Development of αGlcN(1↔1)αMan-Based Lipid A Mimetics as a Novel Class of Potent Toll-like Receptor 4 Agonists

The endotoxic portion of lipopolysaccharide (LPS), a glycophospholipid Lipid A, initiates the activation of the Toll-like Receptor 4 (TLR4)–myeloid differentiation factor 2 (MD-2) complex, which results in pro-inflammatory immune signaling. To unveil the structural requirements for TLR4·MD-2-specific ligands, we have developed conformationally restricted Lipid A mimetics wherein the flexible βGlcN(1→6)GlcN backbone of Lipid A is exchanged for a rigid trehalose-like αGlcN(1↔1)αMan scaffold resembling the molecular shape of TLR4·MD-2-bound E. coli Lipid A disclosed in the X-ray structure. A convergent synthetic route toward orthogonally protected αGlcN(1↔1)αMan disaccharide has been elaborated. The α,α-(1↔1) linkage was attained by the glycosylation of 2-N-carbamate-protected α-GlcN-lactol with N-phenyl-trifluoroacetimidate of 2-O-methylated mannose. Regioselective acylation with (R)-3-acyloxyacyl fatty acids and successive phosphorylation followed by global deprotection afforded bis- and monophosphorylated hexaacylated Lipid A mimetics. αGlcN(1↔1)αMan-based Lipid A mimetics (α,α-GM-LAM) induced potent activation of NF-κB signaling in hTLR4/hMD-2/CD14-transfected HEK293 cells and robust LPS-like cytokines expression in macrophages and dendritic cells. Thus, restricting the conformational flexibility of Lipid A by fixing the molecular shape of its carbohydrate backbone in the “agonistic” conformation attained by a rigid αGlcN(1↔1)αMan scaffold represents an efficient approach toward powerful and adjustable TLR4 activation.


■ INTRODUCTION
Toll-like Receptor 4 (TLR4) is a mammalian transmembrane receptor protein which, in complex with a myeloid differentiation factor 2 (MD-2), detects picomolar concentrations of Gram-negative bacterial endotoxin (or lipopolysaccharide, LPS) ( Figure 1A) and initiates an inflammatory signaling cascade aimed at the eradication of bacterial infection. 1 Activation of the innate immune response through TLR4·MD-2-LPS complex was shown to contribute to the pathogenesis of numerous inflammatory, autoimmune, and chronic diseases, such as sepsis syndrome, asthma, arthritis, and cancer, which highlights the significance of TLR4-MD-2 complex as a therapeutic target. 2−5 Therapeutic modulation of the innate immune response by intervention with TLR4·MD-2 signaling has grown to a "hot" topic in the past decade. 4,6 Moreover, activation of TLR4 has been proposed to bridge the innate and adaptive immunity, 7 emphasizing stimulation of the TLR4·MD-2 complex by nontoxic ligands as a straightforward way to efficient vaccine adjuvants. 8−10 Lipid A, an amphiphilic membrane-bound portion of LPS, represents the major pathogen-associated molecular pattern which drives the activation of TLR4 by binding to the coreceptor protein MD-2 and triggering the dimerization of two TLR4·MD-2·LPS complexes. 11 Generally, the binding of hexaacylated bisphosphorylated Lipid A (such as from Escherichia coli, Figure 1B) by human TLR4·MD-2 complex results in the efficient activation of the innate immune response, whereas underacylated Lipid A variants are either inactive or antagonistic (such as tetraacylated lipid IVa, or the synthetic drug candidate Eritoran). 12,13 The co-crystal structures of the E. coli Re-and Ra-LPS with mouse (m) or human (h) MD-2·TLR4 complex, respectively, unravel that only five long-chain acyl residues of the hexaacylated Lipid A are incorporated into the hydrophobic binding pocket of MD-2 whereas the sixth 2-N-acyl chain is exposed on the surface of MD-2 and is involved in the dimerization interface with the second TLR4*·MD-2*-LPS complex ( Figure 1C). 14,15 The Phe126 residue of MD-2 is proposed to stabilize the presentation of an acyl tail on the surface of the protein and to serve as hydrophobic switch allowing dimerization to occur. 16,17 LPS-driven homodimerization of TLR4·MD-2-LPS complexes initiates recruitment of adaptor proteins to the intracellular TIR (Toll/interleukin-1 receptor) domains of TLR4 which ultimately results in the induction of the intracellular inflammatory signaling cascade. 18 In contrast, submerging of all lipid chains of the ligand into the hydrophobic binding groove of MD-2 results in an efficient binding without initiation of signaling, which is a characteristic feature of TLR4·MD-2 antagonists. 12,13 The presence of both 1-and 4′-phosphate groups of Lipid A was shown to be crucial for the efficiency of the dimerization and the potency of the initiated signaling. 19 The absence of 1phosphate leads to less efficient dimerization 20,21 and dampened cytokine production while maintaining sufficient TLR4-mediated immune activation and full adjuvant activity, which guided the development of monophosphoryl Lipid A (MPLA), a licensed vaccine adjuvant ( Figure 1B). 8,10 Despite tremendous intensive research on the interaction of TLR4·MD-2 complex with isolated, 22,23 genetically engineered, 24 and synthetic Lipid A's and analogues, 25−29 the structure−activity relationships of the LPS-triggered TLR4 activation are not unambiguously established. Minor variations in the length and distribution pattern of fatty acyl chains in Lipid A typically result in dramatic amendment of TLR4mediated immune signaling 25−27 which cannot be rationally predicted.
We have addressed the challenges associated with the exploration of structural basis of LPS-induced TLR4 activation by development of a novel type of Lipid A mimetics wherein the flexible βGlcN(1→6)GlcN backbone of Lipid A is replaced by the conformationally restricted (1↔1)-connected disaccharide scaffolds. Notably, all Lipid A analogues synthesized so far were based either on the native β(1→6)-diglucosamine or on the more flexible backbones wherein one or both GlcN residues were replaced by a linear aglycon. 9,10,27,30 Previously we reported on the synthesis and potent anti-endotoxic activity of tetraacylated Lipid A mimetics derived from the β,α(1↔1)linked diglucosamine representing an "antagonistically" shaped scaffold. 31 Taking advantage of a striking similarity between the conformation of the nonreducing sugar trehalose [αGlc(1↔1)αGlc] and the molecular shape of the β(1→6) diglucosamine backbone of TLR4·MD-2-bound agonist E. coli Lipid A disclosed in the X-ray structure, we have developed novel agonistic conformationally confined Lipid A mimetics based on the two-bond-linked rigid trehalose-type αGlcN(1↔1)αMan scaffold ( Figure 2). ■ RESULTS AND DISCUSSION X-ray Structure-Guided Design of αGlcN(1↔1)αMan-Based Lipid A Mimetics. The β(1→6) diglucosamine backbone represents the most conserved part of Lipid A, whereas its acylation and phosphorylation pattern varies within bacterial species. 22,32 The overall three-dimensional conformation of the intrinsically flexible three-bond-linked βGlcN(1→6)GlcN backbone of Lipid A is determined by the values of the dihedral angles ω, ϕ, and ψ about (1→6) glycosidic and oxymethyl linkages ( Figure 3A). Thus, the relative orientation of GlcN rings can be easily adapted by rotation about glycosidic and oxymethyl linkages via altering the corresponding torsion angles. This permits spontaneous adjustment of the shape of Lipid A to the geometry of the binding pocket of MD-2, which complicates the estimation of the "active" conformation of the ligand in the [Lipid A-MD-2· TLR4] 2 complex. As seen in the co-crystal structures, the proximal (reducing) GlcN ring of MD-2-bound hexaacylated Lipid A adopts an inclined (or "twisted") orientation which, as we assume, is essential for the exposure of the long-chain 2-Nacyl residue on the surface of MD-2 followed by dimerization with the second MD-2·TLR4 complex ( Figure 3A,B). 31 To explore the structural prerequisites needed for an effective receptor complex homodimerization, we have manipulated the flexibility of the carbohydrate backbone of Lipid A by fixing its molecular shape in an "agonistic" conformation. Since the relative spatial arrangement of the two GlcN rings of MD-2bound agonist E. coli Lipid A disclosed in the co-crystal structures 14,15 resembles the arrangement of α,α-(1↔1)connected glucoses in the nonreducing disaccharide trehalose ( Figure 3B,C), we have developed predictably agonistic Lipid A mimetics based on the conformationally confined α,αtrehalose-like αGlcN(1↔1)αMan scaffold ( Figure 3D).
The values for the torsion angles Φα and Φ′α, representing rings orientation about the α,α-(1↔1) glycosidic linkage ( Figure 3D), are governed mostly by the anomeric and exoanomeric effects and are only marginally dependent on the nature of functional groups in variably substituted α,αtrehaloses. 33 The existence of a single conformational minimum with respect to the dihedrals about glycosidic linkage in α,αtrehaloses was confirmed by molecular dynamics simulations, 33,34 whereas the preferred gauche−gauche conformation of the substituted α,α-trehalose 35 and its αGlc(1↔1)αMan analogue 36 was corroborated by X-ray and conformational analysis, respectively. Thus, conformationally restrained α,α-(1↔1) glycosidic linkage in αGlcN(1↔1)αMan-based Lipid A mimetics would impose a specific relative orientation of sugar rings resembling the molecular shape of the diglucosamine backbone of the agonistic MD-2-bound Lipid A. αGlcN(1↔1)αMan-based Lipid A mimetics (α,α-GM-LAMs) 1−3 were designed such that the acylation and phosphorylation pattern of the nonreducing (distal) GlcN residue of E. coli Lipid A remains unaffected, whereas the β(1→6) glycosidic linkage is substituted by an α,α(1↔1) glycosidic bond and the reducing (proximal) GlcN moiety of natural Lipid A is exchanged for a nonreducing sugar (mannose) having a specific acylation and phosphorylation pattern ( Figure 3D). The location of the phosphate functionality at C-6 of the Man residue was selected to closely resemble the positioning of a 1-phosphate group of E. coli lipid A at the secondary dimerization interface of the TLR4-MD-2-Lipid A complex (PDB code 3FXI). The site of acylation at the mannose moiety was chosen such that the attachment of the  long-chain (R)-3-acyloxyacyl residue at Man C-4 would provide a sufficient hydrophobic patch to support the homodimerization and the interaction with the second TLR4*·MD-2*-ligand complex.
Upon interaction with the receptor complex, the tetraacylated GlcN unit of α,α-GM-LAMs 1−3 was supposed to be fully accommodated within the hydrophobic pocket of MD-2, whereas the "twisted" mannose ring should be excluded from the binding site on MD-2 such that the two lipid chains at Man C-4 are presented onto the surface of the protein and involved in the secondary dimerization interface ( Figure 3D). The axial configuration at C-2 of mannose should provide, according to the crystal structures, a better fitting to the geometry of the binding pocket of MD-2 and, simultaneously, ensure easier stereocontrol in the 1,2-trans glycosylation step to α,α(1↔1)linked disaccharide.
Synthetic Strategy. The assembly of αGlcN(1↔1)αMan, a 1,1-glycosidically connected (nonreducing) disaccharide, represents a formidable synthetic challenge with regard to simultaneous stereocontrol at two anomeric centers. Typically, approaches involving conventional glycosylation procedures for the synthesis of trehalose provide moderate stereoselectivity and low yields. 38 Since we aimed to establish the α,α-(1↔1) glycosidic linkage between an amino sugar and a mannoconfigured monosaccharide, we could hardly rely on the intramolecular aglycon delivery approach 39 or on the versatile synthetic desymmetrization of the natural trehalose. 40 For the synthesis of bis-and monophosphorylated Lipid A mimetics 1−3 ( Figure 2) having non-symmetrically distributed acyloxyacyl functional groups, a convergent approach involving first the preparation of the orthogonally protected αGlcN(1↔1)αMan disaccharide scaffold followed by regioselective phosphorylation and acylation with (R)-3-acyloxyacyl fatty acids of variable chain lengths was envisaged (Scheme 1).
For the assembly of the αGlcN(1↔1)αMan backbone, a 2-Ncarbamate-protected glucosamine-based lactol was chosen as acceptor, and a 2-O-levulinoyl (Lev, 4-oxopentanoyl)-protected mannose was selected to serve as glycosyl donor. The participating protecting group at C-2 (Lev) should allow for a preferable 1,2-trans mannosylation. Since the 2-O-Lev protection had to be exchanged for a 2-O-Me group later in the synthesis, an alternative glycosylation approach using nonparticipating methyl protection at C-2 of the mannose-based donor was planned to be explored as well.
The 2-N-carbamates of variably protected GlcN-based lactols revealed the highest α/β ratio (up to 9:1) of the anomeric 1-OH group, which highlighted these intermediates as the most "stereoselective" glycosyl acceptors. Apparently, the presence of a carbamate N−H capable of hydrogen bonding with the axial oxygen at C-1 is responsible for the substantial enrichment with the α-anomer.
The synthesis of the required mannose-based donors commenced with Zempleń deacetylation of the peracetylated thioethyl glycoside The key point in our initial approach was to obtain a good double α-stereoselectivity in the glycosylation reaction between reducing acceptor 8 (α/β = 9:1) and the 2-O-levulinoylprotected imidate donors 18 or 19. The participating levulinoyl group at C-2 of the manno-configured donors 18 and 19 should allow for a selective 1,2-trans glycosylation. A survey of the literature revealed that a complete α-manno selectivity upon application of 2-O-Lev-protected mannosyl donors could be obtained with a variety of acceptors. 44,45 In an initial glycosylation attempt comprising coupling the TCA donor 18 and acceptor 8 using trimethylsilyl trifluoromethanesulfonate (TMSOTf) as promoter, an orthoester 23 was obtained as the major product (23%) along with a minor proportion of the target α,α-disaccharide 22 (8%) and a concurrently formed donor self-coupling product 24 (Scheme 4A, Supplementary SI- Table 1). With the less reactive NPTFA donor 19 or with the thioethyl donor 14 the formation of the orthoester was not observed; however, the desired product 22 was isolated only in trace amounts.
We have hypothesized that the diminished reactivity of the torsionally disarmed 4,6-di-O-cyclic-protected mannose donor was responsible for the glycosylation failure. 46 Accordingly, the 4,6-di-O-DTBS group in 14 was cleaved and substituted for two acetates to provide 25, wherein the C5−C6 bond was unlocked from the disarming trans−gauche conformation. 46,47 To provide consistency with the previously performed imidate-mediated glycosylations, the thioglycoside at C-1 was exchanged for a TCA group to furnish 27 (Scheme 4B). The coupling of 8 and the torsionally unlocked 4,6-di-O-acetyl donor 27 resulted in the isolation of the α,α-configured disaccharide 28, albeit in a similarly low yield (8%).
Thereafter, our attention was turned to the apparently low reactivity of the lactol acceptor 8 affected by the H-bonding between the α-1-OH and the carbamate NH groups. To increase the nucleophilicity of the lactol acceptor and to reduce steric constraints, the disarming acetate at C-3 was exchanged for a TBDMS group, and the sterically demanding 2-N-Troc group was replaced by Fmoc protection, which provided α-lactol acceptor 33. To this end, compound 30, made by a highly selective reductive opening of benzylidene acetal in the allyl glycoside 29, 31 was phosphorylated at C-4 to give 31 (Scheme 4C). Reductive cleavage of 2-N-Troc protection with Zn in acetic acid followed by reaction with Fmoc chloride in the presence of a EtN(iPr) 2 furnished 32, which was anomerically deprotected using [Ir(COD)(Ph 2 MeP) 2 ]PF 6 -catalyzed isomerization of the allyl group followed by oxidative cleavage of the 1-propenyl group with I 2 in THF−H 2 O to afford the "armed" acceptor 33. Reaction of the 2-O-levulinoyl TCA donor 18 with 33, however, reproducibly resulted in the formation of the orthoester 34 as the major product (Scheme 4C).
In retrospect, we assume that the failure of the 2-Olevulinoyl-protected mannose-based donors to provide the desired αGlcN(1↔1)αMan compound in a glycosylation reaction with GlcN-based lactol acceptors was rather related to a particular conformation of the arising trehalose-type α,αdisaccharide. A successful glycosylation would lead to a sterically hindered coupling product 22, having overlapping bulky 2-N-Troc/Fmoc groups at the GlcN moiety and a linear  ; (e) (CF 3 CO) 2  4-oxopentanoate (levulinate) ester group at C-2 of mannose, while reciprocal repulsion of the N-carbamate and electron-rich levulinoyl groups could contribute as well.
Given the failure of 2-O-levulinoyl-protected mannosyl donors to provide the desired coupling products, the participating protecting group at C-2 of mannose was exchanged for the 2-O-Me group as in the donors 20 and 21 (Scheme 3). Successful application of non-participating groups in the α-selective mannosylation has been extensively reported. 48,49 Gratifyingly, the coupling of the 2-O-Me-Man imidate donor 21 with the GlcN-lactol acceptor 8 allowed for a much higher isolated yield (51%) of the α,α-disaccharide 35 (Scheme 5, Table 1, entry 2). Owing to the low reactivity of acceptor 8, the NPTFA donor 21 was found to be superior to the TCA donor 20 due to the propensity of the latter to form substantial amounts of the rearranged glycosylamide 45 (52%), which is characteristic for glycosylations involving acceptors of diminished reactivity (Table 1, entries 1 and 2). The isolation of the α,α-disaccharide 35 was complicated by the concomitant formation of the co-migrating αGlcN−βMan 38 and βGlcN−αMan 41 byproducts. Glycosylation of the 2-N-Fmoc-protected acceptor 10 by the NPTFA donor 21 afforded a similar isolated yield (52%) of the target α,α-configured disaccharide 36 (Scheme 5, Table 1, entry 3).
Enhancement of the acceptor reactivity by the use of 3-O-TBDMS-protected "armed" lactol 33 did not improve the yields in the coupling reactions with either donor 20 or 21 (Table 1, entries 4 and 5), furnishing αGlcN(1↔1)αMan disaccharide 37 in 25% and 50% yield, respectively. The yield of the αGlcN(1↔1)αMan disaccharide was related to the ease of its chromatographic purification, which, in turn, was strongly dependent on the protection group pattern. Since isolation of the 2-N-Fmoc-3-O-Ac-protected 36 from the mixture of anomeric products was the most straightforward, this disaccharide was chosen for further transformation to the target α,α-GM-LAMs 1−3.
The configurations at the anomeric centers of the (1↔1)linked disaccharides were assigned on the basis of 1 H and 13 C NMR shifts at the anomeric positions and the 1 J C1,H1 coupling constants 50 (SI- Table 2). To circumvent the severe peak broadening in the 1 H NMR spectrum, the Fmoc group in 42 was replaced by a 2-N-acetate to give acetamide 44 appropriate for the unambiguous signal assignment. The α,α-linkage in 35− 37 was confirmed by the large 1 J C1,H1 coupling constant values of the anomeric carbons ( 1 J C1,H1 = 170−174 Hz for α-mannoand 1 J C1,H1 = 173−178 Hz for α-gluco-anomers) and by the downfield shifts and the corresponding vicinal proton coupling constants 3 J H1,H2 of the anomeric H-1 signals (δ 5.02−5.12, 3 J 1,2 = 1.5 Hz for α-mannoand 5.08−5.23, 3 J 1,2 ≈ 3.7 Hz for αgluco-anomers). 51 The β-manno linkage in 38−40 was corroborated by the typically smaller 1 J C,H coupling constants for the anomeric carbons ( 1 J C1,H1 = 155−157 Hz) and by the upfield shifts of the anomeric H-1 signals (4.6 ppm) and H-5 signals (3.28 ppm), characteristic for β-mannosides. 52 Synthesis of Lipid A Mimetics 1−3 Based on αGlcN(1↔1)αMan Scaffold. Having orthogonally protected αGlcN(1↔1)αMan disaccharide scaffold 36 in hand, we next approached the stepwise deprotection and acylation at C-2 and  Table 3). Cleavage of the 3-O-Ac group to furnish 46 was accompanied by the migration of the phosphate from C-4 to the liberated hydroxyl group at C-3 to give 47 (Scheme 6). Besides, a partial hydrolytic loss of one benzyl protecting group in the phosphotriester 46 leading to formation of the phosphodiester 48 was also observed. Further improvements could be achieved through optimization of reaction conditions. Indeed, when the reaction was terminated prior to completion (48 h), the formation of the undesired byproducts 47 and 48 could be largely avoided (2% and 9%, respectively), providing 3-O-deacylated compound 46 in 53% yield (SI- Table 3). Repeated chromatographic purifications of the disaccharide 46 partly account for the relatively low yield.
The first (R)-3-(tetradecanoyloxy)tetradecanoyl residue at C-3 of the GlcN moiety was introduced by reaction of 46 with β-acyloxyacyl acid 49 under the agency of DIC and a catalytic amount of DMAP to provide 52 in 83% yield (Scheme 6). Notably, strictly equimolar amounts of DIC and fatty acid 49, and a catalytic quantity of DMAP at 0°C, had to be applied to suppress the concomitant formation of the co-migrating 3-Otetradecanoyl (53) and 3-O-alkenoyl (54) byproducts (11% and 3%, respectively). Application of higher amounts of DIC and/or fatty acid aimed to accelerate the transformation resulted in augmented formation of 53 and 54, which could be rationalized by a probable β-elimination or rearrangement 53   performed under standard conditions with HF·Py in THF. To introduce acyl and phosphate functional groups at mannose C-4 and C-6, respectively, without additional protecting group manipulation, the regioselectivity of the acylation of the diol 56 with the acyloxyacyl acids 50 or 51 was first examined. Since hydroxyl groups in the substituted trehaloses are known to differ in reactivity due to both steric and electronic effects, 38,40 we expected that, in a heavily substituted αGlcN(1↔1)αMan disaccharide 56, positions C-4 and C-6 at the Man moiety could be discriminated in a subsequent acylation procedure.
Indeed, the major outcome of DIC/DMAP-mediated acylation of 56 with the acids 50 or 51 was not the intrinsically expected primary 6-OH-derived acylation products, but the 4-O-acyloxyacyl derivatives 57 and 58 in 51% and 65% yield, respectively, having a 6-OH group at the mannose unit accessible for the ensuing phosphorylation. Minor amounts of the 4,6-bis-O-acylated derivatives 59 and 60 (24% and 13%, respectively) were isolated as well. Although the 6-OH group of the mannose residue in αGlcN(1↔1)αMan disaccharide 56 is somewhat remote from the 2-NH of the GlcN moiety, the crystal structures of α,α-trehalose-based compounds indicate spatial proximity of the two groups. Thus, it could be assumed that the intramolecular hydrogen bonding (2-NH GlcN −6-OH Man ) exerts an adverse effect on the reactivity of the primary hydroxyl group at Man C-6. Furthermore, the bulkiness of the in situ-formed acyloxyacyl-activated ester resulting from the reaction of the fatty acids 50 or 51 with DIC/DMAP could also explain the limited access to the sterically hindered 6-OH group in the α,α-trehalose-like disaccharide 56.
Next, the free 6-OH group in the hexaacylated disaccharides 57 and 58 was phosphorylated by reaction with dibenzyl-(N,N′-diisopropylamino)phosphoramidite in the presence of a mild acid catalyst, 1H-tetrazole, and subsequent oxidation with mCPBA at −78°C to furnish the bisphosphorylated hexaacylated products 61 and 62 in 91% and 89% yield, respectively. Final debenzylation by hydrogenation of 57, 61, and 62 on Pd-black followed by purification with gel permeation chromatography on Sephadex SX1 in toluene− methanol (2:1) afforded target bisphosphorylated Lipid A mimetics 1 and 2 and a monophosphorylated counterpart 3. In contrast to native Lipid A, compounds 1 and 2 do not possess a labile anomeric phosphate functionality; consequently, they were isolated and biologically assessed as free acids at the phosphates and could be stored in aqueous solution at 4°C for several months without any noticeable sign of degradation, which was confirmed by MALDI-TOF analysis.
Activation of TLR4·MD-2 Complex by α,α-GM-LAMs 1−3. The propensity of the α,α-GM-LAMs 1−3 to stimulate TLR4-mediated immune signaling was first assessed in the hTLR4/hMD-2/CD14 transfected human embryonic kidney (HEK) 293 cells (HEK-Blue). Since we were particularly interested in the molecular recognition mechanisms implicated in the binding of LPS by the MD-2·TLR4 complex wherein the Lipid A/Re-LPS portion of LPS is exclusively involved, we evaluated the activities of α,α-GM-LAMs 1−3 compared to E. coli Re-LPS ( Figure 1A). It has been previously shown that the minimum structural requirement for the expression of the highest cytokine-inducing potency resides in Re-LPS, which entails two 3-deoxy-D-manno-oct-2-ulosonic acid (Kdo) residues in addition to Lipid A (Supporting Information, SI- Figure  1A). 54,55 Moreover, the Re-LPS has a defined molecular weight (MW) similar to the MW range for Lipid A mimetics (1.8−2.2 kDa), in contrast to wild-type LPS having variable MW (10−15 kDa), so that the direct comparison of a dose-dependent response between compounds 1−3 and Re-LPS is more appropriate. The TLR4-stimulating activity of Lipid A mimetics 1−3 was examined over a wide concentration range by monitoring of the activation of the NF-κB regulated signal transduction pathway via measuring the induction of secreted embryonic alkaline phosphatase (SEAP) and compared to the responses elicited by E. coli Re-LPS and SM-MPLA.
Remarkably, the glycosylation at C-6′ of Lipid A with Kdo residues was reported to be responsible for the 10−20-fold enhancement of the activity of the Kdo/Kdo2-Lipid A (Re-LPS) compared to Lipid A alone in the nanomolar concentration range. 28,54 Notably, conformationally confined α,α-GM-LAM 2 based on just a disaccharide scaffold displayed NF-κB activation (EC 50 = 0.08 nM) similar to those of Re-LPS (EC 50 = 0.04 nM) and E. coli LPS (EC 50 = 0.08 nM, SI- Figure  1A), and the TLR4 saturation plateau was reached at a concentration of 1 ng/mL for both ligands (Figure 4).
Compound 1, having a 2×CH 2 -longer acyl side chain at Man C-4 compared with α,α-GM-LAM 2, was a less efficient activator of NF-κB (EC 50 = 0.4 nM), so that its TLR4 saturation plateau was reached at a concentration of 5 ng/mL. Thus, shortening of a lipid side chain at C-4 of the mannose moiety by 2×CH 2 resulted in a 5-fold increase of TLR-4 stimulating activity, which underlines the significance of hydrophobic interactions at the dimerization interface. Along this line, synthetic manipulation of the length of the acyl side chain at Man C-4 could be used for fine-tuning of the hTLR4mediated activity in α,α-GM-LAMs. The monophosphate 3 was, as expected, significantly less active (EC 50 = 31 nM) than its bisphosphorylated counterpart 1, but it showed a more potent activation profile than SM-MPLA at concentrations above 10 ng/mL.
The dose-dependent stimulating activity of synthetic Lipid A mimetics was cytokine-specific, revealing higher potency in the induction of the expression of TNF-α and IL-8 by α,α-GM-LAMs 1 and 2 than by Re-LPS/LPS ( Figure 5A,B, SI-Figures 2   and 3). The release of MCP-1 induced by α,α-GM-LAM 2 was clearly more effective than the production of this chemokine by Re-LPS and compound 1 ( Figure 5C). Expression of MyD88dependent chemokine MCP-1 is associated with the activation of the intracellular TLR4·MD-2 complex. 56 The dampened induction of the expression of cytokines by α,α-GM-LAM 3 correlates to its chemical structure missing a phosphate group at Man C-6. Our results indicate that both α,α-GM-LAMs 1 and 2 are more potent activators of the MyD88 signaling pathway than Re-LPS/LPS ( Figure 5, SI-Figure 4).
The ability of Lipid A mimetics 1−3 to induce the production of TNF-α and IL-6 from bone marrow-derived macrophages (BMDM) in mice was subsequently examined and compared to that of synthetic E. coli Lipid A and E. coli MPLA, which are reliable positive controls due to their chemical purity and homogeneity ( Figure 6). The maximum level of TNF-α (4000 pg/mL) was detected in BMDM cultivated in the presence of 1 nM α,α-GM-LAM 1 or 2, whereas the same quantity of the "parent" Lipid A resulted in release of half the amount of TNF-α (2000 pg/mL). On the other hand, monophosphoryl α,α-GM-LAM 3 exhibited a dampened ability to express TNF-α (500 pg/mL) at a concentration of 1 nM. Likewise, 1 nM (1.8 ng/mL) of 1 or 2 triggered the release of a 3-fold higher amount of IL-6 (1500 pg/mL) compared to Lipid A (500 pg/mL). Monophosphate 3 revealed a dose-dependent cytokine induction profile showing only marginal expression level of IL-6 at a concentration of 1 nM but higher IL-6 release than MPLA at concentrations above 1 nM.
TLR4 Stimulating Activities of α,α-GM-LAMs 1 and 3 in Human Dendritic Cells. To test the impact of selected α,α-GM-LAMs 1 and 3 on maturation of human dendritic cells (hDCs), immature monocyte-derived hDCs were stimulated with 1 and 3 in a wide concentration range or with E. coli LPS as positive control. DCs are able to persistently sense pathogenassociated molecular patterns and present antigens to T lymphocytes, thereby initiating an adaptive immune response. 7 DCs treated with LPS acquired a distinctive morphologic phenotype and, when analyzed by flow cytometry, displayed characteristic markers of mature DCs.
Stimulation of DCs with 1 (1 mg/mL) was as potent as that with LPS in inducing DCs maturation and up-regulation of the co-stimulatory molecules CD86, as well as the antigenpresenting structures MHC class I and MHC class II, which are necessary for the induction of an adaptive immune response (SI- Figure 5). None of the α,α-GM-LAMs exerted cytotoxic effects on DCs, as determined by propidium iodide staining (data not shown).  Activated DCs were examined for the production of proinflammatory cytokines TNF-α, IL-6, and IL-12, which contribute to the modulation of the T cell response and innate effector functions. 57 The release of TNF-α and IL-6 reached nearly maximum levels (attained with 10 ng/mL LPS) at the α,α-GM-LAM 1 concentration of 1 ng/mL ( Figure 7A). Also the expression of IL-12, which promotes the development of adaptive immune cells and is involved in coordinating innate and adaptive immunity, 58 was efficiently induced by α,α-GM-LAM 1, indicating its potential adjuvant capacity. The release of TNF-α, IL-6, and IL-10, a unique cytokine with a wide spectrum of anti-inflammatory effects, in DCs induced by monophosphate 3 was, in agreement with the experiments in the recombinant hTLR4/hMD-2 signaling system, less efficient compared to that induced by 1 ( Figure 7B).

■ CONCLUSIONS AND PERSPECTIVES
We have rationally designed, synthesized, and biologically evaluated an entirely novel type of TLR4 agonists based on the conformationally restrained disaccharide scaffold which mimics the spatial arrangement of (1→6) diglucosamine backbone of MD-2-bound E. coli Lipid A, revealed in the co-crystal structures of the TLR4·MD-2-LPS complexes. A convergent synthetic approach toward the αGlcN(1↔1)αMan scaffold and hexaacylated Lipid A mimetics based thereon has been developed. Orthogonally protected nonreducing αGlcN(1↔1)αMan disaccharide was assembled via imidatemediated glycosylation by taking advantage of the axial orientation of the anomeric OH group in the 2-N-carbamateprotected GlcN-lactol acceptors. The protecting group pattern was fine-tuned to ensure stereospecific α,α-(1↔1) glycosylation and to afford efficient isolation of the α,α-disaccharide from the anomeric mixtures. Replacement of the labile anomeric 1-phosphate functionality (as in natural Lipid A) for a primary phosphate group in α,α-GM-LAMs 1 and 2 provides unambiguous advantages with respect to hydrolytic stability.
In spite of the lack of Kdo moieties which were demonstrated to be responsible for a 10−20-fold enhancement of the activity of Re-LPS (Kdo-Kdo-Lipid A) compared to Lipid A alone, the NF-κB activation efficacy and cytokine inducing capacity of bisphosphorylated Lipid A mimetics 1 and 2 in DCs and human macrophages were comparable to or higher than those of Re-LPS/LPS. The inherent rigidity of the α,α-(1↔1) glycosidic linkage in 1−3 would not allow for extensive conformational adjustment of the αGlcN(1↔1)αMan backbone to the shape of the binding pocket of MD-2, such that the three-dimensional arrangement of an α,α-GM-LAM molecule should remain preserved also in the protein-bound state. Accordingly, our results indicate that restricting the flexibility of the carbohydrate backbone of Lipid A in an "agonistic" conformation, as in α,α-GM-LAMs, allows the prearrangement of the phosphate and acyl groups in the "flipped" Man moiety in a defined conformation, which results in a very potent TLR4 activation (SI- Figure 6). The shortening of a secondary acyl chain at Man C-4 and the presence of a phosphate group at Man C-6 significantly enhanced the TLR-4 stimulating activity, which indicates the involvement of these functionalities at the dimerization interface with the second TLR4*·MD-2*-ligand complex and opens opportunities for fine-tuning the activities of α,α-GM-LAMs by chemical modifications.
Since the molecular shape of αGlcN(1↔1)αMan-based Lipid A mimetics is believed to resemble the conformation of the MD-2-bound E. coli Lipid A in the active [TLR4·MD-2-LPS] 2 complex, further immuno-biological studies of the interaction of α,α-GM-LAMs with TLR4·MD-2 would provide deeper insight into the molecular basis of TLR4 activation by LPS.
Along these lines, application of a conformationally restricted "agonistically" shaped disaccharide scaffold in place of the native βGlcN(1→6)GlcN Lipid A backbone appears to provide a useful tool for modulation of TLR4·MD-2-mediated immune signaling. Thus, synthetic αGlcN(1↔1)αMan-based Lipid A mimetics represent the key structures for the advanced development of pharmaceutically applicable immuno-therapeutics or vaccine adjuvant candidates.

■ EXPERIMENTAL SECTION
General Synthetic Methods. Reagents and solvents were purchased from commercial suppliers and used without further purification unless otherwise stated. Dichloromethane was distilled from CaH 2 and stored over activated 4 Å molecular sieves (MS). THF was distilled over Na/benzophenone directly prior to use. Other solvents were dried by storage over activated MS for at least 48 h prior to use [toluene (4 Å), acetonitrile (3 Å), and DMF (3 Å)]. Residual moisture was determined by colorimetric titration on a Mitsubishi CA-21 Karl Fischer apparatus and did not exceed 20 ppm for dry solvents. Reactions were monitored by TLC performed on silica gel 60 F254 HPTLC precoated glass plates with a 25 mm concentration zone (Merck). Spots were visualized by UV light followed by dipping into a H 2 SO 4 −p-anisaldehyde solution or a ninhydrin−EtOH solution and subsequent charring at 250°C. Solvents were removed under reduced pressure at ≤30°C. Preparative HPLC was performed with linear solvent gradients on a YMC Pack SIL-06 250 × 20 mm, S-5 μm, 6 nm  In the disaccharides, the mannose NMR signals are indicated by primes. The purity (>95%) was determined by LC-MS and HRMS. HPLC-MS was performed by injections of 0.01−0.1% CH 3 CN solutions into a Shimadzu LC-10AD VP system equipped with two gradient pumps, a degasser, a Shimadzu LCMS 2020 detector, and an AllTech 3300 ELSD detector. Analytes were eluted over a Phenomenex Jupiter 5μ C4 300A column using mobile phase A = H 2 O (0.1% HCOOH) and mobile phase B = CH 3 CN (0.1% HCOOH) in linear gradients from 5% B to 100% B and a flow rate of 0.5 mL/min. High-resolution mass spectrometry (HRMS) was carried out on 1−10 mg/L acetonitrile solutions via LC-TOF MS (Agilent 1200SL HPLC and Agilent 6210 ESI-TOF, Agilent Technologies). The mass spectrometer was tuned with Agilent tune mix to provide a mass accuracy below 2 ppm. The data were analyzed using Agilent Mass Hunter Software. MALDI-TOF MS was performed in the negative-ion mode using a Bruker Autoflex Speed instrument with 6aza-2-thiothymine (ATT) as matrix. Optical rotation was measured on a PerkinElmer 243B polarimeter equipped with a Haake water circulation bath and a Haake D1 immersion circulator for temperature control. All , and a solution of TMSOTf (9 μL, 52 μmol) in dry CH 2 Cl 2 (450 μL of a stock solution prepared from 20 μL of TMSOTf in 1 mL of CH 2 Cl 2 ) was added. The mixture was stirred at 0°C for 30 min, and the reaction was quenched by addition of saturated aqueous (sat. aq.) NaHCO 3 (2 mL). The mixture was warmed to r.t. and diluted with EtOAc (100 mL), the solids were removed by filtration over a pad of Celite, and the filtrate was concentrated.