Mechanisms of the TRIF-induced interferon-stimulated response element and NF-kappaB activation and apoptosis pathways.

Toll-like receptor-3 is critically involved in host defense against viruses through induction of type I interferons (IFNs). Recent studies suggest that a Toll/interleukin-1 receptor domain-containing adapter protein (TRIF) and two protein kinases (TANK-binding kinase-1 (TBK1) and IkappaB kinase (IKK)-epsilon) are critically involved in Toll-like receptor-3-mediated IFN-beta production through activation of IFN regulatory factor (IRF)-3 and IRF-7. In this study, we demonstrate that TRIF interacts with both IRF-7 and IRF-3. In addition to TBK1 and IKKepsilon, our results indicate that IKKbeta can also phosphorylate IRF-3 and activate the IFN-stimulated response element. TRIF-induced IRF-3 and IRF-7 activation was mediated by TBK1 and its downstream kinases IKKbeta and IKKepsilon. TRIF induced NF-kappaB activation through an IKKbeta- and tumor necrosis factor receptor-associated factor-6-dependent (but not TBK1- and IKKepsilon-dependent) pathway. In addition, TRIF also induced apoptosis through a RIP/FADD/caspase-8-dependent and mitochondrion-independent pathway. Furthermore, our results suggest that the TRIF-induced IFN-stimulated response element and NF-kappaB activation and apoptosis pathways are uncoupled and provide a molecular explanation for the divergent effects induced by the adapter protein TRIF.

out and chemical mutagenesis approaches (19,20). These studies indicated that TRIF-deficient mice are defective in both TLR3-and TLR4-mediated expression of IFN-␤. Inflammatory cytokine production in response to the TLR4 ligand (but not to other TLR ligands) is severely impaired in TRIF-deficient macrophages. The major signaling pathways leading to production of inflammatory cytokines involve the transcription factor NF-B and various mitogen-activated protein kinases. It was shown that TLR3-mediated NF-B activation is severely impaired in TRIF Ϫ/Ϫ fibroblasts, suggesting that TRIF plays a major role in TLR3-induced NF-B activation. In contrast, TLR4-mediated NF-B and JNK activation is not significantly impaired in either MyD88-or TRIF-deficient fibroblasts. However, double knockout of MyD88 and TRIF completely abolishes TLR4-induced NF-B and JNK activation (19,20), suggesting that MyD88 and TRIF are mutually required for TLR4induced NF-B and JNK activation.
Interestingly, although poly(I⅐C)-induced tumor necrosis factor production is abolished in all macrophages, LPS-induced tumor necrosis factor production is abolished only in a fraction of macrophages (19). These data suggest that LPS response in a subset of macrophages is TRIF-independent. Because LPS response in all macrophages is MyD88-dependent, it has been proposed that an unknown adapter protein functioning downstream of MyD88 may be involved in LPS signaling in TRIFindependent cells (19). In this context, the recently identified protein TIRP (12), which is mostly related to TRIF, provides a candidate adapter for the TRIF-independent signaling in the subset of macrophages.
Although TRIF contains a TIR domain, its amino acid sequence is very divergent from other TIR domain-containing adapter protein sequences. The unique sequence and functions of TRIF suggest that it may interact with distinct downstream signaling proteins. It has been shown that TRIF interacts with IRF-3 to activate IFN-␤ (16). Recently, it has been shown that two serine/threonine protein kinases (IKK⑀ and TANK-binding kinase-1 (TBK1)) can phosphorylate IRF-3 and are involved in TRIF-induced IRF-3 activation (21,22).
Elucidation of the molecular mechanisms of TRIF-induced downstream signaling events is crucial for understanding TLR3-and TLR4-mediated biological effects. In this work, we show that TRIF induces ISRE and NF-B activation and apoptosis through distinct intracellular signaling pathways.

EXPERIMENTAL PROCEDURES
Reagents-Monoclonal antibodies against the FLAG and hemagglutinin (HA) (Sigma) and Myc (Santa Cruz Biotechnology, Santa Cruz, CA) epitopes, the caspase inhibitor ZAD-fm (Enzyme Systems, Livermore, CA), and Sendai virus (American Type Culture Collection, Manassas, VA) were purchased from the indicated manufacturers.
Yeast Two-hybrid Screening-To construct a TRIF bait vector, a cDNA encoding full-length TRIF was inserted in-frame into the Gal4 DNA-binding domain vector pGBT (Clontech, Palo Alto, CA). The human B cell cDNA library (American Type Culture Collection) was screened as described (23,24).
Constructs-The ISRE-luciferase reporter construct was purchased from Stratagene. The NF-B-luciferase reporter construct was provided by Dr. Gary Johnson (University of Colorado Health Sciences Center). The IFN-␤-luciferase reporter construct was made by PCR amplification of the human IFN-␤ promoter fragment (Ϫ300 to ϩ25) and cloning into the pGL3-Basic vector (Promega, Madison, WI). The NF-B site-mutated IFN-␤ promoter-luciferase reporter construct was made by PCR.
All reporter gene assays were repeated at least three times. Data shown are the average from one representative experiment. The S.D. values were Ͻ10% of the average values for all samples.
Co-immunoprecipitation and Western Blot Analysis-For transient transfection and co-immunoprecipitation experiments, 293 cells (ϳ2 ϫ 10 6 ) were transfected for 24 h. Transfected cells were lysed in 1 ml of lysis buffer (15 mM Tris, 120 mM NaCl, 1% Triton, 25 mM KCl, 2 mM EGTA, 2 mM EDTA, 0.1 mM dithiothreitol, 0.5% Triton X-100, 10 g/ml aprotinin, 10 g/ml leupeptin, and 0.5 mM phenylmethylsulfonyl fluoride, pH 7.5). For each immunoprecipitation, a 0.4-ml aliquot of lysate was incubated with 0.5 g of the indicated monoclonal antibody or control mouse IgG and 20 l of a 1:1 slurry of GammaBind G Plus-Sepharose (Amersham Biosciences) for 2 h. The Sepharose beads were washed three times with 1 ml of lysis buffer and 500 mM NaCl. The precipitates were fractionated by SDS-PAGE, and subsequent Western blot analysis was performed as described (12,24). All immunoprecipitation experiments were repeated at least three times, and similar data were obtained.
In Vitro Kinase Assays-Cell transfection and immunoprecipitation were carried out as described above. The immunoprecipitates were washed twice with kinase buffer (25 mM Tris, 5 mM ␤-glycerophosphate, 2 mM dithiothreitol, 0.1 mM Na 3 VO 4 , and 10 mM MgCl 2 , pH 7.5) and then incubated in 30 l of kinase buffer plus 100 M ATP and 10 Ci of [␥-32 P]ATP for 30 min at 30°C. The beads were spun down and washed once with kinase buffer. The proteins were fractionated by SDS-PAGE. The gel was dried and subjected to autoradiography.
Apoptosis Assays-␤-Galactosidase cotransfection assays for determination of cell death were performed as described (23,27). Briefly, cells (ϳ1 ϫ 10 5 /well) were seeded on 12-well dishes and transfected the following day by the standard calcium phosphate precipitation method (26) with 0.1 g of pCMV-␤-galactosidase and 1 g of the indicated expression plasmids. Where necessary, empty control plasmid was added to ensure that each transfection received the same amount of total DNA. In the same experiment, each transfection was performed in duplicate. Twenty-four hours after transfection, the cells were fixed and stained with X-gal as described previously (23,27). The numbers of surviving blue cells from five representative viewing fields were counted under a microscope. The dead cells with a shrunk and condensed morphology were not counted. Data from one of at least three representative experiments are shown.
Sendai Virus Infection-293 cells (ϳ2 ϫ 10 5 ) were seeded in 6-well dishes and transfected the next day with 0.5 g of ISRE reporter plasmid and 0.5 g of pRSV-␤-galactosidase plasmid. Eighteen hours after transfection, cells were washed with medium lacking fetal calf serum (washing medium) and then overlaid with washing medium containing Sendai virus at a multiplicity of infection of 10. After incubation at 37°C for 60 min, non-adsorbed viruses were removed by repeated washing of the cells. Cells were then cultured in fetal calf serum-containing medium for 18 h before luciferase and ␤-galactosidase assays were performed.
DNA Fragmentation Assays-Transfected cells were washed twice with cold phosphate-buffered saline and lysed in 2% Nonidet P-40 containing 0.2 mg/ml proteinase K. DNA in the lysate was precipitated with 2 volumes of ethanol. The pellets were dissolved in H 2 O and analyzed by electrophoresis on 2% gel.

RESULTS
TRIF Interacts with IRF-7 and IRF-3-To unambiguously identify TRIF-interacting proteins, we used the yeast two-hy-brid system to screen a human B cell cDNA library with fulllength TRIF as bait. We screened a total of ϳ2 ϫ 10 6 independent library clones and obtained 20 ␤-galactosidase-positive Co-immunoprecipitation and Western blot analysis were performed as described for A. C, the TIR domain of TRIF is required for interaction with IRF-7. 293 cells (ϳ2 ϫ 10 6 ) were transfected with expression plasmids for HA-IRF-7 and the indicated FLAG-tagged TRIF mutants (Mut). Co-immunoprecipitation and Western blot analysis were performed as described for A. D, TRIF and TIRAP interact with IRF-3. Transfection and co-immunoprecipitation were performed as described for A, except that FLAG-IRF-3 was used to replace FLAG-IRF-7⌬N. E, the TIR domain of TRIF is required for interaction with IRF-3. The experiments were performed as described for C, except that IRF-3 was used to replace  clones. One of the clones encodes the C-terminal domain (amino acids 297-503) of IRF-7. Previously, it has been shown that IRF-3 interacts with TRIF (15,16). We next investigated whether IRF-7 is involved in TRIF signaling.
We first determined whether the IRF-7 fragment obtained from the yeast two-hybrid screening, which lacks its N-terminal DNA-binding domain (IRF-7⌬N), interacts with TRIF. We transfected expression plasmids for TRIF and IRF-7⌬N into 293 cells and performed co-immunoprecipitation experiments.
We then determined the effects of IRF-7 and IRF-3 lacking their DNA-binding domains (mutants IRF-7⌬N and IRF-3⌬N, respectively) on TRIF-induced IFN-␤ activation. In reporter gene assays, both IRF-7⌬N and IRF-3⌬N strongly inhibited TRIF-induced IFN-␤ activation (Fig. 2C). Because TRIF activates NF-B and induces apoptosis (see below), we determined whether CrmA and a non-degradable IB␣ mutant, which inhibit TRIF-induced apoptosis and NF-B activation, respectively (see below), could inhibit TRIF-induced IFN-␤ activation. The results show that CrmA did not inhibit TRIF-induced IFN-␤ activation (Fig. 2C). Surprisingly, the IB␣ mutant completely inhibited TRIF-induced IFN-␤ activation (Fig. 2C). Using an independent ISRE reporter that contains five copies of consensus ISRE-binding sites, we confirmed that TRIF-induced ISRE activation was inhibited by IRF-7⌬N and IRF-3⌬N, but not by CrmA and IB␣(SS/AA). Taken together, our data suggest that IRF-7 and IRF-3 are independently involved in TRIF-induced ISRE activation and that TRIF-induced ISRE activation is uncoupled from TRIF-induced NF-B activation and apoptosis (also see below).
Molecular Order of ISRE-activating Kinases-Because IKK␤-KA could inhibit TRIF-and Sendai virus-induced ISRE activation, we investigated whether IKK␤ could activate ISRE. Consistent with previous reports (20,21), we found that TBK1 and IKK⑀ could activate ISRE (Fig. 4, A and C). In contrast with the previous reports, we found that IKK␤ could also activate ISRE in the same experiments (Fig. 4B). In fact, IKK␤induced ISRE activation was relatively more potent than TBK1-induced ISRE activation in our experiments (Fig. 4, A  and B). In these experiments, IKK␣ did not activate ISRE (data not shown). These data are consistent with our observation that IKK␤-KA can inhibit TRIF-and Sendai virus-induced ISRE activation.
To determine the reason for the discrepancy in the effects of IKK␤ on ISRE activation observed by others and us, we determined whether IKK␤ could activate reporter constructs driven by the IFN-␤ promoter and the NF-B site-mutated IFN-␤ promoter, which are similar to those used in previous studies (20,21). Consistent with the previous reports, IKK⑀ significantly activated these reporters, whereas IKK␤ barely did (Fig.  4D). These studies suggest that IKK␤ may have differential effects on different ISRE reporters.
Differential Phosphorylation of IRF-3 by IKK␤ and IKK⑀-Previously, it has been shown that phosphorylation of IRF-3 and IRF-7 is required for their activation (29,30). We performed in vitro kinase assays to evaluate the abilities of TBK1, IKK␤, and IKK⑀ to phosphorylate IRF-3 and IRF-7. Because full-length IRF-7 is highly degradable (31), we used IRF-7⌬N for these experiments. Previously, it has been shown that the C-terminal domain (but not the DNA-binding domain) of IRF-7 is phosphorylated following viral stimulation (32). We transfected IRF-7⌬N together with TRIF, TBK1, IKK␤, or IKK⑀ into 293 cells and performed in vitro kinase assays. The results indicate that IRF-7⌬N itself was phosphorylated and that overexpression of TRIF, TBK1, IKK⑀, and IKK␤ did not increase its phosphorylation (data not shown). In similar experiments, IKK⑀ potently phosphorylated IRF-3, whereas TRIF, TBK1, and IKK␤ caused weaker phosphorylation of IRF-3 (Fig. 5A). These data suggest that both IKK␤ and IKK⑀ can phosphorylate IRF-3.
Previously, it has been shown that five conserved Ser/Thr residues located in amino acids 396 -405 at the C terminus of IRF-3 are important for IRF-3 phosphorylation and activation induced by viral infection and overexpression of IKK⑀ (22,33). Surprisingly, we found that an active IRF-3 mutant (IRF-3/5D) in which the five serine residues are mutated to aspartic acids could still be phosphorylated by IKK⑀, but not by IKK␤ (Fig.  5B). These data suggest that IKK⑀ can phosphorylate IRF-3 at additional site(s) other than the five Ser/Thr residues, whereas IKK␤ phosphorylates site(s) belonging to the five conserved Ser/Thr residues.
Taken together, our data suggest that TBK1 and IKK⑀ are dedicated to TRIF-induced ISRE activation pathways, whereas IKK␤ is involved in both TRIF-induced ISRE and NF-B activation pathways. In addition, our data suggest that TRIFinduced NF-B activation is uncoupled from TRIF-induced apoptosis (also see below). (34). Although TRIF contains a TIR domain, it does not contain a recognizable death domain. However, overexpression of TRIF potently induced apoptosis in 293 cells (Fig. 7, A  and B). Domain mapping experiments indicated that the Nterminal and middle TIR domains could not induce apoptosis, whereas the C-terminal 181 amino acids (positions 532-712) were sufficient to induce apoptosis (Fig. 7, A and B).

TRIF Induces Apoptosis through RIP/FADD/Caspase-8-mediated Pathways-Infection of cells with Sendai virus induces apoptosis
TRIF-induced apoptosis was not inhibited by IB␣(SS/AA), IRF-3⌬N, and IRF-7⌬N or by their combination (Fig. 7, C and  D). TRIF-induced apoptosis was also not inhibited by kinaseinactive mutants of TBK1, IKK␤, and IKK⑀ (data not shown). These data suggest that TRIF-induced apoptosis is independent of TRIF-induced NF-B and ISRE activation.
Interestingly, TRIF-induced apoptosis was potently inhibited by a dominant-negative mutant of FADD, a caspase-inactive mutant of caspase-8, and the caspase inhibitor CrmA (Fig.  7, C and D). These data suggest that TRIF induces a FADD/ caspase-8-dependent apoptosis pathway, which is also utilized by death receptors.
We next determined whether TRIF interacts with FADD. In transient transfection and co-immunoprecipitation experiments, we failed to detect an interaction between TRIF and FADD (data not shown). However, we found that TRIF could interact with another death domain-containing protein (RIP) in these assays (Fig. 7, F and G). Domain mapping experiments indicated that the TIR and C-terminal domains of TRIF could independently interact with RIP (Fig. 7, F and G). In cell death assays in 293 cells, dominant-negative mutants of FADD and caspase-8 and the caspase inhibitor CrmA inhibited RIP-induced apoptosis (Fig. 7H). Taken together, these data suggest that TRIF induces an RIP/FADD/caspase-8-dependent and mitochondrion-independent apoptosis pathway and that the TRIF-induced apoptosis pathway is uncoupled from TRIF-induced ISRE and NF-B activation. DISCUSSION Recognition of specific patterns of microbial components by TLRs in the body leads to activation of innate immunity against pathogens. TLR3 recognizes double-stranded RNA generated during viral infection and is critically involved in host defense against viruses (35). Activation of TLR3 results in induction of type I IFNs, including IFN-␤ and IFN-␣ family cytokines, which are crucial mediators of the antiviral response.
Recently, it has been shown that a TIR domain-containing adapter protein (TRIF) is critically involved in TLR3-induced IFN-␤ production and NF-B activation (15,16). TRIF interacts with IRF-3, a transcription factor that binds to ISRE. It has also been shown that two Ser/Thr kinases (TBK1 and IKK⑀) can phosphorylate IRF-3 and IRF-7 and that these processes are important for TRIF-induced IFN-␤ activation (20,21). In this study, we further analyzed the mechanisms of TRIF-induced downstream signaling pathways.
In a yeast two-hybrid screening, we found that TRIF could directly interact with IRF-7. This interaction was confirmed by co-immunoprecipitation in mammalian cells (Fig. 1). Domain mapping experiments indicated that the TIR domain of TRIF was required for its interaction with both IRF-7 and IRF-3 (Fig.  1). TRIF could synergize with IRF-7 and IRF-3 to activate the wild-type and NF-B site-mutated IFN-␤ promoters (Fig. 2). In addition, dominant-negative mutants of IRF-7 and IRF-3 could inhibit TRIF-and Sendai virus-induced IFN-␤ and ISRE activation (Fig. 3). These data suggest that TRIF can signal IFN-␤ activation through either IRF-7-or IRF-3-mediated ISRE activation.
Previous reports suggest that TBK1 and IKK⑀ (but not IKK␤) can activate IRF-3 and IRF-7 (20,21). Inconsistent with the previous reports, we found that TBK1 and IKK⑀ (but not IKK␤) could activate the IFN-␤ promoter (Fig. 4D). However, using a commercially available ISRE reporter that contains five copies of consensus IRSE-binding sites, we found that IKK␤ could activate ISRE as potently as TBK1 and IKK⑀ (Fig.  4). That IKK␤ plays a role in virus-induced ISRE activation is further supported by our observations that a kinase-inactive mutant of IKK␤ could block TRIF-and Sendai virus-induced ISRE activation (Fig. 3). In addition, dominant-negative mutants of IRF-3 and IRF-7 completely inhibited IKK␤-induced ISRE activation (Fig. 4B).
In in vitro kinase assays, both IKK␤ and IKK⑀ could phosphorylate IRF-3 (Fig. 5). Previously, it has been shown that five conserved Ser/Thr residues located in amino acids 396 -405 at the C terminus of IRF-3 are important for IRF-3 phosphorylation and activation induced by viral infection (33) and overexpression of IKK⑀ (22). Surprisingly, we found that an IRF-3 mutant (IRF-3/5D) in which the five Ser/Thr residues in amino acids 396 -405 are mutated to aspartic acids could still be phosphorylated by IKK⑀, but not by IKK␤ (Fig. 5B). These data suggest that IKK⑀ can phosphorylate IRF-3 at additional site(s) other than the five Ser/Thr residues, whereas IKK␤ phosphorylates site(s) belonging to the five conserved Ser/Thr residues within amino acids 396 -405. Our data are consistent with a recent study in which it has been shown that IKK⑀ phosphorylates IRF-3 at Ser 386 (36). Our data point to the possibility that IRF-3 is differentially regulated by IKK⑀ and IKK␤.
The IFN-␤ promoter is complicatedly regulated by a combination of various transcription factors (28). Although the IFN-␤ promoter contains an NF-B site, it is barely activated by overexpression of IKK␤ (Fig. 4D) (20,21). However, mutation of the NF-B site in the IFN-␤ promoter dramatically reduced its response to TRIF activation (ϳ60-fold) (Fig. 2, A and B). In addition, IB␣(SS/AA) completely blocked TRIF-induced IFN-␤ promoter activation, but had no effect on TRIF-induced activation of an independent ISRE reporter (Fig. 2, C and D). It is possible that specific IRFs have to collaborate with NF-B and/or coactivators such as p300/cAMP-responsive elementbinding protein-binding protein, TAF II 250, and proteins with histone acetyltransferase activity to activate the IFN-␤ promoter. In this context, it has been shown that the active form of IRF-3 (but not IRF-7) is associated with these coactivators (29).
Previous studies have demonstrated that both IRF-3 and IRF-7 are required for efficient type I IFN production (29,30). However, their roles are different in these processes. In the early phase of viral infection, pre-existing IRF-3 is activated and induces expression of IFN-␤ and IFN-␣4. These early produced IFNs transcriptionally induce IRF-7; and upon viral infection, the induced high level IRF-7 is activated and transactivates multiple IFN genes, leading to robust production of IFNs in response to viral infection (29,30).
In addition to the spatial difference in their functions in virus-induced IFN production, IRF-3 and IRF-7 recognize different DNA-binding sites or have distinct affinities for the same DNA site (29). As mentioned above, the active form of IRF-3 (but not IRF-7) is associated with some coactivators (29). These different properties between IRF-3 and IRF-7 lead to differential gene activation by these two transcription factors. Our studies point to a new possibility that IKK␤ and IKK⑀ can differentially regulate IRFs and mediate activation of distinct ISREs and type I IFNs.
Previously, it has been shown that TRIF can also activate the transcription factor NF-B (15,16). Our studies indicated that the kinase-inactive mutant of IKK␤ (but not those of IKK⑀ and TBK1) could block TRIF-induced NF-B activation (Fig. 6). These data suggest that these kinases have differential roles in the TRIF-induced NF-B and ISRE activation pathways. A dominant-negative mutant of TRAF6 inhibited TRIF-induced NF-B (but not ISRE) activation (Figs. 3B and 6), suggesting that it plays a role in the TRIF-induced NF-B (but not ISRE) activation pathway.
In many cases, viral infection leads to apoptosis of host cells.
When overexpressed, TRIF could potently induce apoptosis (Fig. 7). TRIF-induced apoptosis was mediated by its C-terminal domain. This domain interacted with RIP, a death domaincontaining protein kinase. TRIF-induced apoptosis was blocked by the dominant-negative mutants of FADD and caspase-8 and by the caspase inhibitors ZAD-fm and CrmA, but not by a p53 mutant and Bcl-2 (Fig. 7). Our observations suggest that TRIF induces RIP/FADD/caspase-8-dependent, mitochondrion-independent apoptosis. Previously, it has been suggested that IRF-3 is involved in Sendai virus-induced apoptosis (37). Overexpression of a constitutively active mutant of IRF-3 (IRF-3/5D) induces caspasedependent apoptosis (37). TRIF-induced apoptosis was not inhibited by IRF-3⌬N (Fig. 7), suggesting that TRIF and IRF-3 induce distinct apoptosis pathways. TRIF interacted with RIP and induced apoptosis through death domain-containing FADD (Fig. 7). How IRF-3 induces apoptosis is unknown. In addition, it seems that TRIF is a much more potent inducer of apoptosis than IRF-3. More than 90% of the TRIF-transfected cells died of apoptosis as early as 12 h after transfection (data not shown). In contrast, Ͻ40% of the IRF-3/5D-transfected cells died of apoptosis as long as 48 h after IRF-3/5D induction (37).
Based on our and others' results (see above), we have proposed a working model for TRIF-induced intracellular signaling pathways (Fig. 8). Dominant-negative mutants that block one pathway are not able to block the other two pathways, suggesting that the TRIF-induced ISRE and NF-B activation and apoptosis pathways are uncoupled. The exception is that IKK␤ is probably involved in both TRIF-induced ISRE and NF-B activation pathways. Our studies may provide a basis for elucidation of the complicated regulation of virus-induced cellular effects.