pam 3 CysSK 4 (Pam 3 Cys). Interestingly, TRIF Ϫ/Ϫ macrophages were impaired in their ability to induce Ccl5 in response to Pam 3 Cys (Fig. 1C), suggesting a role for TRIF in TLR2-mediated Ccl5 induction. In contrast, TNF release was normal in TRIF Ϫ/Ϫ macrophages in response to Pam 3 Cys (Fig. 1D). Ccl5 and TNF induction in response to R848 and SV was normal in TRIF Ϫ/Ϫ macrophages ( Fig. 1, C and D), implying no role for the TRIF pathway in these responses, whereas LPS-induced Ccl5 was effectively abolished in the absence of TRIF (Fig. 1C), in agreement with previous reports (6 -9).
To verify a potential role for TRIF in Ccl5 release in response to TLR2 ligands, WT or TRIF Ϫ/Ϫ macrophages were stimulated with the TLR2/6 ligands macrophage-activating lipopeptide 2 (MALP-2), fibroblast-stimulating lipopeptide 1 (FSL-1), or the TLR2/1 ligand Pam 3 CysSK 4 for 18 h. Ccl5 release in the supernatant was subsequently assessed by ELISA. We found that all tested TLR2 ligands induced Ccl5 (Fig. 1E). Furthermore, Ccl5 induction was impaired in TRIF Ϫ/Ϫ cells in response to all tested TLR2 ligands (Fig. 1E), inferring a role for TRIF in mediating Ccl5 induction in response to TLR2 ligands. TNF induction in the same supernatant was normal (data not shown), implying that TNF induction in TLR2-mediated TNF release does not require TRIF.
We found that Ccl5 induction was induced in TLR4 Ϫ/Ϫ cells in response to the tested TLR2 ligands (Fig. 1F), suggesting that TLR4 is not a major contributor to Ccl5 release in response to TLR2 ligands.
TIR Adaptors TRAM and TRIF Mediate Ccl5 Release Induced by TLR2 Ligands-We proceeded to test the role for TRAM in response to different TLR2 ligands. Ccl5 and Tnf induction was assessed in MyD88 Ϫ/Ϫ -, TRAM Ϫ/Ϫ -, and Tlr2 Ϫ/Ϫ -deficient macrophages, in response to MALP-2, FSL-1, and LTA, as well as TLR2/1 ligand Pam 3 Cys. We found that Ccl5 release was markedly reduced in TRAM Ϫ/Ϫ macrophages in response to all TLR2 ligands tested ( Fig. 2A), whereas Tnf induction in TRAM Ϫ/Ϫ macrophages was normal (Fig. 2B). In contrast, the lack of MyD88 led to a sharp decline of both Tnf and Ccl5 release in response to all tested TLR2 ligands (Fig. 2, A and B). Tlr2-mediated Ccl5 induction was also impaired in MAL Ϫ/Ϫ macrophages (data not shown). Ccl5 and Tnf release in response to Pam 3 Cys, FSL-1, and LTA was completely abrogated in Tlr2 Ϫ/Ϫ macrophages, confirming the dependence on TLR2 in mediating these responses and indicating that the TLR2 ligands have high purity (Fig. 2, A and B). As TLR4 does not mediate the lipoprotein-induced Ccl5 (Fig. 1F), this promotes the view that TLR2 alone, and not TLR2/TLR4 heterodimers, triggers responses via TRIF and TRAM. LPS and SV induced normal levels of TNF and Ccl5 in TLR2 Ϫ/Ϫ macrophages (Fig. 2C).

TRAM and TRIF Mediate TLR2 Signaling
TLR2 and TRAM Co-localize in Endosomes, and TLR2-mediated CCL5 Induction Is Dependent on Endocytosis-We proceeded to investigate the subcellular expression of TLR2 and TRAM to verify the involvement of TRAM in TLR2 signaling. TLR2 is expressed in early endosomes, as well as in recycling Rab11a-positive endosomes in human monocytes (30). TRAM has been reported to be expressed at the plasma membrane and in Rab5-positive endosomes (46). We therefore proceeded to investigate whether TLR2 co-localizes with TRAM in early endosomes. Overexpression of fluorescently tagged TLR2 Cherry , TRAM YFP , and the early endosome marker Eea-1 CFP in HEK293 cells revealed that TLR2 and TRAM indeed do co-localize, both at the plasma membrane and in a portion of Eea-1-positive endocytic vesicles (Fig. 3A). Upon stimulation with FSL-1, TLR2 and TRAM accumulated and co-localized in the perinuclear area (Fig. 3B). Overexpression of TRAM YFP and TRIF CFP in HEK293 showed particularly good co-localization in endocytic vesicles (Fig. 3C). Minimal concentrations of plasmid were used to avoid cell activation because expression of high levels of these adaptor molecules alone can drive NFB activation. However, we cannot exclude the possibility that the strong co-localization between TRAM and TRIF may partly be a consequence of transfection-driven cell activation. Nevertheless, we also observed TLR2 co-localizing with TRIF in TRAMpositive vesicles in HEK293 overexpressing TLR2 Cherry , TRIF CFP , and TRAM YFP (Fig. 3C), implying that TLR2 is recruited to a TRAM-and TRIF-containing complex.
Because we observed TLR2 and TRAM co-localizing in endocytic compartments, we investigated the importance of endocytosis for Ccl5 induction by pretreating peritoneal macrophages with the dynamin GTPase inhibitor Dynasore prior to stimuli with LPS and TLR2 ligands. Dynasore efficiently inhibited CCL5 release induced by TLR2 ligands (Fig.  3D) but not TNF release (Fig. 3E), suggesting that endocytosis of ligand is required for Ccl5 induction in response to these ligands. These results suggest that signaling leading to Ccl5 induction likely occurs from early endosomes, where TLR2 and TRAM co-localize.
TLR2-mediated CCL5 Induction Is Impaired in TRIF Lps2/Lps2 Mice in Vitro and in Vivo-TRIF Lps2/Lps2 mice are derived from a C57Bl/6 background and contain a chemically induced base pair-deletion/frameshift in the gene encoding TRIF (25). These animals mimic mice with a gene targeted TRIF deletion. We observed that macrophages from TRIF Lps2/Lps2 mice displayed reduced Ccl5 release in response to TLR2 ligands MALP-2 and FSL-1 (Fig. 4A). In contrast, Tnf release was not impaired in response to these TLR2 ligands (Fig. 4B), which is consistent with previous reports (25).Ccl5 and Tnf release were normal in response to R848 (Fig. 4, A and B).
To determine whether TLR2 ligands can induce Ccl5 in a TRIF-dependent manner in vivo, WT or TRIF Lps2/Lps2 mice received an intraperitoneal injection with Pam 3 Cys. Serum was collected after 4 h and assayed for Ccl5. We found that Ccl5 levels were elevated in wild-type mice following Pam 3 Cys injection (Fig. 4C). This Ccl5 induction in response to Pam 3 Cys injection was impaired in TRIF Lps2/Lps2 mice (Fig. 4C), confirming our in vitro findings and supporting a role for TRIF in TLR2mediated Ccl5 induction.
TLR2-mediated CCL5 Activation Is Impaired upon Overexpression of Signaling Defective Mutant Versions of TRAM and TRIF-Our findings in peritoneal macrophages were verified by transiently transfecting HEK293 cells stably expressing TLR2 with a murine ccl5 luciferase reporter gene. Stimulation of TLR2-expressing HEK293 cells with TLR2 ligands activated the ccl5 promoter (Fig. 5A). Similar results were obtained using a human CCL5 luciferase reporter gene (data not shown). The different TIR-adaptor proteins were also assessed for their ability to activate the CCL5 promoter upon overexpression. This was done by transfecting HEK293 cells with TLR2 and the CCL5 luciferase and co-transfecting with MAL, MyD88, TRAM, or TRIF or the functionally defective adaptor mutants MyD88TIR, TRAMC117H, and TRIF⌬N⌬C. Transfection of MyD88, TRAM, or TRIF all resulted in activation of the CCL5 reporter in this system (Fig. 5B), whereas overexpression of TRAM C117H, TRIF⌬N⌬C, or MyD88TIR impaired CCL5 induction in response to TLR2 ligands Pam 3 Cys and FSL-1 (Fig.  5C). Combined, these results confirm that TLR2 ligands induce CCL5 and that TRAM and TRIF mediate CCL5 induction in response to TLR2 ligands.
The promoters of CCL5, CXCL10, as well as IFN-␤ contain transcription factor-binding elements for both NF-B and IRFs (31,32). We observed that overexpression of NF-B p65 and IRF3, but not of IRF5, efficiently induced activation of the CCL5 promoter in HEK293 cells co-transfected with a CCL5 luciferase reporter (Fig. 5D). TLR2 ligands are considered poor activators of IRF3 (9, 33), so we initially investigated whether the observed TLR2-mediated CCL5 response could be mediated by NF-B alone. Using HEK293 cells transiently expressing TLR2 and a CCL5 luciferase reporter containing mutated NF-B sites (⌬B CCL5 luciferase), we found that the fold induction activation of this reporter was similar to the activation of WT CCL5 luciferase reporter in response to TLR2 ligands (Fig. 5E). Overexpression of mutant NF-B p65 S536A in HEK293 cells also activated the CCL5 luciferase reporter to the same extent as wild- type p65 (Fig. 5F). These results indicate that phosphorylation of p65 Ser-536 is not essential for TLR2-mediated CCL5 induction and prompted us to investigate a role for IRF3 in TLR2mediated CCL5.
TLR2 Ligands Activate IRF3, and CCL5 Induction Is impaired in IRF3 Ϫ/Ϫ and IRF1 Ϫ/Ϫ Macrophages in Response to TLR2 Ligands-To investigate whether TLR2 ligands can activate IRF3, we initially applied a GAL4 luciferase reporter assay and a vector expressing yeast Gal4 DNA binding domain fused to IRF-3 lacking its own DNA binding domain (IRF3-Gal4). Activation of IRF3 in this assay leads to the expression of the GAL4 luciferase reporter (34,35). We observed that TLR2 ligands activated IRF3 measured by this reporter system, and that the activation was impaired upon overexpression of the dominant negative version of TRAM, TRAMC117H (Fig. 6A). The FSL-1 lipopeptide also induced phosphorylation of IRF3 Ser-396 in RAW264.7 cells (Fig. 6B), confirming that TLR2 ligands can activate IRF3, although to a lesser extent than LPS and poly(I:C) (Fig. 6B). IRF1 also regulates TLR-dependent induction of type I interferon responses in myeloid dendritic cells and macrophages (36). We further investigated the role for IRF3 and IRF1, as well as IRF5 and IRF7, in TLR2-mediated CCL5 induction. Peritoneal macrophages from wild-type (WT), irf1 Ϫ/Ϫ -, irf3 Ϫ/Ϫ -, irf5 Ϫ/Ϫ -, or irf7 Ϫ/Ϫ -deficient mice were stimulated with TLR2 ligands, as well as LPS, poly(I:C), R848, or SV prior to assaying Ccl5 and Tnf release in these cells by ELISA. We found that Ccl5 release was impaired in peritoneal macrophages from both Irf1-and Irf3-deficient mice in response to the TLR2 ligands Pam 3 Cys, Pam 2 Cys, FSL-1, and MALP-2 (Fig.  6C). Tnf was, however, also partially impaired in irf1 Ϫ/Ϫ macrophages in response to TLR2 ligands (Fig. 6D), suggesting that irf1 plays a role in both Tnf and Ccl5 release in response to lipopeptides. LPS-and poly(I:C)-induced Ccl5 was normal in irf1 Ϫ/Ϫ macrophages, in line with previous reports (37). Importantly, IRF3 was found to be a key player in mediating the release of Ccl5 (Fig.  6C), but not Tnf (Fig. 6D), in response to TLR2 ligands. This correlates with our observations in TRAM Ϫ/Ϫ and TRIF Ϫ/Ϫ macrophages ( Figs. 1 and 2). Ccl5 induction in response to TLR2 ligands was found to be normal in irf5 Ϫ/Ϫ peritoneal macrophages, whereas both Ccl5 and TNF were partially impaired in irf7 Ϫ/Ϫ cells in response to TLR2 ligands (Fig. 6, C and D). These results show that TLR2 ligands activate Irf3, and TLR2-mediated Ccl5 is mediated by Irf3, as well as by Irf1, and they suggest that TLR2induced Irf3 activation is mediated by the TRAM/TRIF pathway.
TLR2-mediated Ccl5 Induction Is Impaired in TBK1-deficient Cells-The noncanonical IB kinases IKK⑀ and TBK1 are important mediators of TRIF-dependent IRF3 activation lead- ing to type I IFN and CCL5 induction in response to viral components and LPS (38). To determine the role of IKK⑀, as well as the type I interferon receptor I (IFNAR) in TLR2-mediated Ccl5, we assayed Ccl5 induction in response to TLR2 ligands in peritoneal macrophages from WT, ikke Ϫ/Ϫ (IKK⑀) and ifnar Ϫ/Ϫ mice, as well as irf3 Ϫ/Ϫ mice. Ccl5 release in response to TLR2 ligands was normal in IKK⑀-deficient cells after 18 h of stimulation (Fig. 7A). The receptor for type I IFNs, Ifnar, did not appear to be important for TLR2-mediated Ccl5 release either (Fig. 7A), in contrast to Irf3 (Fig. 7A), indicating that TLR2mediated Ccl5 induction is a primary response that occurs independent of type I IFN signaling. TLR2-induced ccl5 mRNA induction was also unaffected by inhibition of protein synthesis with cycloheximide (data not shown), further suggesting that this Ccl5 induction is a primary response.
TLR2 Ligand-induced Ccl4 and IFN-␤ Are Partially Impaired in the Absence of TRIF Ϫ/Ϫ -To determine whether the TRAM/TRIF pathway is involved in the regulation of other genes in response to TLR2 ligands, we performed gene expression analysis of selected inflammatory genes on total RNA from peritoneal macrophages from WT or TRIF Ϫ/Ϫ mice treated Graph shows quantification of band intensities of blot shown below stained with anti-IRF3 Ser-396 (pIRF3) (top blot) and total IRF3 (TOT IRF3) (bottom blot) relative to ␣-tubulin staining. TLR2-mediated Ccl5 induction is impaired in IRF1 Ϫ/Ϫ and IRF3 Ϫ/Ϫ mice. Peritoneal macrophages from wild-type (WT), irf1 Ϫ/Ϫ , irf3 Ϫ/Ϫ , irf5 Ϫ/Ϫ , and irf7 Ϫ/Ϫ mice were exposed to medium (0), Pam 3 Cys (1000 -100 -10 ng/ml), Pam 2 Cys (1000 -100 -10 ng/ml), MALP-2 (1000 -100 -10 ng/ml), FSL-1 (1000 -100-10 ng/ml), LPS (10 ng/ml), poly(I:C) (pIC) (50 g/ml), R848 (100 ng/ml), or Sendai virus (SV) (100 HAU/ml) for 18 h before supernatant was harvested and analyzed for Ccl5 content (C) and Tnf content (D) by ELISA. Results show mean with range of duplicate samples and are representative of two experiments. FEBRUARY 6, 2015 • VOLUME 290 • NUMBER 6 with medium or FSL-1 lipopeptide for 4 h. The results confirmed that Ccl5 is induced in response to TLR2 ligand FSL-1 and that this response is impaired in TRIF Ϫ/Ϫ macrophages (Fig. 8A). The interferon-inducible genes cxcl10, ifit1, and ifit2 were very weakly induced in response to FSL-1, although FSL-1 failed to induce ifn␤1 induction above the detection limit in these cells using this nonamplified system (Fig. 8A). The induction of the chemokine ccl4 was partially impaired in TRIF Ϫ/Ϫ macrophages in response to FSL-1 (Fig. 8B), suggesting that TRIF participates in mediating this response. We proceeded to evaluate Ccl4 protein induction in response to TLR2 ligands by stimulating peritoneal macrophages from TRIF Ϫ/Ϫ mice with lipopeptides and LPS, poly(I:C), or Sendai virus. TLR2-induced Ccl4 protein was found to be partially impaired in TRIF Ϫ/Ϫ cells after 20 h of stimulation (Fig. 8C). Ccl4 release was also impaired in TRIF Ϫ/Ϫ cells in response to LPS and poly(I:C), but not in response to Sendai virus (Fig. 8C). Overexpression of TRAM and TRIF, as well as MAL and MyD88, in HEK293 cells activated the CCL4 promoter-driven reporter luciferase (data not shown), supporting the view that these adaptors are involved in the regulation of CCL4. Overexpression of MAL and MyD88 induced a more potent activation of the CCL4 reporter luciferase (data not shown), suggesting that the MAL/ MyD88 pathway is more important in regulating CCL4, and possibly explaining why TLR2-mediated Ccl4 induction in TRIF Ϫ/Ϫ macrophages was only partial impaired (Fig. 8, B and  C). To further confirm a role for TRAM in the regulation of CCL4, HEK293 cells were transiently transfected with TLR2, CCL4 luciferase reporter, and decreasing concentrations of dominant-negative TRAMC117H, prior to stimulation with lipopeptides or Sendai virus. TLR2 ligands potently induced the CCL4 reporter luciferase, and this induction was impaired upon overexpression of mutant adaptor TRAMC117H (Fig.  8D), indicating a role for TRAM in mediating CCL4. In contrast, CCL4 induction in response to Sendai virus was normal upon overexpression of TRAMC117H (Fig. 8D).

TRAM and TRIF Mediate TLR2 Signaling
TLR2 has been suggested to induce IFN-␤ in an IRF-3-dependent manner in inflammatory monocytes (41). We consequently assayed immortalized bone marrow-derived macrophages from WT, TRAM Ϫ/Ϫ , TRIF Ϫ/Ϫ , and TLR2 Ϫ/Ϫ mice stimulated with FSL-1 by quantitative RT-PCR to determine whether these cells can induce IFN-␤ in a TRAM/TRIF-dependent manner. Induction of ifn-␤ was observed in wild-type cells following 3 h of stimulation (Fig. 8E). This ifn-␤ induction was further found to be partially impaired in TRAM Ϫ/Ϫ and TRIF Ϫ/Ϫ cells (Fig. 8E), implying a role for TRAM and TRIF in TLR2-mediated IFN-␤ induction. TNF induction was normal in TRAM Ϫ/Ϫ and TRIF Ϫ/Ϫ cells in response to FSL-1 (Fig. 8F). These results suggest that the TRAM/TRIF signaling pathway plays an important role in certain TLR2-mediated responses.

DISCUSSION
TLR4 has been considered the only TLR that can activate both the MyD88-dependent pathway and the TRAM/TRIF-dependent signaling pathway in macrophages. In this study, we show a new role for TRAM and TRIF in TLR2 signaling. In resemblance to TLR4 signaling, we propose that TLR2 can utilize the TRAM/TRIF signaling pathway, in addition to the MyD88-dependent pathway, to mediate certain responses.
We initially observed that LPS-induced up-regulation of surface TLR2 was partially impaired in TRIF Ϫ/Ϫ macrophages, suggesting that TRIF contributes to this response. TLR2 has previously been shown to be differentially regulated by both TRIF and MyD88 in response to LPS at the gene level by microarray (42). Here, we show that the TRIF signaling branch appears to be more important for correct regulation of surface TLR2 in response to LPS, because this expression was impaired in the absence of TRIF but not in the absence of MyD88 (14). The mechanisms regulating LPS-induced TLR2 expression have yet to be elucidated, but they appear to differ from LPSinduced CD86 surface expression, which has been shown to be TRIF-, type I IFN-, and IFNARI-dependent (25).
Interestingly, we observed that TLR2-mediated induction of Ccl5, Ccl4, and IFN-␤ are impaired in TRIF-deficient macrophages. TLR2-dependent induction of IFN-␤ has previously been reported in murine inflammatory monocytes in response to vaccinia virus (41), although this response was not found to be regulated by a TRIF-dependent mechanism. Bacterial lipopeptides were later shown to induce IFN-␤, as well as the IFNinducible genes cxcl10, mx-2, il-6, and nos2 (inos) in a TLR2dependent manner in BMDM at high concentrations of ligand (43). The IFN-␤ induction observed by Dietrich et al. (43) in response to TLR2 ligands was shown to be a primary response mediated by MyD88, by a mechanism resembling TLR7/9 signaling. In this report we confirm that TLR2 ligands induce IFN-␤ in macrophages. We further show that TLR2-mediated IFN-␤, Ccl5, and Ccl4 are TRAM-and TRIF-dependent, in addition to MyD88-dependent. The TLR2-mediated Ccl5 induction we observed was also a primary response occurring independent of IfnarI and protein synthesis. We observed that high concentrations of 0.1-1 g/ml lipopeptide were required to induce Ccl5, whereas Tnf release is induced at considerably lower concentrations of ligand. Involvement of TRAM and TRIF was observed regardless of the magnitude of Ccl5 that was induced by either TLR4 or TLR2 ligands. Importantly, Tnf release in response to TLR2 ligands was not inhibited in TRIF Ϫ/Ϫ or TRAM Ϫ/Ϫ TRIF Ϫ/Ϫ macrophages, consistent with previous reports (25,44).
We have previously shown that TLR2 localizes to early and recycling endosomes (30,45) and that inhibiting endocytosis did not affect NF-B activation and TNF release. Dietrich et al. (43) found that IFN-␤ induction in response to TLR2 ligands Results are representative of three experiments. * denotes below detection limit. TLR2-mediated Ccl4 activation is partially mediated by the TRAM/TRIF pathway. C, Ccl4 induction in wild-type (WT) or TRIF Ϫ/Ϫ peritoneal macrophages treated with medium (0), Pam 3 Cys (P 3 C) (1-0.1 g/ml), Pam 2 Cys (P 2 C) (1-0.1 g/ml), MALP-2 (1-0.1 g/ml), FSL-1 (1-0.1 g/ml), LPS (100 ng/ml), poly(I:C) (5 g/ml), or Sendai virus (10 HAU/ml) for 20 h. Supernatant was analyzed for Ccl4 by ELISA. Results show mean Ϯ S.E. of duplicates from two independent experiments. ***, p Ͻ 0.001; **, p Ͻ 0.01; *, p Ͻ 0.05 (one-way ANOVA with Bonferroni post-test). D, HEK293 cells were transfected with TLR2 and CCL4 reporter luciferase and co-transfected with empty vector (EV) and decreasing amounts of TRAMC117H (20 -2-0.2 ng) overnight. Cells were subsequently stimulated with Pam 3 Cys (100 -10 ng/ml), FSL-1 (100 -10 ng/ml), or Sendai virus (SV) (300 HAU/ml) for 12 h before cells were lysed and assayed for luciferase activity. Results were normalized for Renilla activity and show mean Ϯ S.D. of triplicates. F, IFN-␤ induction in BMDM in response to FSL-1 is mediated by TRAM and TRIF. Induction of Ifn-␤ (E) and Tnf (F) mRNA in BMDM from C57BL/6 wild-type (WT), TRAM Ϫ/Ϫ , TRIF Ϫ/Ϫ , and TLR2 Ϫ/Ϫ mice that were left untreated (gray bars) or stimulated with FSL-1 (10 ng/ml) (black bars) for 3 h. Ifnb1 and tnf mRNA in the samples were determined by RT-qPCR and are presented as relative induction with each of the nontreated cell lines as reference sample. Gapdh served as internal control. Results show mean fold induction Ϯ S.D. of triplicates and are representative of five experiments. FEBRUARY 6, 2015 • VOLUME 290 • NUMBER 6 was efficiently blocked by inhibiting endocytosis. We also observe that TRAM/TRIF-dependent TLR2-mediated Ccl5 induction is impaired upon inhibition of endocytosis, in line with Dietrich et al. (43), whereas TNF release occurs independent of ligand internalization as we previously reported (30). We further show that TLR2 co-localizes with TRAM in early endosomes, in a manner resembling TLR4 (46). TLR2 and TRIF co-localization was observed in TRAM-positive vesicles suggesting that endocytosis is required to deliver TLR2 to endosomes where TRIF resides. Subsequently, TRAF3 may be recruited to initiate downstream signaling leading to IRF3 activation in resemblance with TLR4 (46). Expression of TRAF3 at the plasma membrane has indeed been shown to potentiate TLR2-mediated IFN-␤ induction from the plasma membrane (46).

TRAM and TRIF Mediate TLR2 Signaling
Our results indicate that TLR2 appears to resemble TLR4 with regard to some aspects of localization and the requirement for ligand internalization to activate the TRAM/TRIF pathway. Important differences between TLR2 and TLR4 were, however, also observed in our experimental settings. In contrast to TLR4mediated Ccl5, which is heavily TRIF-dependent, TLR2-mediated Ccl5 is dependent on both the MyD88-and TRIF-dependent pathway. TLR4-mediated Tnf release is controlled by both signaling branches, whereas TLR2-mediated Tnf release is strictly MyD88-dependent and is not influenced by TRAM and TRIF. Cross-talk between the MyD88-and the TRIF-dependent pathway appears to be regulated differently downstream of TLR2 and TLR4. TLR2 ligands also induce lower levels of Ccl5 and IFN-␤ than TLR4 ligands, and the induction of the interferon-inducible genes cxcl10, ifit1, and ifit2 were only weakly induced in response to FSL-1. More studies are needed to investigate the role of TRIF or TRAM in these responses. The levels of IFN-␤ induced by TLR2 ligands may be too low to induce these genes in peritoneal macrophages, although BMDM may induce higher levels of IFN-␤ (43). It was recently reported that the TLR2 ligand Pam 3 Cys induces cxcl10, ifit1, and type I IFNs in a TRIF-dependent manner (47). In that report, Pam 3 Cys was, however, also shown to induce Tnf, Il6, and Il10 by a TRIF-dependent mechanism, which is different from our results. Notably, the latter is in contradiction with early studies of TRIF Lps2/Lps2 mice, which contain a loss-offunction mutation in the gene encoding TRIF (25). We consistently observed normal induction of Tnf, as well as Il6 and Il10 in response to TLR2 ligands in TRIF-deficient cells, which is in line with Hoebe et al. (25). Our results support however the finding by Petnicki-Ocwieja et al. (47), showing that TLR2 ligands induce type I IFNs in a TRIF-dependent manner, but we observed that only a subset of genes, including Ccl4, Ccl5, and IFN-␤, were impaired in TRIF Ϫ/Ϫ cells in response to TLR2 ligands.
We speculate that some of the differences between TLR2 and TLR4 signaling could be due to the relative subcellular location of the receptors. IFN-␤ induction in response to LPS is proposed to originate in endosomes where TRAM is expressed, whereas MyD88 activation and NF-B activation are proposed to originate at the plasma membrane. We have shown that TLR2 is expressed in early endosomes where the receptor colocalizes with TRAM, and that Ccl5 release is impaired upon inhibition of endocytosis, in likeness with TLR4. In contrast to TLR4, which is down-regulated at the cell surface and targeted to endosomes (48), we observe that TLR2 is up-regulated at the plasma membrane. Although TLR2 is expressed in endosomes, we speculate that up-regulation of TLR2 at the plasma membrane in response to stimuli could still limit the amount of TLR2 delivered to the endosomes where activation of TRAM and TRIF seems to occur. This could, in turn, lead to lower levels of TRIF-mediated responses induced by TLR2 ligands, relative to TLR4 ligands. The precise mechanisms differentiating TLR2 and TLR4 signaling and responses have yet to be elucidated.
The IFN-␤ induction observed by Dietrich et al. (43) in response to TLR2 ligands was shown to be mediated by MyD88, Irf1, and Irf7 in BMDM, in a manner resembling TLR7/9 signaling. Irf3 and Irf7 were, however, required for the TLR2-dependent induction of IFN-␤ in response to vaccinia virus, which was observed by Barbalat et al. (41). IFN induction has been proposed to be controlled by different IRFs in different cell types (36), possibly explaining these variations. We observed that IFN-␤ induction was TRAM-and TRIF-dependent, in addition to MyD88-dependent. TLR2-mediated Ccl5 induction was further found to require Irf3, in addition to Irf1, and to some extent Irf7, but not Irf5, in peritoneal macrophages. Thus, TLR2-induced Ccl5 induction mimics TLR4-induced Ccl5 induction with regard to Irf3 dependence (9); however, similarities between TLR2 and TLR7/9 signaling were also observed with regard to Irf1 dependence in this response. IRF1 has been reported to be involved in TLR7/9-MyD88-mediated IFN-␤, CCL5, and IL-12 induction in dendritic cells, but not in the induction of TNF. We observed that TNF release was also partially impaired in response to TLR2 ligands in IRF1 Ϫ/Ϫ cells. The role of IRF1 in TLR signaling is still incompletely understood, although it has been suggested that IRF1 can cooperate with STAT1. It has also been proposed that IRF1 can be recruited to promoter elements of TNF, IL6, IL-12, and other inflammatory genes following cellular exposure to LPS (45). Our results show that TLR2 can utilize both the MyD88-dependent pathway and the TRAM/TRIF pathway leading to induction of cytokines such as CCL5 and IFN-␤. We cannot exclude the possibility that the relative importance of the adaptor molecules in signaling may differ somewhat between different cell types.
In addition to Ccl5 and IFN-␤, we also observed that TLR2 ligands induce Ccl4 in a manner partially dependent on TRIF, suggesting that TRIF plays a role in several TLR2-mediated responses. LPS-induced Ccl4 was also partially impaired in TRIF Ϫ/Ϫ macrophages, suggesting that both the MyD88 and the TRIF pathways contribute to this response.
A role for TRIF in TLR5 signaling in intestinal epithelial cells has also been proposed (49). In that study TRIF was found to mediate NF-B and mitogen-activated protein kinases (MAPK) activation and the induction of Il6, Cxcl1, and Ccl20. We did not observe impaired IL-6 induction in response to TLR2 ligands in TRIF Ϫ/Ϫ cells or impaired Cxcl1 release (data not shown), suggesting differences in the role of TRIF in TLR5 and TLR2 signaling. Choi et al. (49) also reported that flagellin failed to activate IRF3 and did not induce IFN-␤, while we found that TLR2 ligands can activate IRF3 and can induce IFN-␤ in a TRAM/TRIF-dependent manner in BMDM.
Both the TRIF and the MyD88 pathway mediate TLR2-induced Ccl5 release in macrophages, in contrast to Tnf induction, which is tightly controlled by the MyD88-dependent pathway. The reason for this redundancy is still unclear, but it is likely important for induction of certain responses, even when the MyD88 pathway is compromised. Cytokine induction by whole microorganisms are typically mediated by several pattern-recognition receptors. It has recently been reported that a mutant strain of Listeria monocytogenes induces IFN-␤ by a TLR2-TRIF-dependent mechanism (50); however, this response was also shown to be mediated by TLR3, which is known to utilize the adaptor TRIF.
In conclusion, our results provide new insight into the contribution of the MyD88-and TRIF-dependent signaling pathways in response to different TLR ligands, and they show a novel role for TRAM and TRIF in TLR2 signaling. Induction of gene expression by both MyD88 and TRIF pathways may be necessary for optimal host responses toward certain infections. These results have implications for our understanding of TLR-mediated innate immune responses against infectious organisms.