Limulus amebocyte lysate
Bacterial lipopolysaccharides (endotoxins, LPS) belong to the most potent immunostimulators in mammals. The endotoxic principle of LPS is located in its lipid A moiety, which for Escherichia coli-type LPS consists of a hexaacylated diphosphoryl diglucosamine backbone. This lipid A adopts a cubic inverted aggregate structure from which a conical shape of the molecule can be deduced, whereas the tetraacyl lipid A precursor IVa adopts a cylindrical shape and is endotoxically inactive, but antagonizes active LPS. We hypothesize that non-lipid A amphiphiles with similar physicochemical properties of amphiphilicity, charge, and shape, might mimic the respective lipid A. To test this hypothesis, phospholipid-like amphiphiles with six acyl chains attached to a bisphosphorylated serine-like backbone of varying length replacing the diglucosamine backbone were synthesized. The compound with a short backbone fulfills all criteria of an endotoxic agonist, and that with longbackbone fulfills those of an antagonist. This holds true for the human as well as for the murine system. Interestingly, these compounds are inactive in the Limulus amebocyte lysate test which is specific for LPS diglucosamine backbone. These results define a general endotoxic principle and, furthermore, provide new insights into an understanding of early steps of endotoxin action.
Lipopolysaccharides (LPS), the endotoxins of Gram-negative bacteria, consist of an oligo- or polysaccharide chain covalently linked to a lipid moiety termed lipid A. The endotoxically most active Escherichia coli-type lipid A consists of a diphosphoryl diglucosamine acylated with six hydroxylated fatty acid residues in asymmetric distribution bound in ester or amide linkage to the non-reducing (four) and reducing (two) glucosamine, respectively, and has been shown to constitute the ‘endotoxic principle’ of LPS 1. Each change in chemical composition such as a reduction of the number of charges and/or the number and distribution of fatty acids causes dramatic decreases of the bioactivity down to a complete loss of endotoxicity 2, 3. These observations provoked questions concerning the structural requirements for the induction of endotoxicity. The systematic investigation of synthetic lipid A analogues and partial structures 4, 5 exhibited a broad pattern of biological activities, ranging from high endotoxicity for compound ‘506’, synthetic analogue of natural hexaacyl lipid A from Escherichia coli, to complete inactivity for compound ‘406’, corresponding to tetraacyl lipid A precursor Ia or IVa 6–8. The latter, however, was able to inhibit the action of agonistical LPS, i.e. to act antagonistically.
In aqueous environments, amphiphilic molecules such as endotoxins form supramolecular aggregate structures, and the type of aggregate depends strongly on the chemical structure of the composing molecules. Lipid A of enterobacterial LPS adopts a cubic inverted structure under near physiological conditions, from which a conical shape of the individual lipid A molecule was deduced 9. Extension of these investigations to other lipid A structures indicated that a conical molecular shape is required for endotoxicity (agonism) and a cylindrical one for antagonism 10, 11. It is important to emphasize that the aggregate structure of the lipid A part of LPS, although not occurring in biological systems, and its molecular conformation are the decisive determinants. LPS with longer sugar chain of course tends to adopt more and more lamellar aggregate structures 12 from which, however, it is not possible to deduce the molecular conformation of the lipid A moiety.
A necessary condition for agonism as well as antagonism is the presence of at least two negative charges at the lipid A backbone, which is usually fulfilled by the two phosphates in 1 and 4′ position of the diglucosamine backbone. Based on these findings and other observations, we propose, as a model of cell activation, that the lipid A moiety of agonistic and antagonistic endotoxins intercalate into the cell membrane of mononuclear cells, on their own or mediated by the action of serum or membrane-bound proteins such as soluble or membrane-bound lipopolysaccharide-binding protein (sLBP and mLBP) and sCD14 or mCD14 13–17, and interact within the membrane with the Toll-like receptor TLR4 and K+-channels, which have been shown to be parts of the transmembrane signaling complex 18–19. According to this model, antagonistic lipid A would inhibit endotoxic action by occupying these binding sites, whereas the agonistic lipid A would provoke a conformational change of the binding proteins in the receptor cluster, likely by sterical stress as known for the K+ channel MaxiK 20, triggering transmembrane signaling.
According to these considerations, any amphiphilic molecules with clearly separated hydrophobic and hydrophilic moieties and negative charges in the hydrophilic backbone are candidates for agonistic or antagonistic endotoxin-like action as far as they can adopt a conical or a cylindrical shape, respectively, under physiological conditions. To test this hypothesis and with the aim of defining a ‘generalized endotoxic principle', we have designed and synthesized phospholipids with a constant hydrophobic moiety composed of six acyl chains which are ester- or amide-linked to a hydrophilic part that has a serine-like structure and carries two phosphates. To allow a potential variation of the hydrophilic cross-section and with that of the molecular conformation, the distance between the two phosphates was varied by different spacers: for ER803022 a 7-atom linker and for ER805046 a 15-atom linker was introduced (Fig. 1).
2 Results and discussion
In the following, data on physico-chemical parameters and biological reactivities of the two structurally related phospholipids are presented and compared with respective data characteristic for endotoxins, in particular for lipid A, the ‘endotoxic principle’ of bacterial LPS: (i) the conformation of the molecules under near physiological conditions; (ii) their intercalation into phospholipid membranes mediated by LBP; (iii) their capacity to induce cytokine production (TNF-α) in human mononuclear cells or to antagonize cytokine production by agonistic LPS; (iv) their ability to stimulate the clotting cascade in the Limulus test (Limulus amebocyte lysate, LAL activity); and (v) their capacity to activate CHO reporter cells transfected with the Toll-like receptor TLR4. For compounds with a capacity to induce cytokine production, (vi) the dependence of cytokine induction on the membrane-anchored endotoxin receptor protein CD14, (vii) the ability of K+ channel blockers to inhibit the cytokine production, and (viii) the binding and cytokine-modifying characteristics of the polycationic decapeptide polymyxin B (PMB) were tested.
Synchrotron radiation small-angle diffraction was used to establish the aggregate structure at high water content (80–90%). In Fig. 2, the data are summarized for ER803022. The patterns (a) in the absence of Mg2+ indicate the existence of a cubic phase, most likely of space group Q224 at 20 to 60°C (at the higher temperature, a conversion into another phase, the inverted hexagonal phase HII, starts), and (b) in the presence of Mg2+ clearly point to a multilamellar phase at 20 to 40°C and an inverted hexagonal HII phase at 50 to 70°C. For ER805046, in contrast, only lamellar structures are observed under the same conditions of water and Mg2+ content and of temperature (data not shown). The data for ER803022 are in complete accordance with those found earlier for bioactive enterobacterial lipid A, where under physiological water contents in the absence or at low Mg2+ concentration cubic structures, in particular of space group Q224, were found and in the presence of Mg2+ a multilamellar structure at 20 to 50°C and an inverted hexagonal HII structure above 21. The results for ER805046 agree well with those for non-bioactive lipid A from non-enterobacterial origin such as that from Rhodobacter capsulatus, Chromobacterium violaceum, or Rhodospirillum fulvum, or enterobacterial mutants or lipid A precursors with reduced numbers of acyl chains 11. From these findings, a conical molecular conformation of ER803022 and a cylindrical of ER805046 can be deduced.
Monitoring of the LBP-mediated intercalation into liposomes made from a phospholipid mixture corresponding to that of the macrophage membrane by fluorescence resonance energy transfer (FRET) spectroscopy showed that both compounds are transported very effectively, the intercalation being qualitatively similar to that determined for reference LPS (Re-LPS) (data not shown).
The TNF-α-inducing capacity shown with human mononuclear cells was nearly identical for ER803022 and deep rough mutant Re-LPS, but about two orders of magnitude lower for lipid A (Fig. 3). ER805046, in contrast, exhibited no TNF-α-inducing capacity at any concentration up to 10 μg/ml, but showed antagonistic activity, i.e. it blocked cytokine induction by the active compounds Re-LPS and ER803022. The administration of ER805046 in tenfold excess caused a reduction of the Re-LPS-or ER803022-induced TNF-α production by 20 to 60% (data not shown).
The well-known ability of the polycationic decapeptide polymyxin B (PMB) to inhibit the immunostimulatory action of LPS 22 was tested with ER803022. At a concentration of 10 ng/ml of this compound, human macrophages were activated to produce 1,110 pg/ml of TNF-α, while the simultaneous addition of 100 ng/ml of PMB reduced the TNF-α production to 320 pg/ml, a decrease by approximately 75%. Parallel to this, the binding characteristic of PMB to ER803022 corresponded to that of PMB to lipid A 23. The negative surface charge is neutralized by PMB binding to the phosphates of ER803022, which is accompanied by a fluidization of the acyl chains and a change of the aggregate structure from inverted cubic to multilamellar (data not shown).
To investigate the potential involvement of CD14 and TLR4 in the recognition of the ER compounds, their stimulatory activity was analyzed in a CHO (Chinese hamster ovary) cell reporter system 24. These cells are natural TLR2-knockouts and express the human CD25 antigen on the surface upon induction of NF-κB translocation 25. The data clearly indicate that the expression of TLR4 alone (EL1) is already sufficient to enable the cells to respond to ER803022 (Fig. 4). The additional co-expression of human CD14 (3E10) dramatically increases the sensitivity for LPS but reduces the activation by ER803022. Very recently, for a structurally related synthetic compound (ER112022) it has been shown that it associated functionally with TLR4 in the absence of CD14 26, in accordance with the presented result. Not surprisingly, the addition of ER805046 did not cause any activation (data not shown). To test a possible species specificity of cell activation of the ER compounds, the mouse cell line RAW264.7 was stimulated with the respective compounds and compared to LPS. The experiments exhibit high reactivity in the concentration range 100 ng/ml down to 100 pg/ml of ER803022 indistinguishable from LPS with respect to nitic oxide release and TNF-α production (data not shown).
The lack of dependence on CD14 may be seen in the light of an ‘CD14-independent receptor cluster’ as has been proposed for LPS 27. The detailed investigation of the dependence of the cytokine induction on CD14, however, showed a clear positive response (Fig. 5a): the addition of the anti-CD14 antibody BiG14 (10 μg/ml) to LPS and to ER803022 at both concentrations, 100 ng/ml and 10 ng/ml, caused a dramatic decrease of the TNF-α production. From these and the above data it can be concluded that the blocking of CD14 by the monoclonal antibody also comprises proteins of the receptor complex such as TLR4.
It has been reported that the activation of the K + channel MaxiK by LPS is an initial event in endotoxin-mediated macrophage signaling and that the channel activation, and thus TNF-α induction, can be inhibited by the specific MaxiK blocker, paxilline 19. In fact, the addition of paxilline led to a significant reduction of TNF-α production induced by ER803022 as well as by Re-LPS (Fig. 5b). Finally, the ability of ER803022 to activate the LAL-clotting cascade was investigated. Results obtained at concentrations between 1 μg/ml down to 1 ng/ml yielded values of 0.1 to 0.4 endotoxin units (EU)/ ml lying slightly above the value for pure water (0.09 EU/ ml), but considerably below those for Re-LPS (3.29 for 1 ng/ml and >125.6 for 100 ng/ml). The absence of LAL reactivity may be surprising, since the activation of the Limulus coagulation cascade is thought to be a typical endotoxin reaction. Based on the observation that also lipid A part structures with a reduced number of acyl chains, which have no cytokine-inducing capacity, are highly active in the LAL test 28, it is proposed that the lipid A backbone is the epitope being recognized in the LAL test. Thus, although the LAL test is useful for quantifying LPS and LPS-like structures for example in blood and serum, it is not a measure of endotoxicity. This finding may also be discussed in the context of antibody recognition: The monoclonal antibody, which recognizes the bisphosphoryl diglucosamine lipid A epitope 1, looses this ability when one phosphate is replaced by a carboxymethyl group (L. Brade, personal communication), although the cytokine-inducing capacity remains unchanged 29.
Except for the LAL test, all data obtained in the various analytical systems for the two synthetic phospholipid structures are characteristic for endotoxins and its endo-toxic principle, lipid A. Consequently, we hypothesize that a more general principle governs endotoxicity: prerequisites for agonistic action are amphiphilic molecules with clearly separated polar and apolar moieties and a conical conformation of the molecules with a larger cross-section of the hydrophobic than of the hydrophilic moiety. For antagonistic action, a cylindrical rather than a conical conformation of the molecules is required, usually guaranteed by a less acylated apolar moiety. For both, agonism as well as antagonism, the presence of at least two negative charges in the polar headgroup is an important prerequisite, no matter whether these are provided by phosphate or carboxylate groups.
It should be noted that a stereochemical change of the substitution pattern at the positions 1, 6, 22 and 27 (see Fig. 1) from R,R,R,R (ER803022) to S,S,S,S (ER805259) reduces the cytokine-inducing capacity by one order of magnitude as well as the molecular conformation from strictly conical to weakly conical (unpublished results). Thus, beside the pure number of acyl chains the kind of their linkage is of high impact.
4 Materials and methods
The phospholipid-like structures ER803022 and ER805046 were synthesized at ESAI Research Institute of Boston, Andover, MA based on synthetic work done by Krisovitch and Regen 30. The recipes for the syntheses of the compounds are published as US patent 31. Alternatively, the recipes can be made available on request. The structures were confirmed by 1H-, 13C-, 31P-NMR, mass spectroscopy, and elemental analysis.
4.2 X-ray diffraction
X-ray diffraction experiments on the lipids’ aggregate structures were performed at the European Molecular Biology Laboratory (EMBL) outstation at the Hamburg synchrotron radiation facility HASYLAB using the double-focussing monochromator-mirror camera X33 32. X-ray diffraction patterns, obtained with exposure times of 2 min using a linear gas proportional detector with delay line readout, were evaluated according to previously described procedures 9. These allow to assign the spacing ratios to defined three-dimensional aggregate structures, from which the conformation of the individual molecules are deduced.
4.3 Fluorescence resonance energy transfer spectroscopy
Transport of the lipids into liposomes made from a phospholipid mixture corresponding to that of the macrophage membrane, mediated by LBP was determined by fluorescence resonance energy transfer spectroscopy (FRET) spectroscopy using the dyes NBD phosphatidylethanolamine (NBD-PE) and Rhodamine-PE 33. To the doubly labeled liposomes first the lipids and then LBP were added, all at a final concentration of 1 μM, and the NBD-fluorescence intensity at 531 nm was monitored as function of time.
4.4 Biological assays
4.4.1 Stimulation of macrophages
Monocytes were isolated from peripheral blood from healthy donors by the Hypaque-Ficoll gradient method. To differentiate the monocytes to macrophages cells were cultivated in Teflon bags in the presence of 2 ng/ml M-CSF in RPMI 1640 medium (endotoxin ≤0.01 EU/ml; Biochrom, Berlin, Germany) containing 100 U/ml penicillin, 100 μg/ml streptomycin, 2 mM L-glutamine, and 4%heat-inactivated human serum type AB at 37°C and 6% CO2. On day 6 cells were washed with PBS, detached by Trypsin-EDTA treatment and seeded at 1×106 cells/ml in complete medium in 96-well tissue culture dishes (Nunc, Wiesbaden, Germany). After stimulation of cells with various lipids for 12 h cell-free supernatants of duplicate samples were collected, pooled, and storedat –20°C until determination of cytokine content. Data shown are representative of at least three independent experiments. The concentration of TNF-α in supernatants was determined by using an immunological sandwich-ELISA, in which 96-well plates were coated with a monoclonal antibody against TNF (clone 6b from Intex AG, Muttant, Switzerland). In a color reaction, the substrate is cleavedenzymatically, and the product can be measured photometrically on an ELISA reader at 450 nm.
4.4.2 Activation of CHO reporter cells
The CHO/CDE14 reporter line, clone 3E10, is a stably transfected CD14-positive CHO (Chinese hamster ovary) cell line that expresses inducible membrane CD25 (Tac antigen) under transcriptional control of the human E-selectin promoter pELAM.Tac 24. The CHO reporter cell line EL1 (CD14-free) was obtained by stable cotransfection of CHO-K1 cells with the plasmid pCEP4 and pELAM.Tac and was a kind gift of Dr. Egil Lien. Both cell lines react sensitively to the activation of the nuclear factor NF-κ B. CHO cells were cultivated in Ham's F12 medium (Biochrom, Berlin, Germany) containing 100 U/ml penicillin, 100 μg/ml streptomycin, 2 mM L-glutamine, 400 U/ml Hygromycin, and 10% heat-inactivated fetal calf serum (Linaris, Bettingen, Germany) at 37°C and 6% CO2. For stimulation experiments cells were washed with PBS, detached by Trypsin-EDTA treatment and seeded at a density of 2.5×105 cells/well in complete medium in a 24-well dish. Following overnight culture cells were stimulated for 18 h, detached, and stained with a FITC-conjugated anti-CD25 antibody (Becton Dickinson, Heidelberg, Germany). Cells were analyzedby flow cytometry on a FACSCalibur (Becton Dickinson, Heidelberg, Germany) using Cellquest software.
4.4.3 Limulus amebocyte lysate
Limulus clotting activity of the lipids in the concentration range 1 μg/ml down to 10 pg/ml was determined by a quantitative kinetic assay based on the reactivity of Gram-negative endtoxin with LAL 34, using test kits from BioWhittaker (Walkersville, MD).
4.5 Anti-CD14 antibody
The anti-CD14 antibody BiG14 was purchased from Biometec (Greifswald, Germany) and added to macrophages at a concentration of 10 μg/ml prior to stimulation with LPS Re or ER803022.
We thank G. von Busse, C. Hamann, and U. Diemer for performing the FTIR-and FRET-spectroscopic and LAL measurements, respectively, E. Lien for EL1 CHO reporter cell line, and S. F. Carroll for the kind gift of LBP. This work was financially supported by the Deutsche Forschungsgemeinschaft (SFB 367, project B8) and the European Union (project ANEPID).