A Complex of Soluble MD-2 and Lipopolysaccharide Serves as an Activating Ligand for Toll-like Receptor 4*

MD-2, a glycoprotein that is essential for the innate response to lipopolysaccharide (LPS), binds to both LPS and the extracellular domain of Toll-like receptor 4 (TLR4). Following synthesis, MD-2 is either secreted directly into the medium as a soluble, active protein, or binds directly to TLR4 in the endoplasmic reticulum before migrating to the cell surface. Here we investigate the function of the secreted form of MD-2. We show that secreted MD-2 irreversibly loses activity over a 24-h period at physiological temperature. LPS, but not lipid A, prevents this loss in activity by forming a stable complex with MD-2, in a CD14-dependent process. Once formed, the stable MD-2 (cid:1) LPS complex activates TLR4 in the absence of CD14 or free LPS indicating that the activating ligand of TLR4 is the MD-2 (cid:1) LPS complex. Finally we show that the MD-2 (cid:1) LPS complex, but not LPS alone, induces epithelial cells, which express TLR4 but not MD-2, to secrete interleukin-6 and interleukin-8.

Lipopolysaccharide (LPS), 1 a component of the outer membrane of Gram-negative bacteria, stimulates an exceptionally potent innate immune response in mammals that can result in septic shock and death (1). The LPS response is mediated by four proteins (2): LPS binding protein extracts single molecules from LPS micelles and transfers them to CD14, a glycosylphosphatidylinositol-anchored cell surface receptor that also exists as a serum protein. In turn, the CD14⅐LPS complex activates two proteins that comprise the essential signaling complex of the LPS response, Toll-like receptor 4 (TLR4) and MD-2 (3)(4)(5)(6). TLR4 is a type I integral membrane glycoprotein and is one of 10 TLR paralogs that activate NF-B, mitogen-activated protein kinases, and other transducers of inflammatory signals in response to pathogen-specific structural motifs. MD-2, a small cysteine-rich glycoprotein, binds to the ectodomain of TLR4 (7) in the endoplasmic reticulum and then transits to the cell surface in an active TLR4⅐MD-2 complex. However, MD-2 is also secreted into the medium as a soluble, active protein (sMD-2) by primary cells such as immature dendritic cells (iDC), and by MD-2-transfected cell lines (8). The activity of sMD-2 was shown by its ability to bind to TLR4 and confer LPS responsiveness to cells that express TLR4 but lack MD-2 (8,9). Forward genetic and gene deletion studies have demonstrated that both MD-2 and TLR4 are required for normal responsiveness to LPS in vitro and in vivo (3)(4)(5)10).
Analyses of species specificity differences for various forms of LPS provide strong evidence that LPS interacts directly with the TLR4⅐MD-2 complex (11)(12)(13)(14)(15)(16). However, the molecular events leading to LPS binding and TLR4 activation are only partially understood. Photoaffinity labeling (17) and binding (9,18,19) studies have shown that LPS binds directly to MD-2 and TLR4, and that binding of LPS to TLR4 requires MD-2 (17)(18)(19). In addition, it has been shown that CD14 is required for LPS binding to MD-2 (18) or the TLR4⅐MD-2 complex (17,19). Taken together, these data have suggested a current model in which CD14 delivers LPS to a complex between TLR4 and MD-2, and that this interaction promotes signal transduction (18 -20).
The purpose of the current investigation is to define the role of soluble MD-2 in the LPS response. We show that sMD-2 is a labile protein that loses most of its activity over a 24-h period at physiological temperature. Treatment with LPS stabilizes sMD-2 in a process that requires CD14, and when purified, the stable MD-2⅐LPS complex, in the absence of CD14 or free LPS, directly activates TLR4. Finally, we show that the MD-2⅐LPS complex has the capacity to trigger the release of cytokines by epithelial cells, which express TLR4 but not MD-2, and therefore fail to respond to LPS alone. We conclude that the activating ligand of TLR4 is a complex between LPS and MD-2, and that the soluble form of this complex can play a role in regulating the innate response to LPS. the NF-B reporter plasmid, ELAM-1-luciferase, were a gift from Dr. Jesse Chow (Eisai Research Institute, Andover, MA). HEK293 cells were maintained in Dulbecco's modified Eagle's medium containing 10% low endotoxin fetal bovine serum (FBS), 2 mM glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin. The protein-free medium HyQ® PF 293 (Hyclone, Logan UT) was used for serum-free conditions. Human conjunctival epithelial cells (HCEC) (21) were cultured in Earle's minimal essential medium containing L-glutamine, 10% FBS, 15 mM HEPES, and antibiotics. Human monocyte derived iDC and culture supernatants were produced as described (8,22).
Expression Vectors-pEFBOS-MD-2-FLAG-His, encoding human MD-2 with C-terminal FLAG and His 6 sequences, was provided by Dr. Miyake (7) and used to produce sMD-2. Human TLR4 and TLR9 fused at their C-terminal ends with enhanced green fluorescent protein (TLR4-GFP and TLR9-GFP), were generated by inserting the TLR in-frame into pEGFP-N1 (Clontech).
Production of sMD-2-Two million HEK293T cells in 10-cm tissue culture dishes were transfected with 5 g of pEFBOS-MD-2-FLAG-His using calcium phosphate. Control supernatants (mock) were prepared in the same way, except that the plasmid was omitted. After overnight transfection, cells were washed once and 10 ml of fresh medium was added. For production of serum-free sMD-2, the transfected cells were washed twice with Hanks' balanced salt solution and 10 ml of serumfree medium was added. Supernatants were collected 24 h later, centrifuged, and stored at 4°C. Where indicated, sMD-2 was generated in medium containing 100 ng/ml LPS.
MD-2 Activity Assay-TLR4 reporter cells (5 ϫ 10 4 cells per well) were allowed to adhere in a 96-well tissue culture plate. To initiate the assay, medium was removed and 100 l of MD-2 supernatant and, where required, 100 ng/ml LPS were added to each well. The cells were incubated 8 -16 h at 37°C, lysed in Reporter Lysis Buffer (Promega, Madison, WI), and luciferase activities were determined by using a Luciferase Chemiluminescent assay kit (Promega). Each sample was measured in triplicate, and results are reported as average luciferase units (ϫ10 Ϫ3 ) Ϯ S.D. In one experiment, HCEC cells (8 ϫ 10 4 cells per well in a 48-well plate) were allowed to adhere 2 days, then incubated for 24 h with either LPS (100 ng/ml) or MD-2⅐LPS complex (1:4 dilution of purified sample) after which IL-6 and IL-8 were determined by ELISA (23).
MD-2 Preincubation Conditions-In some experiments, sMD-2 was either preincubated with 100 ng/ml LPS for 24 h, or produced in medium containing 100 ng/ml LPS, and in these cases, no additional LPS was added at the beginning of the assay. In other experiments, MD-2 samples were preincubated at 37°C in the absence of LPS, and in these cases 100 ng/ml LPS was added at the beginning of the assay. For background stimulation, LPS was omitted altogether. and tested for binding to cells expressing TLR4-GFP (thick lines) or TLR9-GFP (thin lines). Cells were labeled with anti-FLAG (or an isotype control) and stained with an allophycocyanin-conjugated secondary antibody. Shown are histograms gated on GFP ϩ (transfected) cells. Surface expression of TLR4 was similar in all samples (data not shown).
Immunoprecipitations-For anti-FLAG immunoprecipitations, 1-2 ml of MD-2 containing supernatant was incubated at 4°C overnight with 4.9 g of anti-FLAG M2 mAb and 20 l of protein A-Sepharose beads (Amersham Biosciences). For the biotin-LPS precipitations, 10 ml of MD-2 containing supernatant were incubated with 100 ng/ml biotin-LPS at 37°C for 8 h. Supernatants were then incubated with 50 l of monoclonal anti-biotin-agarose overnight at 4°C. For all samples, beads were washed, spun, and boiled in either reducing or non-reducing SDS sample buffer. Samples were resolved by SDS-PAGE, immunoblotted with anti-FLAG M2 mAb peroxidase conjugate, and developed by using enhanced chemiluminescence (Amersham Biosciences).
TLR4 Binding Assay-HEK293T cells (3 ϫ 10 5 cells/well of 6-well plate) were transfected overnight with 1 g of either TLR4-GFP or TLR9-GFP using calcium phosphate. After overnight transfection, cells were washed once and 2 ml of fresh medium was added. After 24 h, medium was replaced with MD-2 supernatants and incubated for 4 h at 37°C. Cells were washed twice with Hanks' balanced salt solution, 0.1% bovine serum albumin, 0.1% NaN 3 , and labeled with 20 g/ml anti-FLAG M2 mAb, 2 g/ml anti-TLR4, or 2 g/ml control antibody (MOPC300) followed by 2 g/ml allophycocyanin-conjugated secondary antibody. The stained cells were analyzed using a FACSCalibur flow cytometer and FlowJo software.
Purification of sMD-2-sMD-2 was purified by metal affinity chromatography using a 1-ml HiTrap chelating HP column (Amersham Biosciences) charged with nickel sulfate according to the manufacturer's protocol. MD-2-containing supernatants were passed through the column in 20 -30-ml aliquots at 4°C. The column was washed with 10 ml of 20 mM Tris, 500 mM NaCl, 15 mM imidazole, pH 8.0, and the MD-2 was eluted in 3 ml of the same buffer but containing 250 mM imidazole. Purified sMD-2 was dialyzed at 4°C against medium and filter steril-ized before use. MD-2 concentrations were determined as previously described (8). In brief, one sMD-2 sample was resolved by SDS-PAGE and intensities were compared with known concentrations of a Cterminal FLAG-tagged standard protein (FLAG-BAP fusion protein, Sigma). The calculated molar concentration of this sample was used as the standard in an ELISA to determine molar concentrations of all sMD-2 samples.

MD-2 Is
Unstable at Physiological Temperature-MD-2, an essential component of the LPS signaling pathway, is secreted as a soluble form (sMD-2) that confers LPS responsiveness to cells expressing TLR4 (8,9). However, we found that preincubation of sMD-2 at 37°C prior to addition to TLR4 reporter cells resulted in a dramatic loss in activity. As seen in Fig. 1A, after 24 h at 37°C the capacity of sMD-2 to confer LPS responsiveness to TLR4, as measured by NF-B activation, decreased by ϳ90% when compared with sMD-2 that had been kept at 4°C (zero time point). During this same period, the amount of sMD-2 in solution remained constant as determined by ELISA and there was no change in apparent molecular weight by SDS-PAGE (data not shown), suggesting that the loss in activity was because of a non-covalent structural transition rather than proteolytic degradation or precipitation. We next asked whether sMD-2 lost the ability to bind either LPS or TLR4 at 37°C. To examine LPS binding, we incubated sMD-2 with biotin-labeled LPS and immunoprecipitated with anti-biotin; FIG. 2. LPS prevents the inactivation of MD-2 at 37°C. A, supernatants containing sMD-2 were preincubated at 37°C for 24 h with or without LPS. After the preincubation, LPS was added to the samples that were preincubated without LPS. All samples were added to TLR4 reporter cells and assayed for activity. Supernatants not containing MD-2 were used as a control. Data are representative of 10 separate experiments. B, supernatants from iDC were tested for MD-2 activity using TLR4 reporter cells, as in panel A. iDC supernatants (iDC sup.) were either maintained at 4°C or were preincubated at 37°C in the presence or absence of LPS. Medium alone was used as a negative control. Data are representative of iDC from 3 different donors. C, supernatants from MD-2-transfected cells were preincubated at 37°C for 24 h with or without the indicated concentration of lipid A. After the preincubation, lipid A was added to those samples preincubated without lipid A and all samples were added to TLR4 reporter cells and assayed for activity. Data are representative of two experiments.
LPS-bound MD-2 was detected by SDS-PAGE and immunoblotting. Fig. 1B (upper panel) shows that sMD-2 maintained at 4°C binds LPS, confirming previous results (9, 18) (lane 2). By contrast, sMD-2 that had been preincubated at 37°C failed to bind LPS (lane 4). Immunoprecipitations with anti-FLAG confirmed that similar amounts of sMD-2 protein were present in each sample (Fig. 1B, lower panel). To determine whether the ability of sMD-2 to bind TLR4 was also lost at 37°C, TLR4 expressing cells were incubated with sMD-2, and cell-bound MD-2 was detected by flow cytometry. As a negative control, sMD-2 was also incubated with cells expressing TLR9. Fig. 1C shows that sMD-2 maintained at 4°C stained the TLR4 transfectants brightly, but gave weak staining on the TLR9 transfectants, as expected. However, after a 24-h preincubation at 37°C the sMD-2 exhibited a substantial (85%) decrease in TLR4 binding. Thus, at 37°C sMD-2 rapidly loses the ability to confer LPS responsiveness to TLR4 reporter cells, accompanied by a loss in LPS and TLR4 binding capacities.
LPS Stabilizes MD-2-Because ligand binding often protects proteins from denaturation, we asked whether LPS would stabilize sMD-2 against loss of activity at 37°C. Accordingly, we preincubated sMD-2 for 24 h at 37°C with LPS and then assayed the samples for activity. As a control, sMD-2 was preincubated for 24 h without LPS, and LPS was added at the beginning of the activity assay. As seen in Fig. 2A, sMD-2 that had been preincubated with LPS was at least 10-fold more active than the control sample preincubated without LPS, and was as active as sMD-2 that had been kept at 4°C. Treatment of inactivated (24 h, 37°C) sMD-2 with LPS for 24 h at 37°C failed to restore activity, indicating that inactivation at 37°C is irreversible (data not shown). To compare recombinant sMD-2 with its endogenously expressed counterpart, we collected conditioned medium from human iDC, a potent source of sMD-2like activity (8,22). Fig. 2B shows that iDC supernatants lost activity at 37°C and addition of LPS prevented this loss. Thus, recombinant sMD-2 closely resembles the endogenously produced protein. We also asked whether lipid A, the toxic portion of LPS (1), would stabilize sMD-2. Interestingly, lipid A activated TLR4 when added with sMD-2 to reporter cells but failed to stabilize sMD-2 during a 24-h preincubation at 37°C (Fig.  2C). This observation suggests that the lipid A portion of LPS is responsible for TLR4 activation, whereas the carbohydrate portion mediates MD-2 stabilization.
Because sMD-2 is derived from cells cultured for 24 h at 37°C, a substantial amount of activity must be lost during production. To prevent this loss in activity, LPS was added to the culture medium of MD-2-transfected cells, and after 24 h supernatants were collected and assayed for activity. Parallel cultures not containing LPS were treated in the same way. The left two clusters in Fig. 3A show that immediately after culture, sMD-2 from the LPS containing supernatant was about 4 -8fold more active than sMD-2 produced in the absence of LPS, as estimated from sMD-2 dilutions giving similar activities. Conversely, these same two samples bound TLR4 to similar extents (Fig. 3B, left 2 clusters), indicating that sMD-2 can bind to TLR4 regardless of whether or not it has associated with LPS. The same sMD-2 samples were then incubated an additional 48 h at 37°C. Following this incubation, the activity of the sMD-2 produced in LPS had decreased only slightly, whereas sMD-2 cultured in medium lacking LPS was inactive (Fig. 3A,  right 2 clusters). The TLR4 binding capacity of sMD-2 also decreased after 48 h incubation at 37°C, and LPS, in part, prevented this loss (Fig. 3B, right 2 clusters). From these results we can conclude that interaction with LPS prevents the loss of sMD-2 function that would otherwise occur at 37°C.
The MD-2⅐LPS Complex Triggers TLR4 -In view of our find-ing that LPS binds and stabilizes sMD-2, we asked whether the sMD-2⅐LPS complex, by itself, would activate TLR4. Accordingly, sMD-2 produced in LPS-containing cultures was purified by metal affinity chromatography and added in graded concentrations to TLR4 reporter cells. As seen in Fig. 4A, the purified sMD-2⅐LPS complex activated TLR4 at nanomolar concentrations in a dose-dependent manner, with the same activity as the sample prior to purification. As expected, purified sMD-2 from a culture not containing LPS failed to activate the TLR4 reporter cells, although this sample did activate TLR4 when LPS was added at the time of the assay (data not shown). To control for the possibility that our purification procedure failed to remove free LPS, we subjected supernatants from mock transfected cultures containing LPS (mock/LPS) to the same purification procedure. The purified mock sample was added to TLR4 reporter cells along with supernatants that contained unpurified sMD-2, which was shown to be active in the presence of LPS (Fig. 4B, right bar). If free LPS contaminated the purified mock sample, then combining it with sMD-2 would result in TLR4 activation. However, no activation was observed (Fig. 4B, middle bar), indicating that the amount of LPS re-

FIG. 3. LPS stabilizes sMD-2.
A, sMD-2 was produced in medium Ϯ LPS. Half of each sample was kept at 4°C and the remainder was incubated at 37°C for an additional 48 h. 2-Fold serial dilutions were prepared and all samples were assayed for activity as in Fig. 2A. B, the same samples as those used in panel A were added to TLR4-GFP-or TLR9-GFP-transfected HEK293T cells and tested for TLR4 binding as in Fig. 1C. Data are presented as geometric mean MD-2 fluorescence on TLR4-GFP ϩ cells minus geometric mean MD-2 fluorescence on TLR9-GFP ϩ cells, as a measure of TLR4-specific MD-2 binding. In all samples, the starting sMD-2 concentrations were the same within experimental error by ELISA. LPS did not alter the rate of cell growth or sMD-2 production. Data are representative of two separate experiments. maining after purification was below our limit of detection. Thus, the MD-2⅐LPS complex activates TLR4 in the absence of free LPS. To demonstrate a physiological relevance for this finding, we treated HCEC with either LPS alone or with the MD-2⅐LPS complex and measured IL-6 and IL-8 secretion. Similar to other types of epithelial cells, the HCEC express TLR4, but not MD-2 (21,24). Fig. 4C shows that HCEC in fact, respond to the sMD-2⅐LPS complex but not to LPS alone, indicating that the complex is essential for LPS recognition in some cell types.
CD14 Is Required for MD-2 Stabilization, but Not for TLR4 Activation by MD-2⅐LPS-The studies described above were performed in medium containing FBS, which contains soluble CD14. CD14 is known to be required for LPS responsiveness at the low LPS levels used in the current study (25). To examine the role of CD14 in the stabilization of sMD-2 by LPS, sMD-2, produced in serum-free medium, was preincubated with LPS for 24 h at 37°C. Recombinant human CD14 (rCD14) was added either before or after the preincubation. As seen in Fig.  5A, sMD-2 displayed activity only when rCD14 was present during the preincubation with LPS. To provide further evidence for the involvement of CD14, we used an anti-bovine CD14 (anti-bCD14) mAb to neutralize the CD14 in serum-containing sMD-2 supernatants (Fig. 5B). When the anti-bCD14 was added before the preincubation with LPS, the stabilization of sMD-2 was inhibited. However, if the anti-bCD14 was added after the preincubation with LPS, there was no change in activity. Taken together, these results suggest that CD14 is required for the interaction between LPS and MD-2, as previously reported (17,18), but is not required for the activation of TLR4 by the sMD-2⅐LPS complex (20). To provide direct evidence for the lack of involvement of CD14 in the interaction between the sMD-2⅐LPS complex and TLR4, anti-bCD14 mAb was added to purified sMD-2⅐LPS complex prior to transfer to TLR4 reporter cells. As seen in Fig. 4C, anti-bCD14 failed to block activation, ruling out the possibility that the purified complex contained a CD14 contaminant that was required for the activation of TLR4 by sMD-2⅐LPS. DISCUSSION In the current study we demonstrate that the interaction of LPS with sMD-2 produces a stable, activating ligand for TLR4. Our results, summarized in Fig. 6, show that MD-2 is secreted as a labile molecule that, over a relatively short period of time at 37°C, irreversibly decays into an inactive form (1) that is unable to bind either LPS (2) or TLR4 (3). If, however, freshly FIG. 4. The MD-2⅐LPS complex activates TLR4. A, activity of purified sMD-2 produced in the presence (filled squares) or absence (filled circles) of LPS. Open symbols represent activities of samples prior to purification. sMD-2 was generated in medium Ϯ 100 ng/ml LPS, purified by metal affinity chromatography, and assayed for MD-2 activity without additional LPS in medium containing 10% FBS. Data are representative of two separate experiments. B, the purification procedure removes free LPS. Supernatant from a mock transfected culture containing 100 ng/ml LPS (mock/LPS) was subjected to the metal affinity chromatography procedure and added to reporter cells alone (left bar) or with supernatants containing sMD-2 (middle bar). As a positive control, the activity of an equal amount of sMD-2 supplemented with 100 ng/ml LPS is shown (right bar). Data are representative of two separate experiments. C, MD-2⅐LPS complex activates epithelial cells. HCEC, an example of a cell type that expresses TLR4 but not MD-2, were treated with either LPS or MD-2⅐LPS complex. The MD⅐LPS complex, but not free LPS, stimulates IL-6 and IL-8 secretion.
FIG. 6. Model for the activation of TLR4 by sMD-2⅐LPS. MD-2 is secreted as a labile protein (sMD-2) that will normally decay within a 24-h period (1) into an inactive form, incapable of binding LPS (2) or TLR4 (3). However, if LPS is present, signaling a bacterial infection, sMD-2 binds the LPS in a CD14 dependent manner (4), forming a stable MD-2⅐LPS complex. This complex directly activates TLR4, thereby initiating an inflammatory response (5).
FIG. 5. CD14 is required for sMD-2 stabilization, but not for TLR4 activation by the sMD-2⅐LPS complex. A, LPS does not stabilize sMD-2 in serum-free medium; CD14 restores stabilization. Serum-free supernatants containing sMD-2 were preincubated for 24 h at 37°C with LPS (100 ng/ml). rCD14 (100 ng/ml) was added either before or after the preincubation, and all samples were assayed for activity. Data are representative of three separate experiments. B, CD14 is required for the stabilization of sMD-2 by LPS, but not for later events in TLR4 triggering. An anti-bovine CD14 (bCD14) mAb or an IgG1 isotype control mAb was added to sMD-2 supernatants (containing FBS) before or after a 24-h preincubation at 37°C with LPS. After preincubation, samples were assayed for activity. Data are representative of two separate experiments. C, activation of TLR4 by purified MD-2⅐LPS is independent of CD14. Purified MD-2⅐LPS was incubated for 30 min at 4°C with 10 g/ml anti-bCD14 mAb and assayed for activity. synthesized sMD-2 is exposed to LPS and CD14, it converts to a stable MD-2⅐LPS complex (4) that is capable of activating TLR4 (5). Purified MD-2⅐LPS complex activates TLR4 in the absence of free LPS or CD14. Thus, the MD-2⅐LPS complex by itself is competent to activate TLR4, and the principal role of CD14 is to aid in the formation of the MD-2⅐LPS complex, and not to facilitate the transfer of MD-2⅐LPS to TLR4, or to transfer LPS directly to TLR4. Of interest, a recent report indicates that sMD-2 produced by insect cells also binds LPS and activates TLR4 (20).
Both freshly secreted sMD-2 and sMD-2⅐LPS complex bind to TLR4, but only the complex triggers a response. What, then, is different about the complex that permits it to activate TLR4? One possibility is that LPS induces a conformational change in sMD-2 and this conformation, rather than LPS itself, triggers TLR4. A similar situation occurs with the TLR prototypic receptor, Drosophila Toll, which recognizes protease-induced structural changes in the upstream protein, Spaetzle, rather than the pathogen itself (26,27). However, this model fails to explain the results of gene transfer studies showing that both TLR4 (11)(12)(13) and MD-2 (14 -16) discriminate between lipid A variants, suggesting that TLR4, as well as MD-2, interacts directly with the lipid A portion of LPS. To explain these findings, we prefer a model in which MD-2 "presents" LPS to TLR4, perhaps by directing the lipid A portion of LPS to one of the atypical leucine-rich repeats of TLR4 thought to be involved in ligand binding (28). Interestingly, lipid A is able to activate TLR4 in the presence of MD-2, but, unlike LPS, does not protect MD-2 from inactivation at 37°C (Fig. 2C). This suggests that MD-2 contains two functionally discreet binding sites for LPS, one for the lipid A portion of the LPS molecule that is involved in TLR4 activation, and a second site for the carbohydrate portion that mediates MD-2 stabilization. Subsequent to the binding of sMD-2⅐LPS, TLR4 transduces a signal that initiates inflammatory responses. Previous data suggest that the signaling event itself is the aggregation of TLR4⅐MD-2 complexes (18,19), but how the interaction of MD-2⅐LPS with TLR4 would bring this about at the molecular level remains to be determined.
Although many cells, such as monocytes, dendritic cells, and B cells express both TLR4 and MD-2, other cells, such as epithelial cells express TLR4 but little or no MD-2 (21, 29). As we have shown, these cells depend upon sMD-2 for the generation of an inflammatory response to LPS. Thus, in microenvironments that are populated with appreciable numbers of TLR4 ϩ /MD-2 Ϫ cells, sMD-2 would play a crucial role in shaping the inflammatory response. Based upon our results with the HCEC cells, we would expect, for example, that MD-2⅐LPS would induce epithelial cells to secrete IL-8, a chemokine that initiates potent anti-microbial responses by recruiting neutrophils to the site of inflammation. An important prediction for tissues that contain TLR4 ϩ /MD-2 Ϫ cells is that the concentration of active sMD-2 in the interstitial fluids controls their responses to LPS. Presumably sMD-2 concentrations are maintained at low basal levels to guard against overwhelming inflammatory responses to Gram-negative bacteria, and the instability of sMD-2 at physiological temperature may provide a means for limiting its activity. However, at sites of inflammation local concentrations of sMD-2 may be increased, for example, by the recruitment of cells such as immature dendritic cells that secrete high amounts of sMD-2.
Although we have dealt with the soluble form of MD-2 in this study, many types of cells co-express TLR4 and MD-2, and in these cells a portion of the MD-2 reaches the cell surface bound to TLR4. An important question posed by our studies is whether the soluble and TLR4-bound forms of MD-2 are similar in structure and function. Specifically, is the TLR4-associated form of MD-2, like its soluble counterpart, unstable at physiological temperature? In addition, does the sMD-2⅐LPS complex activate cells that express both MD-2 and TLR4? If so, is this mediated by an exchange of the entire MD-2⅐LPS complex, or of the LPS alone? Studies currently underway to answer these questions should provide important new insights into the biology of the innate response to LPS.