Novel Roles in Human MD-2 of Phenylalanines 121 and 126 and Tyrosine 131 in Activation of Toll-like Receptor 4 by Endotoxin*

Potent mammalian cell activation by Gram-negative bacterial endotoxin requires sequential protein-endotoxin and protein-protein interactions involving lipopolysaccharide-binding protein, CD14, MD-2, and Toll-like receptor 4 (TLR4). TLR4 activation requires simultaneous binding of MD-2 to endotoxin (E) and the ectodomain of TLR4. We now describe mutants of recombinant human MD-2 that bind TLR4 and react with E·CD14 but do not support cellular responsiveness to endotoxin. The mutants F121A/K122A MD-2 and Y131A/K132A MD-2 react with E·CD14 only when co-expressed with TLR4. Single mutants K122A and K132A each react with E·CD14 ± TLR4 and promote TLR4-dependent cell activation by endotoxin suggesting that Phe121 and Tyr131 are needed for TLR4-independent transfer of endotoxin from CD14 to MD-2 and also needed for TLR4 activation by bound E·MD-2. The mutant F126A MD-2 reacts as well as wild-type MD-2 with E·CD14 ± TLR4. E·MD-2F126A binds TLR4 with high affinity (Kd ∼ 200 pm) but does not activate TLR4 and instead acts as a potent TLR4 antagonist, inhibiting activation of HEK/TLR4 cells by wild-type E·MD-2. These findings reveal roles of Phe121 and Tyr131 in TLR4-independent interactions of human MD-2 with E·CD14 and, together with Phe126, in activation of TLR4 by bound E·MD-2. These findings strongly suggest that the structural properties of E·MD-2, not E alone, determine agonist or antagonist effects on TLR4.

The ability of mammalian cells to respond to minute (pM) concentrations of endotoxins, unique and abundant surface glycolipids of Gram-negative bacteria, is needed for optimal host defense against many invading Gram-negative bacteria. This remarkable sensitivity depends on sequential protein-en-dotoxin and protein-protein interactions involving at least four extracellular and cell surface host proteins: lipopolysaccharidebinding protein (LBP), 2 soluble (s) and GPI-linked membraneassociated forms of CD14, secreted and Toll-like receptor (TLR) 4-associated forms of MD-2 and TLR4 (1)(2)(3)(4). LBP promotes extraction and delivery of individual molecules of endotoxin from endotoxin-rich interfaces (e.g. the Gram-negative bacterial outer membrane or aggregates of purified endotoxin) to CD14, yielding monomeric complexes of endotoxin (E)⅐CD14 that are the preferred substrate for MD-2 (2,(5)(6)(7). Transfer of endotoxin from CD14 to MD-2 coupled to binding of MD-2 to TLR4 triggers TLR4-dependent cell activation (2,7). Thus, MD-2 has a pivotal role in endotoxin-induced TLR4 activation, bridging endotoxin recognition initiated by LBP and CD14 to receptor activation.
The requirement for simultaneous engagement by MD-2 of endotoxin and TLR4 for receptor activation strongly suggests that MD-2 contains structurally and topologically distinct binding sites for endotoxin and TLR4. Studies of the effects of mutagenesis of MD-2 have provided experimental support for this concept, showing that discrete mutations of MD-2 could markedly impair cellular endotoxin responsiveness without apparently affecting the ability of MD-2 to engage TLR4 (8 -14). A subset of these MD-2 mutants also apparently retains normal endotoxin binding properties but fails to efficiently induce TLR4 oligomerization and receptor activation (15). These findings suggest that, in these select MD-2 mutants, the simultaneous engagement of endotoxin and TLR4 is somehow insufficient to trigger subsequent molecular events within the endotoxin⅐MD-2⅐TLR4 complex needed for receptor and cell activation.
The studies carried out to date seeking to characterize the structural requirements in MD-2 for endotoxin binding and TLR4 activation have presented endotoxin as aggregates Ϯ serum (8 -20). Under these conditions, it is likely that the added endotoxin is present in a variety of physical and biochemical states, including as aggregates of endotoxin Ϯ LBP Ϯ sCD14 and as complexes with lipoproteins (21,22) with little or no endotoxin present as a monomeric endotoxin⅐CD14 complex. This may explain the relatively high and variable apparent K d values (ϳ3-65 nM) that were estimated for endotoxin-MD-2 interactions (15)(16)(17)(18)(19)(20), in contrast to the pM reactivities (apparent "K d " ϳ 100 -200 pM) we have observed for transfer of endotoxin from CD14 to MD-2 (23). With wild-type recombinant human MD-2, efficient transfer of endotoxin monomers from CD14 to MD-2 occurs in solution with or without co-expression of TLR4, thus making it possible to assess separately direct E⅐CD14-MD-2 interaction and E⅐CD14 interaction that follows prior engagement of MD-2 with the ectodomain of TLR4. Once formed, a monomeric complex of hexa-acylated endotoxin with wild-type human MD-2, unlike either endotoxin alone or the E⅐CD14 complex, is a high affinity ligand (K d ϳ 300 pM) for the ectodomain of TLR4 (TLR4 ECD ) and a potent agonist for cells expressing TLR4 without MD-2 (7,23,24). In stark contrast to free MD-2, E⅐MD-2 is functionally stable (2,7,23,25), greatly facilitating isolation of E⅐MD-2 and subsequent functional testing (e.g. interaction with TLR4 ECD and activation of TLR4). Hence, comparison of the ability of wild-type and mutant MD-2 species to react with E⅐sCD14 (ϮTLR4 ECD ) and of the functional properties of isolated E⅐MD-2 complexes (wild-type and variant) that are formed should provide a more direct and complete appraisal of the structural requirements for MD-2 function.
We now describe, using this approach, the identification and characterization of two different classes of MD-2 mutants that bind TLR4 and react with E⅐CD14 but do not support TLR4-dependent cell activation by endotoxin. One class, exemplified by the mutants F121A/K122A and Y131A/K132A can react with E⅐CD14 but only when co-expressed with TLR4. A second class, F126A, reacts normally with E⅐CD14 in the absence of TLR4 to form an E⅐MD-2 complex that binds avidly to TLR4 ECD (K d ϳ 200 pM) but does not activate TLR4. The stability and functional properties of E⅐MD-2 F126A should make it a valuable tool for understanding the mechanism of TLR4 activation by endotoxin.
Production of Recombinant Proteins-Human embryonic kidney (HEK) 293T cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. Cells (ϳ80% confluency in 6-well plates or T75 flask for preparative amounts of material) were transfected with 4 g (6-well plate) or 20 g (T75 flask) of DNA using PolyFect reagent (Qiagen). After 12-16 h, plates were rinsed in PBS and 1 ml per well for 6-well plates or 10 ml for T-75 flask of serum-free medium (293 SFM, Invitrogen) ϩ 0.4% HSA Ϯ [ 3 H]LOS⅐sCD14 (1 nM; see below) were added. Media containing expressed proteins were collected 24 -48 h later. Conditioned medium containing secreted TLR4 ECD Ϯ MD-2 wild type (wt) or indicated mutants maintained activity to react with [ 3 H]LOS⅐MD-2 or [ 3 H]LOS⅐sCD14 for at least 6 months when stored at 4°C. Expression vectors containing DNA of interest for production of FLAG-TLR4 ECD , amino acids 24 -634, (pFLAG-CMV-TLR4) and MD-2-FLAG-His (pEF-BOS) as well as MD-2 containing the indicated mutations have been previously described and characterized (27) or have been generated according to that protocol.
Preparative amounts of wt and F126A MD-2 were generated from infections of High Five insect cells with baculovirus containing the cDNA for either wt human MD-2-His 6 or human MD-2 F126A inserted into pBAC11 as described previously (7). H]-labeled compounds in "spiked" media. Reaction products were analyzed by Sephacryl HR S200 or S300 (1.6 ϫ 70 cm) chromatography in PBS, pH 7.4, 0.03% HSA as indicated (23). Fractions (1.0 or 0.5 ml) were collected at a flow rate 0.5 or 0.3 ml/min at room temperature using AKTA Purifier or Explorer 100 FPLC (GE Healthcare). Radioactivity in collected fractions was analyzed by liquid scintillation spectroscopy (Beckman LS liquid scintillation counter). Recoveries of [ 3 H]LOS were Ն 70% 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.
To measure the apparent K d of [ 3 H]LOS⅐MD-2 F126A interactions with TLR4 ECD , [ 3 H]LOS⅐MD-2 F126A was incubated with concentrated (8 -10ϫ) conditioned medium containing TLR4 ECD in a final volume of 0.5 or 1 ml in PBS, pH 7.4, for 30 min at 37°C. The same conditioned medium (containing secreted TLR4 ECD ) was used with all concentrations of [ 3 H]LOS⅐MD-2 F126A tested for Scatchard analysis. Scatchard analysis was done using GraphPad Prism 4.
Immunoblotting-Polyhistidine-labeled wt and mutant MD-2 were detected by SDS-PAGE/immunoblot, using an anti-polyhistidine antibody (Tetra-His antibody, Qiagen, Valencia, CA) as previously described (7). Samples were electrophoresed (Bio-Rad minigel system) through a 4 -15% gradient acrylamide gel (Tris/HEPES/SDS buffer) and transferred to nitrocellulose. The nitrocellulose was washed with Tris-buffered saline (TBS), pH 7.5, containing 0.05% Tween-20 and 0.2% Triton X-100 (TBSTT), blocked to reduce nonspecific background with 3% bovine serum albumin in TBSTT for 1 h at 25°C, and incubated with the anti-His 4 antibody in TBSTT overnight. After washing with TBSTT, the blot was incubated with donkey anti-mouse IgG conjugated to horseradish peroxidase (Bio-Rad) for 1 h at 25°C in TBS containing 3% goat serum and washed with TBSTT exhaustively. Blots were developed using the Pierce SuperSignal substrate system.
HEK293 Cell Activation Assay-HEK293/TLR4 cell lines have been extensively characterized and were cultured as has been previously described (30) or according to recommendations of Invivogen, Inc. For cell activation assays by LOS⅐sCD14, cells were grown to confluency in a 6-well plate and then transfected with vector (4 g of DNA) containing wt or mutant MD-2 as described above. After transfection, serum-free medium was added for 24 h at 37°C in 5% CO 2 , and 95% humidity. The supernatants were removed; cells were dislodged and seeded in a 96-well plate (1 ϫ 10 5 cells/well) in triplicate. The cells were then stimulated with 1 nM LOS⅐sCD14 for 3 h in DMEM, 0.1% HSA. The short incubation time (3 h) was used to preclude significant secretion of MD-2 that could compete with cell surface MD-2/TLR4 for reaction with LOS⅐sCD14. Supernatants were removed and evaluated for extracellular accumulation of IL-8 by ELISA.
For cell activation by LOS⅐MD-2 containing either wt or F126A MD-2, HEK293 cells Ϯ TLR4 were seeded in a 96-well plate in triplicate (1 ϫ 10 5 cells/well) and stimulated with increasing concentrations of LOS⅐MD-2 complex in DMEM, 0.1% HSA for 20 h. Activation of HEK293 cells was assessed by measuring accumulation of extracellular IL-8 by ELISA (BD Clontech, Inc., Palo Alto, CA).

Assay of Binding of [ 3 H]LOS from [ 3 H]LOS⅐sCD14 to HEK293 Cells Expressing TLR4 Ϯ wt or Mutant MD-2-Parental
HEK293T cells and cells transiently transfected with pFLAG-CMV-TLR4 Ϯ pEFBOS-MD-2-FLAG-His, wt or mutant, (6 g of each plasmid DNA) were grown for 24 h in DMEM/10% fetal bovine serum in T75 flasks followed by 24 h in serum-free medium. Cells were dislodged with PBS, sedimented at 1000 rpm for 5 min, washed once with 3 ml of PBS/0.1% HSA, and then aliquoted so that each sample contained 4 ϫ 10 6

Presence of Monomeric E⅐sCD14 during Secretion of MD-2 from HEK293T Cells Results in High Yields of Bioactive E⅐MD-2-
We have previously demonstrated that bioactive monomeric endotoxin (meningococcal lipopoligosaccharide, LOS or Escherichia coli lipopolysaccharide, LPS)⅐MD-2 can be efficiently generated by incubation of monomeric endotoxin⅐sCD14 with conditioned medium harvested from insect cells expressing and secreting recombinant human MD-2 (7). However, initial efforts to reproduce these findings with harvested conditioned medium from HEK293T cells transiently expressing human MD-2 were unsuccessful ( Fig. 1A), despite secretion and extracellular accumulation of MD-2 ( Fig. 1C). Kennedy et al. (17) had previously shown that recombinant human MD-2 secreted into serum-free culture medium lost activity (i.e. ability to support serum/LPS-dependent activation of HEK/TLR4 cells) in a time (24 h) and temperature (37°C)-dependent manner unless serum (LBP and sCD14) and LPS were also present in the medium. These findings are consistent with the much greater stability at 37°C of monomeric LPS⅐MD-2 (as compared with MD-2 alone) (2, 7) that could be formed rapidly by serum (LBP and sCD14)-dependent conversion of LPS aggregates to LPS⅐sCD14 and reaction of LPS⅐sCD14 with secreted MD-2. In an attempt to produce monomeric endotoxin⅐MD-2 from MD-2 secreted by transiently transfected HEK293T cells, the medium was supplemented ("spiked") with MD-2 Mutants That Bind TLR4 but Do Not Support TLR4dependent Cell Activation by Endotoxin:Comparison of Reactivity with [ 3 H]LOS⅐sCD14-We made use of the same experimental design to re-examine the functional properties of three previously described human MD-2 double mutants: F121A/ K122A, K125A/F126A, and Y131A/K132A (11). These mutants had been previously shown to interact with TLR4, as judged by co-precipitation and/or FACS-based analyses, but did not support robust cellular endotoxin responsiveness (11). These results suggest either a defect in LPS binding (i.e. reactivity with endo-toxin⅐CD14) and/or activation of TLR4 by bound endotoxin⅐MD-2. Gel filtration chromatography analysis of medium from cells "spiked" with [ 3 H]LOS⅐sCD14 and producing the mutant MD-2s indicated only the K125A/F126A MD-2 reacted with [ 3 H]LOS⅐sCD14 to produce monomeric [ 3 H]LOS⅐MD-2 ( Fig. 2A). Even though MD-2 K121A/K122A and MD-2 Y131A/K132A were secreted in amounts comparable to MD-2 K125A/F126A , neither medium containing MD-2 F121A/K122A nor MD-2 Y131A/K132A (Fig. 2D) produced MD-2-dependent alteration of [ 3 H]LOS⅐sCD14 (Fig. 2, B and C), implying no interaction of these two MD-2 mutant species with E⅐sCD14.
Reaction of Human MD-2 F121A/K122A and MD-2 Y131A/K132A Mutants with LOS⅐sCD14 Depends on Co-expression of TLR4 Ectodomain-In many cells expressing MD-2, TLR4 is also expressed resulting in formation of MD-2/TLR4 heterodimer complex that reacts with monomeric E⅐CD14 to trigger cell activation (2,3,7,9,23,29). The functional stability (i.e. reactivity with E⅐CD14) of wild-type human MD-2 in serum-free medium at 37°C is markedly enhanced by coexpression of the predicted ectodomain of TLR4 (residues 24 -634) (23), raising the possibility that for certain mutant MD-2 species co-expression of the TLR4 ectodomain might be necessary to maintain MD-2 in a form reactive with E⅐sCD14. To test this possibility, the mutants of MD-2 that did not react with [ 3 H]LOS⅐sCD14 (i.e. MD-2 F121A/K122A and MD-2 Y131A/K132A ) and, for comparison, wild-type MD-2, were co-expressed and secreted with the TLR4 ectodomain into a culture medium "spiked" with [ 3 H]LOS⅐sCD14. After 24 h, the culture medium was harvested and analyzed by gel filtration chromatography. A Sephacryl S200 gel filtration system that resolves  (23). Co-expression of TLR4 ECD with either MD-2 F121A/K122A (Fig. 3B) or MD-2 Y131A/K132A (Fig. 3C) did not yield [ 3 H]LOS⅐MD-2, as expected (Fig. 2, B and C), but did yield ([ 3 H]LOS⅐MD-2/TLR4 ECD ) 2 . These findings indicate that MD-2 F121A/K122A and MD-2 Y131A/K132A react with [ 3 H]LOS⅐sCD14 but only when co-expressed and pre-associated with the ectodomain of TLR4.
The region encompassing residues 121-132 in MD-2 forms a rim surrounding the opening to a deep hydrophobic cavity in which the acyl chains of endotoxin are buried when bound to MD-2 (5,8,31,32). This region in human MD-2 contains several lysines (Lys 122 , Lys 125 , Lys 128 , Lys 130 , and Lys 132 ) that have been implicated in E-MD-2 interactions (8,11,13,14). To test whether the TLR4-dependence of the reactivity of MD-2 F121A.K122A and MD-2 Y131A/K132A reflected the effects of substitution of lysine 122 or 132, we expressed and tested the single mutants MD-2 K122A and MD-2 K132A . Strikingly, both MD-2 K122A and MD-2 K132A , albeit MD-2 K132A Ͻ MD-2 K122A , retained the ability to react with [ 3 H]LOS⅐sCD14 in a TLR4-independent manner as manifest by the generation of [ 3 H]LOS⅐MD-2 when these mutants were expressed and secreted into culture medium containing [ 3 H]LOS⅐sCD14 with (Fig. 3, B and C) or without TLR4 ECD (data not shown).
Because LOS⅐sCD14 is the substrate for both MD-2 and MD-2/TLR4 ECD , differences in TLR4-independent reactivity of different MD-2 species could affect the yield of ([ 3 H]LOS⅐MD-2/TLR4 ECD ) 2 by differentially affecting the availability of LOS⅐sCD14 to react with MD-2/TLR4 ECD . To eliminate the potentially complicating factor of formation of LOS⅐MD-2, we repeated co-expression of the various wt and mutant MD-2 species with TLR4 ECD in the absence of Role of Phe 121 and Tyr 131 in MD-2 in Activation of TLR4 by Bound Endotoxin-The ability of MD-2 F121A/K122A and MD-2 Y131A/K132A to react with LOS⅐sCD14 when co-expressed with TLR4 ECD suggested that the inability of these mutants to support TLR4-dependent cell activation by endotoxin might still reflect, at least in part, a defect in the ability of these MD-2 mutants to trigger TLR4 activation when bound to endotoxin and TLR4. To test this hypothesis, full-length TLR4 Ϯ wt or selected MD-2 mutants were expressed in HEK293 cells and cell binding of  (Fig. 4). Increased [ 3 H]LOS binding was also seen when MD-2 F121A/K122A or MD-2 Y131A/K132A were co-expressed with TLR4 but, as expected from earlier observations (11), these cells were not activated by LOS⅐sCD14 (Fig. 4). In contrast, cells co-expressing TLR4 and either MD-2 K122A or MD-2 K132A were activated by LOS⅐sCD14 (Fig. 4B), indicating that substitution of alanine for either Phe 121 or Tyr 131 in MD-2 reduces activation of TLR4 by endotoxin bound to MD-2.
Monomeric Complex of LOS⅐MD-2 F126A Is a Potent TLR4 Antagonist-The ability of a mutant MD-2 to bind both TLR4 and endotoxin without conferring TLR4 activation suggests that a complex of endotoxin with this MD-2 mutant should act as an effective TLR4 antagonist. Inability of MD-2 F121A/K122A and MD-2 Y131A/K132A to react with LOS⅐sCD14 unless these MD-2 mutants were pre-associated with TLR4 precluded testing this prediction with these mutants. However, the ability of MD-2 K125A/F126A to react efficiently with [ 3 H]LOS⅐sCD14 in the absence of TLR4 ( Fig. 2A) and to bind TLR4 without supporting robust TLR4-dependent cell activation by endotoxin (11) suggested that the LOS⅐MD-2 K125A/F126A mutant complex would act as a TLR4 antagonist rather than a TLR4 agonist. To test thishypothesis,weexaminedthedosedependent effects of the purified LOS⅐MD-2 K125A/F126A on HEK/ TLR4 cells, both alone and in the presence of wild-type LOS⅐MD-2. Fig.  5A shows that, in comparison to wild-type LOS⅐MD-2, LOS⅐MD-2 K125A/F126A produced much more limited activation of HEK/TLR4 cells and, in molar excess, reduced cell activation triggered by wild-type LOS⅐MD-2 to levels closely similar to that produced by the mutant complex alone (Fig. 5B). These findings strongly suggest that engagement of TLR4 by LOS⅐MD-2 K125A/F126A triggers only limited receptor activation and, in excess, inhibits activation of TLR4 by the wild-type complex.
To further define the importance of lysine 125 and/or phenylalanine 126 in MD-2 function, single site mutants of human MD-2 were produced and tested. As expected, both MD-2 K125A and MD-2 F126A reacted readily with [ 3 H]LOS⅐sCD14 to form  monomeric [ 3 H]LOS⅐MD-2 (data not shown). LOS⅐MD-2 K125A was as potent as wild-type LOS⅐MD-2 in activation of HEK/ TLR4 cells but LOS⅐MD-2 F126A produced essentially no activation of HEK/TLR4 cells (Fig. 6A). Instead, in molar excess, LOS⅐MD-2 F126A produced virtually complete dose-dependent inhibition of cell activation by wild-type LOS⅐MD-2 (Fig. 6B). Thus, the single amino acid alteration, F126A, in human MD-2 was sufficient to convert a potent TLR4 agonist (wild type LOS⅐MD-2) to a potent TLR4 antagonist.
High Affinity Interaction of LOS⅐MD-2 F126A with the Ectodomain of TLR4-The potency of LOS⅐MD-2 F126A as an antagonist of TLR4-dependent cell activation strongly suggested that this mutant complex retained ϳnormal binding properties toward the ectodomain of TLR4 where interactions with MD-2 and MD-2 complexed to endotoxin take place. To test this hypothesis more directly, we took advantage of the very high specific radioactivity of the purified LOS⅐MD-2 complexes (ϳ25,000 cpm/pmol) and gel filtration chromatography to measure specific, saturable, high affinity interactions between [ 3 H]LOS⅐MD-2 F126A and TLR4 ECD . Incubation of [ 3 H]LOS⅐MD-2 F126A with harvested culture medium containing secreted TLR4 ECD yielded a supra-molecular complex containing LOS, MD-2 and TLR4 ECD of M r ϳ 190,000, representing ([ 3 H]LOS⅐MD-2/TLR4 ECD ) 2 (Fig. 7, A and C) (23) that is identical to the product of the reaction of [ 3 H]LOS⅐sCD14 with wild-type MD-2/TLR4 ECD (Fig. 3, A and D) (23). This reaction was specific, saturable, and of high affinity with an apparent K d of ϳ200 pM (Fig. 7B), similar to that between wild-type LOS⅐MD-2 and TLR4 ECD (23). Thus, interactions of TLR4 ECD with wild-type LOS⅐MD-2 and LOS⅐MD-2 F126A were virtually identical, as judged both by the product formed (Fig. 7A) and apparent K d (Fig. 7B), consistent with the ability of LOS⅐MD-2 F126A to act as a potent TLR4 antagonist.

DISCUSSION
We have described in this study the application of novel experimental approaches to address the identification of specific structural determinants in MD-2 for activation of TLR4 by bound E⅐MD-2 and the role of TLR4 in the transfer of endotoxin from CD14 to MD-2. These approaches have led to several new insights concerning the structure and function of human MD-2. First, our findings indicate that the structural requirements in human MD-2 for transfer of E from CD14 to MD-2 are more stringent when MD-2 is presented as a soluble extracellular protein in the absence of TLR4 than when pre-associated with the ectodomain of TLR4 (Fig. 3). Several mutants of MD-2 are described (i.e. F121A/ K122A, Y131A/K132A, and K132A) that retain much greater reactivity with LOS⅐sCD14 when co-expressed with the TLR4 ectodomain (Fig. 3, B and C) or full-length TLR4 (Fig. 4A)   MD-2 K132A co-expressed with full-length TLR4 supported potent cell activation by LOS⅐sCD14 (Fig. 4B), presumably reflecting reactivity of cell surface MD-2 K132A /TLR4 with LOS⅐sCD14.
Second, we demonstrate that several aromatic residues situated either just inside the opening of the hydrophobic cavity in MD-2 (i.e. Phe 121 and Tyr 131 ; (32); Fig. 8) or extending outward from the rim of this cavity (Phe 126 ; Fig. 8), are important for activation of TLR4 by bound E⅐MD-2. This role was demonstrated most clearly for Phe 126 but also strongly suggested for Phe 121 and Tyr 131 , where substitution with alanine nearly ablated cell activation by endotoxin (Fig. 4B) (11) despite main-tenance of MD-2 binding to TLR4 (11) (data not shown) and significant reactivity of MD-2/TLR4 ECD with LOS⅐sCD14 (Figs. 3 and 4A). Our findings extend earlier observations by other investigators who showed a role for each of these three residues in cellular responsiveness to endotoxin (9) and, in the case of phenylalanine 126, a role in receptor clustering induced by endotoxin binding to MD-2/TLR4 (15). The conservation of Phe 121 , Phe 126 , and Tyr 131 in almost all of the mammalian MD-2 species reported to date (15) is consistent with a crucial role for each of these three aromatic residues in cellular responsiveness to endotoxin.
Third, we have demonstrated that simply by substituting phenylalanine 126 with alanine, we could produce an E⅐MD-2 complex that was a potent TLR4 antagonist, despite the fact that the bound endotoxin was normally a potent activator of TLR4. Previous studies have demonstrated contrasting agonist versus antagonist effects of tetraacylated lipid A in mouse and human cells, respectively. These contrasting species-dependent effects are due, substantially, to discrete structural differences between murine and human MD-2 (10,12,18,(33)(34)(35), indicating that activation or blocking of activation of TLR4 can be determined not only by the structural properties of endotoxin/lipid A bound by MD-2 but also by the structure of MD-2 itself. The properties we have described for human MD-2 F126A support this view and demonstrate for the first time that a single amino acid substitution in MD-2 is sufficient to convert a potent TLR4 agonist (hexa-acylated LOS⅐MD-2 wt ) to a potent TLR4 antagonist (hexa-acylated LOS⅐MD-2 F126A ). This finding underscores the remarkably discrete structural variables that distinguish TLR4 agonists from TLR4 antagonists and demonstrates that what specifies TLR4 agonist properties is not the structural properties of endotoxin alone but rather that of the monomeric endotoxin⅐MD-2 complex.
MD-2 mutants defective in endotoxin binding have been previously described (8,(12)(13)(14). These studies have relied on assays of recombinant MD-2 either expressed in and purified from bacteria or from conditioned culture medium of insect cells or present in impure form in harvested conditioned medium from mammalian cells (e.g. HEK293 cells) (8,(12)(13)(14). MD-2-endotoxin interactions have been measured by assays of protein interaction to immobilized LPS and/or by co-precipitation with biotinylated LPS. Previously, we have emphasized that the very high affinity (pM) interactions demonstrable between endotoxin and MD-2 or MD-2/TLR4 ECD require presentation of endotoxin as a monomeric complex with CD14 (2,7,23), paralleling the molecular requirements for potent MD-2/TLR4-dependent cell activation by most endotoxin species (2,7,23,36,37). The functional relevance of direct MD-2-endotoxin interactions and the composition of the product(s) of  these interactions require further study. The experimental approaches described in this study may be used to juxtapose the structural requirements in MD-2 and MD-2/TLR4 ECD for reaction with endotoxin aggregates and monomeric E⅐CD14.
The functional instability of secreted MD-2 in the absence of the TLR4 ectodomain adds an additional layer of complexity to interpretations of the effects of changes in MD-2 structure on MD-2 function. By spiking the cell culture medium with [ 3 H]LOS⅐sCD14 (25,000 cpm/pmol), we could measure transfer of [ 3 H]LOS from [ 3 H]LOS⅐sCD14 to secreted MD-2 before inactivation of MD-2 and thus assay directly the reactivity of ng amounts of wild-type and mutant MD-2 with LOS⅐sCD14 in the absence of TLR4. In the case of wild-type MD-2 and certain mutant MD-2 species (e.g. K122A, K125A, F126A), the ratio of the products ([ 3 H]LOS⅐MD-2/([ 3 H]LOS⅐MD-2/TLR4 ECD ) 2 ) formed roughly corresponded to the ratio of expression and secretion of MD-2 and TLR4 ECD , i.e. ratio of MD-2 to MD-2/ TLR4 ECD , as determined by immunoblot (data not shown), suggesting that for these secreted MD-2 species the reactivity of MD-2 with LOS⅐sCD14 is not significantly altered by prior engagement of MD-2 with TLR4 ECD . In contrast, for MD-2 F121A.K122A , MD-2 Y131A.K132A , and MD-2 K132A , interaction with TLR4 ECD was apparently crucial for reactivity of these MD-2 species with LOS⅐sCD14. Whether or not binding of MD-2 to TLR4 ECD in these MD-2 mutants is needed to induce a conformational change in MD-2 or simply to preserve longer the functionally active conformation of MD-2 is not known. MD-2, when secreted in molar excess of TLR4, can form an array of oligomers as well as persist as a monomer (27,38). 3 It is generally believed that the functionally reactive form of MD-2, both with respect to endotoxin and TLR4 binding, is monomeric MD-2 (20,27) and that the state of the resting MD-2/ TLR4 heterodimer (i.e. before binding of endotoxin) may also be monomeric (14,18,39,40). Thus, it is possible that, in the case of MD-2 F121A.K122A , MD-2 Y131A.K132A , and MD-2 K132A , the functional half-life of monomeric MD-2, unless co-expressed with full length TLR4 or TLR4 ectodomain , is too short to have a chance to react with extracellular LOS⅐sCD14. The orientation of the side chains of Phe 121 and Tyr 131 , as revealed in the crystal structure of human MD-2 (32) (Fig. 8), suggest that these aromatic side chains may be needed to increase the stability of monomeric MD-2, in the absence of TLR4, by partially shielding the opening to the hydrophobic cavity.
In contrast, the normal (wild-type) reactivity of MD-2 F126A with endotoxin⅐CD14 Ϯ TLR4 suggests that Phe 126 has no essential role in maintaining MD-2 in monomeric form in the absence of endotoxin and TLR4. Instead, the orientation of the side chain of Phe 126 (references in Fig. 8) suggests a more likely role in intermolecular protein-protein interactions, such as between MD-2 and MD-2 or MD-2 and TLR4, that may be induced when MD-2 is simultaneously engaged with activating endotoxin species and TLR4 (7,14,18,25,39,40). One possibility is that occupation of the hydrophobic cavity of MD-2 by the multiple acyl chains of lipid A of endotoxin species that are normally potent TLR4 agonists induces displacement of side chains within the hydrophobic cavity (e.g. Phe 121 and Tyr 131 ) and, secondarily, Phe 126 leading to the downstream proteinprotein interactions needed for signal/transduction. The fact that LOS⅐MD-2 wt /TLR4 ECD and LOS⅐MD-2 F126A /TLR4 ECD are structurally indistinguishable by gel sieving (Figs. 3 and 7) and immunocapture (data not shown) analyses suggest more subtle conformational differences and/or a role of other (intramembrane, cytosolic) regions of TLR4 or other cellular proteins in determining the agonist versus antagonist action of wt and variant E⅐MD-2 complexes. The remarkably stable and soluble properties of monomeric LOS⅐MD-2 wt and LOS⅐MD-2 F126A , and of LOS⅐MD-2 wt or F126A /TLR4 ECD , in contrast to MD-2 alone, should make these complexes valuable reagents for better defining the structural basis of TLR4 activation by bound E⅐MD-2.