The instantaneous immune response to infection by Gram-negative bacteria depends on the structure of lipopolysaccharide (LPS, also branded as Endotoxin), a large (MW ca. 20 kDa) complex heterogeneous glycan constituting the outer leaflet of the bacterial outer membrane (1). A small amphiphilic terminal motif of LPS known as lipid A is responsible for induction of the host innate immune responses by interacting with several pattern recognition receptors (PRRs) normally expressed by mammalian immune cells. Detection of picomolar quantities of pathogenic LPS by a particular host PRR—the germline-encoded transmembrane Toll-like Receptor 4 (TLR4) complex—results in the initiation of diverse intracellular pro-inflammatory signaling cascades (2). An acute inflammatory response triggered upon activation of TLR4 by endotoxic lipid A involves downstream expression of pro-inflammatory cytokines, adhesive proteins, prostaglandins, and reactive oxygen species and is directed at elimination of bacterial disease (3). Thus, activation of the innate immune response through LPS-driven engagement of TLR4 during infection inherently contributes to spontaneous healing in immunocompetent host. However, uncontrolled upregulation of TLR4—mediated signaling can lead to an unrestrained cytokine storm and sepsis which provoked the development of TLR4 inhibiting strategies (4–6). On the other side, it has been shown that the inefficiency of the TLR4-driven inflammation could be responsible for the pathogenicity of Gram-negative sepsis (7–9). Besides, activation of TLR4 links the innate and adaptive immunity (10, 11) justifying application of less toxic TLR4 agonists as vaccine adjuvants (12, 13). The pivotal “endotoxic principle” of LPS resides in its rather small (MW ca. 2 kDa) terminal amphiphilic motif—a glycophospholipid lipid A which can be directly bound by the hydrophobic binding pocket of myeloid differentiation factor 2 (MD-2)—a TLR4 co-receptor protein (14). Binding of lipid A provokes dimerization of two TLR4/MD-2/LPS complexes which results in the activation of adaptor proteins and formation of a MyDDosome—an intracellular supramolecular organizing centre (SMOCs) coordinating the induction of the pro-inflammatory signaling cascades (2, 15–17). LPS-induced assembly of the dimeric [TLR4/MD-2/LPS]₂ complexes triggers conformational changes in the intracellular Toll/IL-1R (TIR) domain of TLR4 (18). The latter event leads to the activation of diverse signaling molecules: MyD88 (myeloid differentiation primary response 88) (19) and TRIF [TIR (Toll IL-1R)-domain containing adaptor inducing Interferon-β (20)], along with the respective adaptor proteins MAL [MyD88-adaptor-like, also known as TIR-domain containing adaptor protein TIRAP (21)] and TRAM [TRIF-related adaptor molecule (22)]. TLR4 is able to induce different sets of pro-inflammatory responses: rapid MyD88-dependent induction of cytokines which starts at the plasma membrane, and TLR4/MD-2/LPS internalization-driven induction of type I interferons which depends on TRIF and originates from endosomal membranes. Although TLR4 has long been considered an important therapeutic target, very few structurally defined TLR4 - activating molecules have been developed so far. The reason for this deficiency resides in rather intricate and multifaceted TLR4-LPS structure-activity relationships which adds to the complexity in designing suitable TLR4 activating ligands.

The chemical structure of lipid A relies on a β(1 → 6) linked disaccharide backbone (βGlcN(1 → 6)GlcN) which is highly conserved in most bacterial species. The diglucosamine skeleton possesses one or two phosphate groups at positions 1 and/or 4′ and bears varying number long-chain (R)-3-hydroxyacyl- or (R)-3-acyloxyacyl residues which are symmetrically or asymmetrically allocated within the diglucosamine backbone at positions 2,3- and 2′,3′ (Figure 1A) (1). Hexaacylated lipid A of E. coli and N. meningitidis possessing two (R)-3-acyloxyacyl and two (R)-3-hydroxyacyl residues variably distributed within β(1 → 6) diglucosamine backbone (4+2 and 3+3 acylation pattern, respectively) exhibit the utmost TLR4-activating capacity, whereas tetra- or pentaacylated lipid A variants can act as antagonists at human (h)TLR4 [such as lipid IVa (23), or non-pathogenic lipid A from R. sphaeroides (24)].

The co-crystal structures of hexaacylated E. coli Re- and Ra-LPS with mouse (m) or (h)MD-2/TLR4 complex, respectively, reveal that five long-chain acyl residues of lipid A (out of six) are incorporated into the hydrophobic Leu- and Phe-reach binding pocket of MD-2, while the sixth lipid chain is presented on the surface of the co-receptor protein MD-2 (Figure 2A) (14, 25). Binding of lipid A induces conformational rearrangement in MD-2 shifting the Phe126 loop inwards which, together with exposure of a lipid chain, creates a major hydrophobic interface for the dimerization with the second TLR4^*/MD-2^*/LPS complex (26, 27). Juxtapose, all lipid chains of hTLR4 antagonist ligands such as tetraacylated lipid IVa (23) (a biosynthetic precursor of E. coli lipid A) and synthetic drug-candidate Eritoran (28) are housed within the binding groove of hMD-2 (Figure 2C) and the Phe126 residue of MD-2 is exposed to solvent (23, 28) which prevents dimerization and abolishes the induction of pro-inflammatory signaling. Deduced from the co-crystal structures and molecular dynamic simulations, the Phe126 residue of MD-2 stabilizes presentation of the 6^th acyl chain of lipid A on the surface of the protein and serves as “hydrophobic switch” which either promotes (for agonists) or blocks (for antagonists) the dimerization and the formation of a hexameric receptor complex (14, 29). Ionic interactions between lipid A phosphate groups and positively charged side chains in MD-2 and TLR4^* (Lys, Arg) also substantially contribute to binding and dimerization (30). Remarkably, agonist lipid A variants are bound by MD-2 in inverted (by 180°) orientation compared to antagonist ligands. The position of the lipid A (±180°) within the binding cleft of MD-2 was demonstrated to be crucial for the expression of TLR4-mediated biological activity. Some lipid A variants exhibit species-specific TLR4-dependent activities which is largely determined by the binding orientation of lipid A in the binding groove of MD-2. For instance, tetraacylated lipid IVa acts as antagonist at hTLR4 but performs as an agonist at mTLR4 wherein it binds in an inverted by 180° orientation (similar to E. coli lipid A bound by hMD-2) and exposes one (out of four) lipid chain on the surface of mMD-2 (Figure 2B) (26, 31). The reason for exposure of a particular lipid chain (always 2N-acyl chain linked to the proximal GlcN moiety), as well as the basis for “inverted” positioning of the agonist lipid A in the binding pocket of MD-2 (rotation of the lipid A backbone by 180°) compared to antagonist lipid A variants is not yet understood.

Since the structure of the hydrophobic fatty chain region of lipid A belongs to primary determinants of “endotoxicity,” modulation of the length, number and distribution pattern of acyl chains was typically employed for alteration of cytokine-inducing capacity of genetically engineered (32, 33), isolated (34), and synthetic lipid A variants (35). Shortening or lengthening of only one acyl chain in the hydrophobic region of lipid A for 2xCH₂ atoms frequently results in the accidental effects on TLR4-mediated signaling. This well-known dependency of TLR4 activation on the minimal aberrations in the length of lipid chains complicates prediction of biological activity of natural lipid A/LPS variants. Additionally, structural heterogeneity of lipid A isolates in respect to number of lipid residues and acylation pattern, along with potential contaminations by other bacterial components makes an application of lipid A/LPS isolated from bacterial cultures rather challenging for therapeutic use. As example, a pharmacological lipid A preparation—licensed vaccine adjuvant MPL® (clinical-grade MPL®, monophosphoryl lipid A manufactured by partial hydrolysis of S. enterica serovar Minnesota Re595 LPS)—represents a heterogeneous mixture of differently acylated counterparts exerting partial TLR4 antagonist activity which thwarts the adjuvant's efficacy (36). Besides, species—specific differences (human vs. mice) in lipid A recognition by the TLR4/MD-2 complex (26, 37) add to discrepancies in estimating therapeutic effect using experimental data obtained from in vivo models. To address these issues, we pursued a multidisciplinary approach combining synthetic glycochemistry, structural biology, and immunology.

Hydrophobic interaction between lipid chains are commonly dominant so that the six acyl chains of E. coli lipid A tend sticking together to form a solitary hydrophobic cluster. The exposure of a single 2N-acyl lipid chain on the surface of MD-2 must be, therefore, energetically unfavored unless there are some specific factors which assist in destabilization of hydrophobic interactions. We proposed that the exposure of 2N-acyl chain of E. coli lipid A on the surface of human or mouse MD-2 is enabled through intrinsic flexibility of a three-bond β(1 → 6) glycosidic linkage connecting two GlcN residues of the diglucosamine backbone of lipid A (Figure 2D). Protein driven rearrangement of the carbohydrate backbone of lipid A is facilitated by rotation about the β(1 → 6) linkage which results in a dramatic change in a 3D-molecular shape of the lipid A backbone seen in the co-crystal structures (14, 26). Accordingly, upon interaction with MD-2/TLR4, the proximal GlcN ring of lipid A (GlcN ring facing secondary dimerization interface) is relocated in a skewed orientation which might assists in weakening of intermolecular hydrophobic interactions between long-chain lipid residues (Figure 2E). This could allow a protrusion of 2N-acyl lipid chain (attached at the proximal sugar ring) on the surface of MD-2 without significant entropic loss (Figure 2F) (38). Thus, the inherently flexible β(1 → 6)-linked disaccharide backbone of lipid A readily adjusts its molecular shape upon interaction with the MD-2/TLR4 complex which is unveiled in the co-crystal structures. Taking into account intrinsic plasticity of MD-2 (39), the major conformational changes in both the ligand and the protein take place upon binding of the lipid A portion of LPS. This reduces the predictability of ligand-protein interaction such as the binding pose of a lipid A-like ligand (±180°), the deepness of insertion of lipid A (or a lipid A analog) into the binding groove of MD-2 (that significantly varies for agonist and antagonist lipid A), and the rearrangement behavior of Phe126 loop of MD-2 (29). Since the major lipid A recognizing protein MD-2 reveals multiple conformational states which exist in ligand-dependent equilibrium, the designing of TLR4 activating molecules using common structure-based approaches seems unsuitable for pharmacological manipulation of the TLR4 system.

We hypothesized that fixing the molecular shape of the diglucosamine backbone of lipid A in a skewed conformation found in the agonist TLR4/MD-2/lipid A co-crystal structures (PDB Code: 3FXI, 3VQ1, 3VQ2) using chemical methodologies could furnish lipid A mimetics that would be predictably and specifically bound by the TLR4/MD-2 complex in an “agonist” orientation (i.e., without ±180° uncertainty). Such lipid A mimetics should retain the major binding characteristics of E. coli lipid A, and simultaneously induce structural rearrangements in MD-2 which are required for inducing the receptor complex dimerization, i.e., moving the Phe126 loop inwards (Figure 3A) (26, 27, 40).

Using glycochemistry approaches we have synthetically assembled a non-reducing two-bond linked α,α-(1↔1′)-connected disaccharide to apply as surrogate for intrinsically flexible three-bond linked βGlcN(1 → 6)GlcN backbone of lipid A (Figure 3B) (41). High conformational rigidity of the α,α-(1↔1′)-linked disaccharides and their peculiar twisted molecular shape imitating the tertiary structure of the glucosamine backbone of TLR4/MD-2 bound E. coli lipid A (PDB Code: 3FXI, 3VQ2) were the key features our design relied on. The TLR4-mediated activity of disaccharide lipid A mimetics (DLAMs) relies, therefore, on both the 3D-tertiary structure of the non-reducing disaccharide backbone and on the number/length/distribution of lipid chains. Of note, all lipid A analogs and mimetics developed so far were based on either natural β(1 → 6)-linked diglucosamine backbone or even more flexible skeletons where one or both glucosamines were replaced by a linear aglycon (as in RCX 527 or E6020) (Figure 1B) (42). Also, a MD-2-specific TLR4 agonist having no similarity to LPS has been developed, although this peptidomimetic showed specificity to mouse TLR4 only (43).

The DLAMs are based on the conformationally restricted αGlcN(1↔1)αMan scaffold and therefore abbreviated α,α-GM-DLAMs (α,α-(1↔1)-linked Glucosamine-Mannose-Disaccharide-Lipid A-Mimetics). Here we validate the ability of synthetic TLR4 agonist α,α-GM-DLAMs to induce tailored activation of cellular pro-inflammatory responses in human and mouse TLR4/MD-2/CD14(±)-transfected cells, TPA-primed THP-1 cells, mouse macrophages (mBMDM), human mononuclear cells (MNC), and several types of lung epithelial cells. We demonstrate that DLAMs can induce robust TLR4-mediated release of cytokines in human and murine immune cells (MNC and BMDM, respectively) at picomolar concentrations, whereas the activation of TLR4-dependent responses in human airway epithelial cell lines and TPA-primed THP-1 macrophages required nanomolar DLAMs concentrations. We also assess the affinity of DLAMs to MD-2/TLR4 by competitive cellular assays using a known TLR4 antagonist which competes with DLAMs for the same binding site on MD-2. We demonstrate that graded and adjustable modulation of the pro-inflammatory signaling induced by DLAMs in human and murine cell lines can be achieved in a concentration-dependent manner through modification of the chemical structure of DLAMs.

Thus, disaccharide lipid A mimetics offer several advantages compared to natural lipid A variants in regard to (1) chemical stability: in DLAMs the inherently labile glycosidic phosphate group P-1 is replaced by a stable secondary phosphate group; (2) chemical purity: the highest purity (99.9%) and chemical homogeneity of DLAMs is confirmed by multiple analytical techniques such as ¹H-, ³¹P, ¹³C-NMR (Bruker 600 MHz), HRMS and MALDI-TOF; (3) biological purity: DLAMs are produced by chemical synthesis which is performed starting from glucosamine by application of organic solvents and chemical reagents exclusively; all synthetic intermediates and target compounds are carefully isolated and thoroughly purified using chromatography in organic solvents which guaranties the absence of any biological contaminations. This renders the immunobiological data obtained with DLAMs to absolutely reliable, highly trustable and reproducible. Importantly, modulation of cellular pro-inflammatory responses induced by synthetic DLAMs has been independently studied and confirmed by four European laboratories.

The affinity of lipid A to MD-2 is largely determined by its primary chemical structure and depends on the number, lengths and structure of lipid chains and the phosphorylation status of its β(1 → 6)-linked diglucosamine backbone (52, 53). Likewise, the efficiency and tightness of ligand-induced TLR4 complex dimerization is governed by the acylation pattern of lipid A and the presence/absence of negative charges (e.g., phosphate groups). Since the 3D-conformation of the diglucosamine backbone of lipid A can be easily amended by rotation about highly flexible β(1 → 6) glycosidic linkage (54), the disaccharide backbone of lipid A readily changes its molecular shape upon binding by the MD-2/TLR4 complex. Consequently, the GlcN rings of the diglucosamine backbone adopt a co-planar relative arrangement for MD-2/TLR4-bound antagonist lipid A variants (Figures 14A,B) and a skewed relative arrangement for MD-2—bound agonist ligands (Figures 14D,E). We have recently shown that the 3D tertiary structure of the MD-2—bound lipid A is decisive for the expression of specific biological activity and that the particular binding orientation of the lipid A molecule in the binding grove of MD-2 (±180°) can be impelled by fixing the 3D-molecular shape of the disaccharide skeleton of lipid A in a pre-defined conformation (41, 45). Following this concept, we have developed potent synthetic TLR4 antagonists derived from a co-planar arranged β,α-(1↔1)-linked diglucosamine scaffold (Figure 14C) (38, 55) and TLR4 agonists based on a twistedly shaped disaccharide scaffold (Figure 14F, this paper) (41). Notably, our approach allows to surmount species-specificity in ligand recognition by the TLR4 complex for all types of biological activities (endotoxic/antiendotoxic) and to predictably and gradually modulate TLR4 activation.

Thus, in contrast to native lipid A derived from a flexible three-bond linked βGlcN(1 → 6)GlcN backbone which can spontaneously adjust its molecular shape upon binding to the MD-2/TLR4 complex (Figures 2D, 3A), the TLR4 agonist DLAMs are based on the extraordinary rigid α,α-1,1′-linked spaciously twisted disaccharide backbone mirroring the molecular shape of the MD-2 bound E. coli lipid A. The relative orientation of two sugar rings in the α,α-(1↔1) linked disaccharides depends on the conformation of glycosidic linkage (more precisely, on the torsion angles around glycosidic (1↔1) linkage) and is governed by specific carbohydrate related effects (the anomeric and exo-anomeric effects). The unique “skewed” 3D-molecular shape of α,α-(1↔1) linked disaccharides was determined by X-ray crystallography: in particular, it was demonstrated that the conformation of α,α-(1↔1) linkage is conserved in all circumstances and that the characteristic “skewed” molecular shape of α,α-(1↔1) linked disaccharides is sustainably maintained and does not depend on the nature of functional groups attached at either sugar ring (56–59). The exceptional rigidity of an α,α-(1↔1) glycosidic linkage and a restrained skewed arrangement of α,α-(1↔1)-linked sugar rings was also confirmed by molecular dynamics simulations that corroborated the existence of a single conformational minimum (54, 60–62). Thus, conformationally restrained αGlcN(1↔1)αMan disaccharide with its twistedly oriented sugar rings mirroring the 3D-tertiary structure of the diglucosamine skeleton of MD-2/TLR4–bound E. coli lipid A provides a versatile scaffold for synthetic TLR4—activating lipid A mimetics.

Initially we assessed the specificity of synthetic lipid A mimetics toward human and mouse TLR4/MD-2 complexes by examining the ability of DLAMs to induce the pro-inflammatory signaling in hTLR4/hMD-2/CD14(±) and mTLR4/MD-2 transfected human embryonic kidney (HEK)293 cells, respectively. DLAMs 1 and 4 having 2 × CH₂ longer acyl side chain at the Man residue facing secondary dimerization interface were less efficient activators of NF-κB compared to DLAMs 2 and 5. Thus, shortening the secondary acyl chain by 2 × CH₂ resulted in enhanced TLR4-mediated signaling, whereas the sites of attachment of acyl chains and phosphate group (position 4 or 6) did not exert significant effect on the efficiency of SEAP induction (Figure 5).

We also examined the memCD14-dependency of DLAMs by comparing the IL-8 release induced by DLAMs in HEK293 cells transfected with hTLR4/hMD-2/hCD14 or hTLR4/hMD-2 (Figure 6). memCD14 enhanced the sensitivity of hTLR4/MD-2 complex to DLAMs diphosphates at picomolar concentration range (0.1–10 ng/mL) (Figure 6A). At higher concentrations (100–1000 ng/mL) the hTLR4 activation induced by DLAMs was not significantly influenced by the presence of CD14. In contrast, DLAMs monophosphates, similarly to MPLA, were CD14—dependent in the whole concentration range showing no activity below 1000 ng/ml in the CD14 deficient cells. Since DLAM-monoP 3 is a 6-dephosphorylated counterpart of the DLAM-diP 1 and DLAM-monoP 6 is a 4-dephosphorylated counterpart of the DLAM-diP 4, and both DLAMs-monoP are “mimetics” of MPLA, the CD14—dependency is closely related to the presence of the second phosphate group (phosphate group exposed at the secondary dimerization interface) in the molecule. CD14 is known to enhance the sensitivity of the TLR4 complex to endotoxin by transferring LPS from LBP to the binding pocket of MD-2 (48, 63), as well as to be involved in the CD14—dependent TLR4/MD-2/LPS endocytosis (64, 65) and in the regulation of tightness of the TLR4 complex dimerization (66).

Monophosphorylated DLAMs, as expected, showed reduced activity compared with DLAMs diphosphates, but still overperformed MPLA in inducing NF-κB regulated signal transduction pathway both at h- and mTLR4. For DLAMs-monoP the TLR4 activation profiles were similar in human and mouse systems highlighting DLAM 6 as the most potent, followed by DLAMs 3 and 7 (Figures 6B, 7). Thus, modulation the activity of DLAMs-monoP was achieved by varying acylation sites at the Man moiety (acyloxyacyl chain at position 6 for DLAM 6 and at position 4 for DLAM 3) and two 4,6-hydroxyacyl chains for DLAM 7.

To examine the impact of αα-GM-DLAMs on cytokine induction in MNC, mononuclear cells were stimulated with a wide concentration range of DLAMs or with E. coli LPS and MPLA as controls. Picomolar doses of DLAMs-diP were sufficient to induce potent induction of TNF-α and IL-6 in human MNC (Figures 8A,B). Dose-dependent release of IL-6 and TNF-α induced by higher concentrations of DLAMs allowed for determination of EC50 values which were in the picomolar range for DLAMs diphosphates (Figures 8C,D). The induction of cytokines by αα-GM-DLAMs correlated to their chemical structure revealing shorter-chain 6-lipidated DLAM 5 as the most potent in inducing the release of TNF-α and IL-6, followed by shorter-chain 4-lipidated DLAM 2, whereas DLAMs 4 and 1 having longer secondary acyl chain at Man (secondary dimerization interface) were somewhat less efficient in the picomolar range (below 500 pg/ml). At higher concentrations, the maximum responses induced by DLAMs-diP differed for release of IL-6 and TNF-α disclosing DLAMs 4 and 5 as most efficient in respect to IL-6 secretion, whereas the DLAMs 5 and 2 revealed higher potency (lower EC50 values). Accordingly, synthetic manipulations of the length of acyl side-chain as well as the site of attachment of acyloxyacyl residue and the phosphate group at Man moiety allowed for the fine-tuning of the TLR4-mediated cytokine release.

The cytokine-inducing potential of DLAMs-monoP was associated with their primary chemical structure missing a phosphate group at the dimerization interface and dropped from picomolar to nanomolar concentration range. The structure-activity relationships among DLAMs-monoP were cell-line and species-independent and revealed the 6-acyloxyacyl DLAM 6 (a monophosphorylated counterpart of DLAM 4) as the most potent, followed by 4-acyloxyacyl DLAM 3 (a monophosphorylated counterpart of DLAM 1) and 4,6-lipidated DLAM 7 (Figures 8E,D). Notably, DLAM 6 overplayed S. minnesota MPLA in inducing the release of both TNF-α and IL-6 in MNC.

Remarkably, E. coli Re-LPS (Kdo₂-lipid A)—which has at least a 10-fold higher potency in TLR4-mediated activation of cytokine release than the corresponding lipid A (67)—was less powerful in inducing TNF-α and MCP-1 in TPA-primed THP-1 macrophages compared to DLAMs-diP (Figures 10A,B). Two Kdo residues attached at position 6 of lipid A (Re-LPS) are sufficient to endorse lipid A with a full E. coli LPS-like activity (68). Also the presence of only one Kdo residue is known to substantially enhance the TLR4-dependent activity of Kdo-lipid A compared to lipid A alone (35). The induction of NF-κB signal transduction pathway by diphosphate DLAMs-diP was as strong as the response elicited by E. coli Re-LPS indicating that the unique structural features of Kdo-deficient DLAMs render lipid A mimetics exceptionally powerful TLR4 agonists. This has been confirmed also in mouse BMDMs where DLAMs 2 and 5 showed 10-fold higher potency in inducing TNF-α compared to E. coli lipid A and were also more potent than Re-LPS (Figure 12A). Monitoring the secretion of IL-1β in TPA-primed THP-1 macrophages revealed an inverted activation profile, with longer-chain αα-GM-DLAM 1 being most active, indicating the involvement of different structural factors in ligand recognition for induction of IL-1β pathway (Figure 10C) (69). Indeed, intracellular LPS-mediated IL-1 release in human monocytic cell lines is dependent on caspase-4/5 which regulates the secretion of inflammasome-activated IL-1β (70, 71). Activation of the NLRP3 inflammasome in macrophages requires two simultaneous signals: the LPS-induced activation of the NF-κB pathway as a priming stimulus which activates the transcription of NLRP3 and pro-IL-1β, as well as a second signal to induce the assembly of inflammasome complex via recruitment of pro-caspase-1 and adaptor molecule ASC (apoptosis-associated speck-like protein containing a caspase-recruitment domain) (72, 73). Although macrophages are believed not being able to release IL-1β in response to LPS without exogeneous ATP as 2nd stimulus (74), it has been demonstrated before that differentiation of THP-1 cells using high concentrations of TPA (100–200 nM) or prolonged differentiation times induces a more activated macrophage phenotype (75, 76). Such macrophage phenotype is associated with secretion of IL-1β in response to LPS alone as a result of spontaneous release of endogenous ATP and a moderate activation of caspase-1 (77). Similar mechanisms may be involved in our experimental set-up as well. Additionally, CD14-mediated internalization of TLR4/MD-2/LPS complexes into endosomal compartments in TPA-primed THP-1 macrophages by a process mediated by tyrosine kinase Syk and its downstream effector phospholipase C (PLC) γ2 could follow the LPS or DLAMs challenge (64, 65). Therefore, we suppose that the DLAM-mediated caspase-4/5 activation in TPA-primed THP-1 cells culminating in elevated IL-1β secretion may occur by a physiological process of DLAM-induced CD14—dependent TLR4/MD-2/DLAM internalization and activation of cytosolic LPS-sensing receptors. The specificity and high affinity of DLAMs-diP to CARD of caspase-4 has been previously confirmed in in-vitro studies (44).

We also assessed the cytokine-inducing capacity of DLAMs in two human bronchoepithelial cell lines, BEAS-2B and Calu-3. The immune response of lung epithelium toward LPS has gained a lot of interest over the last years, in particular with respect to allergy and asthma (78). Tailored activation of immune signaling in lung epithelium through lipid A-like compounds may therefore provide possible therapeutic interventions for skewing the early immune response away from Th2 bias. The responsiveness of human airway epithelial cells toward LPS is limited due to low expression of MD-2, Jia et al. (51) therefore, nanomolar concentrations of DLAMs were required to induce cytokine release in bronchoepithelial cells. Regarding secretion of IL-8, lipid A mimetics behaved similarly in both cell lines revealing DLAMs 5 and 4 the most potent inducers of pro-inflammatory signaling. Longer chain 4-lipidated DLAM 1 showed substantially diminished IL-8 release. At higher DLAMs concentrations (100–1000 ng/mL) the activity abruptly dropped which could be related to a change in the aggregation state of DLAMs in aqueous solutions requiring the action of mem-CD14 for efficient transfer of a single DLAM molecule to MD-2. Membrane-bound CD-14 that augments cellular response to LPS is not sufficiently expressed in bronchoepithelial cell lines (79) which results in poorer activity of CD14-dependent TLR4 ligands. Along these lines, the monophosphate DLAMs 3 and 7 were nearly inactive at the concentrations tested, which is consistent with the full CD14 dependency of DLAMs-monoP-induced TLR4 activation. An almost identical pattern of agonist activity was monitored for the induction of IL-6 release with DLAM 5 as the most potent and DLAM 1 as the least efficient.

Generally, we assessed two essential properties of DLAMs to interact with TLR4 complex, namely, affinity and efficacy. The affinity of DLAMs to TLR4/MD-2 complex can be defined as the ability to bind to MD-2, whereas efficacy can be determined as an inherent property of the MD-2-bound DLAM to induce the dimerization and activation of the TLR4 complex. Along these lines, the EC50 values could describe the affinity of DLAMs to MD-2/TLR4 complex, whereas efficacy is more related to maximum response. Different DLAMs can bind to MD-2/TLR4 with similar affinity but differ in terms of relative efficacy (maximum response): for example, DLAMs 1 and 4 or DLAMs 2 and 5 in BMDMs (Figure 12A) or DLAMs 2 and 5 in MNC (Figures 8C,D). We measured the activity by secretion of cytokines (that may also lead to auto-self-activation) as well as by other parameters which by themselves can be influenced by additional factors. Hence, the whole test system for assessment of TLR4 ligands affinity, activity, and potency is very complex and that was the reason for using many different in-vitro systems to examine the potency of DLAMs to activate TLR4-mediated signaling. Since the recognition of LPS or DLAMs is a multifaceted multistep process where many innate immune proteins are involved, specific structural features of DLAMs could have enormous impact on the mode of their binding to both CD14 and MD-2. This can also affect the tightness and the mode of the TLR4 complex dimerization which, in turn, could result in differential cytokine release. CD14 has been shown to bind a single LPS molecule (49) and shuttle it to the binding pocket of MD-2 (48), although the atomic structure of LPS/CD14 complex has not yet been solved and the structural basis of LPS binding by CD14 is largely unknown. We suppose that the mode of interaction of DLAM with CD14 might also have significant impact on the way how DLAMs are “inserted” into the binding pocket of MD-2 and, consequently, on the following signaling events. Also, the DLAM “load” must have substantial effect on the recognition process since cellular responses to low concentrations of DLAMs are fully memCD14 dependent, whereas the induction of the pro-inflammatory signaling at higher concentrations of DLAMs can proceed in memCD14 independent manner.

Due to their specific structural features such as rigidified disaccharide backbone mimicking the molecular shape of MD-2-bound lipid A, acylation pattern mirroring the one of E. coli LPS, homogeneity, stability and defined chemical structure, DLAMs represent versatile tools for studying the structural basis of lipid A recognition by proteins of the LPS transfer cascade and the molecular basis of TLR4 activation.

In addition to primary chemical structure, we manipulated the three-dimensional molecular shape of lipid A and designed a novel class of TLR4/MD-2-specific glycolipids based on the conformationally restricted synthetic α,α-(1↔1)-linked disaccharide scaffold reflecting the 3D-tertiary structure of TLR4/MD-2-bound E. coli lipid A. These specific structural features of DLAMs, that were created using approaches of synthetic glycochemistry, ensured picomolar affinity to TLR4/MD-2, species-independency in respect to human and mouse systems, and the possibility to predictably modulate the TLR4-mediated activity by chemical modifications. The TLR4/MD-2-mediated cellular pro-inflammatory responses were induced by αα-DLAMs in a CD-14 dependent manner and could be readily modulated through modification of the primary chemical structure of synthetic glycolipids. Both the position of phosphate group (at C-4 or C-6 of Man) facing secondary dimerization interface and the length of acyl chains were accountable for modulation of human and murine TLR4—stimulating activity. DLAMs having shorter secondary acyl chain (C₁₀) were generally more active than DLAMs having longer secondary acyl chain (C₁₂). As to phosphorylation pattern, the 4-phosphate DLAMs (lipidated at position 6 of Man) were more potent than the corresponding 6-phosphates (lipidated at position 4 of Man; compounds having lipid chains of equal length are compared). DLAMs lacking the phosphate group at the sugar moiety closest to the secondary dimerization interface (monophosphate DLAMs) induced diminished inflammatory responses and performed as weak TLR4 agonists showing similar tendency toward higher activity of 6-O-lipidated counterpart. For tailored modulation of TLR4-mediated signaling induced by DLAMs, the positioning of phosphate groups and acyloxyacyl residues decorating the disaccharide backbone, as well as the length of lipid chains can be manipulated by chemical modifications. The chemical synthesis would allow to create an even more broad diversity of conformationally restricted DLAMs varying in the sites of acylation and phosphorylation as well as in the length of acyl chain bordering secondary dimerization interface (C₆ to C₁₆ in length) to warrant customized modulation of TLR4-mediated cellular responses. All these features render DLAMs promising candidates for further development into TLR4-dependent immunomodulating therapeutics which can enhance the innate immune responses as well as into potential vaccine adjuvants.

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

AZ conceived the project. AZ, HH, RB, and RJ analyzed literature and designed detailed research. All authors performed their respective experiments and interpreted results. AZ, HH, RB, SI, and TP analyzed the data and wrote the manuscript. All authors read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.