FADD Negatively Regulates Lipopolysaccharide Signaling by Impairing Interleukin-1 Receptor-Associated Kinase 1-MyD88 Interaction

Abstract

Lipopolysaccharide (LPS) engages Toll-like receptor 4 (TLR4) on various cells to initiate inflammatory and angiogenic pathways. FADD is an adaptor protein involved in death receptor-mediated apoptosis. Here we report a role for FADD in regulation of TLR4 signals in endothelial cells. FADD specifically attenuates LPS-induced activation of c-Jun NH₂-terminal kinase and phosphatidylinositol 3′-kinase in a death domain-dependent manner. In contrast, FADD-null cells show hyperactivation of these kinases. Examining physical associations of endogenous proteins, we show that FADD interacts with interleukin-1 receptor-associated kinase 1 (IRAK1) and MyD88. LPS stimulation increases IRAK1-FADD interaction and recruitment of the IRAK1-FADD complex to activated MyD88. IRAK1 is required for FADD-MyD88 interaction, as FADD does not associate with MyD88 in IRAK1-null cells. By shuttling FADD to MyD88, IRAK1 provides a mechanism for controlled and limited activation of the TLR4 signaling pathway. Functionally, enforced FADD expression inhibited LPS- but not vascular endothelial growth factor-induced endothelial cell sprouting, while FADD deficiency led to enhanced production of proinflammatory cytokines induced by stimulation of TLR4 and TLR2, but not TLR3. Reconstitution of FADD reversed the enhanced production of proinflammatory cytokines. Thus, FADD is a physiological negative regulator of IRAK1/MyD88-dependent responses in innate immune signaling.

Bacterial lipopolysaccharide (LPS) is a potent inflammatory molecule that evokes immune responses by activating various cell types, including endothelial cells, through Toll-like receptor 4 (TLR4). The cytoplasmic portions of all TLRs as well as that of the interleukin-1 receptor (IL-1R) share a highly conserved region known as the Toll-IL-1R (TIR) domain. Upon receptor occupancy, the TIR domain recruits several TIR domain-containing intracellular proteins (MyD88, TIRAP/Mal (MyD88 adaptor-like protein), TRIF (Toll/IL-1R domain-containing adaptor-inducing beta interferon)/TICAM-1 (TIR-containing adapter molecule 2 [TICAM-2]), and TRAM (TRIF-related adaptor molecule)/TICAM-2 (13, 15, 27, 30, 34) via homophilic interactions. Signaling downstream from most TLRs is dependent on the adaptor protein MyD88. In addition to MyD88 and Mal, TLR4 also recruits TRIF and TRAM to the receptor. In contrast, TLR3 does not signal through MyD88 but rather recruits only TRIF (12). The TRIF-dependent pathway activates the transcription factors NF-κB and IRF3 (interferon regulatory factor 3) leading to subsequent induction of genes, such as beta interferon, IP-10 (interferon-inducible protein of 10 kDa), and RANTES (12).

However, the best-characterized TLR adaptor protein is MyD88, which is required for the initial rapid activation of mitogen-activated protein kinases (MAPKs) and NF-κB. In addition to the TIR domain, MyD88 contains a death domain (DD), a highly conserved protein interaction domain, which enables it to recruit other DD-containing signaling transducers involved in innate immune signaling called interleukin-1 receptor-associated kinases (IRAKs). Upon recruitment by MyD88, interleukin-1 receptor-associated kinase 4 (IRAK4) phosphorylates IRAK1, triggering IRAK1 auto- and trans-phosphorylation. Hyperphosphorylated IRAK1 dissociates from the TLR4-MyD88 complex and associates with tumor necrosis factor (TNF) receptor-activated factor 6 (TRAF6), which triggers the oligomerization and polyubiquitination of TRAF6. Functionally, TRAF6 is an E3 ubiquitin ligase which in collaboration with the dimeric ubiquitin-conjugating enzyme (E2) complex Ubc13/Uev1A catalyzes autologous synthesis of lysine 63 (K63)-linked polyubiquitin chains (10). The K63 ubiquitin chains are recognized by the ubiquitin acceptor proteins transforming growth factor β-activating kinase 1-binding protein 2 and 3 (TAB2 and TAB3) which function as adaptors to assemble a signaling complex which phosphorylates and activates transforming growth factor β-activated kinase 1 (TAK1) (21). Activated TAK1 leads to the activation of the IκB kinase (IKK) complex, NF-κB, and the MAPK kinases that activate p38, extracellular signal-regulated (ERK) and c-Jun NH₂-terminal protein kinase (JNK). LPS also activates the phosphatidylinositol 3-kinase (PI3K) pathway downstream of TRAF6 (33).

FADD is a DD-containing adaptor protein that is recruited to death receptors following receptor occupancy. FADD binds directly to Fas/CD95 and to the adaptor protein TRADD in response to Fas ligand and TNF, respectively, to activate caspase 8 and initiate apoptosis (16, 28). Interestingly, others have suggested that TLR2 recruits FADD via MyD88 and thereby also induces apoptosis (2). However, the latter study examined only FADD-MyD88 associations in overexpression systems. Whether FADD is involved in TLR4 signaling to activate the MAPK and PI3K pathways by LPS has not been demonstrated. We report here that FADD, in a manner requiring the DD, negatively regulates LPS signaling by suppressing activation of JNK and PI3K pathways. FADD interacts with IRAK1 and MyD88 and appears to reduce the stability of the MyD88-IRAK1 interaction, thereby attenuating the LPS signal. IRAK1 is essential to allow the MyD88-FADD interaction, as MyD88 and FADD do not associate in IRAK1-null cells. FADD also inhibits IRAK1-MyD88-dependent signaling in response to activation of other receptors, such as TLR2 and IL-1R. We demonstrate further that enforced expression of FADD plays a role in deregulating endothelial function by limiting endothelial sprouting in response to LPS, but not vascular endothelial growth factor (VEGF). Additionally, FADD-deficient cells show significantly enhanced cytokine release in response to TLR4 and TLR2 agonists, but not in response to TLR3. Thus, FADD represents a novel negative regulator of IRAK1/MyD88-dependent innate immune responses.

MATERIALS AND METHODS

Reagents.

LPS and anti-Flag M2 antibody were purchased from Sigma (St. Louis, MO). Anti-JNK1, antiubiquitin, and anti-TRAF6 antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA). The phospho-specific antibodies against JNK and Akt and anti-Akt antibody were purchased from Cell Signaling Technology (Beverly, MA). Protein A and G agarose, anti-IRAK1, anti-IRAK4, and anti-FADD monoclonal and polyclonal antibodies were obtained from Upstate (Charlottesville, VA). Anti-MyD88 antibody was purchased from Alexis Biochemicals (Lausanne, Switzerland). TrueBlot anti-rabbit immunoglobulin immunoprecipitation beads and rabbit immunoglobulin G (IgG) TrueBlot were obtained from eBioscience (San Diego, CA). Recombinant TNF alpha was purchased from R&D Systems. Poly(I)·poly(C) and the synthetic tripalmitoylated lipopeptide Pam3CSK4 were purchased from InvivoGen (San Diego, CA).

Cell culture.

Human microvascular endothelial cells (HMEC) were provided by the Centers for Disease Control and Prevention and were cultured in MCDB131 medium supplemented with 10% calf serum. 293T cells and mouse embryonic fibroblasts (MEF) were cultured in Dulbecco modified Eagle medium supplemented with 10% calf serum. FADD-deficient MEF were obtained from Amgen, Inc. (Thousand Oaks, CA), and IRAK1-deficient and MyD88-deficient MEF have previously been described (1, 32).

Recombinant plasmids and transfection.

Full-length FADD containing a Flag epitope at the N terminus was cloned into the murine stem cell virus-internal ribosome entry site-yellow fluorescent protein (MSCV-IRES-YFP) (MIY) retroviral vector. Transient transfections of the Ampho Phoenix packaging cell line were carried out using Fugene 6. Viral supernatants were used to transduce HMEC, and yellow fluorescent protein-positive cells were selected by flow sorting. 293T cells were transfected with Fugene 6 according to the manufacturer's instructions (Roche Applied Science Laval, QC, Canada).

Coimmunoprecipitation, immunoblotting, and kinase assays.

HMEC, 293T cells, or MEF were lysed for 30 min in 20 mM HEPES (pH 7.4), 150 mM NaCl, 12.5 mM β-glycerol phosphate, 1.5 mM MgCl₂, 10 mM NaF, 1 mM sodium orthovanadate, 2 mM EGTA, 0.5% Triton X-100, and protease inhibitor cocktail to examine noncovalent interactions. Antibodies (3 μg) were added to cell lysates (8 mg) for 3 h at 4°C and captured by the addition of protein A or G or TrueBlot anti-rabbit immunoglobulin immunoprecipitation beads for an additional 12 h at 4°C. The immune complexes were washed three times with lysis buffer followed by the addition of sodium dodecyl sulfate (SDS) sample buffer. The bound proteins were separated by SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose membranes, and analyzed by immunoblotting. HMEC treated without or with LPS were lysed and immunoprecipitated under denaturing conditions (50 mM Tris, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, 20 mM N-ethylmaleimide and protease inhibitors) to examine covalent polyubiquitination of TRAF6. Cell lysates (3 mg) were immunoprecipitated with antibody against TRAF6 (5 μl) for 3 h followed by an additional 12 h of incubation in the presence of TrueBlot anti-rabbit immunoglobulin immunoprecipitation beads. Immunoprecipitates were washed three times with lysis buffer followed by the addition of SDS sample buffer. Proteins were separated on SDS-polyacrylamide gels and subjected to immunoblotting with antiubiquitin and TRAF6 antibodies. For JNK activity, cells were exposed to LPS (100 ng/ml) at the times indicated. At each time point, cells were lysed in buffer containing protease and phosphatase inhibitors, and lysates were obtained by centrifugation. The assay is dependent upon the isolation of JNK with glutathione S-transferase (GST)-c-Jun followed by an in vitro kinase assay using GST-c-Jun as the substrate in the presence of γ-³²P-labeled ATP or nonradioactive ATP. The kinase reaction mixtures were then separated on SDS-polyacrylamide gels and subjected to autoradiography (in the former case). For the latter case, the separated proteins were transferred to nitrocellulose membranes and analyzed by immunoblotting with a phospho-specific c-Jun (Ser 73) antibody.

ELISA.

MEF were seeded in 24-well plates at a density of 200,000 cells per well and cultured for 24 h. Following treatment, the supernatants were recovered and centrifuged (850 × g, 10 min). The supernatants were analyzed using commercially available kits for murine IL-6 (eBioscience) and IP-10 (R&D Systems).

Endothelial sprouting assay.

Endothelial sprouting was assessed as previously described (25). Briefly, microcarrier beads coated with gelatin (Cytodex 3; Sigma) were seeded with HMEC lines and embedded in fibrin gels in 96-well plates. Fibrin gels were supplemented with VEGF (15 ng/ml) or LPS (100 ng/ml). The overlying medium contained either MCDB131 medium supplemented with 2% fetal bovine serum alone or was further supplemented with VEGF (15 ng/ml) or LPS (100 ng/ml). After 2 days of incubation with daily medium changes, the number of capillary-like tubes formed was quantitated by counting the number of tube-like structures longer than 150 μm.

RESULTS

FADD inhibits activation of MAPKs and Akt by LPS.

We first investigated the role of FADD on LPS-induced activation of JNK. Human microvascular endothelial cells were stably transduced with either full-length FADD (HMEC-FADD) or the empty vector MIY (HMEC-MIY), and expression of FADD was confirmed by immunoblotting (Fig. 1E). Cells were stimulated with LPS, and JNK activity was measured by monitoring the phosphorylation of c-Jun. In HMEC transduced with the empty vector, c-Jun phosphorylation was apparent at 30 min after LPS stimulation, whereas FADD-overexpressing cells exhibited a 30-min lag in c-Jun phosphorylation (Fig. 1A). Additionally, at all time points, the magnitude of c-Jun phosphorylation in FADD-overexpressing cells was considerably lower than in the control cells. LPS and TNF signaling pathways share some similarities, such as the activation of MAPKs, including JNK. Moreover, FADD is a critical signaling adaptor molecule which bridges the TNF receptor to the extrinsic apoptotic pathway. We thus tested whether overexpression of FADD also intersected the TNF pathway to JNK activation. As shown in Fig. 1A, TNF treatment led to a similar pattern of c-Jun phosphorylation between FADD-overexpressing and vector-transduced cells. Therefore, FADD-mediated attenuation of JNK activation is specific to the LPS pathway.

FIG. 1.

Enforced expression of FADD inhibits JNK and Akt activation by LPS but not TNF. (A) Endothelial cells expressing full-length FADD (HMEC-FADD) or empty vector (HMEC-MIY) were left untreated or treated with TNF (10 ng/ml) or LPS (100 ng/ml) for the indicated times (in minutes). Cell lysates were subjected to an in vitro JNK protein kinase assay using a GST-c-Jun fusion protein as the substrate. (B) Lysates from unstimulated endothelial cells or endothelial cells stimulated with LPS (100 ng/ml) were subjected to immunoblot analysis with antibodies against phosphorylated JNK (pJNK) and total JNK. Results were quantified using densitometry and are expressed in the bar graph as the change in stimulation (of both p46 and p54 JNK isoforms) above the basal level. Data represent the means plus standard errors of the means (SEMs) (error bars) for three independent experiments. Values that were significantly different (P < 0.05) are indicated (*). (C) Lysates from the blots in panel B were immunoblotted with antibodies against phosphorylated Akt (pAkt) (Ser 473) and total Akt. (D) Lysates from unstimulated endothelial cells or endothelial cells stimulated with TNF (10 ng/ml) were immunoblotted for phosphorylated JNK (pJNK) and total JNK. JNK (p46 and p54) activation was quantified by densitometric scanning and is presented in the bar graph as the ratio of phosphorylated to unphosphorylated JNK. Data represent the means plus SEMs for six independent experiments. (E) Lysates from HMEC-MIY and HMEC-FADD were immunoblotted (IB) with anti-FADD antibody to monitor the expression of FADD. Molecular mass (in kilodaltons) is indicated. to the left of the blot. (F) HMEC cells were pretreated for 20 min with vehicle (dimethyl sulfoxide [DMSO]) or 50 μM Z-IETD-fmk and then stimulated with LPS for the indicated times (in minutes). Activation of JNK was monitored by immunoblotting for phospho-JNK (pJNK) and total JNK.

The effect of FADD on JNK activity could arise either from the inhibition of JNK phosphorylating activity directly or through inhibition of JNK activation by upstream effectors. In order to determine whether FADD was inhibiting JNK directly or indirectly via a different molecule, we determined the phosphorylation status of JNK. Figure 1B shows the phosphorylation of JNK as determined by a JNK phospho-specific antibody. Overexpression of FADD diminished LPS-induced phosphorylation of JNK, suggesting that FADD does not interfere with JNK activity directly but rather targets upstream molecules involved in JNK activation by LPS.

In addition to JNK, LPS simultaneously activates the PI3K pathway. LPS-stimulated phosphorylation of Akt at Ser 473, a downstream effect of the activation of PI3K, was impaired in cells overexpressing FADD (Fig. 1C). In contrast to LPS, TNF-mediated activation of JNK was not inhibited in cells overexpressing FADD (Fig. 1D). LPS-induced activation of ERK1/2 MAPKs was also suppressed by enforced expression of FADD (data not shown). We also observed inhibition of LPS-mediated activation of NF-κB by FADD, consistent with a previous report (data not shown) (3). FADD and caspase 8/10 are essential for the induction of apoptosis through death receptors. To determine whether caspase 8/10 were also involved in suppression of LPS-mediated JNK activation, we used a caspase 8/10-specific inhibitor, Z-IETD-fmk. Inhibition of caspase 8/10 by Z-IETD-fmk (50 μM) did not affect activation of JNK by LPS (Fig. 1F), although Z-IETD-fmk was able to inhibit TNF-induced apoptosis in the same cells (data not shown). Together these results indicate that the site at which FADD impinges on the LPS pathway lies upstream of MAPK, NF-κB, and Akt activation and is independent of caspase 8/10 activation.

The death domain of FADD-DD is sufficient and necessary to inhibit LPS signaling.

The FADD death domain (FADD-DD) has previously been reported to inhibit LPS-induced apoptosis (8). Moreover, JNK has been shown to be involved in endothelial cell apoptosis in response to LPS (17). Therefore, we examined the effect of FADD-DD on LPS-induced activation of JNK in HMEC. HMEC were stably transduced with AU1-tagged FADD-DD or the empty vector, and the expression of FADD-DD was verified by immunoblotting (Fig. 2D). Similar to full-length FADD, expression of FADD-DD was sufficient to inhibit LPS stimulation of JNK activity (Fig. 2A). The expression of FADD-DD also impaired the phosphorylation of JNK by its activators as determined by immunoblotting with a JNK phospho-specific antibody (Fig. 2B). As with FADD, FADD-DD also inhibited Akt (Fig. 2B) and ERK1/2 activation (data not shown).

FIG. 2.

The death domain of FADD is sufficient and necessary to inhibit LPS signaling. Endothelial cells expressing the death domain of FADD (HMEC-FADD-DD) or the empty vector (HMEC-LNCX) were either left untreated or treated with LPS (100 ng/ml) for the indicated times (in minutes). (A and B) Cell lysates were subjected to JNK in vitro protein kinase assay using GST-c-Jun as the substrate (A) or immunoblot analysis with antibodies against phosphorylated and total JNK and Akt (B). pJNK, phosphorylated JNK; pAkt, phosphorylated Akt. (C) Endothelial cells expressing the empty vector (MIY) or mutant FADD (FADDmt) with the DD point mutation (V121N) were left untreated or exposed to LPS (100 ng/ml). Cell lysates were immunoblotted for phosphorylated and total JNK and Akt. (D). Lysates were immunoblotted (IB) with anti-AU1 or anti-Flag antibodies to monitor the expression of FADD-DD and FADDmt, respectively. Molecular mass (in kilodaltons) is shown to the left of the blots.

To determine whether the DD of FADD-DD was necessary for the inhibitory function of FADD, we introduced a point mutation in FADD to replace the codon for valine 121 (V121) with that for asparagine (N) (FADDmt), which disrupts the interaction of FADD with Fas (7). As shown in Fig. 2C, FADDmt was not able to inhibit LPS-induced activation of JNK and Akt. Thus, the integrity of the DD of FADD is necessary for its modulation of LPS signals. Figure 2D shows expression of Flag-tagged FADDmt as determined by immunoblotting.

FADD-deficient cells show hyperactivation of MAPK and Akt by LPS, but not TNF.

To further explore the biological relevance of FADD inhibition of LPS signaling, we examined LPS-stimulated JNK activity in murine embryonic fibroblasts from gene-targeted mice lacking FADD. We speculated that if indeed FADD serves as a negative regulator of LPS-induced cell signaling, then the absence of FADD should lead to hyperactivation of the specific kinases under the negative regulation of FADD. Consistent with this hypothesis, LPS induced greater JNK activity in FADD-null MEF than in control MEF expressing wild-type FADD (Fig. 3A and B). This effect was not caused by a variation in the level of JNK protein expression (Fig. 3B). Additionally, lack of FADD markedly enhanced the activation of Akt in response to LPS (Fig. 3C). As expected, JNK activation by TNF was unaffected in FADD-deficient cells (Fig. 3D).

FIG. 3.

FADD-deficient cells show hyperactivation of JNK and Akt in response to LPS but not TNF. Wild-type (FADD-WT) and FADD-deficient (FADD-KO [KO stands for knockout]) MEF were treated with LPS (100 ng/ml) for the times indicated. Cell lysates were analyzed by a JNK in vitro protein kinase assay using GST-c-Jun as the substrate (A) or by immunoblotting for phospho- and total JNK (B) and phospho- and total Akt (C). pJNK, phosphorylated JNK; pAkt, phosphorylated Akt. The bar graph in panel A shows the change in phosphorylation of c-Jun relative to untreated samples. Data represent the means plus standard errors of the means (error bars) for three independent experiments. Values that were significantly different (P < 0.05) are indicated (*). (D) Lysates from MEF stimulated with TNF (10 ng/ml) were immunoblotted for phospho- and total JNK. Data are representative of three independent experiments.

FADD acts upstream of TRAF6.

TRAF6 is a key intermediary molecule in the LPS signaling pathway that serves as a convergence point for coordinate activation of NF-κB, JNK, and Akt (14, 33). As mentioned above, TRAF6 is ubiquitinated in a K63-linked manner upon stimulation which is required for activation (10). We thus investigated the role of FADD on TRAF6 ubiquitination. We treated HMEC-MIY and HMEC-FADD with LPS for various times. TRAF6 was then immunoprecipitated followed by immunoblotting with a ubiquitin-specific antibody. Within 5 min of LPS treatment, a ladder of polyubiquitinated TRAF6 appeared in vector control cells (HMEC-MIY). This ladder was still evident after 30 min of LPS treatment (Fig. 4A). In contrast, HMEC-FADD exhibited transient and weaker TRAF6 ubiquitination that reached a peak at 20 min and had decayed by 30 min of LPS treatment (Fig. 4A). This result was not due to a variation in the protein level of TRAF6, since the amounts of immunoprecipitated TRAF6 were comparable in the different samples. Together, these findings indicate that FADD negatively regulates LPS signaling upstream of TRAF6 ubiquitination.

FIG. 4.

FADD acts upstream of TRAF6 and associates with MyD88 and IRAK1. (A) Endothelial cells transduced with the empty vector (HMEC-MIY) or FADD (HMEC-FADD) were either left untreated or stimulated with LPS (100 ng/ml) for the indicated times (in minutes). Whole-cell lysates were prepared and immunoprecipitated with anti-TRAF6 antibody under denaturing conditions followed by immunoblotting with antiubiquitin and anti-TRAF6 antibodies. Lanes A are a negative control which contained the immunoprecipitating antibody with lysis buffer. Lanes L are a positive control and contain 30 μg of unstimulated protein lysate. Molecular mass (in kilodaltons) is shown to the left of the blots. IP, immunoprecipitation; WB, Western blotting or immunoblotting; (Ub)n TRAF6, polyubiquitinated TRAF6. Data are representative of three independent experiments. (B) (Top) Coimmunoprecipitation analysis of lysates of 293T cells cotransfected with the indicated combinations of epitope-tagged expression plasmids (Myc-tagged MyD88 [Myc-MyD88] and Flag-tagged FADD [Flag-FADD]) and immunoprecipitated (IP) with an anti-Myc antibody. (Bottom) Immunoblot (IB) of whole-cell lysates, using anti-Flag or anti-Myc antibodies. NS, nonspecific band recognized by anti-Flag antibody. (C) (Top) Coimmunoprecipitation analysis of lysates of 293T cells cotransfected with the indicated combinations of epitope-tagged expression plasmids (Flag-tagged IRAK1 [Flag-IRAK1] and Flag-FADD) and immunoprecipitated with an anti-IRAK1 antibody. (Bottom) Immunoblot of whole-cell lysates using anti-Flag antibody. NS, nonspecific band recognized by anti-Flag antibody. (D) (Top) Coimmunoprecipitation analysis of lysates of 293T cells cotransfected with the indicated combinations of epitope-tagged expression plasmids (Flag-IRAK1 and AU1-tagged FADD [AU1-FADD]) and immunoprecipitated with an anti-FADD antibody or control IgG. (Bottom) Whole-cell lysates were immunoblotted with anti-FADD or anti-IRAK1 antibodies. (E) (Top) Coimmunoprecipitation analysis of lysates of 293T cells transfected with empty vector (pcDNA3) or the indicated epitope-tagged expression plasmid (AU1-FADD, Flag-FADD-DED, AU1-FADD-DD, or Flag-IRAK1) and immunoprecipitated with an anti-IRAK1 antibody or control IgG. (Bottom) Whole-cell lysates were immunoblotted with anti-FADD or anti-Flag antibodies. Molecular mass (in kilodaltons) is shown to the left of the blots.

FADD interacts with IRAK1 and MyD88.

Given that the action of FADD lies upstream of TRAF6, a mechanistic model involving direct interactions between FADD, MyD88, and IRAKs can be postulated, since all of these proteins contain a DD. We used coimmunoprecipitation assays to probe for these associations. We first tested whether FADD interacts with MyD88. 293T cells were transiently transfected with Flag-tagged FADD and Myc-tagged MyD88 and immunoprecipitated with an anti-Myc antibody. Coimmunoprecipitated Flag-tagged FADD was detected by immunoblotting with an anti-Flag antibody. Under these conditions, FADD interacted with MyD88, confirming the results of a previous study (2) (Fig. 4B). We then addressed whether FADD could also interact with IRAK1. 293T cells were transiently transfected with Flag-tagged FADD and Flag-tagged IRAK1. Cell lysates were immunoprecipitated with an antibody specific for IRAK1, and coimmunoprecipitated FADD was detected with an anti-FADD antibody. As shown in Fig. 4C, FADD associated with both the transfected Flag-tagged IRAK1 and endogenous IRAK1. In reverse immunoprecipitations with an antibody against FADD, transfected FADD associated with transfected IRAK1 (Fig. 4D). Furthermore, transfected FADD also interacted with endogenous IRAK1 (Fig. 4D). Control immunoprecipitations with rabbit polyclonal IgG did not immunoprecipitate FADD or IRAK1. We also examined the interaction of FADD with another member of the IRAK family, IRAK4, which also participates in LPS signaling. In contrast to IRAK1, there was no association between FADD and IRAK4 (data not shown).

To confirm that the DD of FADD was the crucial motif in the IRAK1-FADD interaction, we transfected 293T cells with the empty vector, various epitope-tagged FADD constructs or IRAK1, and immunoprecipitated IRAK1. Immunoprecipitation of endogenous IRAK1 pulled down endogenous FADD, as well as overexpressed FADD and overexpressed FADD-DD, but not FADD-death effector domain (DED) (Fig. 4E). IRAK1 immunoprecipitation also pulled down overexpressed IRAK1 (Fig. 4E); however, control immunoprecipitations with rabbit polyclonal IgG did not immunoprecipitate IRAK1 or FADD. Thus, the FADD-IRAK1 interaction requires the DD, but not the DED motif, of FADD.

FADD interferes with IRAK1-MyD88 interaction.

We demonstrated above that FADD interacts with IRAK1 and MyD88. Next we wanted to investigate the consequences of these associations with respect to LPS signaling. We performed time course experiments to examine IRAK1-MyD88 and IRAK1-TRAF6 interactions in the presence of endogenous or overexpressed FADD. HMEC transduced with the empty vector (MIY) or full-length FADD were either left untreated or stimulated with LPS for different times, and cell lysates were immunoprecipitated with the anti-IRAK1 antibody followed by immunoblot analyses with antibodies against MyD88, TRAF6, FADD, and IRAK1. As expected, endogenous IRAK1 interacted with MyD88 immediately after LPS stimulation (Fig. 5A). The association was evident by 1 min and was sustained for up to 10 min of LPS stimulation in vector control cells. The kinetics of IRAK1-MyD88 interaction was similar in both control and FADD-overexpressing cells; however, the association was considerably weaker and transient in the FADD-overexpressing cells (Fig. 5A).

FIG. 5.

FADD interferes with IRAK1-MyD88 interaction. (A) Endothelial cells expressing the empty vector (HMEC-MIY) or FADD (HMEC-FADD) were either unstimulated or stimulated with LPS (100 ng/ml) for the indicated times (in minutes). Cell lysates were immunoprecipitated (IP) with anti-IRAK1 antibody and then immunoblotted with antibodies against MyD88, TRAF6, FADD, and IRAK1. (B) Lysates from endothelial cells were prepared as described above for panel A and subjected to immunoprecipitation (IP) with anti-FADD antibody, followed by immunoblotting for IRAK1, MyD88, and FADD. (C) MEF expressing wild-type FADD (MEF-FADD-WT) or MEF deficient in FADD (MEF-FADD-KO) were stimulated with LPS (1 μg/ml) for the indicated times (in minutes). Cell lysates were subjected to anti-IRAK1 immunoprecipitation, followed by immunoblotting for MyD88, TRAF6, ubiquitin, FADD, and IRAK1. Molecular mass (in kilodaltons) is shown to the left of the blots. Ubn-TRAF6, polyubiquitinated TRAF6.

As expected, endogenous IRAK1 also associated with TRAF6. In the absence of stimulation, TRAF6 interacted weakly with IRAK1 in control cells (Fig. 5A). However, after stimulation, there was an increase in the amount of TRAF6 that associated with IRAK1 as expected. In contrast, FADD-overexpressing cells exhibited stronger IRAK1-TRAF6 association prior to LPS stimulation with no evident changes in IRAK1-TRAF6 association following LPS stimulation (Fig. 5A). There was little association of endogenous FADD with IRAK1 in unstimulated cells. However, this association increased significantly upon LPS stimulation (Fig. 5A). Overexpressed FADD clearly interacted with IRAK1 before stimulation, and this association increased further with stimulation. In reciprocal immunoprecipitations, FADD interacted with both MyD88 and IRAK1 (Fig. 5B). The interaction between FADD and MyD88 peaked at 5 min following LPS stimulation in HMEC-MIY. In contrast, in HMEC-FADD, FADD interacted with MyD88 prior to LPS stimulation, and LPS induced a further increase in the FADD-MyD88 interaction.

To confirm and extend our findings, we examined the effects of FADD deficiency on IRAK1-MyD88 and IRAK1-TRAF6 associations. MEF expressing wild-type FADD or MEF deficient in FADD were stimulated with LPS or left untreated. Protein lysates were immunoprecipitated with an antibody against IRAK1 followed by immunoblotting with antibodies against MyD88, TRAF6, FADD, and IRAK1. MyD88 interacted with IRAK1 immediately after stimulation. In wild-type MEF, this IRAK1-MyD88 association peaked at 1 min and became weaker thereafter (Fig. 5C). In contrast, FADD-deficient MEF exhibited a robust IRAK1-MyD88 association which lasted up to 10 min, consistent with a role for FADD in inhibiting IRAK1-MyD88 interactions. Formation of IRAK1-TRAF6 complexes was also dependent on LPS in both cells expressing wild-type FADD and cells deficient in FADD, but TRAF6 ubiquitination was stronger in the FADD-deficient cells. Taken together, the temporal course of the IRAK1-FADD-MyD88 interactions above suggests that LPS stimulation recruits an IRAK1-FADD complex to MyD88, which reduces the stability of the IRAK1-MyD88 interaction and attenuates TRAF6 activation.

IRAK1 is required for FADD-MyD88 interaction.

To define further the sequence of events during IRAK1-FADD-MyD88 interactions following LPS stimulation, we performed coimmunoprecipitations using cells deficient in either MyD88 or IRAK1. We first tested whether MyD88 was required for FADD-IRAK1 interaction by examining the association of FADD and IRAK1 in MyD88-deficient cells. As shown in Fig. 6A, FADD interacted with IRAK1 in MyD88-deficient cells, which implies that the FADD-IRAK1 interaction does not depend on the presence of MyD88. In contrast, the FADD-MyD88 interaction was impaired in cells lacking IRAK1 (Fig. 6B), suggesting that IRAK1 plays a critical role in shuttling FADD to MyD88 to activate the LPS signaling cascade in a controlled and limited fashion.

FIG. 6.

IRAK1 is required for FADD-MyD88 interaction. (A) MEF expressing wild-type MyD88 (MEF-MydD88-WT) or MEF lacking MyD88 (MEF-MydD88-KO) were stimulated with LPS (1 μg/ml) as indicated, and cell lysates were immunoprecipitated with anti-FADD antibody. Coimmunoprecipitating IRAK1 and MyD88 were detected by immunoblotting with anti-IRAK1 and anti-MyD88 antibodies, respectively. (B) Lysates from IRAK1-deficient MEF (MEF-IRAK-1-KO) or MEF expressing wild-type IRAK1 (MEF-IRAK-1-WT) controls were prepared and subjected to immunoprecipitation with anti-FADD antibody. Coimmunoprecipitating MyD88 and IRAK1 were detected using anti-MyD88 and anti-IRAK1 antibodies, respectively. IP, immunoprecipitation, NS, nonspecific band recognized by antibody.

FADD attenuates signaling only through MyD88/IRAK1-dependent receptors.

Our model would predict that FADD modulates pathways which depend on IRAK1 and MyD88 for signaling. We thus examined the ability of FADD to modulate signaling by other MyD88/IRAK1-dependent receptors, such as TLR2 and IL-1R, and the MyD88/IRAK1-independent receptor, TLR3. Activation of JNK and Akt by Pam3CSK4, a TLR2/TLR1 ligand, was enhanced in FADD-deficient cells compared to the wild-type cells (Fig. 7A). Similarly, IL-1β stimulation led to the hyperactivation of JNK and Akt in FADD-deficient cells consistent with activation of a MyD88/IRAK1-dependent pathway (Fig. 7B). Activation of JNK or Akt by poly(I)·poly(C), a TLR3 ligand, was weak in HMEC and MEF. In contrast to TLR2, TLR4, and IL-1R, this weak TLR3-mediated JNK and Akt activation was not enhanced in FADD-deficient cells (Fig. 7C). Taken together, these data confirm a role for FADD as a negative regulator of MyD88/IRAK1-dependent signaling.

FIG. 7.

FADD attenuates signaling only through IRAK1/MyD88-dependent receptors. MEF expressing wild-type FADD (FADD-WT) or deficient in FADD (FADD-KO) were stimulated with Pam3CSK4 (500 ng/ml) (A), IL-1β (50 ng/ml) (B), and poly(I)·poly(C) (Poly I:C) (100 μg/ml) (C) for the indicated times (in minutes), and lysates were immunoblotted for phospho- and total JNK and Akt. Results are representative of three independent experiments. pJNK, phospho-JNK; pAkt, phospho-Akt.

FADD inhibits LPS-induced endothelial sprouting.

We have previously shown that LPS induces endothelial sprouting in vitro and angiogenesis in vivo by directly activating endothelial cells (31). Activation of both JNK and NF-κB are essential for LPS-initiated endothelial sprouting (31). Given that enforced expression of FADD in endothelial cells inhibits both JNK and NF-κB activation, we expected that a functional outcome of FADD overexpression would be to block endothelial sprouting. We and others have used a microcarrier-based endothelial tube-forming assay that has shown remarkable correlation between in vitro and in vivo findings (23, 25, 29, 31). We thus tested whether FADD overexpression would block endothelial sprouting. As seen in Fig. 8A and B, endothelial cell sprouting in response to LPS, but not serum or VEGF, was completely blocked in the presence of FADD.

FIG. 8.

FADD suppresses LPS-induced endothelial sprouting and IL-6 production, but not IP-10 production. Microcarrier beads seeded with endothelial cells were embedded in fibrin gels and exposed to various stimuli. (A) Micrographs of endothelial cell-coated microcarrier beads stimulated with 5% fetal bovine serum (FBS) (negative control) (i) or LPS plus 5% FBS (100 ng/ml) (ii). Arrows indicate endothelial sprouts formed after LPS stimulation. (B) Quantitation of the number of sprouts formed following stimulation by 5% FBS, VEGF (15 ng/ml), or LPS (100 ng/ml) using HMEC overexpressing FADD (MIY-FADD) or vector control (MIY). Each condition was evaluated in triplicate, and the data are the means plus standard errors of the means (error bars) from three independent experiments. Values that were significantly different (P ≤ 0.001) from the values for vector-transduced cells stimulated with LPS are indicated (*). (C) MEF expressing wild-type FADD (FADD-WT) and FADD-deficient (KO) MEF were treated for 6 h with medium alone (control) or stimulated with LPS (100 ng/ml), poly(I)·poly(C) (PIC) (100 μg/ml), or Pam3CSK4 (500 ng/ml), and cell culture supernatants were analyzed for IL-6 by an enzyme-linked immunosorbent assay (ELISA). Data are means plus standard errors of the means (SEMs) (error bars) from four independent experiments. Values that were significantly different (P ≤ 0.05) from the values for MEF expressing FADD-WT stimulated with LPS, poly(I)·poly(C), or Pam3CSK4 are indicated (*). (D) FADD-KO MEF reconstituted with the empty vector (MIY) or full-length FADD (MIY-FADD) were treated for 6 h with LPS or Pam3CSK4. Cell culture supernatants were analyzed for IL-6. Data are means plus SEMs (error bars) from five independent experiments. Values that were significantly different (P ≤ 0.05) from the values for vector-transduced cells stimulated with LPS or Pam3CSK4 are indicated (*). (E) MEF expressing FADD-WT and FADD-KO were exposed to LPS or poly(I)·poly(C) (PIC) for 6 h. Supernatants were analyzed for IP-10 by ELISA. Results are means plus SEMs (error bars) from two independent experiments performed in triplicate. Values that were significantly different (P ≤ 0.05) are indicated (*). (F) Lysates from cells in panel D were immunoblotted (IB) with anti-FADD antibody to confirm reconstitution of FADD in FADD-KO MEF. Molecular mass (in kilodaltons) is indicated to the left of the blot.

We also performed RNA interference using lentivirus-mediated transfer of two short hairpin RNAs targeting distinct FADD sequences into HMEC and performed endothelial sprouting assays. Despite not being able to completely knock out FADD, we found that targeted knockdown of FADD by RNA interference slightly increased endothelial sprouting (data not shown). The increased sprouting was best seen when both short hairpin RNA constructs were used to target FADD in the same population (data not shown). As the effect of targeting FADD was only slight, these findings suggest that even small amounts of FADD may be sufficient to attenuate LPS signaling. Thus, FADD likely plays a physiological role in modulating angiogenesis in response to TLR4 signals.

FADD differentially modulates cytokine induction in response to activation of TLR4, TLR2, and TLR3.

We next analyzed IL-6 and IP-10 responses of MEF expressing wild-type FADD and FADD-deficient MEF following LPS, poly(I)·poly(C), and Pam3CSK4 stimulation. LPS induced an 18-fold increase of IL-6 in the medium of wild-type MEF over that of untreated MEF cultures. As would be predicted, FADD-deficient MEF displayed a robust 48-fold induction of IL-6 production in response to LPS (2.7-fold induction over wild-type MEF) (Fig. 8C). Similarly, Pam3CSK4 stimulation led to enhanced secretion of IL-6 in FADD-deficient cells compared with wild-type MEF (twofold induction over wild-type MEF) (Fig. 8C). In contrast, poly(I)·poly(C) led to an eightfold stimulation of IL-6 production in wild-type MEF, whereas FADD-deficient MEF showed almost no responsiveness to the same stimulation (Fig. 8C). This finding confirms previous data showing a requirement for FADD in TLR3-mediated signaling (18). To confirm that the enhanced IL-6 production in LPS- and Pam3CSK4-stimulated FADD-deficient MEF was due solely to the absence of FADD, FADD was transduced into FADD-deficient MEF, and IL-6 concentration was measured in the medium. Reconstitution of FADD in FADD-deficient MEF reversed the enhanced IL-6 production following stimulation by LPS and Pam3CSK4 (Fig. 8D).

Poly(I)·poly(C) activates the IRAK1/MyD88-independent TLR3 pathway using alternative adaptor proteins, such as TRIF. LPS is able to activate both IRAK1/MyD88-dependent and -independent pathways. Our findings thus far would predict that the IRAK1/MyD88-independent pathway should not be enhanced in FADD-deficient cells. To confirm that the IRAK1/MyD88-independent, TRIF/IRF3-dependent pathway was not enhanced in FADD-deficient MEF, IP-10 production was examined (22). Poly(I)·poly(C) led to robust IP-10 production in wild-type MEF, but IP-10 production in FADD-deficient MEF was significantly impaired. Although a similar trend was observed between wild-type and FADD-deficient MEF following LPS stimulation, the reduced IP-10 response was not statistically significant (P = 0.47) (Fig. 8E). Nevertheless, the lack of enhanced IP-10 production by LPS in FADD-deficient MEF proves the differential effect of the role of FADD in IRAK1/MyD88-dependent and -independent events. Together, our data indicate that FADD negatively regulates cytokine production by attenuating the IRAK1/MyD88-dependent pathway, but not the IRAK1/MyD88-independent pathway.

DISCUSSION

TLRs function as primary sensors of microbial products. TLR activation is necessary to provide host defense against invading pathogens. However, an excessive inflammatory response to LPS, for example, can be harmful and in some cases leads to fatal septic shock. Consequently, cytokine production must be tightly controlled by mechanisms which temper the TLR response. Negative regulation of the TLR response occurs at many levels and can be achieved by a combination of different mechanisms, which include degradation of positive signaling components, such as IRAK1, TIRAP, and TLR4 (9, 26, 35). Recently, several diverse intracellular proteins (such as RP105, ST2, IRAK-M, MyD88s, and A20) have been identified as negative regulators of TLR signaling (4-6, 11, 24). Many of these are inducible by LPS and therefore embed the pathway with a negative-feedback loop for desensitization. However, there exists a considerable lag period for induction of some of these inhibitors. The negative regulators Dok1 and Dok2 in contrast are constitutively expressed, and this allows for the control of the initial burst of the inflammatory response. However, Dok1 and Dok2 target only activation of ERKs by LPS (31a).

In this study we demonstrate that FADD interacts with IRAK1 and MyD88 to inhibit MyD88-dependent (LPS, Pam3CSK4, and IL-1β), but not MyD88-independent [poly(I)·poly(C) and TNF], signaling pathways. Our data support a model where LPS stimulation recruits a FADD-IRAK complex to MyD88, whereupon increasing accumulation of FADD-bound IRAK1 to MyD88 destabilizes the ternary complex, resulting in dissociation of IRAK1 from MyD88. Indeed, overexpression of FADD significantly reduces detectable MyD88-IRAK1 interactions (Fig. 5A), whereas complete loss of FADD enhances and prolongs the MyD88-IRAK1 interaction (Fig. 5C). Importantly, the IRAK1-FADD interaction is required for MyD88-FADD interaction to occur, suggesting that IRAK1 shuttles FADD to MyD88 at the receptor complex. Although IRAK4 also contains a DD, we were unable to detect an association between FADD and IRAK4 (data not shown). It is conceivable that the binding specificity is conferred by other determinants which are present in IRAK1 but are absent in IRAK4. Although previous studies have demonstrated the ability of FADD to inhibit LPS-induced NF-κB activation (3), the mechanism of this activity has until now remained unknown. Our findings demonstrate that FADD acts at a very early point in the MyD88-dependent TLR cascade and thus serves as an immediate mechanism to temper signaling from the activated receptor.

We also show that enforced expression of FADD in endothelial cells suppresses LPS-induced activation of multiple signaling pathways. This inhibition also occurs in cells expressing only the DD of FADD, indicating that the DD alone is sufficient to confer this negative role. These findings are consistent with previous data showing that the FADD DD can inhibit LPS-induced apoptosis, which is mediated by JNK activation (8, 17). In contrast, a point mutation in the DD (V121N) abrogates the effects of overexpressed FADD on LPS signaling, indicating that the integrity of the DD is critical for the inhibitory function of FADD. In the absence of FADD, LPS leads to hyperactivation of JNK, Akt, and ERK, further supporting the role of FADD as a negative regulator. Interestingly, although we were able to knock down FADD protein to less than 20% of normal levels by RNA interference (data not shown), we were not able to detect an effect on downstream MAPK signaling, suggesting that FADD is present in vast excess with respect to its function in modulating IRAK1/MyD88-dependent pathways. Together, these results suggest that under normal conditions, FADD provides an early phase safety check of the innate immune response by conferring a proper magnitude of cell activation, rather than a complete abrogation of MyD88-dependent TLR signaling. Alternatively, it may be necessary for innate immune cells to reduce the expression of FADD to allow for a more vigorous response against invading pathogens.

Regulation of the innate immune response by FADD likely plays a significant physiological role and may in part explain the major defects in T- and B-cell development in FADD-null mice that appear to be independent of death receptor pathways (19, 36, 37). FADD-deficient embryos also show significant heart defects, are hemorrhagic, and die on embryonic day 10.5 (36). In the context of the vasculature, LPS-induced activation of the endothelium promotes angiogenesis (31). We have previously shown that LPS-induced angiogenic pathways lie downstream of TRAF6 and that JNK and NF-κB activity are essential for LPS-induced endothelial sprouting (31). Our finding that FADD overexpression inhibits endothelial sprouting suggests that FADD likely modulates the angiogenic response in the innate immune pathway.

FADD-deficient cells also displayed enhanced cytokine production following activation of MyD88-dependent (TLR4 and TLR2) but not MyD88-independent (TLR3 and TLR4) pathways. The enhanced IL-6 production correlated with hyperactivation of MAPK and NF-κB signaling pathways which are critical in both transcriptional and posttranscriptional control of IL-6 production. Reconstitution of FADD reversed the enhanced IL-6 production, providing further functional confirmation of the role of FADD in regulating MyD88-dependent pathways.

Although most TLRs use the adaptor protein MyD88 leading to the activation of IRAK, TLR4 and TLR3 use an alternate adaptor protein, TRIF. In the case of TLR4, both MyD88- and TRIF-dependent pathways are engaged, whereas TLR3 functions only through TRIF (12). This latter pathway proceeds without IRAK participation and induces expression of genes, such as IP-10 (18). TLR3 stimulation with poly(I)·poly(C) led to robust IP-10 production in wild-type cells but was impaired in FADD-deficient cells, thus indicating a positive role for FADD in TLR3-regulated IP-10 gene expression. This finding was expected, since a role for FADD in TLR3-mediated responses, such as B-cell proliferation, as well as in apoptosis, has previously been reported (18, 20). In contrast to the IL-6 response, LPS-induced IP-10 production was not affected in the absence of FADD, thereby confirming that FADD diminishes only IRAK1/MyD88-dependent signaling. Given that IP-10 can also be regulated by NF-κB, which is hyperactivated—through the IRAK1/MyD88-dependent pathway—in the absence of FADD, this may explain the lack of suppression of IP-10 induction in response to LPS stimulation. Taken together, our data indicate that whereas FADD is required to propagate the TRIF-dependent response, FADD acts to attenuate IRAK1/MyD88-dependent signaling.

In conclusion, our findings demonstrate that FADD is a novel negative regulator of LPS signaling through its interaction with IRAK1 and inhibition of IRAK1-MyD88 interactions. Our data also suggest that IRAK1 acts as a shuttle to bring FADD in contact with activated MyD88, which serves as a switch to attenuate the LPS signal. Given the ubiquitous expression of FADD in immune cells, our findings implicate FADD in playing a critical role in fine-tuning the innate immune response and may explain some of the death receptor-independent functions postulated for FADD (19, 36, 37).