TIR-containing Adapter Molecule (TICAM)-2, a Bridging Adapter Recruiting to Toll-like Receptor 4 TICAM-1 That Induces Interferon-β*

Lipopolysaccharide (LPS) is an agonist for Toll-like receptor (TLR) 4 and expresses many genes including NF-κB- and interferon regulatory factor (IRF)-3/IFN-inducible genes in macrophages and dendritic cells (DCs). TICAM-1/TRIF was identified as an adapter that facilitates activation of IRF-3 followed by expression of interferon (IFN)-β genes in TLR3 signaling, but TICAM-1 does not directly bind TLR4. Although MyD88 and Mal/TIRAP adapters functions downstream of TLR4, DC maturation and IFN-β induction are independent of MyD88 and Mal/TIRAP. In this investigation, we report the identification of a novel adapter, TICAM-2, that physically bridges TLR4 and TICAM-1 and functionally transmits LPS-TLR4 signaling to TICAM-1, which in turn activates IRF-3. In its structural features, TICAM-2 resembled Mal/TIRAP, an adapter that links TLR2/4 and MyD88. However, TICAM-2 per se exhibited minimal ability to activate NF-κB and the IFN-β promoter. Hence, in LPS signaling TLR4 recruits two types of adapters, TIRAP and TICAM-2, to its cytoplasmic domain that are indirectly connected to two effective adapters, MyD88 and TICAM-1, respectively. We conclude that for LPS-TLR4-mediated activation of IFN-β, the adapter complex of TICAM-2 and TICAM-1 plays a crucial role. This results in the construction of MyD88-dependent and -independent pathways separately downstream of the two distinct adapters.

LPS, 1 a constituent of the outer membrane of Gram-negative bacteria activates macrophages and dendritic cells (DCs) in the innate immune system. TLR4 together with MD-2 and CD14 play major roles in the recognition of LPS on membranes and initiate signaling that activates NF-B and interferon-regulatory factor (IRF)-3 (1). The IRF-3 activation is followed by IFN-␤ induction in the cells (2). TLR4 signaling is mediated by an adapter molecule MyD88 for activation of NF-B (2, 3), but MyD88-dependent signaling pathways are not responsible for the DC maturation or IFN-␤ induction caused by LPS (2,3). This point was confirmed by analyses of cells from MyD88deficient mice (4). The cellular responses not induced by MyD88 are called MyD88-independent responses (2, 3) and presumably involve unknown adapters. Thus, DC maturation and IFN-␤ induction should be attributable to the activation of a putative MyD88-independent pathway. TIRAP (5), a Toll-interleukin-1 receptor domain (TIR)-containing adapter also called Mal (6), was initially considered to account for the MyD88-independent pathway. Indeed, a dominant-negative form of TIRAP and a synthetic TIRAP peptide reportedly inhibited IFN-␤ induction and double strand RNAdependent protein kinase (PKR) phosphorylation (5,7). However, TIRAP knockout studies showed that DC maturation and type 1 IFN induction are largely independent of TIRAP and the phenotype of TIRAP Ϫ/Ϫ cells resembled those of MyD88 Ϫ/Ϫ cells (8,9). Thus, the TLR4-TIRAP-mediated activation of the IFN-␤-inducing pathway remain inconclusive, if any, in mouse cells and most human cells in culture.
TLR3 recognizes poly(I-C) (10,11) and activates the IFN-␤ promoter, and to a lesser extent NF-B (12). Type 1 IFN production and DC maturation with distinct properties from the LPS-induced phenotype are representative cellular responses induced by TLR3-mediated DC activation (13). A recent study on TLR3 and its adapter suggested that an as yet unidentified adapter molecule, later named TICAM-1 (12) or TRIF (14), directly binds the TIR of TLR3 and activates IRF-3 independent of MyD88. TICAM-1, however, barely binds the TIR of TLR4 in the yeast two-hybrid system and in direct immunoprecipitation from mammalian cells (12). Nevertheless, both □ S The on-line version of this article (available at http://www.jbc.org) contains Annex Figs. 1-3.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AB091054 and AB100441 for the human and mouse TICAM-2 cDNA sequences, respectively.
¶ Both authors contributed equally to the results of this article. ʈ Present address: Dept. of Biology, Graduate School of Science, Osaka University, Toyonaka, Osaka 560-0043, Japan.
TLR3-and TLR4-mediated IRF-3 activation and induction of IFN-inducible genes were abrogated in TICAM-1-deficient or -mutated cells (15,16). This suggested that another factor or mechanism must be involved in the MyD88-independent pathway of TLR4.
Here, we report the identification of a protein that directly binds TLR4. This protein harbored a TIR domain that has significant homology to TICAM-1, and conferred heterophilic dimerization with TICAM-1. We have named this protein TIRcontaining adapter molecule-2, TICAM-2. Based on its functional and physical association analyses, we demonstrated that TICAM-2 helps TLR4 bind TICAM-1 that subsequently induces LPS-mediated IRF-3 activation in the TLR4-mediated MyD88independent cellular response.

MATERIALS AND METHODS
Cell Culture and Reagents-The human lung fibroblast cell line MRC-5 was obtained from the Riken Cell Bank (Tsukuba, Japan) and the epithelial cell line HeLa was supplied by Tokyo Metropolitan Institute (Tokyo, Japan). Both cell lines were maintained in minimal essential medium supplemented with 5 (HeLa) or 10% (MRC-5) heat-inactivated fetal calf serum (JRH Biosciences) and antibiotics. The mouse macrophage cell line RAW267.4 was maintained in RPMI 1640, 10% fetal calf serum (17). HEK293 cells (obtained from RIKEN Institute, Japan) were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and antibiotics. The HEK293 line used expressed TICAM-2 mRNA but no TLR3 mRNA and an extremely low level of TICAM-1 mRNA (data not shown). Poly(I-C) was purchased from Amersham Biosciences. Polymyxin B, LPS from Escherichia coli serotype 0111:B4, and mouse IgG1 were from Sigma. The mycoplasma lipopeptide MALP-2 was prepared as described (18). These reagents, except for LPS, were treated with polymyxin B (10 g/ml) for 1 h at 37°C before stimulation of cells. MAbs against human TLR3 (TLR3.7) and TLR2 (TLR2.45) were produced in our laboratory as described previously (10,19). MAb against human TLR4 (20) was a gift from Dr. K. Miyake (University of Tokyo). Anti-IRF-3 antibody was used for determination of IRF-3 activation as noted previously (21). Polyclonal antibodies against human TICAM-1 and TICAM-2 were produced in our laboratory according to the reported method (22).
Identification and Cloning of TICAM-2 cDNA-TICAM-2 expressed sequence tags were identified by BLAST search analysis with the human TICAM-1 protein sequence in the NCBI BLAST server using the tblastn program. The full-length TICAM-2 cDNA sequence was predicted by searching for human expressed sequence tags in the DDBJ data base, and the two expressed sequence tag clones, BQ438847 and AW450623, were found to cover the putative full sequence of TICAM-2. The BLAST search revealed that the mouse cDNA sequence (accession number AK080879) showed high similarity to human TICAM-2. Fulllength human TICAM-2 cDNA was obtained from MRC-5 cDNA by RT-PCR using primers F1 and R1 (see Table I). Mouse TICAM-2 cDNA was isolated by RT-PCR using cDNA of NIH3T3 cell lines with the mF1 and mR1 primers ( Table I). The alignment of the adapter molecules was carried out using ClustalW. Expression vectors for TICAM-2 and its mutants were prepared as follows. The cDNA encoding full-length human or mouse TICAM-2 was placed between the XhoI-NotI sites of the pEFBOS plasmid. Regarding human TICAM-2, pEFBOS-TICAM-2 (TIR) was made by inserting the 68 -235-amino acid region of TICAM-2 following the Kozak sequence into the XhoI-NotI site of pEFBOS. pEFBOS-TICAM-2(P116H) and -(C117H) were made from pEFBOS-TICAM-2 by site-directed mutagenesis changing Cys-117 to His and Pro-116 to His. The plasmids were prepared with an endotoxin-free Plasmid Maxi kit (Qiagen).
Northern Blotting and RT-PCR-Human 12-Lane MTN Blot and Blot III membranes (Clontech) were hybridized under stringent conditions with a 32 P-labeled probe of full-length human TICAM-2 cDNA and rehybridized with a labeled ␤Ϫactin probe. The mRNA signals were detected using a BAS2000 image analyzer (Fuji, Japan) with 4 -24 h exposure.
RT-PCR for mRNA identification and semiquantitation were performed with primers for human TICAM-2 or ␤-actin. The primers used are summarized in Table I.
Yeast Two-hybrid Method-Yeast media were made as described previously (23). Yeast AH109 strain (Clontech) was transformed with the bait and prey plasmids. The transformants were streaked and incubated for 3 to 5 days. BD and AD in figures represent the bait and prey plasmids, respectively. BD-TLR2, TLR3, TLR4, TLR5, TLR6, TLR7, TLR8, and TLR9 were made by inserting the TIR domain of TLRs into the pGBKT7 vector (Clontech). The mutant P714H TLR4 was produced according to the previous report (18) and BD-TLR4 P714H (TLR4 PH) was made for this assay. BD-TICAM-2 and AD-TICAM-2 were made by inserting the full-length wild-type TICAM-2 cDNA into the pGBKT7 (bait) or pGADT7 (prey). BD-TICAM-1 was made by inserting a TICAM-1 cDNA fragment (from the ϩ1102to ϩ2142-bp region relative to the start codon). AD-TIRAP and AD-MyD88 were made by inserting the cDNA fragment encoding the full-length protein into the pGADT7 vector. SD-WLH is yeast synthetic dextrose medium lacking Trp, Leu, and His amino acids. SD-WLHA lacks adenine as well. SD-WLH plates are mild stringency plates and SD-WLHA plates are high stringency plates (24).
Assay for IRF-3 Activation-Two methods were employed to assess IRF-3 activation. Native gel assay was performed according to the report by Fujita et al. (21). Reporter gene assay for IRF-3 was performed as follows (26). HEK293 cells (1 ϫ 10 5 ) were transfected with the indicated amounts of the plasmids for TICAM-1 or TICAM-2 together with 5 ng of pCMV-␤, 100 ng of p55 UASG-Luc, and/or 25 ng of Gal4 DBD/IRF-3 in LipofectAMINE 2000 reagent (Invitrogen). Gal4 DBD only was used as a control. At timed intervals (typically 24 h), cells were harvested and the reporters were determined.
RNAi-The mouse macrophage-like cell line RAW264.7 was employed in this study because no good human cell lines were found for RNAi knockdown analysis. Mouse RAW cells (5 ϫ 10 6 ) were plated onto 6-well plates and incubated for 24 h. Then cells were transfected with buffer alone, control siRNA (for human TLR3 or human lamin A/C) (11), siRNA for mouse TICAM-2 or siRNA for mouse TICAM-1 using Oligo-fectAMINE reagent (Invitrogen). Twenty h after transfection, cells were collected after treatment with EDTA (5 mM), saline and plated onto 24-well plates. Cells were 30 -50% confluent the next day. The cells were transfected again with siRNAs using OligofectAMINE, and incubated for 2 days. The culture supernatant was removed and the medium containing 100 ng/ml LPS was added to each well. Before and after 6 h of stimulation, the total RNA from cells was extracted using TRIZOL. The RNA was treated with DNase I (Promega, Madison, WA) to eliminate possible contaminating genomic DNA. No contamination was confirmed by RT-PCR (see Fig. 6). In some experiments, TICAM-2 or TICAM-1 knockdown cells were rescued by transfection with TICAM-1 (20 ng) or TICAM-2 plasmid (50 ng), respectively.
The siRNA for human TLR3 was used as a control. Because the target TLR3 sequence is not conserved between humans and mice, the control siRNA did not down-regulate the mouse TLR3. The oligonucleotides used for RNAi were purchased from JBioS (Saitama, Japan) (Table II).
IRF-3 reporter assay was performed with siRNA-treated cells as follows. RAW cells were twice transfected with siRNA (Table II) in OligofectAMINE (Invitrogen) for gene silencing of TICAM-1 or TI-CAM-2 as described above. Cells were then transfected with reporter genes (10 ng of pCMV-␤, 100 ng of p55 UASG-Luc, and/or 50 ng of Gal4 DBD/IRF-3) by FuGENE 6 (Roche Applied Science). One day after the transfection, cells were stimulated with LPS (100 ng/ml) or polymixin B-treated poly(I-C) (10 g/ml), and 6 h later, the reporters were measured as described above.

Molecular
Cloning of TICAM-2-Because TICAM-1 was identified as an adapter that induces IFN-␤ (12), we searched for sequences that contained a TIR similar to that of TICAM-1 in expressed sequence tag databases and identified a relevant partial cDNA for humans and mouse. The complete sequence of the human protein was confirmed by RT-PCR using the primers covering the open reading frame and named TICAM-2 (TIR-containing adapter molecule-2) (Fig. 1a). This protein consisted of 235 amino acids with a short (ϳ70 amino acids) N-terminal Ser/Thr-rich domain, TIR motif, and a C-terminal 20-amino acid stretch. The N-and C-terminal domains of TI-CAM-2 did not resemble those of TICAM-1. The TIR and Cterminal portion were highly conserved in human and mouse TICAM-2 ( Fig. 1a). Although the TIR motif of TICAM-2 had low similarity to the TIR motifs of the reported adapters TIRAP (5, 6) and MyD88 (4), the N-terminal domain of TICAM-2 shared tandemly arranged basic amino acids and serines with that of TIRAP. Similar to TICAM-1, TICAM-2 lacked the conserved sequences present in other TIR-containing proteins (27), (F/Y)D in box 1, RD in box 2, and FW in box 3 (Fig. 1b). As in MyD88, TIRAP, and TICAM-1, the proline in the BB loop (27), which is essential for the TIR-mediated signaling, was situated in front of the conserved position or replaced with cysteine in TICAM-2 ( Fig. 1b). Identities in amino acid composition among the comparable TIR domains of the reported adapters are shown in Fig. 1c. Fig. 1d is a comparative depiction of TICAM-2 with the known adapters. The TICAM-2 gene was located in chromosome 5q22.
TICAM-2 message expression was analyzed using a TICAM-2 cDNA probe. mRNA species of 3.3-3.6 kb was detected in many tissues, including lymph nodes, prostate, stomach, thyroid, trachea, heart, skeletal muscle, and placenta (Fig. 1e). RT-PCR analysis of immature DCs, macrophages, and NK cells prepared from human peripheral blood and the indicated cell lines also revealed the presence of the mRNA for TICAM-2 (Fig. 1f).
Physical Association of TICAM-2 with TLR4 -Physical binding of TICAM-2 to the TIR domain of TLR4 was confirmed by immunoprecipitation in HEK293 cells expressing TLR2, TLR3, or TLR4 ϩ MD-2, and TICAM-2 with HA-tag (Fig. 3a). A a "r" and "d" represent ribo-and deoxyribonucleotide, respectively.  13. c, alignment of amino acids for identities among the TIR domains of the adapters. d, schematic representation of the four human adapter molecules. The TIR domains are shaded. e, Northern blot of TICAM-2 mRNAs from specified tissues. Northern blots for human tissues (Clontech) were probed with the TICAM-2 or control ␤-actin cDNA. Notice the two band profiles of 3.6 and 3.3 kb in tissues such as lymph node, prostate, stomach, thyroid, muscles, and lung. An additional 1.8-kb band was specific only for placenta. f, TICAM-2 mRNA in various human cells and cell lines. Total RNA was isolated from the indicated types of human cells/cell lines, and RT-PCR was performed using primers for TICAM-2 (35 PCR cycles) or control ␤-actin (20 cycles). iDC, immature DC. molecular complex containing both TLR4 and TICAM-2 was detected by immunoprecipitation using anti-FLAG antibody and immunoblotting using anti-HA antibody (Fig. 3a, lane 6). Association of TICAM-2 with TLR2 or TLR3 was barely detectable under these conditions (Fig. 3a, lanes 1-4). Even when co-expressed with MD-2, TLR2 failed to co-precipitate TICAM-2 (Fig. 3a, lane 3 and 4). These results were confirmed with specific mAbs against TLR2, TLR3, and TLR4 (data not shown). Stimulation of TLR4-expressing cells with LPS had no affect on this molecular association (Fig. 3b). The binding of TICAM-2 to TICAM-1 and TLR4 was tested next using TICAM-1 (HA-tag), TICAM-2 (with HA tag or FLAG tag), and TLR4 (FLAG tag). Results show that TLR4 recruited TICAM-1 only in the presence of TICAM-2 (Fig. 3c, lane 6). Under the same conditions, TLR3 (12) but not TLR4 (data not shown) directly recruited TICAM-1 into a complex. TICAM-2 formed a dimer because TICAM-2-HA and TICAM-2-FLAG were co-precipitated (Fig. 3c, lane 1). TICAM-2⅐TICAM-1 complex was only marginally detected in the absence of TLR4 (Fig. 3c, lane 4). The two TICAM-2 mutants (P116H and C117H) (see Fig. 4b) lost the ability to dimerize with TICAM-1 and TICAM-2 (data not shown) as shown in yeast. Thus, the TICAM-2 mutants, although still bind the tail of TLR4, cannot be coupled with intact TICAM-1, which may reflect its dominant-negative properties. These results are essentially in accord with those observed in yeast (Fig. 2). The P714H mutant of TLR4 did not recruit TICAM-2, thereby disassembling TICAM-1 (Fig. 3d). A similar TLR4-TICAM-1 relationship was confirmed with the specific mAb against TLR4 (data not shown). Thus, simultaneous expression of normal TLR4 is crucial for assembling TICAM-2 and TICAM-1.

FIG. 2. Association between TICAM-2 and TLRs or adapters. a,
interactions between TICAM-2 and the TIRs of TLR2-9 in the yeast two-hybrid system. Strong association was observed between TICAM-2 and TLR4 (SD-WLHA plate), whereas weaker TICAM-2-TLR3 association was observed in SD-WLH plates. b, interaction between TICAM-2 and the tail of TLR4 or its P714H mutant. Strong association was observed between TICAM-2 and the normal TIR of TLR4. The association was not detected between TICAM-2 and the mutant TIR of TLR4 (TLR4PH). TICAM-1 was used as a control and no association was observed between TICAM-1 and two TLR4 TIRs, whereas TICAM-1 strongly bound TLR3. c, interactions between the TICAM-2 or TICAM-2 mutant (TICAM-2 C117H, TICAM-2CH) and other adapter molecules in yeast. TICAM-2 homodimerization and weak TICAM-1-TICAM-2 heterodimerization were observed. TICAM-2CH failed to couple with other intact TICAM-2 or TICAM-1. The TICAM-1-TICAM-2 heterodimerization was confirmed in the WLHA plate (lower panel). The yeast growth was usually observed Ͻ3 days, whereas in this case 6 days were required to confirm the growth. d, interactions between TICAM-2 and the other adapter molecules in yeast. No association was observed between TICAM-2 and MyD88 or Mal/TIRAP under the conditions where TICAM-2 strongly homodimerized. AD-Mal/TIRAP and AD-MyD88 constructs were functional, because we observed the interaction among these prey and some TLR bait plasmids (data not shown). SD-WLH, reflecting weak interaction; SD-WLHA, reflecting strong interaction. BD and AD represent bait and prey plasmids, respectively.
TICAM-2 Is Involved in TLR4-mediated IFN-␤ Production-To clarify the involvement of TICAM-1/TICAM-2-mediated IFN-␤ production, TICAM-2-knockdown cells were generated with mouse RAW264.7 macrophage-like cells by siRNA transfection and tested for IFN-␤ induction in response to LPS (Fig. 6). RAW cells were transfected twice with buffer, nonspecific siRNAs, or two siRNAs for mouse TICAM-2 site A or B (Fig. 6a). The transfected cells were stimulated with LPS and total RNAs were extracted before or 6 h after stimulation, and IFN-␤ levels induced by LPS were assessed by quantitative PCR. Results show that the mRNA levels of TICAM-2 were suppressed by ϳ60 or 26% by the site A or site B RNAi targeting, respectively (Fig. 6b). TICAM-2 knockdown was accompanied by specific inhibition of IFN-␤ induction by ϳ60 and 35% by site A or site B siRNA, respectively (Fig. 6c). Functional uncoupling of TICAM-2 with TLR3 was again confirmed in TICAM-2-knockdown cells (Supplemental Materials Annex 3).
LPS-stimulated induction of IFN-␤ in TICAM-1-depleted RAW cells was examined by a similar RNAi method. The level of TICAM-1 mRNA was suppressed by ϳ70% by the mouse TICAM-1-specific siRNA in RAW cells (data not shown). Such TICAM-1 knockdown suppressed LPS-mediated IFN-␤ induction irrespective of the level of TLR3 (Fig. 6d). TICAM-2 site B knockdown also reduced the IFN-␤ mRNA level by Ͼ60% (Fig.  6d). Thus, both TICAM-1 and TICAM-2 are required for full induction of LPS-mediated IFN-␤ in macrophage-like cells. LPS-mediated activation of IRF-3 was severely impaired in TICAM-1-or TICAM-2-knockdown cells paralleling suppression of IFN-␤ induction (Fig. 6e). In TICAM-2 knockdown cells, transfection of TICAM-1 resulted in recovery of IRF-3 activation that is presumably its dimer formation, but in TICAM-1 knockdown cells, transfection of TICAM-2 virtually did not (Fig. 6f). These results reinforce our interpretation of the TICAM-1/2 dominant-negative overexpression analyses (Fig.   4): the TICAM-2 is required for the LPS-TLR4-mediated MyD88-independent pathway and TICAM-1 is the critical factor for IRF-3 activation in this pathway. Of note, the tumor necrosis factor-␣ levels were reduced at most by 15% in TICAM-2-knockdown cells (not shown), whereas Ͼ60% were reduced in TIRAP-or MyD88-knockdown cells made by the method previously reported (Ref. 12, data not shown). These results are consistent with those obtained with MyD88-deficient and TIRAP-deficient mouse cells (8,9). Absolute requirement of TICAM-1 for the MyD88-independent pathway was later confirmed by TRIF (TICAM-1)-deficient or -mutant mice (15,16). Our results revealed the missing link between the essential TICAM-1 for LPS-TLR4-mediated signaling (15,16) and no physical link of TICAM-1 to TLR4 (12). This, together with the biochemical results (Fig. 3), implies that TICAM-2 is an adapter that physically bridges the TIR domain of TLR4 and TICAM-1 for LPS-mediated activation of IRF-3 followed by induction of IFN-␤ as part of the MyD88-independent pathway. DISCUSSION We have identified a novel adapter TICAM-2 that participates in LPS-TLR4-mediated IFN-␤ promoter activation. TICAM-2 is not ubiquitously expressed and TICAM-2 per se even overexpressed barely activates AP-1 and only weakly activates NF-B and the IFN-␤ promoter. TICAM-2 binds tightly to the TIR domains of TLR4 and TICAM-1, but has weak or no binding affinity for other TLRs and adapters. Requirement of both TICAM-2 and TICAM-1 to complete the LPS-TLR4-triggered IFN-␤ production was finally confirmed by the RNAi-knockdown analysis (Fig. 6). The point that TICAM-1 is the crucial effector in TLR3 and TLR4 was con-firmed after the submission of this study by two groups using gene disruption and forward genetics (15,16).
TICAM-2 physically binds TLR4 and TICAM-1. In the presence of endogenous TICAM-1, TICAM-2 causes IRF-3 activation followed by activation of IFN-␤ promoter. Virtually no activation of the IFN-␤ promoter was observed when TICAM-2 was supplemented to TICAM-1-knockdown cells. These, together with the profile of TICAM-2-interacting molecules and  Table II. b, gene silencing of TICAM-2 with siRNA. Mouse RAW264.7 cells were transfected twice with buffer, the control siRNA, or the indicated siRNA for mouse TICAM-2 (site A or site B). Four days after the first transfection, the cells were stimulated with LPS for 0 or 6 h, and then the total RNA of the cells was extracted. The isolated RNA was treated with DNase I, and probed with random primer-labeled cDNA. PCR were carried out using cDNA as templates and the primers for mouse IFN-␤ (mIFN-F1 and mIFN-R1, 30 PCR cycles), ␤-actin (mBACT-F1 and mBACT-R1, 22 PCR cycles), or mouse TICAM-2 (mF1 and mR1, 27 PCR cycle). PCR products were separated by electrophoresis on a 1% agarose gel and stained with ethidium bromide. Primer sequences are shown in Table I. c, impaired IFN-␤ induction in RAW cells with TICAM-2 depletion. The cDNA prepared as described in b was used as the template for determination of the IFN-␤ mRNA levels by quantitative PCR. Values were normalized by the division of the copy numbers of IFN-␤ mRNA and ␤-actin. Relative copy numbers were calculated by dividing each value by the mean value of the mock transfection. Data are a mean of three experiments. d, suppression of IFN-␤ message levels by silencing of TICAM-1 or TICAM-2 mRNA in RAW cells. RNAi gene silencing was performed for mouse TICAM-1 as in b. The levels of mRNA for mouse IFN-␤ were reduced by ϳ70% by TICAM-1 silencing and ϳ80% by TICAM-2 site B silencing (data not shown). Six h after LPS stimulation, the induction of IFN-␤ was examined by quantitative PCR. Relative copy numbers were calculated as in c. SiRNAs used are indicated at the bottom panels. Human lamin A/C and human TLR3 siRNAs (12) were used as controls. One of three similar experiments is shown. e, suppression of LPS-mediated IRF-3 activation by gene silencing of either TICAM-1 or TICAM-2 in RAW cells. RAW cells were twice transfected with siRNA (Table II) for gene silencing of TICAM-1 or TICAM-2. Cells were then transfected with reporter genes (see the "Materials and Methods"). One day after the transfection, cells were stimulated with LPS and 6 h later, the reporters were measured. f, no recovery of IFN-␤ induction by transfer of TICAM-2 into TICAM-1-knockdown cells. RAW cells treated with the TICAM-2 or TICAM-1 siRNA were transfected with murine TICAM-1 (20 ng) or TICAM-2 plasmid (50 ng). Six h later, the induction of IFN-␤ was examined by quantitative PCR. Relative copy numbers were calculated as in c. The level of IFN-␤ in mock (instead of siRNA) transfected cells was evaluated as 1.0. Control, vector only was added. functional analyses including the IRF-3 assay, suggest that the TICAM-2⅐TICAM-1 complex is the adapter that mediates TLR4-triggered activation of IRF-3. Collectively, the most marked properties of TICAM-2 are to physically bridge TLR4 and TICAM-1 and transmit TLR4 signaling to the downstream TICAM-1. TICAM-1 is the effecter that exerts potent IFN-␤ promoter activation in TLR4, as well as TLR3 (12,14). If at all, the induction of IFN-␤ is largely supported by TICAM-1-mediated IRF-3 activation in these TLRs. The IFN-␤ promoter appears to be more potently activated by transfection with TICAM-2 and TICAM-1 than the IRF-3-responsive element (Fig. 4g versus Fig. 5b). It has been reported that in the TLR3 pathway NF-B and MAP kinases are activated through an IRAK-independent pathway that contains PKR (28). Thus, other factors than IRF-3 may be involved in the TICAM-1-dependent IFN-␤ induction. At least, TIRAP and MyD88 are not responsible for relevant IFN-␤ induction (8,9), although they induce robust NF-B activation with virtually no IRF-3 activation.
Direct binding of TICAM-1 to the tail of TLR3 has been demonstrated in a previous report (12,14). We have not observed any involvement of TICAM-2 in poly(I-C)-TLR3-mediated IRF-3 activation (Fig. 5c, and Supplemental Materials Annex 3). Thus, at least two modes, direct and indirect binding, sustain the molecular association between TLRs and TICAM-1 (29). Currently, TLR3, TLR4, TLR7, and TLR9 have been reported to induce type 1 IFN. TLR3 provides a typical instance of the direct binding mode and TLR4 represents the indirect binding mode. TICAM-1 is phosphorylated during TLR3 stimulation (12) and we noticed that TICAM-2 is also phosphorylated during TLR4 stimulation. Signal transduction may be retarded and vulnerable for modulation at multiple checkpoints in the indirect mode (30,31). The modulation of LPS signaling would be related to an unknown part of LPS tolerance (4,32). Requirement for TLR signaling in the indirect mode remains to be further investigated.
An adapter protein TIRP (33) was reported after completing this study. TIRP is structurally identical to our reported protein TICAM-2. TIRP has a TRAF6-binding motif PXEXX and actually interacts with TRAF6 (33). Overexpression and dominant-negative analysis in HEK293 cells suggested that TIRP is associated with interleukin-1 receptors and also interacts with kinase-inactive mutants of IRAK-4, IRAK-2, and IRAK-M. The authors reasoned that TIRP functions upstream of IRAK and TRAF6 in the interleukin-1-induced NF-B activation pathway (33). If this is the case, TIRP/TICAM-2 acts encompassing the interleukin-1-NF-B pathway and the TLR-IFN-␤ pathway. Their results opposing those of our TICAM-2 are that TIRP could not activate the IFN-␤ promoter (data not shown) and their dominant-negative TIRP (TIRP-(78 -171)) did not inhibit TLR4-mediated NF-B activation. Perhaps, the point mutation at Pro-116 or Cys-117 is essential for the relevant dominant-negative function (Supplemental Materials Annex 2). However, we have no idea in terms of the discrepancy on the TIRP/TICAM-2 function of IFN-␤ promoter activation. TIRP/ TICAM-2 may be a regulatory adapter that bridges TIR domains and other molecules. Gene disruption studies will further clarify the unique bimodal function of TIRP/TICAM-2.
Thus far, Gram-negative bacterial LPS, lipid A, the antitumor agent Taxol, and the respiratory syncytial virus (respiratory syncytial virus) F protein are agonists for TLR4 (34,35). The cell wall skeleton of bacillus Calmette-Guérin (BCG-cell wall skeleton) (36) and listeriolysin O (37) activates TLR2 and TLR4. It is of interest whether the TLR4 ligands other than LPS (including the F protein of respiratory syncytial virus) can induce the genes regulated by IRF-3 and IFN-␤ (38). Some TLR4 agonists such as BCG-cell wall skeleton do not up-regulate the IFN-inducible genes mediated via TICAM-1 but activate NF-B via the MyD88-dependent pathway in macrophages and DCs (7,39). Hence, the properties of agonists may affect the selection of the downstream signaling in TLR4. If this is the case, LPS rather than TLR4 itself is fastidious in IFN-␤ signaling.
The intermolecular complex formed by TLRs, TICAM-2, and TICAM-1 appears analogous to the intermolecular association among TLRs, TIRAP, and MyD88 (29). TIRAP Ϫ/Ϫ cells show a loss-of-function phenotype similar to MyD88 Ϫ/Ϫ cells (8,9). TIRAP probably serves as a linker between TLR2/4 and MyD88, and MyD88 acts as an effecter for activation of downstream signaling. Structural findings also support this idea. MyD88 possesses the death domain (40) that adheres to the death domain of IRAK4 in the IRAK⅐TRAF6 complex leading to the activation of NF-B and p38 MAPK/AP-1 (41). On the other hand, TIRAP resembles TICAM-2 in its overall features and has no particular domains in its N and C termini (8,9), endorsing their common properties as bridging adapters. Unlike TICAM-2, TICAM-1 has unique proline-rich motifs in its N and C termini, and novel molecules bind these stretches. 2 Based on the topology of TIRAP and MyD88, we can delineate the molecular association between TICAM-2 and TICAM-1 and thus propose the two important pathways downstream of TLR4 (29). TLR4 recruits two bridging adapters, making it possible to affect NF-B/MAPK (AP-1) and IRF-3 activation. This is why LPS induces a larger variety of cellular responses than other agonists of TLRs.
LPS is highly toxic to mammals. The first Toll discovered in humans was TLR4 (42). Identification of an LPS-resistant gene as a mutated TLR4 by positional cloning (43) is an important landmark for its physiological importance. In fact, the P714H mutant of TLR4 failed to recruit both TICAM-2/TICAM-1 (Fig.  3d) and TIRAP/MyD88 (3,44). Molecular complexing involving TLR4 is the most complicated compared with other TLRs. A stable TLR4 dimer is present in conjunction with MD-2 (45) and signals the presence of LPS complexed with lipopolysaccharide-binding protein or bactericidal/permeability increasing protein (46). CD14 on the same cell membrane binds the LPS complex and in turn transfers LPS to TLR4 (47). MD-2 intrinsically confers species specificity on the recognition of lipid A (48), the active principle of LPS. To our knowledge on the TLR4 scenario, LPS responsiveness is highly sophisticated, which is sustained by a receptor complex encompassing intra-and extracellular regions. At least, the fish lacks TLR4 (49) and TICAM-2 3 and LPS sensitivity is incomplete in reptiles, amphibians, and fish (50). The indirect mode of TLR4 signaling probably emerged as a result of evolution of the TLR system.
Many molecules other than those in the TICAM-1-related pathway are implicated in IRF-3 activation. Proteins with RNA-binding motifs, such as PKR (21) and PACT (51), can sense poly(I-C) to induce IFN-␤ inside cells and are activated secondarily to sensing viral replication. Virus-associated kinases (52) have not yet been identified but are present as activators of IRF-3. At present it is not possible to fully depict the IRF-3 activation pathway for IFN-␤ induction at the molecular level (53). However, TICAM-1-related pathways are being elucidated through recent topics: two reports implicated in the TICAM-1/2-related pathways were recently published (39,54). The first report is that two TANK-binding kinases, TBK-1 and iB kinase ⑀, are involved in TICAM-1-mediated activation of IRF-3 and NF-B (54). The TANK-binding kinase complex act upstream of IRF-3 (54) and second report (55) suggested that it is a component of virus-associated kinase. A tantalizing point is whether TICAM-1 can directly bind and activate TBK-1 and iB kinase ⑀. Although TRIF (14), i.e. TICAM-1, reportedly coprecipitated with IRF-3, we believe that some other molecules are positioned between TICAM-1 and IRF-3. Some of the TICAM-1-binding proteins we collected would be candidates to satisfy the missing link between TICAM-1 and the TANK-binding kinases.