Divergence of bacterial lipopolysaccharide pro-apoptotic signaling downstream of IRAK-1.

The vascular endothelium is a key target of circulating bacterial lipopolysaccharide (LPS). LPS elicits a wide array of endothelial responses, including the up-regulation of cytokines, adhesion molecules, and tissue factor, many of which are dependent on NF-kappa B activation. In addition, LPS has been demonstrated to induce endothelial apoptosis both in vitro and in vivo. Although the mechanism by which LPS activates NF-kappa B has been well elucidated, the signaling pathway(s) involved in LPS-induced apoptosis remains unknown. Using a variety of dominant negative constructs, we have identified a role for MyD88 and interleukin-1 receptor-associated kinase-1 (IRAK-1) in mediating LPS pro-apoptotic signaling in human endothelial cells. We also demonstrate that LPS-induced endothelial NF-kappa B activation and apoptosis occur independent of one another. Together, these data suggest that the proximal signaling molecules involved in LPS-induced NF-kappa B activation have a requisite involvement in LPS-induced apoptosis and that the pathways leading to NF-kappa B activation and apoptosis diverge downstream of IRAK-1.

Gram-negative bacterial sepsis is a life-threatening event that afflicts ϳ200,000 patients annually in the United States (1). A deleterious outcome of sepsis is the development of septic shock. Endothelial cell (EC) 1 injury and/or dysfunction is a commonality among several key complications associated with septic shock, including systemic vascular collapse, disseminated intravascular coagulation, and vascular leak syndromes (2,3). The pathogenesis of septic shock and its attendant vascular complications is believed to be mediated, at least in part, by lipopolysaccharide (LPS), a component of the outer envelope of all Gram-negative bacteria (4 -6).
LPS directly activates the vascular endothelium and elicits an array of EC responses including the up-regulation of pro-inflammatory cytokines, adhesion molecules, and tissue factor. Many of these responses are mediated by NF-B (7). The signaling pathway by which LPS activates NF-B in EC has only recently been elucidated (8). Circulating LPS binds to the acute phase protein, LPS-binding protein (LBP), and this complex is recognized by soluble CD-14 (sCD14) (9). Although the exact mechanism of interaction remains unknown, this LPS⅐LPSbinding protein⅐soluble CD14 complex is recognized by the transmembrane receptor, Toll-like receptor (Tlr)-4 (10). The extracellular domain of Tlr-4 contains repeating leucine-rich motifs, and the cytoplasmic portion contains a Toll receptorinterleukin-1 receptor (TIR) domain with homology to the intracellular signaling domain of the type 1 interleukin-1 receptor (11). After Tlr-4 activation, another TIR domain-containing protein, MyD88, is recruited to Tlr-4 through respective TIR-TIR interactions (12,13). MyD88 also contains a death domain (DD), a highly conserved protein binding domain that facilitates its interaction with another DD-containing signaling molecule, interleukin-1 receptor-associated kinase-1 (IRAK-1) (11). IRAK-1 subsequently autophosphorylates, dissociates from MyD88, and interacts with TNF receptor-associated factor-6 (14,15). This activates a downstream kinase cascade involving NF-B-inducing kinase and IB kinase, resulting in the phosphorylation and degradation of the NF-B inhibitor, IB, and the nuclear translocation of NF-B (11).
Several reports indicate that LPS elicits vascular endothelial apoptosis both in vitro (2, 16 -20) and in vivo (21)(22)(23). We have previously elucidated an anti-apoptotic signaling pathway involving FLICE-like inhibitory protein (FLIP) that protects human EC from LPS-induced apoptosis (17). FLIP is a cytoprotective protein that is rapidly degraded by the proteosome. In the absence of new protein synthesis, FLIP levels decrease, resulting in sensitization of human EC to LPS-induced apoptosis. Recently, Tlr-4 has been reported to have a requisite involvement in LPS pro-apoptotic signaling (24), similar to its role in mediating LPS-induced NF-B activation. However, the intracellular signaling pathway through which LPS activates apoptosis remains unknown. We, therefore, decided to investigate whether downstream signaling molecules involved in Tlr-4-mediated activation of NF-B were similarly involved in mediating apoptosis.

EXPERIMENTAL PROCEDURES
Materials-LPS from Escherichia coli serotype 0111:B4 and polymyxin B were purchased from Sigma. Recombinant human TNF-␣ was purchased from R&D Systems, Inc. (Minneapolis, MN). The caspase inhibitor peptide, z-VAD-FMK (zVAD) and the protein synthesis inhibitor, cycloheximide, were purchased from Calbiochem-Novabiochem.
Cloning and Stable Expression of cDNA Constructs-cDNA encoding either the DD or TIR domain of MyD88, the DD of IRAK-1 (generous gifts of Dr. Marta Muzio, Mario Negri Institute, Milan, Italy), or the DD of Fas-associated death domain (FADD) (generous gift of Dr. Vishva Dixit, Genentech, Inc., South San Francisco, CA) was cloned into the bicistronic retroviral expression plasmid, pBMN-IRES-enhanced green fluorescent protein (EGFP) (kindly provided by Dr. Gary Nolan, Stanford University, Stanford, CA) (26). High titer retrovirus was prepared from the Phoenix amphotropic packaging cell line (ATCC, Manassas, VA) transfected with 24 g of the expression plasmid by calcium phosphate precipitation. Recombinant retroviral supernatants were collected 48 h after transfection and filtered through a Millex-HV 0.45-m filter (Millipore Corp., Bedford, MA). For infection, 4 ϫ 10 5 EC were seeded/well of a 6-well plate for 24 h to achieve ϳ80% confluence. The growth medium was replaced with 2.5 ml of retroviral supernatant supplemented with 32 g/ml Polybrene and 10 mM HEPES, and the plate was centrifuged for 2 h (1430 ϫ g; 32°C). The cells were then incubated for 10 h (5% CO 2 , 37°C), after which the retroviral supernatant was replaced with normal growth medium. Cells were analyzed and sorted on the basis of EGFP expression using a FACVantage SE cell sorter (Becton Dickinson Corp., Franklin Lakes, NJ).
Luciferase Assay-A recombinant adenovirus (KZ142) system (gift of Dr. James Kelly, ZymoGenetics Inc., Seattle, WA) was used to transfect cells with a luciferase reporter construct. The adenoviral construct was created as follows. An oligonucleotide encoding a consensus NF-B binding site, the tandem NF-B binding sites of the human immunodeficiency virus-1 long terminal repeat (27), two copies of the collagenase AP-1 element, and a single copy of the c-Jun TRE (28) were ligated into a luciferase reporter cassette and then placed in the pACCMV.pLpA adenoviral shuttle vector for construction of recombinant adenovirus as described (29). For transfection of the luciferase reporter construct, ECs were seeded into 96-well black view plates (Corning Inc., Corning, NY) at a density of 50,000 cells/well for 24 h and subsequently incubated for 16 h at a multiplicity of infection of 2000 in RPMI supplemented with 1% fetal bovine serum. After infection, ECs were exposed to experimental treatment in Ham's F-12 medium supplemented with 2.5% fetal bovine serum, 20 mM HEPES, and 0.5% bovine serum albumin for 4 h at 37°C. Luciferase activity was determined using a commercially available assay kit and a TopCount NXT luminescence counter (both from Packard Instrument Co.).
Caspase Assay-ECs were seeded into 96-well plates at a density of 60,000 cells/well, cultured for 24 h, and treated, and caspase activity was measured with the homogenous caspases assay according to the manufacturer's instructions (Roche Molecular Biochemicals). The plates were analyzed on a Cytofluor Series 4000 fluorescence plate reader (Perseptive Biosystems Inc., Framingham, MA) at 485-nm excitation and 530-nm emission, and caspase activity was expressed relative to simultaneous medium control. Because human EC sensitization to LPS-and TNF-␣-induced apoptosis is dependent upon the inhibition of new protein synthesis by a mechanism that we have previously characterized (17) (30). Construction and purification of IB␣M and control (␤-galactosidase) recombinant adenovirus were performed as previously described (20). For adenoviral transduction, ECs were seeded into 96-well plates at a density of 50,000 cells/well for 24 h, washed once with medium, and incubated for 48 h at a multiplicity of infection of 1000 with control or IB␣M adenovirus in complete EC medium. ECs were washed once with complete medium and subjected to experimental treatment.
Statistical Methods-A t test or analysis of variance was used to compare the mean responses between a single experimental group or multiple experimental groups, respectively, and the control group. For experiments analyzed by analysis of variance, the Tukey post hoc comparison test was used to determine between which groups significant differences existed. All statistical analyses were performed using GraphPad Prism version 3.00 for Macintosh (GraphPad Software, Inc., San Diego, CA). A p value of Ͻ0.05 was considered significant.

RESULTS
Polymyxin B Inhibits LPS-induced EC Apoptosis-As the name implies, LPS is comprised of both a polysaccharide and a lipid moiety. It is well established that the lipid A moiety of LPS is responsible for it pro-inflammatory properties (31). Furthermore, this portion of LPS alone activates the Tlr-4 receptor, resulting in NF-B activation and the up-regulation of pro-inflammatory cytokines (32). To determine whether LPSinduced apoptosis is similarly mediated by lipid A, ECs were exposed to medium or LPS preincubated with or without polymyxin B (Fig. 1). Polymyxin B, derived from the bacterium Bacillus polymyxa, binds and neutralizes the lipid A moiety of LPS (33,34). Consistent with a role for lipid A in cell activation, neutralization with polymyxin B completely inhibited LPSinduced NF-B activation (Fig. 1A). Similarly, polymyxin B was able to completely block LPS-induced apoptosis (Fig. 1B), suggesting that the pro-apoptotic properties of LPS are local- ized to the lipid A region as well.
Expression of Dominant Negative (D/N) MyD88 Inhibits LPS-induced Apoptosis in EC-MyD88, by virtue of its direct recruitment to Tlr-4, is one of the most proximal intracellular signaling molecules involved in LPS-induced activation of NF-B. MyD88 contains two distinct protein interaction domains, a TIR domain and a DD, which mediate its ability to bind to Tlr-4 and IRAK, respectively (12). The expression of either domain alone blocks Tlr-4 mediated NF-B activation in a dominant negative fashion (8,12). To determine whether MyD88 mediates LPS-induced apoptosis, ECs were stably transfected with cDNA encoding either the TIR domain or the DD. Expression of these two D/N forms of MyD88 was confirmed by Western blot analysis ( Fig. 2A). Furthermore, the functional efficacy of these D/N constructs was confirmed by assaying for the ability of each to inhibit LPS-induced NF-B activation (Fig. 2B). Ex-pression of either the TIR domain or the DD alone inhibited LPS-induced NF-B activation by Ͼ80%. In experiments conducted in parallel, expression of either MyD88 D/N also inhibited LPS-induced caspase activity, albeit to a lesser extent (Fig.  2C). EC expressing either the TIR domain or the DD demonstrated a Ն30% reduction in caspase activity after LPS exposure than EC-expressing vector alone.
Similar to LPS, TNF-␣ is a well described activator of EC NF-B (35). Distinct receptor membrane complexes and proximal intracellular signaling molecules mediate LPS-and TNF-␣-induced NF-B activation; farther downstream, however, the signaling pathways leading to NF-B activation converge at the level of NF-B-inducing kinase and IB kinase (36). Consistent with the role of MyD88 in mediating signaling elicited by LPS, but not TNF-␣, expression of the MyD88 D/N encoding the DD had no effect on TNF-␣-induced apoptosis (Fig. 2C). ECs expressing the MyD88 D/N encoding the TIR domain, however, demonstrated ϳ15% less caspase activity after TNF-␣ exposure than ECs expressing vector alone. Although this decrease in caspase activity was subtle, it was statistically significant.

IRAK-1 D/N Inhibits LPS-induced EC Apoptosis-After
Tlr-4 activation and MyD88 recruitment, IRAK-1 transiently binds MyD88 and undergoes an autophosphorylation step leading to its activation (12,14,15). Optimal activation of NF-B by LPS is dependent upon functional IRAK-1 (8,12,15). To determine whether IRAK-1 has a similar involvement in promoting LPS apoptotic signaling, a truncated form of IRAK-1, reported to function as a D/N, was expressed in EC. Consistent with previous reports (12, 20), expression of the DD alone of IRAK-1 significantly blocked LPS-induced NF-B activation (Fig. 3A). Expression of this IRAK-1 D/N construct blocked Ͼ60% of the To confirm the functional efficacy of the D/N constructs, these cells were treated for 4 h with either medium or LPS (100 ng/ml), lysed, and assayed for luciferase activity (B). In other experiments, ECs expressing EGFP vector, the TIR domain of MyD88, or the DD of MyD88 were treated for 8 h with medium, LPS (100 ng/ml), or TNF-␣ (10 ng/ml) and assayed for caspase activity (C). The vertical bars represent the mean Ϯ S.E. of NF-B (B) or caspase (C) activity relative to simultaneous media controls. * ϭ significantly decreased compared with EGFP vector-transfected ECs exposed to identical treatment. IB, immunoblot.

FIG. 3. Expression of IRAK-1 D/N inhibits LPS-induced apoptosis.
ECs were stably transfected with either EGFP vector alone or cDNA encoding the DD of IRAK-1 (IRAK-1-D/N) (A and B). The functional efficacy of this D/N construct to inhibit LPS-induced NF-B activation was assayed (A). ECs were treated for 4 h with either medium or LPS (100 ng/ml), lysed, and assayed for luciferase activity. In other experiments, ECs expressing EGFP vector or IRAK-1-D/N were treated for 8 h with medium, LPS (100 ng/ml), or TNF-␣ (10 ng/ml) and assayed for caspase activity (B). The vertical bars represent the mean Ϯ S.E. NF-B (A) or caspase (B) activity relative to simultaneous media controls. * ϭ significantly decreased compared with EGFP vector-transfected ECs exposed to identical treatment.
FADD Is Not Required for LPS Pro-apoptotic Signaling-FADD is a pro-apoptotic signaling molecule that couples death receptors to initiator caspases. FADD has been clearly established to mediate apoptosis initiated by several receptors, including those for TNF (37,38), TRAIL (TNF-related apoptosisinducing ligand (39,40)), and CD95L (37,41). FADD has also been implicated in mediating Tlr-2-induced apoptosis (42). Both Tlr-2 and Tlr-4, which recognize bacterial lipoproteins and LPS, respectively, are highly homologous receptors belonging to the larger family of Toll-like receptors (43). Although each recognizes distinct ligands, Tlr-2 and Tlr-4 utilize a common intracellular signaling pathway that activates NF-B. To determine whether FADD is required for LPS-induced apoptosis, ECs were transfected with cDNA encoding a truncated form of FADD that has previously been demonstrated to function in a D/N manner (37,38). Western blot analysis confirmed efficient expression of the FADD D/N construct (Fig. 4A). ECs expressing the FADD D/N displayed equivalent amounts of caspase activity after LPS stimulation as EC-expressing vector alone (Fig. 4B). Consistent with previous studies (37,38), expression of the FADD D/N construct significantly inhibited TNF-␣-induced apoptosis.
LPS-induced EC Apoptosis Is Independent of NF-B Activation-To determine whether NF-B activation is required for LPS-induced apoptosis, ECs were transduced with a gene encoding mutations in the inhibitor of NF-B, IB␣, that render it resistant to phosphorylation and degradation (20,30). Accordingly, expression of the IB␣M significantly inhibited LPSinduced NF-B activation compared with expression of a vector control encoding ␤-galactosidase (Fig. 5A). ECs expressing the IB␣M, however, demonstrated equivalent levels of caspase activity after LPS exposure as those ECs expressing vector alone (Fig. 5B). To determine whether caspase activation contributes to NF-B signaling, EC were exposed to LPS in the presence or absence of the caspase inhibitor, zVAD. At a concentration that could block 100% of the LPS-induced caspase activity (Fig. 5C), zVAD failed to inhibit LPS-induced NF-B activation (Fig. 5D). Interestingly, zVAD actually enhanced LPS-induced NF-B activation by ϳ40%. DISCUSSION We have previously established that LPS induces human EC apoptosis by several different criterion including poly(ADPribose) polymerase cleavage, nuclear histone release, and DNA laddering (17,44). To quantify relative changes in apoptosis, we have used a caspase activity assay. Caspases are highly specific effector proteases that are activated during apoptosis and cleave cellular substrates (45). The observed increase in caspase activity after LPS exposure is consistent with previous reports that LPSinduced EC apoptosis is mediated, in part, by caspase-dependent proteolysis of key EC substrates (16,17,44).
In the present report, we have established that the lipid A moiety of LPS is responsible for its pro-apoptotic signaling properties. Consistent with its role in activation, neutralization of the lipid A moiety of LPS with polymyxin B inhibited LPSinduced NF-B activation (Fig. 1A). Similarly, polymyxin B was able to completely abrogate LPS-induced apoptosis (Fig.  1B), suggesting that the lipid A portion of the molecule is responsible for its pro-apoptotic signaling properties. It has previously been reported that LPS-induced apoptosis is dependent on Tlr-4 signaling (24). In that study, the authors demonstrate that macrophages derived from C3H/HeJ mice, which have a missense mutation in the third exon of Tlr-4 (46), were resistant to LPS-induced apoptosis. It has further been established that the domain of LPS recognized by Tlr-4 is the  A and B). An anti-AU1 antibody and an antibody raised against FADD that recognizes its DD were used to confirm expression of the FADD D/N (A). In other experiments, these cells were treated for 8 h with medium, LPS (100 ng/ml), or TNF-␣ (10 ng/ml) and assayed for caspase activity (B). The vertical bars represent the mean Ϯ S.E. caspase activity relative to simultaneous media controls. * ϭ significantly decreased compared with EGFP vector-transfected ECs exposed to identical treatment. IB, immunoblot.

FIG. 5. Effect of the inhibition of NF-B activation on LPSinduced apoptosis.
ECs were infected with either control (␤-galactosidase (␤-gal)) or IB␣M adenovirus for 48 h, subsequently treated with medium or LPS (100 ng/ml) for 4 or 8 h, and assayed for luciferase (A) or caspase activity (B), respectively. In other experiments, ECs were pretreated for 1 h with Me 2 SO or the caspase inhibitor peptide zVAD (100 M), treated as above, and assayed for caspase (C) and luciferase activity (D). The vertical bars represent the mean Ϯ S.E. of NF-B (A and D) or caspase (B and C) activity relative to simultaneous media controls. * ϭ significantly decreased compared with ␤-galactosidase vector control. ** ϭ significantly different from LPS alone. lipid A moiety (32). Together, these data imply that the lipid A moiety confers pro-apoptotic signaling through a Tlr-4-dependent pathway. This is consistent with our finding that neutralization of lipid A completely protects against LPS-induced apoptosis.
There has been some controversy in the past regarding whether Tlr-2 or Tlr-4 mediates LPS-induced signaling. The genetic evidence clearly indicates that Tlr-4 is the putative receptor for LPS (46). Other studies using overexpression conclude that Tlr-2 recognizes LPS (47,48). It was later ascertained that certain commercial preparations of LPS were contaminated with bacterial lipoproteins, the latter of which activate Tlr-2 signaling (49). In the present report, highly purified LPS, which was phenol-extracted and purified by ion exchange chromatography, was used for all experiments. The finding that polymyxin B completely blocks LPS-induced NF-B activation and apoptosis rules out that the EC responses studied were influenced by contaminating lipoproteins.
To investigate whether signaling molecules that link Tlr-4 to NF-B activation are also involved in mediating LPS-induced apoptosis, D/N versions of MyD88 and IRAK-1 were expressed in EC, and apoptosis was assayed. LPS-induced apoptosis was significantly inhibited by expression of either one of two MyD88 D/N constructs (Fig. 2C) or an IRAK-1 D/N (Fig. 3B). These constructs had minimal or no inhibitory effect on TNF-␣-induced apoptosis, consistent with previous reports that TNF-␣ signaling occurs independently of MyD88 and IRAK-1 (50,51). Expression of a D/N form of FADD, a pro-apoptotic adapter molecule that links death receptors to initiator caspases, failed to block LPS-induced apoptosis (Fig. 4B). In accordance with the well described role of FADD in promoting TNF-␣-induced apoptosis (37), FADD D/N expression significantly inhibited TNF-␣-induced caspase activation. It has previously been reported that both MyD88 and FADD contribute to Tlr-2-induced apoptosis. Because LPS has been shown to initiate apoptosis via Tlr-4 (24), the present data suggest differences in pro-apoptotic signaling between different members of the Tlr family. Similar to Tlr-2, MyD88 promotes Tlr-4induced apoptosis; in contrast, Tlr-4-induced apoptosis occurs independently of FADD.
The MyD88 and IRAK-1 D/N constructs were able to inhibit Ն30 and 50% of the LPS-induced caspase activation, respectively. Expression of the FADD D/N, which has been demonstrated to have a requisite involvement in TNF-␣-induced apoptosis (37), inhibited ϳ50% of the caspase activity elicited by TNF-␣. The inability to completely inhibit apoptosis using any of the D/N constructs is not surprising considering endogenously expressed full-length MyD88, IRAK-1, and FADD are still expressed. There is evidence suggesting that Tlr-4 activation of NF-B can occur through a MyD88-and IRAK-1-independent pathway. First, LPS-induced NF-B DNA binding activity in macrophages derived from either MyD88 or IRAK-1 knockout mice was delayed, but not inhibited, indicating that cellular activation by LPS can occur in the absence of these signaling molecules (15,52). Second, a MyD88-like protein has been recently described by two independent groups named MyD88-adapter-like protein (MAL) or TIR domain-containing adapter protein (TIRAP), which can promote LPS-induced NF-B signaling through IRAK-2 (53,54). Whether MAL/TI-RAP and/or IRAK-2 mediate LPS-induced apoptosis remains unknown. The existence of alternative activation pathways, which may also be involved in LPS pro-apoptotic signaling, could explain the lower efficacy of the MyD88 D/N in abrogating LPS-induced caspase activation.
Because both MyD88 and IRAK-1 contribute to LPS-induced NF-B activation and apoptosis, we investigated whether these two events were mutually dependent. NF-B activation has been classically viewed as a cytoprotective event based on its ability to up-regulate anti-apoptotic proteins (55,56); however, there are reports that NF-B signaling has a role in pro-apoptotic signaling as well (57,58). In the present study, inhibition of NF-B activation had no effect on LPS-induced apoptosis, and inhibition of caspases failed to down-regulate LPS-induced NF-B activation, suggesting that these two responses occur independent of one another (Fig. 5). Interestingly, caspase inhibition augmented LPS-induced NF-B activation. Recently, the p65 subunit of NF-B has been identified as a substrate of EC caspases (59). Because zVAD-exposed EC demonstrated 30% less caspase activity than EC exposed to medium alone, it is possible that zVAD prevents the degradation of NF-B under basal conditions, thus increasing the amount of functional NF-B in the cell. This may account for the enhanced LPSinduced NF-B activation in EC pretreated with the caspase inhibitor.
In summary, we have identified a pro-apoptotic signaling pathway involved in LPS-induced apoptosis. First, the lipid A moiety of LPS was identified as the portion of the molecule that confers pro-apoptotic signaling. Second, a role for MyD88 and IRAK-1 in mediating LPS-induced apoptosis was established. Third, LPS-induced apoptosis was determined to occur through a FADD-independent pathway. Finally, despite the involvement of MyD88 and IRAK-1 in promoting both LPS-induced NF-B signaling and apoptosis, these cellular responses were determined to occur independent of one another. We propose that LPS-induced apoptosis is mediated through Tlr-4 and involves the recruitment of MyD88 and IRAK-1. Furthermore, the data presented here suggest that the signaling pathways leading to NF-B activation and apoptosis diverge downstream of IRAK-1.