NMR Studies of Hexaacylated Endotoxin Bound to Wild-type and F126A Mutant MD-2 and MD-2·TLR4 Ectodomain Complexes*

Background: Phenylalanine 126 of MD-2 is essential for endotoxin-induced TLR4 activation. Results: NMR of 13C-labeled endotoxin bound to wt or F126A MD-2 ± TLR4 ectodomain reveals effects of Phe126 on interactions of MD-2 ± TLR4 with endotoxin. Conclusion: Phe126 acts as a “hydrophobic switch” promoting agonist-dependent TLR4 dimerization. Significance: This study describes a novel approach to further define the structural requirements for endotoxin-induced TLR4 activation. Host response to invasion by many Gram-negative bacteria depends upon activation of Toll-like receptor 4 (TLR4) by endotoxin presented as a monomer bound to myeloid differentiation factor 2 (MD-2). Metabolic labeling of hexaacylated endotoxin (LOS) from Neisseria meningitidis with [13C]acetate allowed the use of NMR to examine structural properties of the fatty acyl chains of LOS present in TLR4-agonistic and -antagonistic binary and ternary complexes with, respectively, wild-type or mutant (F126A) MD-2 ± TLR4 ectodomain. Chemical shift perturbation indicates that Phe126 affects the environment and/or position of each of the bound fatty acyl chains both in the binary LOS·MD-2 complex and in the ternary LOS·MD-2·TLR4 ectodomain complex. In both wild-type and mutant LOS·MD-2 complexes, one of the six fatty acyl chains of LOS is more susceptible to paramagnetic attenuation, suggesting protrusion of that fatty acyl chain from the hydrophobic pocket of MD-2, independent of association with TLR4. These findings indicate that re-orientation of the aromatic side chain of Phe126 is induced by binding of hexaacylated E, preceding interaction with TLR4. This re-arrangement of Phe126 may act as a “hydrophobic switch,” driving agonist-dependent contacts needed for TLR4 dimerization and activation.


Host response to invasion by many Gram-negative bacteria depends upon activation of Toll-like receptor 4 (TLR4) by endotoxin presented as a monomer bound to myeloid differentiation factor 2 (MD-2). Metabolic labeling of hexaacylated endotoxin (LOS) from Neisseria meningitidis with [ 13 C]acetate allowed the use of NMR to examine structural properties of the fatty acyl chains of LOS present in TLR4-agonistic and -antagonistic
binary and ternary complexes with, respectively, wild-type or mutant (F126A) MD-2 ؎ TLR4 ectodomain. Chemical shift perturbation indicates that Phe 126 affects the environment and/or position of each of the bound fatty acyl chains both in the binary LOS⅐MD-2 complex and in the ternary LOS⅐MD-2⅐TLR4 ectodomain complex. In both wild-type and mutant LOS⅐MD-2 complexes, one of the six fatty acyl chains of LOS is more susceptible to paramagnetic attenuation, suggesting protrusion of that fatty acyl chain from the hydrophobic pocket of MD-2, independent of association with TLR4. These findings indicate that re-orientation of the aromatic side chain of Phe 126 is induced by binding of hexaacylated E, preceding interaction with TLR4. This re-arrangement of Phe 126 may act as a "hydrophobic switch," driving agonist-dependent contacts needed for TLR4 dimerization and activation.
Recognition of endotoxin (E), 2 unique glycolipids present on the surface of Gram-negative bacteria (GNB), is pivotal for host responses to many GNB (1,2). The ability to respond to minute (pM) concentrations of E depends on the ordered action of four extracellular and cell surface host proteins: lipopolysaccharidebinding protein (LBP), soluble (s) and GPI-linked membrane (m)-associated forms of CD14, secreted and TLR4-associated MD-2 and TLR4 (1,(3)(4)(5). Among the various TLRs, TLR4 plays the major role in recognition and response to E and is unique in that its activation leads to both MyD88-dependent (NFB-mediated) and TRIF-dependent (interferon-mediated) cellular responses (6 -11). The concerted action of LBP, CD14 and MD-2 dramatically alters the physical presentation of E, such that individual E monomers are extracted from GNB or purified E aggregates to form monomeric E⅐protein complexes that, at pM E concentrations, indirectly (E⅐CD14 with MD-2⅐TLR4) or directly (E⅐MD-2 with TLR4) engage TLR4 and cause receptor and cell activation (3,(12)(13)(14).
Recent biochemical and structural studies have significantly advanced understanding of the molecular and structural basis of ligand-receptor interaction, and how this interaction results in TLR4 activation (3,(12)(13)(14)(15)(16)(17). MD-2 forms a heterodimer with TLR4 by agonist-independent binding to sites within the N-terminal and central leucine-rich repeat (LRR) domains of the TLR4 ectodomain (TLR4ecd) (15,17). Comparison of the published crystal structures of MD-2 bound to TLR4 antagonists (lipid IVA and eritoran) to that of the agonist-driven heterooligomer, [TLR4ecd⅐MD-2⅐LPS] 2 has revealed that binding of TLR4-activating LPS to MD-2 induces additional agonist-dependent contacts of MD-2 and LPS with TLR4 involving sites within the C-terminal LRR domain of the ectodomain of a second TLR4 molecule (15)(16)(17). When tetraacylated antagonists, such as lipid IVA and eritoran, are bound to MD-2, each of the four tightly packed fatty acyl chains fits within the deep, hydrophobic cavity of MD-2 and is protected from the aqueous solvent (15,16). In contrast, when hexaacylated LPS is bound to the MD-2⅐TLR4ecd heterodimer, the tail of one fatty acyl chain of LPS protrudes from the pocket of MD-2 of one ternary com-plex leading to contacts between this fatty acyl chain and TLR4ecd of a second ternary complex (15,17).
An apparently critical determinant for agonist (i.e. hexaacylated lipid A/endotoxin)-dependent activation of TLR4 is the presence of an aromatic or hydrophobic amino acid, most often phenylalanine 126, at the mouth of the hydrophobic cavity of MD-2 (18 -20). F126A mutants of both human and murine MD-2 do not support endotoxin-dependent activation of TLR4 (21)(22)(23). Monomeric complexes of endotoxin⅐human (h)MD-2 wild-type (wt) and endotoxin⅐hMD-2 F126A display the same high affinity (K d ϳ100 -200 pM) for TLR4, but the mutant complex has Յ 1% the TLR4 agonist activity of the wt complex and is a TLR4 antagonist (12,13,22). Phe 126 in hMD-2 clearly plays an essential role in agonist-dependent contacts between two ligand⅐MD-2⅐TLR4 ternary complexes. Comparison of the crystal structures of MD-2 complexed with non-activating ligands (lipid IVA, eritoran, myristic acid) with the crystal structure of [TLR4ecd⅐MD-2⅐LPS] 2 revealed a re-arrangement of the Phe 126 -containing loop ("Phe 126 loop") of MD-2 apparently driven by re-orientation of the Phe 126 aromatic side chain from outside the pocket when bound to a non-activating ligand, to inward toward the hydrophobic cavity when activating hexaacylated LPS was bound (15)(16)(17). As suggested by Park et al. (17), this local conformational re-arrangement in MD-2 involving Phe 126 could be critical for LPS-triggered TLR4 activation by influencing the position of the fatty acyl chains of bound activating LPS and fatty acyl:TLR4 contacts that drive TLR4 dimerization and activation. In addition, the re-arrangement of the Phe 126 loop could promote MD-2:TLR4 contacts at the dimerization interface.
Because the structure of a potent TLR4-activating endotoxin bound to MD-2 in the absence of TLR4 has not been reported, it is not known whether the inability of the fatty acyl chains of hexaacylated agonist lipid A to fit completely into the MD-2 pocket helps drive dimerization of LPS⅐MD-2⅐TLR4 or whether contact with the second TLR4 stabilizes the extrusion of the sixth fatty acyl chain and subsequent conformational changes in MD-2. It also is not clear if agonist-induced re-orientation of Phe 126 in MD-2 is triggered by the binding of a TLR4-activating lipid A/endotoxin to MD-2 independent of TLR4 or if binding to TLR4 is required to stabilize this altered conformation.
We now report NMR studies of the acyl chain properties of hexaacylated endotoxin bound to wt or mutant F126A hMD-2 and as part of a monomeric ternary complex of E⅐MD-2, wt or F126A⅐TLR4ecd. These studies were designed to address two questions: 1) Are protrusion of a single fatty acyl chain and re-orientation of the aromatic side chain of Phe 126 intrinsic properties of TLR4-activating hexaacylated E⅐MD-2 complexes induced before contact with TLR4? 2) Does Phe 126 of MD-2 affect positioning of the acyl chains of bound hexaacylated E that may drive contacts with TLR4 at the dimerization interface leading to TLR4 activation? Key to these studies was our ability to metabolically label uniformly and efficiently the acyl chains of endotoxin with alternating [ 13 C] and [ 12 C] atoms and generate mg quantities of purified monomeric, functional, and stable endotoxin⅐MD-2, TLR4ecd, and endotoxin⅐MD-2 wt and F126A⅐TLR4ecd complexes. Results of the NMR studies indicate that protrusion of one of the six fatty acyl chains of endo-toxin bound to MD-2 precedes interaction with TLR4 when endotoxin is bound to either wt or F126A MD-2. Consequently, the presence of a protruding fatty acyl chain is not a distinguishing feature of TLR4-activating E⅐MD-2 complexes and so is not sufficient for driving TLR4 activation. However, the position and surrounding environment of the bound fatty acyl chains in MD-2 are altered when Phe 126 is substituted with alanine. In the endotoxin⅐MD-2 wt complex, the aromatic side chain of Phe 126 is re-oriented toward the hydrophobic cavity preceding interaction with TLR4. Local rearrangements of the Phe 126 loop in MD-2 may promote TLR4 dimerization and activation, by affecting both contacts between the protruded fatty acyl chain and TLR4 as well as MD-2:TLR4 contacts.  2D) and had sufficient radioactivity (15 cpm/pmol LOS) to monitor purification and facilitate quantitative analysis. To minimize effects of possible endotoxin heterogeneity, a single preparation of purified [ 13/14 C]-LOS was used. [ 3 H]LOS⅐MD-2 (25,000 cpm/pmol) was prepared as described (13,24). Reagents used include: human serum albumin (HSA), an endotoxin-free, 25% stock solution (Baxter Health Care, Glendale, CA), bovine serum albumin (BSA), and other chemical reagents from Sigma, and chromatography matrices (GE Healthcare, Piscataway, NJ). Purified Gd(DPTA-BMA) was a gift from Dr. Klaus Zangger, University of Graz, Graz, Austria.

EXPERIMENTAL PROCEDURES
Production of Recombinant WT and F126A hMD-2 and TLR4 Ectodomain-As previously described (12,13), cDNA encoding human MD-2 wt, F126A, or TLR4 ectodomain, amino acids 24 -631, was inserted using SacII and XhoI restriction sites into pBAC3 (Novagen) that provides secreted proteins with a six-residue polyhistidine tag at the N terminus. Baculoviral stocks, generated by transfection of BacVec3000 (Novagen) and plasmid and then amplified in Sf9 cells, were used to infect High Five™ (Invitrogen) insect cells for protein production. Large scale (20 liters) preparations of conditioned insect medium containing MD-2 (wt or F126A) or TLR4ecd were produced by BlueSky Biotech, Worcester, MA. Recombinant proteins in conditioned media were stable at Ϫ80°C.
Preparation of LOS⅐MD-2 Complexes-[ 13/14 C]LOS⅐albumin complexes were generated and purified from [ 13/14 C]LOS aggregates solubilized in buffer A (100 mM Tris-HCl 5 mM EDTA, pH 8.0) as previously described (26). LOS⅐albumin complexes (1 mg LOS, 30 ml) were diluted 10-fold with buffer A and incubated with 2 liters of conditioned insect medium containing His 6 -sMD-2, wt or F126A (ϳ1 g/ml active, monomeric sMD-2) for 3 h at 37°C. An aliquot of the reaction mixture was analyzed on Sephacryl S200 to measure generation of [ 13/14 C]-LOS⅐MD-2. The reaction mixture was dialyzed (Spectrapor1) against buffer B (20 mM phosphate, 0.5 M NaCl, 5 mM imidazole, pH 7.6) and loaded onto Ni FF Sepharose (5 cm ϫ 25 cm) at 4°C at a flow rate of 3 ml/min (GE Healthcare Explorer FPLC). After washing the column with 20 mM imidazole in buffer B, the adsorbed material was eluted by an imidazole gradient in buffer B. Gradient fractions containing radioactive LOS were combined, concentrated to ϳ2 ml (Millipore Centriplus-70), and applied to Sephacryl S100 (1.6 ϫ 100 cm) equilibrated in buffer C (20 mM phosphate, 150 mM NaCl, pH 7.1). Radioactive fractions consistent with the M r of LOS⅐MD-2 were combined and concentrated to ϳ1 ml (Millipore Centricon MWCO 10K). Purity of the samples was confirmed by Coomassie Blue stain of 10 -15% gradient PhastGels (GE Healthcare). Each LOS⅐MD-2 complex analyzed by NMR was derived from reactions of 12-14 liters of conditioned medium containing MD-2. The final samples (1 ml) contained 240 and 260 M of LOS⅐MD-2 wt and F126A, respectively. [ 12 C]LOS⅐MD-2 was prepared by the identical procedure described above.
The functional activity of purified LOS⅐MD-2 complexes was measured by TLR4-dependent cell activation. HEK293 cells, parent, or TLR4-containing, were seeded in a 96-well plate (1 ϫ 10 5 cells/well) in DMEM/10% FBS and incubated overnight at 37°C in 5% CO 2 /95% humidity. The next day, cells were washed twice with PBS, pH 7.4. Increasing doses of purified LOS⅐MD-2 wt and F126A were added in 200 l of DMEM/0.1% HSA. After 18 h at 37°C, cell supernatants were collected and tested for accumulation of extracellular IL-8 by ELISA (BD Biosciences).
LOS⅐MD-2 wt and F126A complexes were separated from Gd(DPTA-BMA) used in PRE NMR experiments on Sephacryl S100. The purified LOS⅐MD-2 complexes were incubated with 1.2ϫ molar excess of HisTLR4ecd in 900 l for 30 min at 37°C. An aliquot was analyzed on Sephacryl S300 to confirm complete conversion of [ 13/14 C]LOS⅐MD-2 complex to a ternary complex containing TLR4ecd.
NMR Spectroscopy-All NMR samples were exchanged into 10 mM sodium phosphate, pH 6.5, in 100% D 2 O. The LOS⅐MD-2 complexes (ϳ250 M) were stable at 4°C for at least 6 months as judged by NMR profile and size exclusion chroma-tography. Aliquots of the LOS⅐MD-2⅐TLR4ecd samples were chromatographed after NMR to check for stability.
NMR spectra were collected at 25°C on a Bruker Avance II 800 MHz NMR spectrometer equipped with a sensitive TCI cyroprobe. High resolution 13 C/ 1 H HSQC spectra (28) of the 13 CH 3 region were collected to provide maximal resolution of the individual LOS 13 CH 3 groups. Relative solvent/surface exposure of individual LOS 13 CH 3 groups in LOS⅐MD-2 complexes was examined using the PRE NMR method (29 -31) by acquiring 1 H T 2 relaxation rates of the LOS 13 CH 3 groups of the [ 13 C]LOS⅐MD-2 complexes in the absence and presence of a neutral chelating gadolinium paramagnetic agent Gd(DPTA-BMA), 2-16 mM, as previously described (29 -31). The 1 H T 2 Carr-Purcell-Meiboom-Gill (CMPG) delays used in these experiments were 0.92, 14.71, and 27.58 ms. NMR spectra were processed with the NMRPipe package (32) and analyzed using NMRView software (33). C]acetate) were added to the growth medium to yield sufficient radiolabeling of LOS (15 cpm/pmol) to facilitate quantitative monitoring of purification and recovery of desired products. All of the experiments described here were done with a single preparation of purified LOS to preclude sample-to-sample differences due to endotoxin heterogeneity.

Production and Purification of [ 13 C]LOS⅐MD-2-To
Monomeric LOS⅐MD-2 complexes are most efficiently generated by LBP-catalyzed extraction and transfer of LOS monomers from LOS aggregates to sCD14 and subsequent transfer of monomeric LOS from sCD14 to (s)MD-2 (3,12,26). The amounts of [ 13 C]LOS⅐MD-2 needed for NMR analysis would have necessitated production of at least 25 mg of recombinant sCD14. Therefore, we took advantage of an alternative method we have recently developed for generation of LOS⅐MD-2 in high yield independent of (s)CD14 (26). This alternative method depends upon generation of monomeric endotoxin⅐albumin complexes during overnight incubation of E aggregates in a divalent cation-depleted environment supplemented with albumin. The resultant E⅐albumin complexes are efficient donors of E monomers to insect cell-derived His 6 -MD-2 as needed for formation of monomeric E⅐MD-2 (26). Reaction of monomeric LOS⅐albumin complexes with insect cell conditioned medium containing His 6 -MD-2 was followed by metal chelation chromatography (Fig. 1A) and size exclusion chromatography (Fig. 1B) to yield purified monomeric (Fig. 1, B and C), bioactive (Fig. 1D) LOS⅐MD-2; nearly 50% of LOS in the LOS⅐albumin complexes was recovered as LOS⅐MD-2. As seen in Fig. 1A, an imidazole gradient separated LOS⅐MD-2 from the bulk of insect protein(s) that also adhered to Ni FF-Sepharose. The identical procedure produced in similar yield and purity LOS⅐MD-2 F126A when insect cell conditioned medium containing His 6 -MD-2 F126A was used. In contrast to LOS⅐MD-2 wt, the purified LOS⅐MD-2 F126A complex did not activate HEK/TLR4 cells (Fig. 1D) and acted as a TLR4 antagonist (data not shown).

NMR Analyses of WT and Mutant [ 13 C]LOS⅐MD-2-
High resolution 13 C/ 1 H HSQC spectra acquired on a 800 MHz NMR spectrometer for the 13 CH 3 region of [ 13 C]LOS complexed to wt and F126A MD-2 provided optimal resolution of the six acyl chain 13 CH 3 groups (labeled as M1-M6) (Fig. 2, A and B). Interestingly, M1-M4 signals are clustered together, while M5 and M6 formed a separate cluster, reflecting the distinct local environment surrounding each 13 CH 3 group of LOS when bound to MD-2. To ascertain that these crosspeaks were all derived from [ 13 C]LOS and not from 13 C natural abundance signals of MD-2, [ 12 C]LOS⅐MD-2 wt complex was prepared. NMR spectra collected under identical conditions for this control sample showed that there were virtually no detected crosspeaks in this region (data not shown), thus confirming that the observed signals in Fig. 2, A and B were derived specifically from the fatty acyl chains of [ 13 C]LOS bound to MD-2. Overlay of the spectra of the 13 CH 3 region of [ 13 C]LOS⅐MD-2 wt and F126A revealed significant chemical shift perturbation of all LOS 13 CH 3 groups with the F126A mutation ( Fig. 2C and supplemental Table S1). When the high resolution HSQC spectra of these complexes were plotted at a lower contour level, additional crosspeaks likely corresponding to minor forms of bound LOS were observed. (supplemental Fig. S1, A and B. These minor forms could be due to a reverse orientation of the lipid A backbone of the bound pseudosymmetric LOS (Fig. 2D) as suggested by recent molecular modeling studies of LOS⅐MD-2 complexes (34). Overlay of these lower contour spectra revealed that the peak positions for these minor forms were also significantly different between these two complexes, i.e. they experienced significant chemical shift perturbation by the F126A mutation of MD-2 (supplemental Fig. S1C). Phe 126 is located in a loop adjacent to the entry of the LOS-binding pocket on MD-2. This loop has a large B-factor (17) and is mobile as shown by molecular dynamics simulation (34). The Phe 126 side chain is completely exposed when tetraacylated LOS binds to MD-2 (an inactive complex) (15,16). The observed significant chemical shift perturbation of the hexaacylated LOS 13 CH 3 groups upon F126A mutation strongly suggests that the Phe 126 side chain has reoriented toward the bound LOS to interact with the fatty  Comparison of the crystal structures of hexaacylated LPS⅐MD-2⅐TLR4ecd and tetraacylated lipid A⅐MD-2⅐TLR4ecd complexes revealed that in the complex containing activating hexaacylated LPS, one of the six fatty acyl chains partially protrudes from the hydrophobic cavity of MD-2, permitting fatty acyl:TLR4 contacts that are apparently important in TLR4 dimerization and activation (15)(16)(17). Our ability to identify by high resolution 13 C/ 1 H HSQC the crosspeaks in the methyl region of LOS⅐MD-2 wt and F126A complexes permitted us to probe by paramagnetic relaxation enhancement (PRE), whether a single, more surface exposed fatty acyl chain is: 1) a distinguishing characteristic of TLR4-activating endotoxin (LOS)⅐MD-2 complexes manifest before interaction with and dimerization of TLR4 and/or 2) dependent on Phe 126 of MD-2.
The relative surface exposure of the individual 13 CH 3 groups was determined using PRE by measuring the 1 H T 2 relaxation rates of the LOS 13 CH 3 groups in the presence and absence of a neutral chelated paramagnetic gadolinium compound Gd(DPTA-BMA) (Fig. 3A) (29 -31). Perturbation by Gd(DPTA-BMA) reflects the relative solvent/surface accessibility of the individual fatty acyl chains (i.e. CH 3 groups) in LOS⅐MD-2 wt and F126A complexes as the less sequestered groups will be more readily affected by the paramagnetic reagent (30,35). In both [ 13 C]LOS⅐MD-2 wt and F126A complexes, five of the 13 C/ 1 H LOS methyl crosspeaks (M2-M6) showed similar PRE rates (Fig. 3B) indicating that the methyl groups have similar distances from their respective atomic coordinates to the protein surface of MD-2. However, one (M1) methyl signal in each complex had more significant PRE effects ( Fig. 3B) suggesting that the 13 CH 3 group of this fatty acyl chain was less sequestered and more solvent/surface accessible, i.e. protruding out of the hydrophobic pocket of both wt and F126A MD-2, independent of LOS⅐MD-2 association with TLR4. These findings suggest that protrusion of one of the six fatty acyl chains is a feature of monomeric complexes of hexaacylated lipid A/endotoxin (LOS) with MD-2 that precedes interactions of this complex with TLR4. The apparent similarity of this property in LOS⅐MD-2 wt and F126A complexes suggests that protrusion of a single fatty acyl chain of LOS bound to MD-2 is not dependent on Phe 126 and so not sufficient for TLR4 activation.
Monomeric Ternary Complexes of [ 13 C]LOS⅐MD-2⅐TLR4ecd-According to current theory, TLR4 activation by hexaacylated endotoxin is driven by dimerization of two hexaacylated lipid A/endotoxin⅐MD-2⅐TLR4 ternary complexes. The corresponding monomers of the TLR4-activating ternary complexes must have distinguishing structural features from ternary complexes that do not trigger TLR4 dimerization and activation (e.g. LOS⅐MD-2 F126A⅐TLR4). To test if these distinguishing structural features include differences in the positioning of the fatty acyl chains of bound hexaacylated endotoxin in the ternary complex, the same LOS⅐MD-2 wt and F126A complexes analyzed by NMR were used to produce and purify ternary complexes with soluble TLR4ecd for subsequent NMR analysis. TLR4ecd was expressed in insect cells as a stable, secreted N-terminal His 6 -tagged protein and purified from conditioned insect medium by a combination of metal chelation and size exclusion chromatography. The latter chromatographic step showed that the purified TLR4ecd was exclusively a monomer (Fig. 4A). Reaction of TLR4ecd with either LOS⅐MD-2 complex described above resulted in the generation of a radiolabeled complex that migrated with M r ϳ100,000 as determined by size exclusion on Sephacryl S300 consistent with a monomer of the trimer LOS⅐MD-2⅐TLR4ecd (Fig. 4B). The concentration and functional activity of TLR4ecd was monitored analytically by size exclusion chromatography throughout the purification to assess reactivity with radiolabeled [ 3 H]LOS⅐MD-2 (25,000 cpm/pmol). Nearly complete conversion of [ 3 H]LOS⅐MD-2, wt or F126A, to [ 3 H]LOS⅐MD-2⅐TLR4ecd was observed following the reaction of 0.8 -1 nM [ 3 H]LOS⅐MD-2 with the purified TLR4ecd, indicating pM reactivity of monomeric TLR4ecd with monomeric endotoxin⅐MD-2 (Fig. 4B).
Prior to reaction with [ 13 C]LOS⅐MD-2, 13 C/ 1 H HMQC, and HSQC NMR spectra were collected on the concentrated sample of TLR4ecd to ensure that any crosspeaks detected due to the 13 C natural abundance of the purified TLR4ecd did not overlap with [ 13 C]LOS signals. As shown in supplemental Fig.  S2, the crosspeaks derived from both the methyl (CH 3 ) and methylene (CH 2 ) groups of [ 13 C]LOS were clearly resolved from signals due to 13 C natural abundance of the purified TLR4ecd.
A molar excess (1.2ϫ) of concentrated TLR4ecd was incubated with [ 13/14 C]LOS⅐MD-2 (wt and F126A) for 30 min at 37°C. Complete conversion of [ 13/14 C]LOS⅐MD-2 to [ 13/14 C]-LOS⅐MD-2⅐TLR4ecd was verified by Sephacryl S300 size exclusion chromatography (Fig. 4C). Comparison of 13 C/ 1 H HSQC spectra of the binary [ 13 C]LOS⅐MD-2 and ternary [ 13 C]LOS⅐MD-2⅐TLR4ecd complexes containing wt or F126A MD-2 (Fig. 5, A and  B) revealed a significant chemical shift perturbation upon TLR4ecd binding for all 13 CH 3 groups of LOS bound to MD-2. More severe and selective line broadening was observed for LOS M1-M4 13 CH 3 groups, which includes the 13 CH 3 group M1 that protrudes out of the LOS-binding pocket of MD-2. Overlay of the 13 C/ 1 H high resolution HSQC spectra of the two ternary complexes (Fig. 5C) showed chemical shift perturbation of the LOS 13 CH 3 groups by the F126A mutation of MD-2 in the LOS⅐MD-2 F126A⅐TLR4ecd ternary complex, consistent with the essential role of Phe 126 in the TLR4-activating properties of the LOS⅐MD-2⅐TLR4 complex.
Comparison of 13 C/ 1 H HMQC NMR spectra of the binary [ 13 C]LOS⅐MD-2 versus the ternary [ 13 C]LOS⅐MD-2 wt⅐TLR4ecd complexes showed a remarkable broadening of the    signals derived from the 13 CH 2 (methylene) groups of LOS fatty acyl chains, especially in the ternary complex containing wt MD-2 ( Fig. 5D and supplemental Fig. S3). These changes were not due to dissociation of [ 13 C]LOS from the ternary complex or multimerization of these complexes. After NMR analyses, the complexes were re-examined by size exclusion chromatography. The post-NMR and pre-NMR elution profiles were identical for all complexes. These subtle differences in line broadening of the LOS-derived 13 CH 2 signals in the wt versus F126A MD-2 containing ternary complexes could reflect differential ligand dynamics in these complexes due to the mutation of Phe 126 .

DISCUSSION
Lipid A acyl chains of endotoxin bound to MD-2 were analyzed by NMR using unique reagents comprised of TLR4-activating or -inhibiting endotoxin⅐MD-2 complexes containing 13 C-labeled endotoxin bound either to wt or F126A MD-2, respectively. [ 13 C]Endotoxin, in which each of the lipid A fatty acyl chains of all LOS molecules contained a terminal 13 CH 3 group followed by alternating [ 12 C] and [ 13 C] atoms, was obtained by metabolic labeling of endotoxin (LOS) with [1-12 C], [2-13 C]acetate. The bacterial mutant (NMB Ace1) utilized labeled acetate added to the growth medium as the source of essentially all acetyl precursors needed for fatty acid synthesis (24,25). The stability of sMD-2 in insect cell medium and its reactivity with monomeric endotoxin⅐albumin complexes permitted the generation of monomeric endotoxin⅐MD-2 complexes in high yield (26). The remarkable stability of these monomeric endotoxin⅐MD-2 complexes allowed the purification and subsequent concentration of the complexes as needed for NMR analyses. NMR revealed discrete differences in the 13 C/ 1 H high resolution HSQC spectra of the wt and F126A mutant binary LOS⅐MD-2 and ternary LOS⅐MD-2⅐TLR4ecd complexes. These spectra were monitored to assess the microenvironment of the fatty acyl chains upon engagement of LOS with wt or F126A MD-2 and of these binary complexes with TLR4ecd. These unique reagents and analyses have permitted us to address the following two questions raised by the observations of Park et al. (17): 1) Are protrusion of a single fatty acyl chain of hexaacylated endotoxin and re-orientation of the aromatic side chain of Phe 126 intrinsic properties of TLR4activating hexaacylated endotoxin⅐MD-2 complexes, apparent before contact of this complex with TLR4? 2) Does Phe 126 affect the positioning of the acyl chains of bound endotoxin and potentially promote fatty acyl:TLR4 contacts apparently needed for TLR4 activation at the dimerization interface? 13 C/ 1 H high resolution HSQC spectra of the 13 CH 3 groups of LOS⅐MD-2 wt and LOS⅐MD-2 F126A showed six prominent and distinct signals in each complex (Fig. 2). These six crosspeaks correspond to signals deriving from the six terminal 13 CH 3 groups of the bound hexaacylated LOS and reflect the distinct microenvironment of each fatty acyl chain terminus when LOS is bound to either wt or F126A MD-2. For both wt and mutant complexes, five of the six 13 C/ 1 H methyl crosspeaks showed closely similar paramagnetic relaxation enhancement (PRE) (Fig. 3), a measure of solvent/surface accessibility. One methyl group (M1) in both complexes was more susceptible to quenching (i.e. a larger PRE rate) by Gd(DPTA-BMA), indicating that this fatty acyl chain is less sequestered and presumably protruding out of the hydrophobic pocket of both wt and F126A MD-2 complexes. These findings suggest that protrusion of one of six fatty acyl chains of hexaacylated endotoxin bound to MD-2 is a structural hallmark of hexaacylated endotoxin (lipid A)⅐MD-2 binary complexes, independent of and preceding interactions with TLR4. Both the geometry of fatty acyl chain substitution and the length of the fatty acyl chains differ between the TLR4-activating Escherichia coli LPS used in the crystal structure studies of Park et al. (17) and the TLR4activating meningococcal LOS used in these studies (Fig. 2D). Therefore, consistent with the "pattern recognition" properties of MD-2 and TLR4, the protrusion of one of six fatty acyl chains of bound TLR4-activating endotoxin from MD-2 does not depend on a single structural composition of lipid A. Studies are in progress, following the analytic approaches described here, to further define the structural requirements for this property of endotoxin (lipid A)⅐MD-2 complexes and its relation to TLR4 activation.
As indicated, 13 C/ 1 H high resolution HSQC spectra of LOS bound to F126A MD-2 also showed one methyl group of six that was more susceptible to PRE by Gd(DPTA-BMA), i.e. less sequestered by MD-2. Since LOS⅐MD-2 F126A acts as a TLR4 antagonist rather than an agonist, this finding strongly suggests that exposure of a fatty acyl chain outside of the MD-2 pocket is not sufficient for TLR4 activation. Overlay of the 13 C/ 1 H HSQC spectra of wt and F126A complexes demonstrated differences in each 13 CH 3 group signal between the two binary complexes consistent with a distinct microenvironment of each fatty acyl chain terminus when LOS was bound to wt versus F126A MD-2. Although some of the chemical shift perturbation induced by the F126A mutation in MD-2 could be due to effects of interactions of the aromatic ring of Phe126 with nearby fatty acyl chains, the crystal structure of Park et al. (17) suggested hydrophobic contacts between the re-oriented Phe 126 side chain and only two of the fatty acyl chains of bound LPS. Therefore, our findings suggest a more global effect of Phe 126 on the surrounding microenvironment of each of the fatty acyl chains of bound LOS, i.e. an effect on the positioning of the bound LOS as a whole.
While Phe 126 is apparently not necessary for protrusion of a fatty acyl chain of TLR4-activating endotoxin from MD-2, Phe 126 may be necessary to guide/anchor positioning of a fatty acyl chain(s) in a way that optimizes fatty acyl:TLR4 contacts at the dimerization interface and/or MD-2:TLR4 contacts. The significant chemical shift perturbation observed for the LOS methyl groups upon F126A mutation ( Fig. 2C and supplemental Table S1) seems most compatible with re-orientation of the Phe 126 side chain toward the fatty acyl chains of bound LOS within the LOS⅐MD-2 wt complex, suggesting that this local conformational change in MD-2 is TLR4-independent and may represent a necessary signature of TLR4-activating endotoxin⅐MD-2 complexes. The presence of conserved glycines on either end of the "Phe 126 loop" is consistent with the apparent importance of the conformational mobility of this region and the ability of MD-2 to bind and confer TLR4 agonist (or antagonist) properties on an array of endotoxin species with varying numbers, distribution, and lengths of fatty acyl chains. The flexibility of this region is reflected in the crystal structure of Park et al. (17) where the least resolved structure (i.e. region of the greatest mobility) was observed in the Phe 126 loop between the ␤G and ␤H strands. The NMR data presented here support the conclusions of DeMarco and Woods (34) that utilized molecular dynamics simulations to predict that hexaacylated LOS or LPS bound to MD-2 in a binary complex would have one acyl chain partially accessible to solvent lying near the protein surface and that the acyl chains within the MD-2 pocket would maintain mobility.
In contrast to TLR3 which requires homodimerization to accommodate binding of dsRNA ligand (36), our findings (Fig.  4B) demonstrate that high affinity (pM) binding of E⅐MD-2 with TLR4 occurs between monomeric E⅐MD-2 and monomeric TLR4. Whether monomeric E⅐MD-2 is presented to TLR4 or an E monomer is presented to pre-assembled MD-2⅐TLR4, the initial product is a monomer of the ternary complex E⅐MD-2⅐TLR4. The E⅐MD-2⅐TLR4 product is the most proximal intermediate in endotoxin-induced TLR4 dimerization (i.e. formation of [E⅐MD-2⅐TLR4] 2 ), when it is composed of TLR4-activating hexaacylated E,wt MD-2, and TLR4 or is the final product of TLR4 antagonists when containing either underacylated E and/or mutant (e.g. F126A) MD-2 (15)(16)(17). For reasons not yet clear (see below), reaction of either LOS⅐MD-2 wt or LOS⅐MD-2 F126A with the insect cell-derived recombinant human TLR4ecd yielded exclusively a monomeric ternary complex whether at pM-nM concentrations or Ͼ100 M used for NMR analyses. This is in contrast to the dimeric ternary complexes previously reported when either hexaacylated E. coli LPS or meningococcal LOS was complexed to human MD-2⅐TLR4ecd (13,17). Since the two hexaacylated endotoxin species have closely similar MD-2⅐TLR4-activating properties (12) and the LOS used in earlier studies (13) was the same as that used in this study, we believe the absence of dimerization of the LOS⅐MD-2⅐TLR4ecd ternary complex observed herein reflects structural differences between the recombinant TLR4ecd used in this study versus the earlier studies. Structural differences between the insect cell-derived His 6 -TLR4ecd (aa 24 -631) produced in this study and the recombinant human TLR4ecd produced in earlier studies (aa 24 -631 generated from a transfected HEK293 cell line (13) and aa 27-631 from insect cells (17)) that yielded dimeric ternary complexes include differences in epitope tags and the composition and extent of glycosylation correlating with an insect versus eukaryotic cellderived protein. Whatever the pertinent variable(s), our findings underscore that the minimum structural requirements in TLR4 for agonist-induced TLR4 dimerization are not yet clear (37). For the purposes of this study, the exclusive formation of a monomeric ternary complex was fortuitous. It allowed us to demonstrate that pM interactions between E⅐MD-2 and TLR4ecd are achieved by agonist-independent interactions that occur entirely within a single ternary complex and to compare by NMR the lipid A fatty acyl chains of TLR4-activating and TLR4-inhibiting ternary complexes containing, respectively, wt and F126A MD-2.
Comparison of the 13 C/ 1 H HSQC and HMQC spectra of LOS in the ternary complexes containing wt versus F126A MD-2 (Fig. 5) showed Phe 126 -dependent differences that could be relevant to the different functional properties of these complexes. In both complexes, the engagement of TLR4 resulted in an apparent re-orientation of the fatty acyl chains since the acyl methyl groups overall underwent significant chemical shift perturbations both in position and linewidth of the signal. As indicated by the selective line broadening of the LOS methyl crosspeaks (Fig. 5, A and B) and loss of LOS methylene signals from the acyl chains, particularly in the wt MD-2⅐LOS⅐TLR4ecd ternary complex (Fig. 5D), the effect of TLR4 association was more pronounced overall on the wt MD-2 complex than in the F126A mutant complex. Although some of these changes could reflect altered tumbling properties of the ternary (M r ϳ100,000) versus the binary (M r ϳ25,000) complexes, the fact that these changes were more pronounced in the ternary complex containing wt MD-2 and included changes in the chemical shift of the major crosspeaks of the 13 CH 3 groups suggests a further re-arrangement of the LOS fatty acyl chains within the TLR4-activating ternary complex that could be functionally important.
In summary, our findings demonstrate that Phe 126 of MD-2 affects the position and surrounding environment of the bound fatty acyl chains of TLR4-activating hexaacylated endotoxin. These effects of Phe 126 are manifest before interactions with TLR4 and are likely driven by hydrophobic interactions between the fatty acyl chains of the bound endotoxin and the aromatic side chain of Phe 126 . Phe 126 acts as a "hydrophobic switch" driving agonist-dependent recognition of TLR4-activating E⅐MD-2 by promoting formation of an agonist-dependent composite binding surface made up of optimally extruded endotoxin fatty acyl chain and an altered MD-2 surface conferred by re-arrangement of the Phe 126 loop.