Specific High Affinity Interactions of Monomeric Endotoxin·Protein Complexes with Toll-like Receptor 4 Ectodomain*

Potent Toll-like receptor 4 (TLR4) activation by endotoxin has been intensely studied, but the molecular requirements for endotoxin interaction with TLR4 are still incompletely defined. Ligand-receptor interactions involving endotoxin and TLR4 were characterized using monomeric endotoxin·protein complexes of high specific radioactivity. The binding of endotoxin·MD-2 to the TLR4 ectodomain (TLR4ECD) and transfer of endotoxin from CD14 to MD-2/TLR4ECD were demonstrated using HEK293T-conditioned medium containing TLR4ECD ± MD-2. These interactions are specific, of high affinity (KD < 300 pm), and consistent with the molecular requirements for potent cell activation by endotoxin. Both reactions result in the formation of a Mr ∼ 190,000 complex composed of endotoxin, MD-2, and TLR4ECD. CD14 facilitates transfer of endotoxin to MD-2 (TLR4) but is not a stable component of the endotoxin·MD-2/TLR4 complex. The ability to assay specific high affinity interactions of monomeric endotoxin·protein complexes with TLR4ECD should allow better definition of the structural requirements for endotoxin-induced TLR4 activation.

Potent Toll-like receptor 4 (TLR4) activation by endotoxin has been intensely studied, but the molecular requirements for endotoxin interaction with TLR4 are still incompletely defined. Ligandreceptor interactions involving endotoxin and TLR4 were characterized using monomeric endotoxin⅐protein complexes of high specific radioactivity. The binding of endotoxin⅐MD-2 to the TLR4 ectodomain (TLR4 ECD ) and transfer of endotoxin from CD14 to MD-2/TLR4 ECD were demonstrated using HEK293T-conditioned medium containing TLR4 ECD ؎ MD-2. These interactions are specific, of high affinity (K D < 300 pM), and consistent with the molecular requirements for potent cell activation by endotoxin. Both reactions result in the formation of a M r ϳ 190,000 complex composed of endotoxin, MD-2, and TLR4 ECD . CD14 facilitates transfer of endotoxin to MD-2 (TLR4) but is not a stable component of the endotoxin⅐MD-2/TLR4 complex. The ability to assay specific high affinity interactions of monomeric endotoxin⅐protein complexes with TLR4 ECD should allow better definition of the structural requirements for endotoxin-induced TLR4 activation.
Essential arms of the innate immune system are the Tolllike receptors (TLRs). 2 These receptors link recognition of unique microbial molecules to activation of host defense effector systems by rapidly triggering pro-inflammatory responses (1). Potent host responses toward many Gramnegative bacteria (GNB) are mediated by recognition and response to unique glycolipids (lipopoly-or lipooligosaccharides LOS, endotoxin) of the GNB outer membrane by TLR4. TLR4 does not function alone but requires the accessory protein MD-2, which binds non-covalently to the N-terminal ectodomain of TLR4 (2)(3)(4)(5)(6). Maximally potent endotoxin-induced cell activation also requires the extracellular lipopolysaccharide-binding protein (LBP) and membrane (m) or soluble (s) extracellular CD14 (4,(7)(8)(9). The sequential action of LBP, CD14, secreted or TLR4-associated MD-2, and TLR4 confers the extraordinary sensitivity of mammalian cells to many GNB endotoxins. This ordered action implies differences in endotoxin binding specificity, with LBP having the highest affinity for endotoxin organized at lipid/water interfaces (e.g. purified endotoxin aggregates and endotoxin in the GNB outer membrane), CD14 for LBP-modified endotoxin-rich interfaces, MD-2 for monomeric endotoxin⅐CD14 and TLR4, apparently, for endotoxin presented as a monomeric complex with MD-2 (8). Together, these proteins can convert one GNB (containing ϳ10 6 endotoxin molecules) to 10 6 TLR4-activating monomeric endotoxin⅐protein complexes (i.e. endotoxin⅐CD14 or endotoxin⅐MD-2), greatly amplifying host responsiveness to endotoxin. At pM concentrations, monomeric complexes of endotoxin⅐CD14 or endotoxin⅐MD-2 activate, respectively, mammalian cells expressing MD-2/TLR4 or TLR4 alone, triggering robust cell activation through engagement of Ͻ10 3 TLR4 molecules.
Despite the ability of endotoxin⅐CD14 and endotoxin⅐ MD-2 to activate cells at pM concentrations (half-maximal cell activation is Ͻ50 pM), published estimates of TLR4-dependent cellular interactions of endotoxin have indicated an apparent K d of Ն3 nM (10 -15). It is possible that brief occupation of a small subset of TLR4 is sufficient for robust cell activation. However, an important limitation of earlier studies has been the likelihood that much of the endotoxin added was not presented as the monomeric protein⅐endotoxin complex that is likely the preferred substrate for MD-2/ TLR4 or TLR4 alone. In contrast to the remarkable progress in identification and characterization of the intracellular biochemical machinery responsible for TLR signaling, there is still virtually nothing known about the specificities of ligand-receptor interaction for any of the TLRs. In nearly all cases, TLR recognition has been inferred from the measurement of receptor activation, not binding, properties of specific compounds.
In this study, we have made use of our ability to produce and isolate stable endotoxin⅐protein complexes of high specific radioactivity (ϳ25,000 cpm/pmol) to more rigorously and directly characterize ligand-receptor interactions involving endotoxin and TLR4. For this purpose, we have expressed, in HEK293T cells, the predicted ectodomain of human TLR4 (amino acids 24 -634, TLR4 ECD ) (16) with or without human MD-2 to permit direct assay of endotoxin interactions with TLR4 ECD Ϯ MD-2. We describe, for the first time, the direct binding of endotoxin⅐MD-2 complex to the predicted TLR4 ectodomain in the absence of any other cellular or extracellular co-factors and the direct transfer of endotoxin from CD14 to MD-2/TLR4 ECD . These interactions are highly specific and of very high affinity (K d Յ 300 pM) and fully consistent with the molecular requirements for potent mammalian cell activation by endotoxin. Both binding of endotoxin⅐MD-2 to TLR4 ECD and transfer of endotoxin from endotoxin⅐CD14 to MD-2/ TLR4 ECD result in the formation of a stable complex that is composed of endotoxin, MD-2, and TLR4 ECD without CD14.

EXPERIMENTAL PROCEDURES
Materials-LBP and sCD14 were gifts from Xoma (Berkley, CA) and Amgen Corp. (Thousand Oaks, CA) respectively. Acyloxyacyl hydrolase (AOAH) was a gift from Dr. R. Munford (University of Texas Southwestern, Dallas, TX). Soluble MD-2 containing a hexapolyhistidine tag on the C-terminal end was prepared as previously described (17). Human serum albumin (HSA) was obtained as an endotoxin-free, 25% stock solution (Baxter Health Care, Glendale, CA). [ 3 H]LOS (25,000 cpm/ pmol) from an acetate auxotroph of Neisseria meningitidis serogroup B was metabolically labeled and isolated as described previously (18). Chromatography matrices (Sephacryl HR S200 and S300, Ni 2ϩ FF-Sepharose) were purchased from GE Healthcare (Piscataway, NJ). Anti-FLAG M2-agarose and streptavidin-agarose were purchased from Sigma.
cDNA encoding MD-2 and sCD14-(1-156) were inserted (17) into pBAC11 (Novagen) using XhoI and NotI-sensitive restriction sites. This provides a 5Ј-flanking signal sequence (gp64) to promote secretion of protein and a C-terminal sixresidue polyhistidine tag. Sf9 cells were used for transfection and amplification of baculovirus, whereas High Five cells in serum-free medium were used for protein production (17).
Conditioned medium containing secreted MD-2-His 6 was used directly for generation of LOS⅐MD-2. Conditioned medium containing sCD14-(1-156)-His 6 was dialyzed against 20 mM phosphate, 0.5 M NaCl, adsorbed to Ni 2ϩ FF-Sepharose equilibrated in the same buffer, washed, and the adsorbed protein eluted by an imidazole gradient (Explorer 100 fast protein liquid chromatography, GE Healthcare  4). Reaction products were analyzed by Sephacryl HR S200 or S300 (1.6 ϫ 70 cm) chromatography in PBS, pH 7.4, Ϯ0.5 mM Mg 2ϩ , 1 mM Ca 2ϩ , 0.03% HSA. Divalent cations and HSA were used in columns with LOS agg to improve recoveries of LOS agg . Fractions (0.5 ml) were collected at a flow rate of 0.3 ml/min at room temperature using AKTA Purifier or Explorer 100 fast protein liquid chromatography (GE Healthcare). Radioactivity in collected fractions was analyzed by liquid scintillation spectroscopy (Beckman LS liquid scintillation counter). Recoveries of [ 3 H]LOS were Ն70% in all cases. All solutions used were pyrogen-free and sterile-filtered. After chromatography, selected fractions were sterile-filtered (0.22 m) and kept at 4°C for 3-6 months with no detectable changes in chromatographic or functional properties. The same conditioned medium was used for all concentrations of [ 3 H]LOS⅐sCD14 or [ 3 H]LOS⅐MD-2 used for Scatchard analysis. Multiple preparations of conditioned medium contained similar amounts of TLR4 ECD or MD-2/TLR4 ECD , i.e. 1-2 and 3-4 pmol/ml, respectively.

Soluble TLR4 ECD Binds [ 3 H]LOS⅐MD-2 with Picomolar
Affinity-We have previously described the generation and isolation of a monomeric [ 3 H]LOS⅐MD-2 complex that potently activates cells expressing TLR4 without MD-2 (17,21,22).  (17), suggesting that levels of surfaceexpressed TLR4 in these cells were too low to measure direct binding.
To circumvent these limitations and permit more direct assay of the molecular requirements for endotoxin-TLR4 interactions, we transiently expressed, in HEK293T cells, an N-terminal fragment of recombinant human TLR4 (residues 24 -634) corresponding to the predicted ectodomain of TLR4 (TLR4 ECD ) and containing an N-terminal FLAG tag. Harvested control and TLR4 ECD -containing culture media were incubated with 1 nM of purified [ (Fig. 1). Incubation of [ 3 H]LOS⅐MD-2 with conditioned medium containing TLR4 ECD (but not control medium) yielded a novel [ 3 H]LOS-containing product (Fig. 1A, encircled peak) whose formation depended upon TLR4 ECD and the presentation of [ 3 H]LOS as [ 3 H]LOS⅐MD-2 (Fig. 1, compare A (Fig. 1, B and C), even when added at 200ϫ greater LOS concentrations (data not shown). These findings strongly suggest a specific and direct interaction of Rechromatography on Sephacryl S300 of the newly formed [ 3 H]LOS-containing product recovered after incubation of [ 3 H]LOS⅐MD-2-His 6 with medium containing FLAG-TLR4 ECD (Fig. 1A, encircled fractions) yielded a single symmetrical peak that, by comparison to elution of protein standards, gave a predicted M r ϳ 190,000 (Fig. 1D). [ 3 H]LOS in the product could be captured by anti-FLAG (Fig. 1F) antibodies as well as by Ni ϩ2chelating resin (Fig. 1E) (Fig.  1B) is consistent with the requirements of co-expression of MD-2 and TLR4 for sensitive cellular responses to monomeric endotoxin⅐sCD14 complex (22). To demonstrate more directly the requirement of MD-2 for interaction of [ 3 H]LOS⅐sCD14 with complexes containing TLR4 ECD , we co-transfected HEK293T cells with expression plasmids for FLAG-TLR4 ECD and MD-2-FLAG-His 6 . In contrast to incubations with control medium or medium containing TLR4 ECD alone, incubation of [ 3 H]LOS⅐sCD14 with harvested culture medium containing both secreted MD-2 and TLR4 ECD yielded an [ 3 H]LOS-containing product that eluted at M r Ͼ [ 3 H]LOS⅐sCD14 ( Fig. 2A). In contrast to the efficient reaction of [ 3 H]LOS⅐sCD14 with MD-2 and TLR4 ECD , there was little or no reaction of [ 3 H]LOS⅐MD-2 (Fig. 2B) or [ 3 H]LOS agg (Fig. 2C) under otherwise the same experimental conditions. The product formed by incubation of [ 3 H]LOS⅐sCD14 with the culture medium containing TLR4 ECD and MD-2 ( Fig. 2A) was the same as that observed when the culture medium containing TLR4 ECD alone was incubated with [ 3 H]LOS⅐MD-2, as judged by gel filtration chromatography (Fig. 2D) and co-capture analyses (Fig. 2, E  and F). With limiting amounts of culture medium containing MD-2 and TLR4 ECD , formation of the M r ϳ 190,000 complex increased with increasing concentrations of [ 3 H]LOS⅐sCD14 (Fig. 2G) and was saturable with an apparent K D of ϳ130 pM (Fig. 2H).

Transfer of [ 3 H]LOS from [ 3 H]LOS⅐sCD14 to MD-2/TLR4 ECD -The inability of [ 3 H]LOS⅐sCD14 to react with TLR4 ECD
Production of apparently the same M r 190,000 complex from incubation of [ 3 H]LOS⅐sCD14 with culture medium containing TLR4 ECD and MD-2 as that resulting from the reaction of TLR4 ECD -containing medium with [ 3 H]LOS⅐MD-2 strongly suggested that [ 3 H]LOS⅐sCD14 transferred [ 3 H]LOS to a complex containing MD-2 and TLR4 ECD without CD14 becoming part of this newly formed complex. To test that hypothesis, we took advantage of the fact that the ability of CD14 to bind endotoxin and transfer endotoxin to MD-2 requires only the N-terminal 152 amino acids of CD14 (full-length CD14 is 356 residues) (23)(24)(25)(26).

Presence of TLR4 ECD Stabilizes the Endotoxin Binding Activity of MD-2-It is noteworthy that incubation of [ 3 H]LOS⅐
sCD14 with harvested culture medium containing MD-2 and TLR4 ECD did not yield any [ 3 H]LOS⅐MD-2 (M r ϳ 25,000) despite the secretion of more MD-2 than TLR4 ECD (data not shown). We speculated that this reflected the instability of secreted MD-2 in serum-free culture medium at 37°C, as previously reported by Kennedy et al. (15). To test this hypothesis, we compared the products generated when [ 3 H]LOS⅐sCD14 was incubated with culture medium harvested after transfection (as in Fig. 2A) or was spiked into the culture medium 12 h after transfection so that, immediately upon secretion, MD-2 could potentially react with [ 3 H]LOS⅐sCD14. [ 3 H]LOS⅐MD-2 was generated in significant amounts only when [ 3 H]LOS⅐ sCD14 was present during secretion of MD-2. This was seen whether or not TLR4 ECD was also expressed (compare Fig. 4A with Figs. 2A and 3; Fig. 4, compare B with D). In contrast, formation of the M r ϳ 190,000 complex occurred efficiently when [ 3 H]LOS⅐sCD14 was added either during or after culturing. These findings suggest strongly that, in contrast to free MD-2, the endotoxin binding activity (i.e. reactivity with LOS⅐sCD14) of MD-2 is stable in serum-free medium at 37°C when co-expressed/associated with TLR4 ECD .
Hexaacylated and Tetraacylated LOS⅐MD-2 Complexes Show Similar Reactivity with TLR4 ECD -We and others (21,27) have previously shown that differences in TLR4 agonist activity of underacylated versus hexaacylated endotoxin species reflect differences in structural and functional properties of monomeric endotoxin⅐MD-2 complexes. The ability of underacylated endotoxin⅐MD-2 complexes to inhibit cell activation by hexaacylated endotoxin⅐MD-2 suggested that, even though underacylated endotoxin⅐MD-2 complexes do not efficiently trigger receptor activation, these complexes bind to the TLR4 ECD with an affinity comparable with that of hexaacylated endotoxin⅐MD-2. To test this hypothesis more directly, aggregates of hexaacylated LOS were treated with the deacylating enzyme acyloxyacyl hydrolase (AOAH) and used to generate monomeric complexes with MD-2 ([ 3 H]LOS AOAH ⅐MD-2) containing mainly (ϳ90%) tetraacylated LOS, as has been described previously (21). Incubation of a low concentration  (0.6 nM) of [ 3 H]LOS AOAH ⅐MD-2 with medium containing secreted TLR4 ECD followed by gel filtration of the products revealed a pattern similar to that seen with the hexaacylated [ 3 H]LOS⅐MD-2 complex, i.e. formation of the M r ϳ 190,000 complex (Fig. 5), directly demonstrating the similar reactivity of hexa-and tetraacylated LOS⅐MD-2 complexes with TLR4 ECD .

DISCUSSION
The data presented demonstrate, for the first time, specific high affinity (pM) interactions of endotoxin with the ectodomain of TLR4. These interactions required presentation of endotoxin (meningococcal LOS) as a monomeric complex with MD-2 (LOS⅐MD-2) when TLR4 ECD was present alone (Fig. 1) or as a monomeric complex with CD14 (LOS⅐sCD14) when TLR4 ECD was co-expressed with MD-2 (Fig. 2). The characteristics of these interactions are fully consistent with the molecular requirements for TLR4dependent cell activation by pM concentrations of meningococcal and many other endotoxin species (8,14,17,21,28,29). How certain deep rough species of endotoxin can potently (ϳnM) activate cells expressing TLR4/MD-2 without CD14 remains to be determined (30). The reactivity of LOS⅐sCD14 with culture medium containing TLR4 ECD and MD-2 (but not TLR4 ECD alone) (Fig. 1B) is consistent with an essential role of MD-2 in ligand (endotoxin⅐CD14) recognition by the MD-2/TLR4 receptor complex. Reaction of LOS⅐sCD14 with free MD-2 was not detected in culture medium harvested after 48 h of incubation, indicating that the reactant with LOS⅐sCD14 is preassociated MD-2/TLR4 ECD and that, as previously observed, free MD-2 is unstable. The reactivity of LOS⅐MD-2 with TLR4 ECD when TLR4 ECD was expressed alone (but not when TLR4 ECD was coexpressed with MD-2) also implies that, in the latter situation, TLR4 ECD is preoccupied with co-expressed MD-2. The low reactivity of LOS⅐MD-2 with preformed MD-2/TLR4 ECD also indicates that LOS is not readily transferred from LOS⅐MD-2 to MD-2/TLR4 and that MD-2 associated with TLR4 ECD is not readily exchanged with LOS⅐MD-2.
Previous studies have emphasized the importance of MD-2 in the maturation and surface expression of TLR4 (14,29,31,32). Our findings have revealed a role for TLR4 ECD in stabilizing MD-2, perhaps by stabilizing an otherwise labile monomeric form of MD-2 that may be needed for reactivity with endotoxin⅐CD14 as well as binding to TLR4 (13,16). 3 The apparent stability of MD-2/TLR4 ECD is consistent with the functional activity of this heterodimeric receptor, even when low levels of TLR4 and MD-2 are expressed. Although our stud-  ies do not directly reveal to what protein endotoxin is bound in the M r ϳ 190,000 complex containing MD-2 and TLR4 ECD , the ability of endotoxin⅐CD14 to react avidly with MD-2 (but not to TLR4 ECD ) suggests strongly that the primary reaction of endotoxin⅐CD14 is with MD-2 (ϮTLR4).
Many investigators have speculated that the reaction of endotoxin⅐CD14 with MD-2/TLR4 results in engagement of CD14 as an essential component of an activated oligomeric MD-2/TLR4-containing receptor complex (1,(33)(34)(35)(36). However, the ability of purified endotoxin⅐MD-2 complex to induce robust (MyD88-dependent) signaling in cells expressing TLR4 without MD-2 (or mCD14) (22) has suggested to us that CD14 is not an obligatory component of a TLR4 receptor complex activated by endotoxin. Our findings that reaction of LOS⅐sCD14 with MD-2/TLR4 ECD yields a product that contains LOS, MD-2, and TLR4 ECD (but not CD14 (Fig. 3)) plus the formation of an apparently identical product by reaction of LOS⅐MD-2 with TLR4 ECD (compare Figs. 1 and 2) are consistent with the view that the primary role of CD14 is transfer of endotoxin to MD-2 (ϮTLR4). However, the requirement of mCD14 for MyD88-independent signaling after endotoxin-induced TLR4 activation suggests a more complex role of CD14 in endotoxin-driven TLR4 activation (1,30,37). Our experiments in solution with CD14 and TLR4 ECD do not preclude the possibility of more stable endotoxin-induced interactions of mCD14 with activated receptor complexes containing fulllength transmembrane TLR4.
Both reaction of monomeric LOS⅐sCD14 with MD-2/ TLR4 ECD and of monomeric LOS⅐MD-2 with TLR4 ECD produced a higher order complex of apparent M r of ϳ190,000 containing LOS, MD-2, and TLR4 ECD (Figs. 1 and 2). This nearly matches the predicted M r of a dimer of a 1:1:1 complex of LOS⅐MD-2/TLR4 ECD (M r of LOS ϳ5,000, of MD-2 ϳ20,000, of TLR4 ECD ϳ 75,000). Confirmation of this predicted stoichiometry will require preparative purification of M r ϳ 190,000 complex and chemical and immunochemical analyses. It is not yet clear whether this product reflects endotoxin-induced TLR4 dimerization or reaction of monomeric protein⅐endotoxin complexes with pre-existing TLR4 ECD (MD-2) dimers in the harvested culture medium. In either case, the co-capture results shown in Fig. 1 are intriguing. The relatively efficient capture of the M r ϳ 190,000 complex by an activated Ni 2ϩ -chelating resin, despite inefficient capture of the LOS⅐MD-2-His 6 (or LOS⅐MD-2-FLAG-His 6 ; data not shown) complex suggests that the His tag is shielded in the LOS⅐MD-2 complex from interaction with Ni 2ϩ resin but becomes more accessible upon interaction with TLR4 ECD and formation of the M r ϳ 190,000 complex. This suggests a conformational change of MD-2 or of LOS bound to MD-2 upon binding of LOS⅐MD-2 to TLR4 ECD .
In summary, we have demonstrated for the first time pM interactions of endotoxin with TLR4 ECD . We have succeeded in demonstrating such high affinity interactions because of our use of purified endotoxin⅐protein complexes that are the preferred substrate(s) for (MD-2) TLR4 ECD and that are in radiolabeled form of very high specific radioactivity, making it possible to detect and quantify intermolecular interactions at pM endotoxin concentrations. The reagents and assays described in this study should now make it possible to define the structural requirements of endotoxin and MD-2 interactions with TLR4 ECD in a way that has not been possible before. Our findings are consistent with earlier speculations that the monomeric endotoxin⅐MD-2 complex, not endotoxin itself, is the ligand for TLR4 (13,17). As such, the molecular basis of ligand recognition and receptor activation by mammalian TLR4 may resemble more closely that of Drosophila Toll (38) than has been appreciated before. Whether or not, in comparison to other mammalian TLRs, TLR4 is idiosyncratic or instructive of the molecular basis of ligand recognition by other mammalian TLRs (39) remains to be determined.