We first analyzed the localization of TIRAP in macrophages and mouse embryonic fibroblasts (MEFs) and found TIRAP to be localized to the plasma membrane, where it was concentrated at the leading edge of the MEFs (data not shown). Macrophages do not typically have a leading edge, but instead have membrane ruffles, which are discrete regions of the plasma membrane that are biochemically similar to the leading edge of fibroblasts (Allen et al. 1998; Araki et al. 1996). Accordingly, TIRAP was enriched with filamentous actin at plasma-membrane ruffles in macrophages (Figure 1A). In addition to being concentrated at ruffles, TIRAP-positive, actin-negative vesicles were found throughout the cell (see Figure S1A in the Supplemental Data available with this article online). Identical results were obtained when examining the localization of TIRAP tagged with GFP, FLAG, or HA (Figure 1A; see also below and Figure S1B), indicating that the localization did not result from the tag and represents a property of the TIRAP protein itself. These data indicate that TIRAP resides at both actin-rich membrane ruffles and intracellular vesicles.#

These findings were corroborated by biochemical fractionation of macrophages expressing TIRAP-GFP, which demonstrated that TIRAP cofractionated with membranes but was absent from cytosolic fractions (Figure S1C). Upon detergent extraction, TIRAP was redistributed into the cytosolic fraction. Since TIRAP lacks any obvious membrane-anchoring domains, we next investigated the mechanism of TIRAP localization to membranes and its role in TLR signaling.#

TIRAP regulates signaling by TLR2 and TLR4 (Horng et al. 2002; Yamamoto et al. 2002). To determine whether TIRAP distribution requires TLR2 or TLR4, localization studies were performed in macrophages deficient in both TLR2 and TLR4. No differences in TIRAP localization were observed when comparing wild-type (wt) or TLR2xTLR4 knockout (KO) macrophages (Figure 1B).#

The localization of TIRAP P125H, which is unable to interact with TLR4 (Horng et al., 2001), was examined as an independent approach to determine the role of TLRs in TIRAP localization. The distribution of TIRAP P125H was similar to that of wt TIRAP (Figure S1B). Thus, TIRAP localization to membranes is not controlled by TIR-dependent interactions between TIRAP and TLRs and therefore involves a previously unrecognized property of this adaptor.#

The colocalization between TIRAP and actin at membrane ruffles led us to examine the role of actin in TIRAP localization. Transfected cells were treated with cytochalasin D, an inhibitor of actin polymerization, and TIRAP localization was examined. Within 30 min of cytochalasin D treatment, TIRAP was redistributed from being localized primarily at the plasma membrane to being enriched on intracellular tubules and vesicles (Figures S1D and S1E). Prominent redistribution was observed in 55% ± 6% of the cells examined, with the remaining cells simply losing TIRAP concentration at the cell surface. TIRAP-positive tubules and vesicles were also observed in untreated cells (Figure S1A), suggesting that cytochalasin D treatment did not create a new location for this protein but shifted the equilibrium to favor an intracellular distribution. These results indicate that TIRAP cycles between the plasma membrane and an intracellular location by an actin-dependent process.#

To identify the region that controls TIRAP localization, deletion mutants were generated. Each mutant was GFP tagged, and the resulting proteins were examined for their ability to phenocopy the localization of wt TIRAP. TIRAP mutants lacking the TIR domain (TIRAP 1–85) retained the ability to localize to the cell surface (Figure 1C; Figure S2A). Similarly, the first 40 amino acids of TIRAP (TIRAP 1–40) were sufficient to phenocopy wt TIRAP localization (Figure 1C; Figure S2A). These results provide further evidence that membrane localization of TIRAP is TIR independent.#

TIRAP 1–40 contains a region (residues 15–35) with a high isoelectric point (10.46) compared to the entire protein (7.55) that is enriched in basic and aromatic residues (Figure 1D). These properties are often found in domains that bind PIP2 (McLaughlin et al., 2002). The localization of this putative PI binding motif was identical to wt TIRAP (Figure 1C; Figure S2A). In contrast, residues 1–20 contain no membrane-targeting activity (Figure 1C; Figure S2A). The putative PI binding domain contains six conserved lysines. Since mutation of lysine residues in PI binding domains can abolish lipid binding and disrupt subcellular localization (Stauffer et al., 1998), we generated mutant versions of TIRAP where lysines were substituted with alanines. Membrane localization of TIRAP was abolished by mutation of four of the lysines (TIRAP 4×) (Figure 1C; Figure S2A). These data indicate that the N-terminal region with properties of a PI binding motif is necessary and sufficient to direct TIRAP localization.#

To determine whether TIRAP binds to lipids, we performed several in vitro protein/lipid interaction assays. First, a GST-TIRAP fusion protein was overlaid onto nitrocellulose membranes containing a variety of lipid species (PIP strips). GST-TIRAP, but not GST, bound to all PIs and phosphatidylserine (PS), but not to phosphatidylethanolamine (PE), phosphatidylcholine (PC), or any other lipid examined (Figure 2A). As an independent approach, we examined the ability of GST-TIRAP to bind liposomes containing known concentrations of lipids. GST-TIRAP, but not GST, cosedimented with liposomes containing PIP2 or PI(4)P, but not liposomes containing PC and PE (Figure S3B). To determine which lipid is bound preferentially, quantitative analysis was performed on TIRAP's ability to bind liposomes containing PIP2, PI(4)P, PI, or PS. We focused our attention on these lipids because each is enriched at the plasma membrane at steady state, with the exception of PI. Under all conditions tested, GST-TIRAP interacted preferentially with PIP2-containing liposomes (Figure 2B). Thus, the recruitment of TIRAP to the plasma membrane is likely to be mediated primarily by interactions with PIP2.#

We next analyzed the requirement for the putative PI binding motif of TIRAP for lipid binding. Mutations of two conserved lysines (TIRAP 2×) reduced, and mutation of four lysines (TIRAP 4×) severely diminished, the lipid binding ability of TIRAP (Figure 2A; Figure S3A). Thus, TIRAP 4× is defective in both lipid binding in vitro and cellular localization in vivo, suggesting that PI binding (most likely to PIP2) is required for TIRAP localization.#

We next examined the functional significance of TIRAP's PI binding activity and cellular localization. TIRAP KO macrophages were retrovirally transduced with either wt TIRAP, TIRAP 4×, or TIRAP constructs lacking the TIR domain (TIRAP 1–85). Only macrophages reconstituted with wt TIRAP regained the ability to produce IL-6 in response to LPS treatment (Figure 2C). The minimal activity of TIRAP 4× is likely due to residual ability of TIRAP 4× to bind lipids, as these mutations severely diminish but do not abolish PI binding activity. Interestingly, a construct consisting of the minimal localization motif (residues 15–35) fused to the TIR domain (15–35-TIR) also complemented TIRAP deficiency (Figure 2C). In contrast, a construct lacking the PI binding domain, where the first 20 amino acids were fused to the TIR domain (1–20TIR), was mislocalized (Figure 1C) and nonfunctional (Figure 2C). These data indicate that TIRAP's function is dependent on an intact PI binding domain. The TIRAP protein therefore consists of two functionally distinct domains, a C-terminal TIR domain responsible for the interaction with TLRs and MyD88 and an N-terminal PI binding domain that is responsible for subcellular localization. Both domains are necessary for TLR4 signaling, indicating a requirement of membrane targeting for TIRAP function as a TLR adaptor.#

To determine which PIs are required for TIRAP localization, different classes of PIs were depleted from cells. Treatment of cells with the PI-3 kinase inhibitor wortmannin disrupted the localization of a PI(3)P-specific PX domain (Figures S4A, S4B, and S4E). However, wortmannin did not affect TIRAP localization (Figures S4A, S4B, and S4E), indicating that 3′-PIs are not required for TIRAP localization.#

Salmonella enterica serovar Typhimurium (S. typhimurium) encodes a PI phosphatase, SopB (also known as SigD) (Norris et al. 1998; Terebiznik et al. 2002). SopB dephosphorylates 4′- and 5′-PIs in vivo, and overexpression of SopB is an effective means of depleting cellular levels of PIP2 (Terebiznik et al., 2002). We examined the effect of SopB on TIRAP localization in CHO cells, which tolerate high levels of SopB expression. Expression of SopB, but not the catalytically inactive mutant SopBC460S, disrupted the plasma-membrane localization of TIRAP (Figure 2D; Figure S4E), suggesting that 4′- and/or 5′-PIs are required for TIRAP localization. Furthermore, expression of SopB, but not SopBC460S, blocked TLR4-induced NF-κB activation (Figure 2E), demonstrating a role of 4′- and/or 5′ PIs in TLR4 signaling.#

To corroborate these findings, the localization of TIRAP was examined in cells expressing a phosphatase that acts specifically on 5′-PIs, INP54p (Raucher et al., 2000). Like SopB, INP54p disrupted cell-surface localization of TIRAP (Figure S4C and S4E). In contrast, a construct that contained the same membrane localization motif as the INP54p, but no phosphatase domain, did not effect TIRAP distribution (Figures S4D and S4E). These data indicate a requirement of 5′-PIs for TIRAP localization.#

To determine more specifically which lipid (or lipids) is the primary mediator of TIRAP localization in vivo, we examined the ability of individual PIs to direct localization of this adaptor. This was accomplished by generating chimeric TIRAP constructs where the N terminus (containing the PI binding domain) was replaced by a PI binding domain of a well-defined and restricted specificity. The resulting chimeras contain a heterologous N-terminal PI binding domain and a C-terminal TIR domain. Since PIs (and their interacting proteins) are enriched in distinct subcellular locations (De Matteis and Godi, 2004), only a subset of these chimeras should phenocopy wt TIRAP localization. Thus, using localization as a readout, we eliminated a variety of PIs as mediators of TIRAP localization. TIRAP chimeras whose subcellular distributions are directed by interactions with PI(3)P (PX domain from gp91-phox; PX-TIR), PI(4)P (PH domain from Fapp1; FAPP1-TIR) or PI(3,4,5)P3 (PH domain from GRP1; GRP1-TIR) displayed a distribution that was distinct from that of wt TIRAP (Figures 3B–3E). PX-TIR was found mainly on endosomes, FAPP1-TIR was found on the Golgi, and GRP1-TIR was scattered throughout the cell with minimal membrane staining. These subcellular distributions perfectly match the expected locations of proteins whose transport is dictated by interactions with these respective lipids (De Matteis and Godi, 2004). Thus, the primary mediator (or mediators) of TIRAP localization is not PI(3)P, PI(4)P, or PI(3,4,5)P3. In contrast, the TIRAP chimera whose localization is directed by interactions with PIP2 (PH domain of PLCδ1; PLC-TIR) displayed subcellular distribution identical to that of wt TIRAP; both wt TIRAP and PLC-TIR were enriched at the cell surface and membrane ruffles (Figure 3A).#

We next examined the effect of PI binding specificity on TIRAP function as a TLR4 adaptor. We tested the ability of the chimeras to complement the signaling defect of TIRAP KO macrophages in response to LPS. PX-TIR, GRP-TIR, and FAPP1-TIR were each unable to complement the signaling defect of TIRAP KO macrophages (Figure 3G). In contrast, the PLC-TIR chimera restored LPS responsiveness (Figure 3G).#

Collectively, these data indicate that PIP2 is the primary mediator of TIRAP localization and that PIP2 binding specificity is essential for TIRAP function.#

Intracellular proteins can be localized to the plasma membrane by several different mechanisms, and we therefore asked whether plasma-membrane localization is sufficient for TIRAP functional activity or whether PIP2-mediated targeting was specifically required. To address this question, we generated a TIRAP mutant that retained the ability to target to the plasma membrane by a PIP2-independent mechanism. The SH4 domain of the Src-family tyrosine kinase Fyn is sufficient to direct heterologous proteins to the plasma membrane (van't Hof and Resh, 1999). Membrane localization requires that the domain be both myristoylated and farnesylated. We substituted either the entire N-terminal portion of TIRAP (up to the TIR domain) or the PIP2 binding region of TIRAP with the Fyn SH4 domain (generating SH4-TIR and SH4-link-TIR chimeras, respectively) and examined their localization. Like wt TIRAP, both Fyn chimeras localized to the plasma membrane of macrophages, where they costained with F-actin (Figure 3F and data not shown). Interestingly, while both SH4-TIRAP chimeras localized similarly to wt TIRAP, they could not reconstitute TIRAP function in TIRAP KO macrophages (Figure 3G). These data indicate that TIRAP localization to the plasma membrane is not in itself sufficient for functional activity. Rather, PIP2-mediated recruitment of TIRAP is required.#

PIP2 is a critical hub in many cellular processes, and local concentrations of PIP2 are affected by a variety of signaling pathways (McLaughlin et al., 2002). We hypothesized that PIP2-dependent control of TIRAP function could provide a mechanism for coupling TLR4 signaling with the signaling pathways that regulate PIP2 metabolism. A potential connection with integrin signaling is particularly interesting in this regard because β2 integrins play multiple essential roles in macrophage biology (Hynes, 2002). Signaling by integrins promotes PIP2 synthesis and facilitates the formation of focal adhesions, cell migration, and phagocytosis (Hynes, 2002). CD11b is the only β2 integrin expressed on macrophages (Bhat et al., 1999). Interestingly, CD11b was shown to be involved in LPS signaling, but the mechanism of its role is unknown (Flo et al. 2000; Perera et al. 2001). We reasoned that one function of CD11b may be to promote TIRAP concentration at the plasma membrane by its ability to stimulate PIP2 synthesis. To test this, the localization of TIRAP was examined in CD11b KO macrophages. TIRAP was found primarily in membrane ruffles in 94% ± 5% of wt macrophages (Figure 4A). In contrast, TIRAP localization was evenly dispersed throughout the plasma membrane of CD11b KO macrophages; only 9% ± 3% of cells exhibited a ruffling phenotype (Figure 4A). Time-lapse microscopy confirmed that CD11b KO macrophages were defective for membrane ruffling that is typically exhibited by wt macrophages (Movie S1, Movie S2, Movie S3, and Movie S4). Whereas TIRAP-GFP and PLCδ1-PH-GFP cycled through membrane ruffles in wt macrophages (Movie S1 and Movie S3), an indicator of rapid changes in PIP2 levels, dynamic changes in TIRAP and PLCδ1-PH localization were not observed in CD11b KO macrophages (Movie S2 and Movie S4). These results suggested that CD11b promotes TIRAP enrichment at the plasma membrane by regulating PIP2 production.#

We next examined the role of CD11b in TIRAP-dependent TLR4 signaling. CD11b KO macrophages exhibited a defect in LPS-induced IL-6 production similar to TIRAP KO macrophages (Figure 4B). In contrast, LPS-induced expression of RANTES (a TIRAP-independent gene) did not require CD11b (Figure 4B). Thus, of the two pathways induced by TLR4, only TIRAP-dependent signaling is regulated by CD11b. Moreover, CD11b was not required for IL-6 production induced by CpG DNA, which signals through TLR9 via the TIRAP-independent pathway (Figure 4C).#

Collectively, these results support a model whereby CD11b, by triggering local PIP2 production, facilitates the recruitment of TIRAP to the plasma membrane and positively regulates TLR4-TIRAP-MyD88 signaling.#

Integrins stimulate PIP2 production in part by activation of ARF6. ARF6 is a GTPase that regulates the activity of PI 4-phosphate 5-kinase (PI5K), which generates PIP2 at the plasma membrane and endosomes (Honda et al., 1999). Peripheral-membrane proteins that bind PIP2 are enriched on ARF6-positive endosomes (Brown et al., 2001). Treatment of cells with cytochalasin D, or expression of ARF6 mutants unable to bind GTP, shifts the distribution of these proteins from the plasma membrane to internal compartments (Brown et al., 2001). TIRAP localization is sensitive to cytochalasin D (Figures S1D and S1E), suggesting that it is controlled by ARF6. To examine more directly whether ARF6 regulates TIRAP transport, we determined TIRAP localization in cells expressing wt ARF6 or an ARF6 mutant that is unable to bind GTP (ARF6T27N) (Radhakrishna and Donaldson, 1997). Wild-type ARF6 colocalized with TIRAP at the plasma membrane in both MEFs and macrophages (Figure 5A and data not shown). Expression of ARF6T27N caused the redistribution of TIRAP into vesicles that colocalize with ARF6 (Figure 5B). In agreement with previous studies (Radhakrishna and Donaldson, 1997), the clathrin adaptor α-adaptin did not colocalize with ARF6T27N (Figure 5C). Additionally, TIRAP localization was not affected by expression of Rab5Q79L, a regulator of clathrin-dependent endocytosis (Bucci et al., 1992) (Figure 5D). Thus, TIRAP transport is regulated specifically by ARF6.#

Peptides composed of residues 2–17 of ARF6 interfere with processes regulated by this GTPase (Galas et al., 1997). We designed a peptide containing these residues fused to the cell-permeating domain of the Drosophila antennapedia protein (Derossi et al., 1998) to determine the role of ARF6 in TLR4 signaling. Treatment of MEFs with ARF6 peptides prevented LPS-induced production of the chemokine KC (Figure S5A). LPS-induced KC production was unaffected by control peptides that also contain an antennapedia sequence. The effect of ARF6 peptides was specific in that it did not affect IL-1 receptor (IL-1R) mediated KC production (Figure S5B). This finding is of particular interest because MyD88-dependent signaling pathways downstream of TLR4 and IL-1R are identical, with the exception of TIRAP. Collectively, these results indicate that ARF6 regulates TIRAP localization and thereby controls TLR4 signaling.#

The role of TIRAP in TLR signaling is enigmatic. It is not clear why TLR2 and TLR4 require TIRAP to induce the MyD88-dependent signaling, while other TLRs and IL-1R trigger the same pathway independently of TIRAP. One possibility is that TIRAP activates signaling pathways distinct from MyD88 (Mansell et al., 2006). Another possibility is that TIRAP functions to recruit MyD88 to TLR2 and TLR4, while other TLRs can recruit MyD88 by a mechanism not dependent on TIRAP. This latter possibility is consistent with the fact that TIRAP-dependent signaling is always MyD88 dependent, while MyD88-dependent signaling can be TIRAP independent—for example, downstream of TLR9 or IL-1R.#

To address the possible role of TIRAP in MyD88 recruitment, we examined the subcellular distribution of MyD88. GFP-tagged MyD88 was found in discrete foci scattered throughout the cytosol (Figure 6A). Identical results were obtained with HA- or FLAG-tagged MyD88 (data not shown), indicating that the localization is not a consequence of the tag but represents a property of MyD88 itself. When coexpressed with TIRAP, MyD88 was relocalized to the cell periphery, where it became enriched at the plasma membrane (Figure 6B). TIRAP therefore facilitates MyD88 delivery to the plasma membrane. In contrast, MyD88 was not required for TIRAP delivery to the cell surface, as MyD88 KO macrophages exhibited normal TIRAP distribution (data not shown).#

To examine the mechanism by which TIRAP recruits MyD88, we examined the requirements for the TIR and the PIP2 binding domains of TIRAP. A TIRAP point mutant with a nonfunctional TIR domain (TIRAP P125H) was unable to alter MyD88 localization (Figure 6C), indicating that TIRAP recruits MyD88 by a TIR-dependent process. To address the role of the PIP2 binding domain in MyD88 recruitment, we employed the SH4-TIR chimera. Whereas both wt TIRAP and SH4-TIR were found at the plasma membrane, SH4-TIR could not relocalize MyD88 to the plasma membrane (Figure 6D), indicating that PIP2 binding is necessary for MyD88 recruitment. The inability of the SH4-TIR to recruit MyD88 provides a plausible explanation for the inability of this chimera to reconstitute LPS responsiveness to TIRAP KO cells (Figure 3G). In contrast, the PLC-TIR chimera efficiently relocalized MyD88 to the cell surface (Figure 6E), indicating that PIP2 binding and TIR domains of TIRAP are sufficient for MyD88 recruitment and signal transduction. These data indicate that TIRAP functions to control MyD88 delivery to PIP2-containing membranes to initiate TLR4 signaling.#

The data presented above indicate that TIRAP can recruit MyD88 to PIP2-containing membranes. If this is the only function of TIRAP, then targeting MyD88 to PIP2-containing membranes by other means should bypass the requirement for TIRAP in TLR4 signaling. However, if TIRAP is required for the recruitment of some additional signaling components, then targeting MyD88 to PIP2 membranes would not bypass the requirement for TIRAP in TLR4 signaling. To address these possibilities, we endowed MyD88 with PIP2 binding specificity by the addition of the PIP2-specific PH domain from PLC to the C terminus of MyD88 (MyD88-PLC). As described above, wt MyD88 is not detectable at the plasma membrane and is found in foci scattered throughout the cell (Figure 6A). In contrast, MyD88-PLC was found at the plasma membrane (Figure 6F).#

We next tested whether MyD88-PLC can substitute for wt MyD88 in TLR signal transduction. Wild-type MyD88 and MyD88-PLC, but not TIRAP, restored TLR4 signaling in MyD88 KO macrophages (Figure 6G). Interestingly, MyD88 KO cells reconstituted with MyD88-PLC had enhanced sensitivity to LPS, suggesting that MyD88 recruitment to PIP2-containing membranes is a rate-limiting step in the induction of TLR4 signaling (Figure 6G).#

We next performed complementation experiments in TIRAP KO macrophages. As expected, TIRAP, but not MyD88, rescued the signaling defect of TIRAP KO cells (Figure 6G). Strikingly, MyD88-PLC rescued the TLR4 signaling defect in these cells (Figure 6G). Moreover, MyD88-PLC rescued the signaling defect in MyD88xTIRAP KO MEFs (Figure 6H). Wild-type MyD88 rescued the signaling defect of MyD88xTIRAP KO cells in response to IL-1β treatment (a TIRAP-independent response), but not in response to LPS treatment (Figure 6H). MyD88xTIRAP KO cells reconstituted with MyD88-PLC regained the ability to respond to LPS but, surprisingly, did not regain the ability to respond to IL-1β (Figure 6H), suggesting that TLR4 and IL-1R initiate MyD88-dependent signaling from distinct membrane subdomains, which may explain their differential requirement for TIRAP.#

These data establish that the requirement for TIRAP in TLR4 signaling is bypassed by endowing MyD88 with a PIP2 binding domain. This indicates that TIRAP's function is to recruit MyD88 to PIP2-containing membranes for the initiation of TLR signaling triggered by TLR2 and TLR4. Collectively, these data provide evidence for a cascade of events that is initiated by TIRAP recruitment to the PIP2-containing plasma-membrane subdomains followed by MyD88 recruitment to TLR4 via TIR-dependent interactions with TIRAP.#

MEFs, CHO cells, and 293T cells were transfected using Fugene-6 (Roche) according to the manufacturer's instructions and incubated for 24 hr at 37°C. Where indicated, cytochalasin D (1 μM; Sigma) or wortmannin (1 μM; Sigma) was added to the cells 30 min prior to fixation. Cells were fixed in 2% paraformaldehyde for 20 min at 25°C and permeabilized with 50 μM digitonin for 10 min. Samples were treated with block buffer (1% bovine serum albumin, 50 mM ammonium chloride in PBS) for 30 min and the appropriate antibodies diluted into block buffer. Where indicated, AlexaFluor 594-phalloidin (Molecular Probes) was included in the block buffer to identify F-actin. Antibody binding was detected using AlexaFluor secondary antibodies from Molecular Probes and visualized with an Axioplan 2 epifluorescence microscope (Carl Zeiss). 0.1 micron sections were captured with an Axiocam HRm digital camera, and images were processed using Adobe Photoshop. Luciferase production from transfected 293T cells was measured by the Luciferase Assay System (Promega) according to the manufacturer's instructions. Bone-marrow-derived macrophages were prepared as described previously (Celada et al., 1984). C57BL6 mice were purchased from Jackson Laboratory. MyD88 KO mice were provided by S. Akira. TIRAP KO mice were described previously (Horng et al., 2002). Bone marrow from CD11b KO mice was provided by E. Harvill and B. Kelsall. Macrophage transfections were performed either by nucleofection (AMAXA) using the mouse macrophage transfection reagent or by retroviral transduction (Kagan and Roy, 2002). Staining of macrophages was performed as described above for MEFs except that the cells were processed 4 hr after transfection.#

PIP strips and PIP arrays (Echelon Biosciences) were immersed in block buffer (10 mM Tris [pH 8.0], 150 mM NaCl, 0.1% Tween 20, 0.1% ovalbumin) for 1 hr. Strips were probed for 2 hr at 25°C with the indicated GST fusion protein (50 ng/ml) in the presence of an anti-GST antibody (Sigma). Blots were then washed in block buffer three times for 10 min each and probed with an HRP-conjugated anti-mouse IgG (Amersham) for 30 min in block buffer. Bound protein was detected using ECL (Amersham). Lipids used to create liposomes were purchased from Avanti. PC/PE liposomes contained a 3:1 molar ratio of PC:PE. Liposomes containing PI species retained the 3:1 ratio of PC:PE but also contained the indicated % of PI. The lipid species were dissolved in a 2:1 chloroform/methanol solution in borosilicate tubes and dried by exposure to nitrogen gas to generate a lipid bilayer. Dried lipids were resuspended in 300 mM sucrose to yield a final lipid concentration of 1 mg/ml. Sedimentation assays were performed by incubating 100 μg of liposomes with 0.5 μg of the appropriate GST protein in cytosol buffer (25 mM HEPES [pH 7.2], 25 mM potassium chloride, 2.5 mM magnesium acetate, 150 mM potassium glutamate). Binding proceeded for 15 min at 37°C, and the samples were centrifuged at 100,000 × g for 10 min. The supernatant was then aspirated, and the pellets were examined after SDS-PAGE by silver staining. For quantitative analysis of protein/liposome interactions, liposomes were prepared as described above except that NBD-PC was used. Fifty micrograms of liposomes was mixed with one hundred micrograms of the appropriate GST protein in cytosol buffer and 30 μl of glutathione Sepharose beads. Binding proceeded for 30 min at 25°C, and then the samples were centrifuged at 2000 × g for 2 min. The sedimented material was washed twice with cytosol buffer and then with 100 μl of 10% SDS to release the bound lipids. Fluorescence recovered in the pellet was assessed by spectrofluorimetry. Fluorescence recovered by beads coated with GST represented ∼150 fluorescence units and was subtracted from each sample to account for nonspecific binding of liposomes to the beads.#

Plasmids containing HA-tagged TIRAP, FLAG-tagged TIRAP P125H, GST-TIRAP, CD4-TLR4, and the NF-κB-dependent reporter pBIIX were described previously (Horng et al., 2001). The HA-TIRAP vector was used as a template to clone TIRAP into pEGFP-N1 (Clontech). All mutants of TIRAP and MyD88 were generated by PCR using HA-TIRAP or HA-MyD88 as a template and cloning the product into pEGFP-N1. GST fusions with mutant TIRAP were generated by cloning the cDNA from pEGFP-N1 into pGEX-4Ti (Amersham). Retroviral vectors encoding various TIRAP cDNAs were generated by subcloning the GFP-tagged cDNA from pEGFP-N1 into pMSCV2.2. ARF6 and Rab5 plasmids were provided by C. Roy (Yale). PLCδ1, GRP1, Fapp1, and the PX-domain-containing plasmids were provided by W. Mothes (Yale). INP54p and Lyn plasmids were provided by R. Isberg (Tufts). SopB plasmids were provided by J. Galan (Yale). GST-fusion proteins were purified from BL21 E. coli using glutathione Sepharose 4B (Amersham) according to the manufacturer's instructions. Purity of each GST preparation was confirmed by SDS-PAGE/silver staining.#