Lysines 128 and 132 Enable Lipopolysaccharide Binding to MD-2, Leading to Toll-like Receptor-4 Aggregation and Signal Transduction*

Three cell-surface proteins have been recognized as components of the mammalian signaling receptor for bacterial lipopolysaccharide (LPS): CD14, Toll-like receptor-4 (TLR4), and MD-2. Biochemical and visual studies shown here demonstrate that the role of CD14 in signal transduction is to enhance LPS binding to MD-2, although its expression is not essential for cellular activation. These studies clarify how MD-2 functions: we found that MD-2 enables TLR4 binding to LPS and allows the formation of stable receptor complexes. MD-2 must be bound to TLR4 on the cell surface before binding can occur. Consequently, TLR4 clusters into receptosomes (many of which are massive) that recruit intracellular toll/IL-1/resistance domain-containing adapter proteins within minutes, thus initiating signal transduction. TLR4 activation correlates with the ability of MD-2 to bind LPS, as MD-2 mutants that still bind TLR4, but are impaired in the ability to bind LPS, conferred a greatly blunted LPS response. These findings help clarify the earliest events of TLR4 triggering by LPS and identify MD-2 as an attractive target for pharmacological intervention in endotoxin-mediated diseases.

Invasive Gram-negative infection is a life-threatening medical disorder that accounts for Ͼ300,000 cases of sepsis in the United States annually (1). One-quarter to one-third of afflicted individuals will die. A milestone in understanding the molecular pathogenesis of this life-threatening condition was the discovery and characterization of germ line-encoded immune receptors known as Toll-like receptors (TLRs) 1 (2,3). TLRs are type I transmembrane glycoproteins present in a variety of cell types (4 -6), and they serve as sensors for the molecular signatures present on invading organisms (7). Following the recognition of their cognate ligands, TLRs initiate a complex signal transduction cascade that results in the production of a variety of immunologically active substances, including cytokines and vasoactive lipids (3,6,8).
There is now abundant evidence that TLR4 functions as the mammalian signal transducer for lipopolysaccharides (LPS) from Enterobacteriaceae (9 -11), the family of Gram-negative bacteria that includes many of the most common human pathogenic bacteria. Activation of TLR4 by LPS is absolutely dependent upon the presence of MD-2 (12,13), a secreted 160amino acid glycoprotein that binds to both TLR4 and LPS (14). MD-2 contributes to the fine ligand recognition specificity displayed by the LPS receptor (15)(16)(17), and it has been suggested to function as a chaperone for TLR4 movement to the surface of cells (13), although in human transfected cells, TLR4 expression on the cell surface is MD-2-independent (this study and Refs. 10, 12, 18, and 19). In addition, sensitive responses to LPS depend on two additional proteins, lipopolysaccharide-binding protein (LBP) and CD14. Both of these molecules play a pivotal role in providing LPS to the signal transduction machinery (20).
Using forward genetic screening, we identified an LPS nonresponder cell line derived from CD14-transfected Chinese hamster ovary cells. These cells expressed a point mutation in MD-2 that resulted in the conversion of a conserved cysteine to tyrosine (12). Human MD-2 bearing the same amino acid substitution was unable to confer LPS responsiveness to TLR4expressing cells. These findings implied an absolute requirement for MD-2 in LPS receptor complex function. This conclusion has been confirmed both in vitro (e.g. Refs. 21 and 22) and in vivo (13), as animals with a targeted disruption of the MD-2 gene exhibit the same LPS non-responder phenotype.
To gain insights into the mechanisms by which the LPS receptor complex activates cells, we analyzed the molecular basis for the impaired function of MD-2C95Y and found that it lacks both the ability to bind TLR4 on the cell surface and to interact with LPS. Based on the lipid-binding motifs of other LPS-interacting proteins (23)(24)(25)(26), we hypothesized that a highly positively charged region in human MD-2 (amino acids 122-132) might be involved in LPS recognition. Using targeted point mutagenesis, we discovered that lysines 128 and 132 are critical for binding to LPS, but not to TLR4. As no previous work had demonstrated how CD14 and LBP relate to MD-2, we sought to characterize this interaction. We found that MD-2 binding to LPS is greatly enhanced by both CD14 and LBP, thus explaining their LPS-sensitizing effect. In addition, we found evidence that TLR4 and LPS physically interact, consistent with a previous study from our group concerning the phar-macology of TLR4 and lipid A (27). The formation of these stable complexes requires the presence of MD-2. The functional outcome of this interaction is the formation of surface aggregates of TLR4 that can be visually observed on the cell surface following LPS binding to TLR4 on transfected HEK293 cells. We found that the resultant cytoplasmic surface formed by this cluster rapidly recruits toll/IL-1/resistance domain-containing adapter molecules via homophilic interactions and propagates the activation signal to downstream effectors, culminating in the activation of NF-B-dependent gene expression. We propose that the induction of large aggregates of TLR4 promoted by MD-2 is the critical event that initiates LPS signal transduction.

Cells and Reagents
HEK293 cell lines stably expressing fluorescent protein-tagged TLR4 have been described previously (19). HEK293 cells stably expressing MD-2 or MD-2C95Y were generated by retroviral transduction (18). Escherichia coli 0111:B4 LPS was purchased from Sigma and repurified to remove contaminating TLR2 ligands as described (28). Monoclonal antibody (mAb) HTA125 (IgG2a), which recognizes TLR4, was a gift from Dr. Kensuke Miyake (29). The mAb and polyclonal antiserum against green fluorescent protein (GFP) were from Clontech and Molecular Probes, Inc. (Eugene, OR), respectively. The horseradish peroxidase-conjugated anti-biotin polyclonal antibody was from New England Biolabs Inc. (Beverly, MA). Unless otherwise stated, all other reagents were purchased from Sigma.

Cell-surface Biotinylation, Immunoprecipitation, and Western Blotting
Adherent cells from a confluent 10-cm dish were labeled with biotin (Pierce) on ice, solubilized in detergent (1% Triton X-100, 10 mM Tris-Cl (pH 7.4), 137 mM NaCl, 10% glycerol, 2 mM EDTA, and protease inhibitors), subjected to immunoprecipitation with the indicated antibodies (2 g/ml) in 20 l of packed protein A-Sepharose (Amersham Biosciences, Uppsala, Sweden) for 16 h at 4°C, resolved by SDS-PAGE, and electrotransferred onto Hybond-C nitrocellulose membranes (Amersham Biosciences). The membranes were blocked in 5% dry milk in phosphate-buffered saline (PBS) and 0.1% Tween 20 for 30 min at 37°C and probed for an additional 30 min at 37°C with horseradish peroxidase-conjugated anti-biotin polyclonal antibody (1 g/ml). Biotinylated proteins were revealed by enhanced chemiluminescence using a commercial kit for this purpose (Amersham Biosciences). Identical conditions for immunoprecipitation and immunoblotting were used for all of the blots presented in this work. FLAG-tagged MD-2 was revealed by probing the membranes with horseradish peroxidase-conjugated mAb M2, whereas GFP-tagged TLR4 was detected by blotting with anti-GFP mAb, followed by a horseradish peroxidase-conjugated anti-mouse secondary reagent (Bio-Rad). When necessary, membranes were stripped for 30 min in 0.1 M glycine (pH 2.2), 1% Tween 20, and 0.1% SDS and reprobed.

LPS Labeling and Precipitation Assay ("LPS Pull-down" Assay)
The polysaccharide moiety of repurified LPS was biotin-labeled using biotin hydrazide (Pierce) according to the manufacturer's instructions. 1 mg of LPS was treated with 10 mM of sodium metaperiodate on ice for 30 min in coupling buffer (0.1 M sodium acetate at pH 5.5). After quenching in 50 mM glycerol for 5 min, excess oxidant and glycerol were removed by gel filtration using a Sephadex G-25 PD-10 desalting column (Amersham Biosciences), and LPS was eluted in coupling buffer. Biotinylation was performed by adding biotin hydrazide to a 5 mM final concentration for 2 h at room temperature. Unreacted biotin hydrazide was removed by a second purification step with a PD-10 column, and biotinylated LPS was finally eluted in Hanks' balanced saline solution. Biotinyated LPS was subjected to SDS-PAGE and detected by Western blotting using horseradish peroxidase-conjugated anti-biotin polyclonal antibody (see above), which revealed a smeared band ranging from 100 to 20 kDa, as would have been predicted because of the heterogeneous nature of the polysaccharide portion of LPS and the likelihood that the number of biotins per LPS would not be constant (30). Furthermore, the bioactivity of the LPS was retained as assessed by comparing the ability of non-biotinylated LPS with that of biotinylated LPS in an NF-B activation assay of TLR4-expressing HEK293 cells. The binding of LPS to FLAG-tagged soluble MD-2 was examined by incubating 10 ml of supernatant from HEK293 cells that had been transfected with MD-2 with 0.5-1 g of biotinylated LPS/ml either overnight at 4°C or for 1 h at room temperature in the presence of 25 l of packed streptavidin-Sepharose CL-4B. Beads were collected by centrifugation, washed three times with Hanks' balanced saline solution, and resuspended in reducing SDS sample buffer, and anti-FLAG Western analysis was performed as described above. Washing with lysis buffer did not dissociate MD-2⅐LPS complexes.
To study LPS interactions with TLR4, adherent cells from 10-cm confluent dishes were washed extensively with PBS and incubated for 30 min with 5 ml of prewarmed complete medium containing 1 g/ml biotinylated LPS. Cells were then washed with Hanks' balanced saline solution and lysed as described above for immunoprecipitations. Biotinylated LPS⅐protein complexes were collected by addition of 20 l of packed streptavidin beads and processed exactly as described for the immunoprecipitation protocol. The presence of TLR4 in the LPS complexes was assessed by Western blotting with anti-GFP antibody (which also recognizes the spectral variants of GFP such as yellow fluorescent protein (YFP) and cyan fluorescent protein (CFP)).

Microscopy
Confocal Microscopy-To visualize the surface distribution of TLR4 and MD-2, HEK293 cells stably expressing YFP-TLR4 were seeded on 35-mm glass-bottom tissue culture dishes (Mattek Corp., Ashland, MA) and transfected with MD-2 or MD-2C95Y using GeneJuice (Novagen, La Jolla, CA). The following day, cells were washed twice with PBS supplemented with 1% fetal bovine serum and incubated for 30 min on ice with mouse anti-FLAG mAb M2 (10 g/ml). After two washes with PBS/fetal bovine serum, the cells were incubated with Alexa 647-conjugated anti-mouse IgG (2 g/ml) for 15 min on ice. Confocal images were collected from living cells incubated on a warm-stage apparatus at 37°C for 15 min with a Zeiss Axiovert 100-M inverted microscope using an LSM 510 laser scanning unit. YFP was excited using the 514-nm line of a 25-milliwatt argon laser, and Alexa 647 was excited with a helium/ neon laser emitting at 633 nm. Band-pass or long-pass filters were chosen to optimally separate the fluorescence emissions between the different photomultipliers using single-labeled samples of the probes as controls. LPS-induced clustering of TLR4 and MyD88 recruitment was monitored as follows. HEK293 cells stably expressing YFP-TLR4 and MD-2 or YFP-TLR4 and MD-2C95Y were grown on glass-bottom tissue culture dishes and transiently transfected with CFP-tagged MyD88. The cells were then incubated with 250 ng of Cy5-labeled repurified E. coli 0111:B4 LPS/ml of culture medium. After 10 min, the cells were washed twice with prewarmed culture medium and imaged with a Zeiss confocal microscope using a warm-stage apparatus set at 37°C. Sequential scans were taken for CFP, YFP, and Cy5 using excitation at 458, 514, and 633 nm, respectively. z-stacks were acquired by scanning the cells from top to bottom. Side views of the cells were generated by electronically overlaying the images in the xy axis and visualizing the z axis.
Scanning Electron Microscopy-Cells were grown on poly-L-lysinetreated coverslips and treated with 1 g/ml LPS for 5 min at 37°C. Samples were chilled on ice, washed with ice-cold Hanks' balanced saline solution, fixed for 20 min in 4% glutaraldehyde, and stained for 30 min with mAb HTA125 (1 g/sample) in 200 l. Antigen⅐antibody complexes were revealed by a 30-min incubation with a Nanogold TMlabeled anti-mouse F(abЈ) 2 polyclonal antibody, followed by silver enhancement (LIS silver enhancement kit, L-24919, Molecular Probes, Inc.) according to the manufacturer's protocol. The silver-enhanced coverslips were then washed twice for 5 min with ultrapure water, dehydrated through a graded series of ethanol soaks to 100%, and then critical point-dried in liquid CO 2 (31). The coverslips with the dried cells were mounted and gold-coated for scanning electron microscopy. Samples were examined on an Etec Autoscan electron microscope at 20 kV. Atomic contrast imaging was performed on the samples' backscatter to confirm the silver nature of the white particles.

Cell Stimulation Assays
TLR4-induced cell signal transduction was monitored by measuring either NF-B activity or interleukin-8 (IL-8) secretion. To measure NF-B activation, 2 g of a reporter plasmid in which NF-B drives the synthesis of luciferase was transiently cotransfected with 5 g of the indicated cDNA by lipofection (GeneJuice) in 10-cm tissue culture dishes following the manufacturer's recommendations. The following morning, cells were seeded at a density of 50,000 cells/well in a 96-well plate, allowed to recover for 5 h, and stimulated as indicated between 6 and 18 h. Luciferase activity was measured with a plate reader lumi-nometer (Victor 2 , PerkinElmer Life Sciences) using chemicals provided with the luciferase assay system (Promega, Madison, WI). All data are presented as the means Ϯ S.D. of triplicate well readings, normalized to a value of 1.0 in comparison with an unstimulated negative control. Antibody cross-linking experiments were performed by coating the 96well high protein-binding tissue culture plates (catalog no. 3361, Costar, Corning, NY) overnight with 50 l of mAbs (or LPS) in PBS at the indicated concentrations. Excess antibody was removed by washing the plates with pyrogen-free PBS, and cells transfected with the NF-B-luciferase reporter plasmid were seeded. At 6 -18 h post-seeding, NF-B activation was assessed as described above, and IL-8 levels in the supernatants were measured using a commercial IL-8 enzymelinked immunosorbent assay kit (R&D Systems, Minneapolis, MN) following the manufacturer's recommendation. As a negative control, antibodies were boiled for 10 min before coating the plates.

TLR4 and MD-2 Form Noncovalently Bound Complexes on the Cell Surface of HEK293 Cells-Substitution of cysteine 95
with a tyrosine residue generates a mutant form of human MD-2 (MD-2C95Y) that binds to TLR4 in the cytoplasm and partially immunoprecipitates with TLR4 (12). MD-2C95Y is unable to confer LPS responsiveness to TLR4 (12). However, substituting cysteine 95 with a serine residue, which is sterically closer to a cysteine, generates a mutant form of MD-2 (MD-2C95S) that has been shown to be slightly active when cotransfected with TLR4 (32). We hypothesized that the failure of MD-2C95Y to enable LPS signaling was due to its failure to properly bind to surface TLR4. Cells stably expressing YFPtagged TLR4 were transfected with FLAG-tagged MD-2, MD-2C95Y, or MD-1. Cell-surface proteins were biotin-labeled, and whole cell lysates were subjected to immunoprecipitation with anti-GFP polyclonal antiserum, which recognizes epitopetagged TLR4. Anti-GFP precipitates were resolved by electrophoresis, blotted, and probed for surface biotin (Fig. 1). The same lysates were also precipitated with anti-FLAG mAb, subjected to SDS-PAGE, and transferred to nitrocellulose. Because of size differences in the proteins of interest, this blot was then bisected, and each portion was developed individually by Western blotting using anti-FLAG antibody for the lower portion or anti-GFP antibody for the upper portion of the blot. We found that TLR4 and MD-2 co-immunoprecipitated from the cell surface (Fig. 1, lane 1). In contrast, neither the MD-2C95Y mutant protein nor MD-1 associated with TLR4 on the cell surface ( Fig.  1, lanes 2 and 3), despite comparable levels of these proteins in the lysates (lanes 4 -9).
MD-2C95Y Does Not Co-localize with TLR4 on the Cell Surface-Western blot analysis of transfected HEK293 cells and their supernatants established that MD-2C95Y is synthesized, glycosylated, and secreted (data not shown). HEK293 cells stably expressing YFP-tagged TLR4 were transiently transfected with MD-2 ( Fig. 2, upper panels) or the MD-2C95Y mutant (lower panels) and stained with anti-FLAG mAb. MD-2 was then cross-linked and visualized by incubating these cells at 37°C with Alexa 647-conjugated anti-mouse IgG polyclonal antibody. Confocal microscopic scans of YFP fluorescence (TLR4) and Alexa 647 fluorescence (MD-2) were obtained. In the absence of antibody treatment, TLR4 is uniformly distributed on the cell surface, as well as in the Golgi apparatus (19). Additionally, quantitative analysis of TLR4 by flow cytometry demonstrated that, in transfected HEK293 cells, the level of TLR4 surface expression was independent of MD-2 or MD-2C95Y coexpression (data not shown). Treatment of wild-type MD-2 with anti-FLAG antibody resulted in the clustering of MD-2 on the cell surface. Because MD-2 is bound to TLR4 on the cell surface, we subsequently observed the reorganization of surface TLR4 in the exact same distribution as MD-2 (yellow signal). Surprisingly, although our immunoprecipitation analysis suggested that MD-2C95Y would not bind TLR4 (and it is not known to bind any other TLR), it was consistently found on the surface of HEK293 cells that had been transfected with the  1-3). The same blot was then stripped, bisected into two nearly equal-sized pieces, and reprobed with either anti-GFP mAb (lanes 4 -6, upper panel) or anti-FLAG mAb (lanes 4 -6, lower panel) to visualize the total amounts of precipitated proteins. A reciprocal immunoprecipitation with anti-FLAG mAb was performed with lysates derived from the same transfectants (lanes 7-9) and processed as described for lanes 4 -6. Note that TLR4 co-immunoprecipitated with non-biotinylated (i.e. intracellular) MD-2C95Y and vice versa. Biotinylated molecular weight markers are shown on the left. TLR4 and MD-2C95Y cDNAs. Antibody cross-linking aggregated the mutant form of MD-2, but no commensurate changes in the distribution of TLR4 were observed (Fig. 2, lower panels).
These results indicate that the MD-2C95Y mutant is not capable of binding to TLR4.
Antibody-mediated Aggregation of MD-2 Efficiently Triggers TLR4 -The co-localization of TLR4 with MD-2 correlates with LPS responsiveness and thus suggests that binding of MD-2 to TLR4 is required for the assembly of an active LPS receptor. It is reasonable to postulate that the MD-2C95Y mutant failed to function because of its inability to mediate the assembly of a signaling complex. If the assembly of a heteromeric complex were truly the initiating event in signal transduction, then cross-linking TLR4 and/or MD-2 with antibodies instead of a ligand should mimic receptor activation. To directly test this hypothesis, we coated the surface of plastic dishes with mAbs and plated transfected cells on the antibody-coated surface. mAb HTA125 (29) recognizes TLR4, and anti-FLAG mAb M2 binds the epitope tag on MD-2 and MD-2C95Y. As shown in Fig. 3, anti-FLAG mAb stimulated IL-8 secretion when wildtype MD-2 was present, but not when the MD-2C95Y mutant was expressed. This indicates that TLR4 is present on the cell surface despite the presence of a mutant form of MD-2 (because the plated antibody can bind only to surface TLR4). Identical antibody-induced activation was obtained in cells that did not express any exogenous MD-2 (data not shown). Isotypematched antibodies and boiled antibodies did not activate cells (see Fig. 5) (data not shown), indicating that activation was not caused by contaminating LPS. We conclude that aggregation is an effective triggering signal for TLR4-mediated cellular activation. The MD-2C95Y mutant is incapable of activating TLR4 when cross-linked with a mAb because it fails to co-aggregate TLR4.
Recombinant Soluble MD-2 Binds to LPS in a CD14-and LBP-dependent Fashion-Although not absolutely required for LPS signaling, the soluble blood proteins CD14 and LBP are known to enhance LPS responses up to 10,000-fold. We therefore sought to determine the effect of CD14 and LBP on LPS binding to MD-2. We developed an LPS binding assay based on the ability of streptavidin to pull-down proteins that are bound to biotin-labeled LPS. The biotinylated probe, consisting of LPS labeled at its polysaccharide moiety, retains its pro-inflammatory activity and can be conveniently captured using streptavidin beads. LPS-interacting proteins can be detected in the precipitate using the appropriate antibody in Western blot analysis.
As shown in Fig. 4A, the binding of LPS to soluble MD-2 was readily demonstrated by LPS pull-down from MD-2-containing supernatants. The specificity of this interaction was demonstrated by the ability of unconjugated LPS to compete completely for MD-2 binding when present at a 20-fold excess. In contrast to MD-2, the homologous protein MD-1 had no ability to bind LPS (Fig. 4C). As might have been expected from the influence of LBP and soluble CD14 on cellular responses, LPS binding to MD-2 was markedly enhanced by whole serum, recombinant soluble CD14 (sCD14), or a combination of sCD14 plus LBP (Fig. 4B). This result correlates with the numerous reports that sCD14 and LBP can shift the LPS dose-response curve in TLR4-responding cells.
As mentioned earlier, MD-2C95Y is incapable of enabling LPS effects in TLR4-bearing cells (12). When we tested MD-2C95Y for binding to LPS using the biotinylated LPS precipitation assay (Fig. 4C), the mutant protein failed to bind to LPS, suggesting that cysteine 95, in addition to being necessary for the binding of MD-2 to extracellular TLR4, confers structural details necessary for LPS binding.
MD-2 Activity Depends on a Highly Positively Charged Cterminal Region-To further define the functional relation- ships between LPS, MD-2, and TLR4, we searched for a mutant form of MD-2 that would interact normally with TLR4, but would be impaired in its ability to bind LPS. Lysines 122 and 132 of MD-2 define a highly positively charged region. Based on the lipid-binding motifs of other LPS-interacting proteins (23-26), we and others (24) hypothesized that this region might be involved in LPS recognition (Fig. 5A). The positive charges at positions 128 and 132 are conserved among different MD-2 orthologs (Fig. 5A, boldface), with the exception of bovine MD-2, which encodes for a proline. Accordingly, we used sitedirected mutagenesis to change these residues to glutamic acid and studied the effects of these mutations on LPS response and binding to TLR4.
Both of the MD-2 mutants were synthesized, secreted, and had the ability to normally interact with TLR4 (Fig. 5, B and  E). When cross-linked by plate-immobilized mAb, the mutant forms of MD-2 and wild-type MD-2 were equally capable of activating NF-B via TLR4 (Fig. 5C), therefore indicating that the mutant MD-2 molecules formed stable TLR4⅐MD-2 complexes on the cell surface. In contrast, the mutants conferred a diminished response to LPS when coexpressed with TLR4 in HEK293 cells (Fig. 5D). The latter observation correlates with the diminished ability displayed by these mutant proteins to bind LPS in the LPS pull-down assay: MD-2 Ͼ MD-2K128E Ͼ MD-2K132E (Fig. 5E). We conclude that MD-2K128E and MD-2K132E possess an intact TLR4-binding domain and suggest that the LPS-binding region of MD-2 depends on lysines 128 and 132.
MD-2 Enables LPS Binding to TLR4 -Previous studies have reported an interaction between LPS and TLR4 using 125 Ilabeled photoactivable derivatives of LPS (33,34). After photoactivation, FLAG-tagged TLR4 and MD-2 were immunoprecipitated and identified based upon their molecular weights and the fact that they bound to LPS, but skepticism about these results (which this study confirms and expands upon) seems to have persisted. The major reason for the skepticism of this previous study was that the probe was of low specific activity. Perhaps as a result of this low specific activity, the assay depends upon the expression of membrane CD14, which invariably results in the movement of endotoxin in large amounts into the membrane lipid bilayer (35). Hence, it was unclear if the interactions of LPS with the TLR4⅐MD-2⅐CD14 complex were indicative of a high affinity binding event or if the results more accurately reflected the formation of the signalosome FIG. 5. MD-2 has a TLR4-binding domain and a second positively charged domain that is necessary for full activation. A, shown is an alignment of a highly basic region of MD-2 orthologs. The asterisks indicate the lysine residues in human MD-2; the lysine residues that were used to prepare mutant MD-2 molecules (MD-2K128E and MD-2K132E) are shown in boldface. B, HEK293 cells stably expressing YFP-TLR4 were transiently transfected with wild-type MD-2 (MD-2 wt ) or mutant MD-2 constructs and an NF-B/luciferase reporter plasmid. TLR4 and MD-2 were immunoprecipitated (IP) from cell lysates or supernatants using anti-GFP or anti-FLAG antibody as indicated. Immunoprecipitates were resolved by 4 -15% reducing SDS-PAGE, transferred to a Hybond membrane, bisected with a razor, and probed individually for TLR4 or MD-2 (upper and middle panels), demonstrating that the wild-type and mutant MD-2 molecules bound TLR4 equivalently. The anti-FLAG immunoblot (lower panel) shows equal levels of MD-2 expression in the same lysates. C, a portion of the cells in B were subjected to cross-linking with immobilized anti-FLAG antibody plated at 10 (gray bars) and 30 (black bars) g/ml, and NF-B activation was assayed by luminometry. As a control, boiled anti-FLAG mAb was plated at 30 g/ml. wt, wild-type MD-2, 128, MD-2K128E; 132, MD-2K132E; RLU, relative light units. D, a second portion of the cells in B were treated with the indicated amounts of LPS overnight, and NF-B activation was measured as described for C. E, HEK293T cells were transiently transfected with the indicated plasmids in 10-cm dishes and 10 ml of complete medium. 24 h later, supernatants were collected and tested for MD-2 secretion (1 ml of conditioned supernatant) (upper panel) and LPS pull-down (9 ml) (lower panel). All experiments shown here were performed three times with similar results. Readings are expressed as the means Ϯ S.D. of triplicate sample wells. present on the surface of uniformly radiolabeled cells. The biotinylated LPS probe differs from this radiolabeled probe because every molecule of LPS appears to be labeled with biotin (data not shown). The assay provides clean and reproducible data without the need for membrane CD14 expression. Hence, we believe that, given the availability of reagents for Western blotting, it is possible to use this assay to ask very specific questions about LPS binding to known proteins.
We incubated adherent cells expressing TLR4, TLR4 plus MD-2, or TLR9 with biotinylated LPS by providing MD-2 as a transgene (Fig. 6, fourth and fifth lanes) or as a soluble molecule (third lane). Under these conditions, only cells that express MD-2 or cells to which soluble MD-2 had been added will respond to LPS. As shown in Fig. 6, biotinylated LPS precipitations of treated cells revealed an unstable interaction between TLR4 and MD-2 when LPS was present. Note that the TLR4 that was bound to LPS was only the mature heavily glycosylated form (18), suggesting that the binding occurred only on the cell surface.
LPS Induces Aggregation of TLR4 -Previous studies have suggested that the initiation of signal transduction in response to LPS begins on the surface of cells (19). This process is dependent upon tight interaction of LPS-bound MD-2 with surface TLR4 because LPS signaling can be mimicked by crosslinking either TLR4 (19) or MD-2 (Figs. 3 and 5) with antibodies. These observations suggest the possibility that LPS itself induces TLR4 aggregation on the cell surface as a means of initiating signal transduction.
We performed ultrastructural studies of TLR4 surface expression on LPS-treated cells using scanning immunoelectron microscopy to determine whether the distribution of TLR4 before and after LPS treatment would change. Following exposure of HEK293 cells expressing TLR4 and MD-2 to LPS or medium for 5 min, we analyzed the surface for TLR4 expression using anti-TLR4 mAb HTA125 as a primary antibody and gold-conjugated anti-IgG antibody as a secondary antibody. Each gold particle is ϳ1.4 nm in diameter, which is below the resolution of scanning electron microscopy. Thus, the cellbound gold particles were enhanced by incubation with silver prior to scanning electron microscopy. Each electron-dense particle on the cell surface represents one or more TLR4 molecules. As a control, we also tested cells that were not labeled with anti-TLR4 mAb, but were counterstained with the secondary reagent and developed in silver. As shown in Fig. 7, the exposure of HEK293 cells expressing TLR4 and MD-2 to LPS resulted in the dramatic rearrangement of TLR4 upon the cell surface. Although the distribution of TLR4 in untreated cells appears to be distributed uniformly as ϳ100 -200-nm silver particles, LPS treatment resulted in the rapid formation of large electron-dense particles close to 2 m in diameter. These large particles, which varied in size and number, are clearly composed of numerous silver-coated gold beads, probably reflecting the formation of large aggregates of signaling molecules.
The presence of the massively large multimerized receptor complexes was observed in roughly one-quarter of all of the cells that had been treated with endotoxin, but was never seen in cells that had not been exposed to LPS. The observation of these large TLR4 complexes was made by examining cells after 5 min. Video electron microscopy is not possible because of the fixation steps involved in processing specimens. However, we hypothesize that, as we proceed to characterize the time course of receptor aggregation, that all of the cells examined will prove to have formed signaling aggregates on their surface; other studies in our laboratory suggest that all of the LPS-treated cells respond and, for example, will make cytokines such as intracellular IL-8 within a few hours after treatment (data not shown). Although a similar degree of resolution is not possible using vital microscopic techniques, as shown below, careful confocal studies suggest that the multimerization of TLR4 is rapid and transient (data not shown).
LPS-clustered TLR4 Forms "Active" Receptosomes in HEK293 Cells-To determine whether the formation of TLR4 aggregates on the cell surface reflects the activation of the receptor, we followed the formation of LPS-induced TLR4 aggregates in living cells using recruitment of MyD88 as an indicator of receptor activation. HEK293 cells expressing YFPtagged TLR4, CFP-tagged MyD88, and FLAG-tagged MD-2 were treated with Cy5-conjugated LPS, and sequential confocal microscopic scans were taken. As soon as 4 min after LPS exposure, TLR4⅐LPS complexes were detected as clusters, which were persistent and mobile on the cell surface. 2 MyD88 recruitment to these clusters of TLR4 and LPS has been observed in living cells and is illustrated in Fig. 8. The presence of detectable aggregates containing TLR4, LPS, and MyD88 2 T. Espevik, unpublished data.
FIG. 6. MD-2 is required for TLR4 binding to LPS. HEK293 cells expressing TLR4, TLR4 ϩ MD-2 or TLR9, were left untreated or treated with 1 g/ml biotinylated LPS (Biotin-LPS) for 30 min at 37°C and solubilized in lysis buffer. Cells expressing TLR4 and MD-2 were extensively rinsed with prewarmed PBS to eliminate soluble MD-2. Cell lysates were then incubated overnight with avidin beads, and the washed pellet was resolved by 4 -15% gradient SDS-PAGE. The presence of TLR4 in the precipitates was assessed as described in the legend to Fig. 1. In the third lane, LPS stimulation was performed in the presence of soluble MD-2 (sMD-2). FIG. 7. LPS rapidly induces cell-surface clustering of TLR4. HEK293 cells expressing TLR4 and MD-2 were left untreated or treated with 1 g/ml LPS, and cell-surface TLR4 was visualized by scanning electron microscopy using anti-TLR4 mAb HTA125. Anti-TLR4 mAb was visualized by counterstaining with gold-conjugated anti-IgG antibody and enhanced with silver as described under "Experimental Procedures." Control cells were stained with the secondary reagent only. Prior to treatment, surface TLR4 appears as ϳ100 -200-nm electrondense white silver particles, whereas after 5 min, clusters of TLR4 appear as substantially bigger particles. The three fields shown in the upper panels were taken at a magnification of ϫ2500, whereas the details in the lower panels were taken at a magnification of ϫ10,000. White scale bar ϭ 1 m; black scale bar ϭ 10 m.
within minutes after the addition of LPS strongly supports the idea that recruitment of toll/IL-1/resistance domain-containing adapters immediately follows TLR4 aggregation. It also parallels the biochemical data presented in Fig. 6. LPS-induced surface redistribution of TLR4 and MyD88 was observed only when LPS was used to treat cells expressing wild-type MD-2, but not mutant forms of MD-2 such as MD-2C95Y (data not shown). DISCUSSION LPS is among the most potent naturally occurring pro-inflammatory molecules known. Its extensive use as a tool to study host defenses is due to its ability to induce a wide spectrum of immunological events (36). The sum of these effects is to prevent the invasion of microorganisms from the skin and mucosal surfaces and, should bacterial invasion occur, to eliminate the invading pathogen and promote healing. Unfortunately, triggering the LPS activation program in phagocytic leukocytes has potential injurious effects as well. These include the paradoxical initiation of the sepsis syndrome and the profound immune dysregulation that accompanies high-grade bacterial invasion.
The characterization of mutant mice that fail to respond to LPS has established a close link between LPS responsiveness and susceptibility to infection. Positional cloning studies, followed shortly thereafter by the description of a TLR4 knockout mouse, provided the first evidence that TLR4 is the LPS signal transducer (9,37,38). In contrast to these labor-intensive studies of LPS non-responder mice, the identification of MD-2 as an LPS receptor component (29) involved a large component of both serendipity and insight.
Certain lipid A-like molecules such as the tetraacyl lipid A precursor known as lipid IVa, inhibit the effects of LPS on human cells (27). The inhibitory effect of these lipid A-like molecules has always been thought to be due to competitive inhibition of a binding site on the LPS receptor. Surprisingly, these compounds have different effects in different species of mammals (27,39). For example, lipid IVa is not an LPS antagonist in mice (or in cells derived from mice), but rather mimics the effects of LPS. The use of molecular genetics has allowed investigators to express human or mouse TLR4 as a transgene in a variety of different cell lines derived from rodents and humans, some of which also express their endogenous forms of MD-2. Other experiments have been done by cotransfecting TLR4 with MD-2 into cells such as HEK293 that ordinarily do not express either protein (15,16). Unexpectedly, both TLR4 (11,40) and MD-2 (15,16) have been reported to genetically encode the element of the LPS receptor that determines the species-specific response. Each of the initial reports was confirmed by another independent group, suggesting that whatever the explanation for the seemingly discordant findings, it was not due to "artifact" or careless experimental work. A unifying explanation is that TLR4 and MD-2 together form a complete lipid A recognition site. Hence, either TLR4 or MD-2 can determine species-specific recognition of lipid A; and depending upon the relative expression levels of each protein, either can influence the specific biological response. However, a major difference between MD-2 and TLR4 with respect to LPS binding is that MD-2 is capable of binding LPS in the absence of TLR4 (this study and Ref. 14), whereas the converse does not appear to be true. This observation explains why MD-2 expression appears to be absolutely required for LPS activation of TLR4.
The identification of the lipid A-binding site on TLR4 and MD-2 is important for any genuine understanding of LPS receptor function. Although binding studies with agonistic and antagonistic compounds have yet to be completed, the evidence is consistent with the idea that lysines 128 and 132 are sites of direct binding to LPS or are essential for stabilizing LPS binding. We infer that this region is specifically dedicated to the interaction with LPS (rather than with TLR4) because mutations of these residues did not alter any of the other important properties of MD-2, including its secretion and tight binding to TLR4. The identification of this region of MD-2 as a participant in LPS binding expands upon a recent study reported Kawasaki et al. (22), who performed alanine scanning mutagenesis of MD-2. Although no LPS binding data was reported, Kawasaki et al. showed that mutations in the region composed of isoleucine 124 and cysteine 133 generate forms of MD-2 that normally associate with TLR4 on the cell surface, but display an impaired LPS responsiveness (approximately one-third of the wild-type response). This is consistent with the idea that this region is exposed and free to interact with the ligand, but is not involved in the interaction with the coreceptor, TLR4.
It is commonly said that TLR4 homodimers are responsible for initiating the LPS response on cells. Based upon the findings in this study, it seems unlikely to be true that TLR4 is activated by forming homodimers. Furthermore, these data imply that other TLRs probably do not signal as dimers either. The dimerization of TLR4 is unlikely to be sufficient to result in activation. Studies done in our laboratory, but not reported here, suggest that TLR4 is probably present on the surface of cells as a dimer (or larger) in the absence of cellular activation. For example, when we cotransfected CD14-expressing Chinese hamster ovary fibroblasts with FLAG-tagged and Myc-tagged TLR4, we found that the differentially tagged receptors coimmunoprecipitated in the absence of LPS. We suggest that a threshold of aggregated TLR4 molecules must be necessary for a functional "signalosome" to be assembled and activated.
Students of biology are well aware of the tendency in nature for systems to be activated via complicated cascades. Complement activation, blood clotting, and electron transport are but a few well known examples of the phenomena. In a more modest way, the activation of immune cells by lipid A may prove to be another example of a biological cascade. LPS-induced activation seems to require multiple transfers of lipid A from the outer leaflet of the Gram-negative bacterial cell wall (perhaps aided by the actions of complement and LBP) and then from LBP 3 CD14 3 soluble MD-2 or membrane TLR4⅐MD-2. Of course, one cannot absolutely exclude the existence of additional receptor components, and such components may represent additional lipid A transfer molecules. Biological cascades probably exist because they represent a highly efficient means of fine-tuning and regulating biological responses. Relatively small perturbations in homeostasis can be amplified, whereas relatively large ones can be blunted. LPS activation, which is intimately tied into bacterial invasion, should be considered to be no exception to these rules, even if the details of LPS activation and the advantage to the host of being able to finely tune responses to bacterial invasion remain only partially defined.
The practical implication of these studies is that developing a pharmacological strategy for interfering with the deleterious effects of LPS might be less difficult to achieve than previously assumed. Compounds that even partially interfere with the function/assembly of the TLR4 receptosome, such as preventing TLR4 and MD-2 interaction, TLR4 and soluble MD-2 interaction, MD-2 binding to LPS, or LPS transfer from CD14 to MD-2, are likely to prove to be potent inhibitors of LPS-induced inflammation and useful as therapeutics for LPS-related disease.