Abstract
A series of novel, synthetic compounds containing lipids linked to a phosphate-containing acyclic backbone are shown to have similar biological properties to lipopolysaccharide (LPS). These compounds showed intrinsic agonistic properties when tested for their ability to stimulate tumor necrosis factor-α in human whole blood and interleukin-6 in U373 human glioblastoma cells without added LPS coreceptor CD14. The presence of the LPS antagonist E5564 completely blocked responses, suggesting that the novel compounds and LPS share a common mechanism of cell activation. Stereoselectivity of the molecules was observed in vitro; compounds with anR,R,R,R-configuration were strongly agonistic, whereas compounds with an R,S,S,R-configuration were much weaker in their activity on human whole blood and U373 cells. We also tested the effect of the compounds in cells transfected with the LPS receptor Toll-like receptor 4 (TLR4), with similar results, further supporting a shared mechanism with LPS. This was confirmed in vivo where the agonists failed to elicit cytokine responses in C3H/HeJ mice lacking TLR4 signaling. Because LPS-like molecules enhance immune responses, the compounds were mixed with tetanus toxoid and administered to mice in an immunization protocol to test for adjuvant activity. They enhanced the generation of specific antibodies against tetanus toxoid. Our results indicate that these unique compounds behave as agonists of TLR4, resulting in responses similar to those elicited by LPS. They display adjuvant activity in vivo and may be useful for the development of vaccine therapies.
Lipopolysaccharide (LPS) is a component of the outer membrane of Gram-negative bacteria, which can stimulate the release of a number of proinflammatory mediators, such as interleukin (IL)-1, IL-6, IL-8, and tumor necrosis factor (TNF)-α by macrophages, monocytes, and other cell types (Glauser et al., 1991). Structurally, LPS is composed of four distinct domains: the membrane-intercalated hydrophobic or lipid A domain, an inner core oligosaccharide (containing heptose and 3-deoxy-d-manno-2-octulosonic acid), an outer core oligosaccharide, and the O-antigenic polysaccharide domain (Rietschel et al., 1994; Zahringer et al., 1994; Schletter et al., 1995). The lipid A moiety of Escherichia coli mimics the biological activities of complete LPS (Galanos et al., 1985). Activation of LPS-responsive cells occurs rapidly after LPS interacts with circulating LPS-binding protein and CD14, a soluble or glycosylphosphatidylinositol-linked cell surface protein necessary for efficient responses to LPS and perhaps interaction with the LPS receptor (Ulevitch and Tobias, 1995). Recently, it was shown that Toll-like receptor 4 (TLR4), a member of a highly conserved family of receptors involved in innate immune responses (Hoffmann et al., 1999), mediates the cellular actions of LPS in vitro (Chow et al., 1999). Consistent with these findings, TLR4 knockout mice or spontaneous TLR4 mutant strains (C3H/HeJ and C57BL/10ScCr) are not responsive to LPS (Poltorak et al., 1998; Hoshino et al., 1999), demonstrating that TLR4 mediates the actions of LPS under physiological conditions.
Molecules mimicking lipid A, the simplest active form of LPS, have been widely reported. With one exception (Pedron et al., 1992), synthetic and naturally derived lipid A-like agonists reported to date contain a saccharide scaffold with lipid substituents (Zahringer et al., 1994,1999; Kusumoto et al., 1999). Such molecules, for example monophosphoryl lipid A, are effective vaccine adjuvants in animal models and in humans (Ulrich and Myers, 1995). Because of the synthetic advantages of a molecule lacking a saccharide scaffold, we have examined simplified agonist structures that would make compound preparation more direct and yield highly purified products. We describe here novel lipid A mimetics that lack a disaccharide backbone yet retain TLR4 stimulatory activity. Structure-activity relationships derived from the study of variants of these synthetic molecules should allow a better understanding of the interactions between TLR4 and its ligands, including LPS. The availability of pure and easily synthesized TLR4 agonists should help define the involvement of TLR4 in biological processes. This goal has been complicated by the difficulty of making defined preparations of bacterial constituents. Finally, the compounds' ease of synthesis promises accessible adjuvants that expand the current options for vaccine development.
Materials and Methods
Reagents.
The LPS antagonist E5564 (Rossignol et al., 1999) and the agonists ER111232, ER111233, ER112040, ER111230, ER112231, ER112093, ER112049, ER112047, ER112066, ER113651, ER119327, ER803022, ER803732, and ER803789 (L. D. Hawkins, manuscript in preparation) were synthesized at Eisai Research Institute of Boston (Andover, MA). The structures were confirmed by 1H NMR,13C NMR, 31P NMR, and mass spectrum. E5564 and the LPS-receptor agonists were solubilized in sterile distilled water at 1 mM concentrations then sonicated with an ultrasonicator (VW-380, Misonix, Inc., Farmingdale, NY) for 2 min immediately before each experiment. Working stock dilutions for the whole blood assay were made using a sterile pyrogen-free 5% dextrose solution (McGaw, Inc., Irvine, CA). Solubilization for other assays was done in PBS. Compounds ER803732, ER803022, ER804059, ER804058, ER804053, and ER804057 test negatively in the limulus amebocyte assay for endotoxin at 10 μM (data not shown). E. coli lipid A was obtained from List Biological Laboratories Inc. (Campbell, CA).
Human Whole Blood Assay for TNF-α.
Blood was collected aseptically from 18- to 50-year old, normal male and female volunteers into sterile tubes containing heparin (20 U of heparin/ml of blood) (Abbott Laboratories, Abbott Park, IL). Aliquots of heparinized blood (400 μl) were added to 48-well plastic tissue culture plates (Invitrogen, Carlsbad, CA) followed by 50 μl of serial 10-fold dilutions of each LPS-receptor agonist. When indicated, antagonist E5564 was added at 1 μM final concentration prior to addition of agonist. The plates were then incubated for 3 h at 37°C, 5% CO2 on a random orbital shaking platform (Bellco Technology, Inc., Vineland, NJ). After incubation, the plates were centrifuged (1000g, 10 min, 4°C); plasma supernatants were removed and then frozen at −80°C. Plasma samples were assayed for TNF-α using the human Quantikine TNF-α ELISA kit (R & D Systems, Minneapolis, MN). Values of 0 pg/ml TNF-α were given to the occasional samples that had absorbance values less than the 15.6 pg/ml point on the TNF-α standard curve. The values presented here are the half-maximal stimulatory concentrations (MS50) or the compound concentration needed to stimulate half the level of TNF-α produced by cells stimulated with 10 ng/ml LPS (E. coli O111:B4, List Biological Laboratories Inc.).
IL-6 Release by U373 Cells.
U373 cells were maintained in minimal essential medium supplemented with 10% FBS,l-glutamine, nonessential amino acids, sodium pyruvate, and antibiotics. Cells were seeded in 96-well plates at a density of 20,000 cells/well and allowed to adhere overnight. Serum-containing medium was removed, the cells were washed three times with Hanks' balanced salt solution, and serum-free medium was added. The indicated compounds, or LPS plus soluble CD14 (sCD14, 10 nM), were incubated with the cells for 6 h at 37°C. Culture supernatants were assayed for human IL-6 by ELISA kit per the manufacturer's recommendations (Pierce Endogen, Rockford, IL).
NF-κB Reporter Activity.
HEK293 cells stably carrying plasmids for TLR4, MD-2, and the NF-κB reporter gene ELAM-1-luciferase (HEK-TLR4/MD-2/ELAM-luc) were generated as described. Cells were plated in 96-well plates at a density of 50,000 cells/well and maintained in Dulbecco's modified Eagle's medium plus 10% FBS for 24 h. The medium was removed and replaced with 100 μl of Dulbecco's modified Eagle's medium plus 0.5% FBS per well, and the cells were stimulated with the indicated concentration of compound, LPS alone (100 ng/ml) or LPS with sCD14 (10 nM) for 6 h. Next, Steady-Glo reagent (Promega Corp., Madison, WI) was added to the wells, and the amount of luciferase activity in each sample was quantified in a Wallac 1450 MicroBeta TriLux counter (PerkinElmer, Gaithersburg, MD). Negative control HEK293 cells not expressing TLR4 and MD-2 were transiently transfected with the NF-κB luciferase reporter construct as previously described (Chow et al., 1999).
In Vivo Studies.
All animal work was performed under protocols approved by the Institutional Animal Care and Use Committee of the facility. To test IL-10 elicitation by the compounds, C3HeB/FeJ or C3H/HeJ mice (Jackson Laboratories, Bar Harbor, ME) were injected s.c. with varying doses of synthetic adjuvant in 200 μl of PBS. At 24 h, the mice were bled by cardiac puncture, and the serum was tested for IL-10 by an ELISA kit (Pierce Endogen).
For adjuvant studies, BALB/c mice (Charles River Laboratories, Inc., Wilmington, MA) were immunized s.c. with 0.2 ml of sterile PBS containing 0.25 μg of tetanus toxoid (Accurate Chemical and Scientific, Westbury, NY) and 3 μg of the indicated compound or alum (Pierce Imject Alum, 32 μl/dose; Pierce Chemical Co., Rockford, IL). Synthetic adjuvant solutions were prepared by dissolution in PBS followed by 2 min of sonication prior to mixing with antigen. Mice were immunized three times at 3-week intervals and bled 2 weeks after the third immunization. Serum antibody levels were determined by tetanus toxoid ELISA, using 150 ng/well toxoid (List Biological Laboratories Inc.) and developed with biotin-labeled rat anti-mouse IgG (Zymed Laboratories, South San Francisco, CA) followed by horseradish peroxidase-streptavidin (Southern Biotechnology Associates, Birmingham, AL) and 3,3′,5,5′-tetramethylbenzidine substrate. IgG concentrations were calculated by reference to a standard curve using purified mouse IgG (Zymed Laboratories) adhered directly to the plate.
Results
The aim of this study was to examine the biological activity of a series of simplified lipid A analogs. Based on the structure of lipid A, we hypothesized that a phospholipid dimer containing a total of six lipid chains, or three lipid chains per monomeric unit, and an amine functionality to mimic the glucosamine of the lipid A structure, would be necessary to obtain biological activity (Fig.1). ER112022 was prepared and found to have agonistic lipid A-like activity in human whole blood, determined by the release of TNF-α (MS50 = 0.42 μM, Table 1). This finding prompted us to generate additional dimeric molecules to investigate the structural requirements for biological activity in vitro. Further simplification provided the more active compounds ER111232 (MS50= 0.03 μM) with a shorter dimer linker, ER112066 (MS50 = 0.04 μM) with saturated lipids, and ER119327 (MS50 = 0.02 μM) with nonfunctionalized lipids (Fig. 1 and Table 1). Synthesis of these initial compounds resulted in mixtures of three diastereomers because the serine portion of the molecule was racemic in origin. To evaluate the individual diastereomers of ER119327, analogs were prepared starting with either l-serine to provide theR,R,R,R-isomer (ER803022, MS50 = 0.01 μM), d-serine to generate theR,S,S,R-isomer (ER803732, MS50 = 8.34 μM), or a combination to produce the R,R,S,R-isomer (ER803789, MS50 = 0.54 μM). Among the three diastereomers, the R,R,R,R-isomer (ER803022) was the most active compound in stimulating TNF-α release in human whole blood cultures. By comparison, the activities of the other two diastereomers diminished as the configuration of the serine portion of the molecule was changed. The same pattern was seen for two other pairs of isomers (ER804057/ER804053 from ER112066 and ER804058/ER804059 from ER113651, Table 1). Compounds containing only the three-lipid monomeric unit, or dimers with greater or less than six lipid chains, proved to be inactive (L. D. Hawkins, manuscript in preparation). In sum, the data indicate that these synthetic molecules possess lipid A-like behavior in blood cell cultures containing a variety of LPS-responsive cell types, although even the most active requires a thousandfold higher concentration than LPS to stimulate a similar level of TNF-α release (Poltorak et al., 2000; data not shown). Thus, the compounds are attenuated in comparison to LPS. Most importantly, a range of compound activity can be attained by changing molecular structure.
To determine whether these agonists can mimic the actions of LPS in a more homogeneous cell culture system, we investigated the IL-6 inductive activity of ER119327 and its diastereomers in the LPS/sCD14-responsive human glioma cell line U373, which does not express surface CD14 (Frey et al., 1992). The synthetic lipid analogs tested here elicited IL-6 under serum-free conditions, in the absence of added sCD14 (Fig. 2A). The strongest stimulation of IL-6 release was seen with ER119327 and itsR,R,R,R-isomer ER803022, in which, at the highest dose, the level of IL-6 reached that seen in LPS/sCD14-treated cells. By comparison, the level of activity for ER113651, ER803789, or ER112066 was about one order of magnitude weaker. ER803732 was by far the least effective activator of IL-6 release among the six compounds tested. This order of activity is similar to that observed when the compounds were used to stimulate TNF-α release in human whole blood cultures (Table 1). Although the compounds can stimulate cytokine release in the absence of soluble or surface CD14, Fig. 2B shows that adding sCD14 enhanced responses to a representative compound, shifting the cytokine release curve by about two logs. Figure 2C confirms that the U373 cells do not respond to lipid A in the absence of CD14, even at concentrations (100–1000 nM) in which the synthetic adjuvants approached maximal stimulation.
To learn whether the synthetic agonists activate cells through the same or similar signaling mechanisms as LPS, two activities were evaluated: 1) the ability of an LPS antagonist to inhibit agonist-induced cytokine generation in human whole blood and 2) the ability of the agonists to activate TLR4 in HEK293 cells expressing human TLR4 and the accessory molecule MD-2.
To accomplish this, we incubated whole blood with varying concentrations of the most active isomer, ER803022, in the presence and absence of the LPS antagonist E5564 (1 μM) (Rossignol et al., 1999). As shown in Fig. 3, E5564 prevented ER803022-induced TNF-α release in these cultures. A similar inhibitory effect on TNF-α release was also observed with all diastereomeric variants of ER119327, ER112066, ER113651, and with the parent diastereomeric mixtures (data not shown). These data suggested that the synthetic agonists activate cells through a mechanism shared with LPS, possibly involving the LPS receptor TLR4. Therefore, we examined whether these agonists are capable of inducing NF-κB activity in HEK293 cells stably expressing TLR4, MD-2, and an NF-κB reporter gene (HEK-TLR4/MD-2/ELAM-luc). In these cells, ER803022 and ER119327 were the strongest inducers of NF-κB activity (Fig.4A), whereas the remaining compounds were less active (ER112066 > ER113651 > ER803789 > ER803732). Figure 4B confirms the activity of LPS in this system and its full dependence on CD14. Neither LPS/sCD14 nor any of the compounds (10 μM) tested induced NF-κB activity in HEK293 cells transfected with only MD-2 and the reporter, despite a robust stimulation of NF-κB by TNF-α in these cells (Fig. 4C).
These data indicated that TLR4 mediates the activation of signaling pathways to NF-κB by the synthetic agonists. This is supported by the fact that C3H/HeJ mice, which carry a mutation in the intracellular signaling domain of TLR4 (Poltorak et al., 1998), do not produce IL-10 in response to a representative synthetic agonist, ER803022 (Table2). This is in contrast to the dose-dependent IL-10 response to agonist seen in a wild-type C3H strain.
Finally, since LPS, lipid A, and lipid A homologs are known to increase immune responses, we examined the ability of these synthetic agonists to enhance antibody response to tetanus toxoid immunization in mice. To assess the correlation between in vitro stimulation of cytokine release and in vivo adjuvanticity, we tested the high- and low-activity diastereomers from three diastereomer sets. The results are shown in Fig. 5. A 3-μg dose of the compounds given subcutaneously with antigen increased antigen-specific antibody concentrations 6- to 16-fold over tetanus toxoid given with PBS alone, with the most active compounds equaling alum in adjuvant effect. The R,R,R,R-isomer in each set induced the largest fold increases in titer. These data are consistent with the greater activity of the R,R,R,R-isomers in the cell-based assays described above and clearly demonstrate that these novel agonists display adjuvant activity in vivo.
Discussion
LPS is perceived by mammalian immune systems as a marker of infection (Hoffmann et al., 1999) that stimulates a cascade of proinflammatory cytokines (Wenzel et al., 1995; Medzhitov et al., 1997;Opal and Cohen, 1999). Given the central role of LPS in the inflammatory response to Gram-negative bacteria, there has been strong interest in the mechanism of action of LPS at the cellular and molecular level. In this study, we examined the activity of a series of novel synthetic nonsaccharide-based phospholipids and found them to have intrinsic lipid A-like properties. This observation may provide new tools for the elucidation of a detailed mechanism of action of LPS and an improved route to the production of a class of compounds already known to have adjuvant activity. The ability to produce purified preparations is useful not only for vaccine technologies but also in assessing the involvement of TLR4 ligands in physiological processes. The challenge in this area has been the difficulty of making pure preparations of bacterially derived molecules, a problem amplified by the high sensitivity of the Toll-like receptor system (Golenbock and Fenton, 2001).
Synthesis of the novel series of compounds presented here was undertaken based on observations regarding supramolecular aggregates that are formed by a variety of lipid As in aqueous media. Studies by Brandenburg et al. (Seydel et al., 1993) demonstrated that lipid As, which are biological antagonists, form lamellar structures, whereas the agonistic lipid As such as E. coli lipid A are found to form cubic and cubic-hexagonal supramolecular structures. In an unrelated study, Krisovitch and Regen (1992) analyzed artificial membranes that were composed of phospholipid dimers that were found to have a lamellar supramolecular form. Based on these observations, we hypothesized that compounds containing the lipid profile established for agonistic and antagonistic lipid As, but with a simplified phospholipid backbone that is structurally similar to the Krisovitch and Regen dimers, might have lipid A-like activities.
A six-lipid chain dimeric molecule with stereochemistry, functionality, and lipid chain lengths similar to that found in the natural lipid As, with a stable linker between the monomeric units, was prepared. Racemic serinol provided the portion of the targeted molecule that would mimic the glucosamine backbone of the lipid As. From antagonist research reported earlier (Christ et al., 1995), an ether linkage is found to be suitable for linkage between the serinol alcohol and the primary fat chains, whereas the remaining fats should be positioned in a fashion similar to that of the natural lipid As. On this basis, we prepared ER112022 (Fig. 1 and Table 1), which demonstrated agonistic LPS-like activity in vitro.
This observation prompted more detailed investigation of the structure-activity relationships of these compounds. Monomeric units of the molecule were found to be inactive. A shorter distance between the two units gave stronger agonistic activity. A comprehensive analysis of this variable will be reported elsewhere, but the distance between the phosphates in ER111232 approximates the distance observed for the lipid moieties of lipid As when using simple modeling techniques (L. D. Hawkins et al., manuscript in preparation).
We also demonstrated that less functionality on the lipid chains, such as saturation and the loss of the β-keto group, yielded a slightly more active agonist. Strikingly, the chirality of the serine portion of the molecule plays a large role in activity. Starting from the most active diastereomeric mixture, ER119327, the R,R,R,R-,R,S,S,R-, and R,R,S,R-isomers (equivalent toR,S,R,R) were prepared. As predicted, if the stereochemistry of the natural lipid As is presumed to be optimal, theR,R,R,R-isomer was found to be the most active of the three diastereomers. The same pattern was seen in other diastereomer series.
The biological behavior of these molecules is consistent with stimulation of the LPS response pathway. They induce TNF-α from human whole blood and IL-6 from an adherent human glioma line, U373, in the absence of added sCD14, although the activity of the synthetic ligands is enhanced by the presence of sCD14 (Fig. 2B; Lien et al., 2000). Their activity in whole blood is blocked by the synthetic LPS antagonist E5564, which prevents mortality in animal models of LPS-mediated sepsis (Rossignol et al., 1999). Like endotoxin, the compounds stimulate NF-κB-driven gene expression in a cell line carrying the TLR4 receptor but are inactive in the absence of TLR4. They do not stimulate cytokine responses in mice that are deficient in TLR4 signaling. Finally, as has been shown for lipid A derivatives such as monophosphoryl lipid A (Ulrich and Myers, 1995), these molecules are able to increase antibody titers against protein antigens when coinjected into mice. Stimulatory activity in all of the in vitro assays followed the same ranking, with ER119327 and itsR,R,R,R-diastereomer ER803022 showing the highest activity and the R,S,S,R-diastereomer ER803732 showing the poorest stimulation. This suggests that the processes measured in whole blood and in cell lines are mediated by the same receptor, defined here as TLR4.
The fact that an LPS antagonist prevents the agonistic effects of these compounds implies that they share some portion of their stimulatory mechanism with LPS. Known proteins involved in the transfer of LPS to TLR4 on the cell surface include LPS-binding protein, CD14, and MD-2. Since the compounds activate cytokine production by CD14-negative U373 cells in the absence of serum or added LPS-binding protein or CD14, variability in binding to these proteins is likely not the source of their differential activity. Instead, it may reflect direct interaction with TLR4 and/or MD-2. This interaction is clearly not dependent on the presence of carbohydrate in the ligand, a finding consistent with previous observations on the critical nature of lipid conformation in LPS activity (Pedron et al., 1992).
The degree to which receptor interaction, measured in terms of cytokine stimulation, is sensitive to changes in the configuration of the backbone of these molecules is also of interest and parallels observations on synthetic lipid A analogs previously reported (Rossignol et al., 1999) and on the stereochemistry of macrophage-activating lipopeptide-2 interaction with TLR2 (Takeuchi et al., 2000). This responsiveness to chiral structure may reflect the significance of lipid arrangement about the backbone and its effects on the overall conformation of the molecule and thereby its interaction with receptor(s). The actual conformation of these novel molecules remains to be addressed by more direct analysis. Finally, the inactivity of monomeric as opposed to dimeric units suggests receptor dimerization as a possible step in activation, as has been reported for TLR2 receptors in vitro (Yang et al., 1999; Ozinsky et al., 2000), or may reflect dependence on a particular lipid structure.
TLR4 has been defined as the predominant receptor for LPS in mice (Poltorak et al., 1998; Hoshino et al., 1999) and transduces signals leading to cellular activation and cytokine release (Yang et al., 2000). Although the basis for in vivo activity of endotoxin-like molecules as adjuvants is not fully defined, it is likely that this signaling pathway leads to effects on the immune system that enhance responses to antigens given concurrently with the adjuvant (De Becker et al., 2000). When used as adjuvants in these studies, the synthetic agonists gave increases in serum antibody titer. For two of the agonist structures, the R,R,R,R-diastereomer gave significantly higher titers than the R,S,S,R-diastereomer, which correlates with their relative in vitro stimulatory capacities in human cells. These simplified synthetic molecules, which can be produced to high purity, will be useful not only as tools for understanding the mechanics of LPS stimulation but also offer the potential for use as enhancers of immune response.
Acknowledgments
We gratefully acknowledge Hongsheng Cheng for technical assistance and Linda Buckley for help with the manuscript.
Footnotes
-
↵1 Present address: BioChem Pharma, 30 Bearfoot Rd., Northborough, MA 01532.
- Abbreviations:
- LPS
- lipopolysaccharide
- IL
- interleukin
- TNF
- tumor necrosis factor
- TLR4
- Toll-like receptor 4
- PBS
- phosphate-buffered saline
- ELISA
- enzyme-linked immunosorbent assay
- MS50
- half-maximal stimulatory concentration
- FBS
- fetal bovine serum
- sCD14
- soluble CD14
- NF-κB
- nuclear factor-κB
- HEK
- human embryonic kidney
- ELAM
- endothelial leukocyte adhesion molecule
- Received July 23, 2001.
- Accepted November 11, 2001.
- The American Society for Pharmacology and Experimental Therapeutics