Physicochemical characterization of carboxymethyl lipid A derivatives in relation to biological activity Ulrich Seydel 1 , Andra B. Schromm 1 , Lore Brade 1 , Sabine Gronow 1 ,Jo ¨ rg Andra ¨ 1 , Mareike Mu ¨ ller 1 , Michel H. J. Koch 2 , Koichi Fukase 3 , Mikayo Kataoka 3 , Masaya Hashimoto 3 , Shoichi Kusumoto 3 and Klaus Brandenburg 1 1 Forschungszentrum Borstel, Leibniz-Zentrum fu ¨ r Medizin und Biowissenschaften, Borstel, Germany 2 European Molecular Biology Laboratory, c ⁄ o DESY, Hamburg, Germany 3 Osaka University, Department of Chemistry, Osaka, Japan Lipopolysaccharide (LPS, endotoxin), as a major amphiphilic component of the outer leaflet of the outer membrane of Gram-negative bacteria, exerts in an isolated form a variety of biological activities in mam- mals [1]. Chemically, LPS consists of a hydrophilic heteropolysaccharide, which is covalently linked to a Keywords endotoxic shock 2 ; inflammation; lipopolysaccharide; signal transduction Correspondence U. Seydel, Forschungszentrum Borstel, Division of Biophysics Parkallee 10, D-23845 Borstel, Germany Fax: +49 4537 188632 Tel: +49 4537 188232 E-mail: useydel@fz-borstel.de (Received 6 September 2004, revised 3 November 2004, accepted 5 November 2004) doi:10.1111/j.1742-4658.2004.04471.x Lipopolysaccharide (LPS) from the outer membrane of Gram-negative bac- teria belongs to the most potent activators of the mammalian immune sys- tem. Its lipid moiety, lipid A, the ‘endotoxic principle’ of LPS, carries two negatively charged phosphate groups and six acyl chain residues in a defined asymmetric distribution (corresponding to synthetic compound 506). Tetraacyl lipid A (precursor IVa or synthetic 406), which lacks the two hydroxylated acyl chains, is agonistically completely inactive, but is a strong antagonist to bioactive LPS when administered to the cells before LPS addition. The two negative charges of lipid A, represented by the two phosphate groups, are essential for agonistic as well as for antagonistic activity and no highly active lipid A are known with negative charges other than phosphate groups. We hypothesized that the phosphate groups could be substituted by other negatively charged groups without changing the endotoxic properties of lipid A. To test this hypothesis, we synthesized carboxymethyl (CM) derivatives of hexaacyl lipid A (CM-506 and Bis- CM-506) and of tetraacyl lipid A (Bis-CM-406) and correlated their physicochemical with their endotoxic properties. We found that, similarly to compounds 506 and 406, also for their carboxymethyl derivatives a par- ticular molecular (‘endotoxic’) conformation and with that, a particular aggregate structure is a prerequisite for high cytokine-inducing capacity and antagonistic activity, respectively. In other parameters such as acyl chain melting behaviour, antibody binding, activity in the Limulus lysate assay, and partially the binding of 3-deoxy-d-manno-oct-2-ulosonic acid transferase 1 , strong deviations from the properties of the phosphorylated compounds were observed. These data allow a better understanding of endotoxic activity and its structural prerequisites. Abbreviations ATR, attenuated total reflectance; CM, carboxymethyl; EU, endotoxin unit; FRET, fluorescence resonance energy transfer; GlcN, D-glucosamine; Kdo, 3-deoxy-d-manno-oct-2-ulosonic acid; LAL, Limulus amebocyte lysate; LBP, lipopolysaccharide-binding protein; LPS, lipopolysaccharide; M-CSF, macrophage colony-stimulating factor; MNC, mononuclear cells; NBD, 7-Nitrobenz-2-oxa-1,3-diazol-4-yl; PE, phosphatidylethanolamine; Rh, rhodamine; TLR, Toll-like receptor; TNF, tumor necrosis factor. FEBS Journal 272 (2005) 327–340 ª 2004 FEBS 327 hydrophobic lipid moiety, termed lipid A, which anchors the molecule to the outer leaflet of the outer membrane. Because free lipid A has been shown to be responsible for the biological activity of LPS in most in vitro and in vivo test systems, it has been termed the ‘endotoxic principle’ of endotoxin [2]. The specific requirements for endotoxin to be biologically active are still only partly defined. It has been shown, how- ever, that for the full expression of biological activity, lipid A must possess a particular chemical composition and primary structure like that found in enterobacteri- al strains. Thus, for example, lipid A from the biolo- gically most potent LPS of the deep rough mutant strain Escherichia coli F515 consists of a b-1,6-linked d-glucosamine (GlcN) disaccharide carrying two negat- ively charged phosphate groups in the 1- and 4¢-posi- tions and six saturated fatty acids in a defined asymmetric distribution – four at the nonreducing and two at the reducing GlcN. Variation of this composi- tion such as a reduction of the number of charges or the number of acyl chains, a change in their distribu- tion, or degree of saturation, results in a dramatic reduction of biological activity [3,4]. These observa- tions could be interpreted to indicate that a variation of the primary structure of endotoxin molecules influ- ences their physicochemical behaviour, and that this physicochemical behaviour is correlated to the biologi- cal activity. Lipid A is an amphiphilic molecule and therefore tends to form multimeric aggregates above a critical concentration, which depends on its hydrophobicity. The structures of lipid A aggregates were found to be either nonlamellar inverted (cubic, Q or hexagonal, H II ) or lamellar (L) depending on the primary chemical structure and the molecular shape of the composing molecules 3 [5]. We have previously shown that a pecu- liar molecular shape of the lipid A portion of endotoxin is a prerequisite for the expression of endotoxic activ- ity. Thus, hexaacyl lipid A (or synthetic compound 506), which adopts a nonlamellar cubic aggregate struc- ture, with a conical molecular shape, is biologically highly active [6]. Pentaacyl lipid A and tetraacyl lipid A (synthetic compound 406) form lamellar structures, which can be related to a cylindrical molecular shape, and are agonistically inactive, but may be antagonistic, i.e., block the action of agonistic LPS [6]. Furthermore, we have reported that nonlipid A amphiphiles with cor- responding conical shape and two phosphates may also induce cytokines. This observation led us to propose a ‘generalized endotoxic principle’ [7]. For all these endotoxically active compounds, cell activation takes place after binding to proteins such as lipopolysaccharide-binding protein (LBP) and CD14 [8,9], which finally transport the compounds to integral membrane proteins such as Toll-like receptor 4 (TLR4) [10] or the K + -channel MaxiK [11]. It has been reported that a prerequisite for endo- toxic activity is a sufficiently high number of negative charges, i.e., essentially the phosphate groups in the lipid A region [12]. However, no systematic study on the role of the kind of charges has been performed. Thus, in the present study the phosphate groups in the synthetic tetraacyl and hexaacyl compounds (406 and 506, respectively) were replaced by carboxymethyl (CM) groups, and the corresponding compounds (Bis-CM-406, CM-506, and Bis-CM-506; chemical structures shown in Fig. 1) were characterized physico- chemically and biologically. We have found that important characteristics such as the aggregate struc- tures and the cytokine-inducing properties remain essentially unchanged or are modified only slightly. Partially clear modifications are observed in the recog- nition by serum and cell-surface binding proteins. However, other important features like Limulus activ- ity and antibody binding change upon replacement of the phosphate groups by the carboxymethyl groups. These findings allow a more general understanding of endotoxin action. Results Phase transition behaviour and intramolecular conformation The b «a gel to liquid crystalline acyl chain melting transition, the conformation and molecular orientation of particular molecular groups with respect to the Te traacyl lipid A (406) Bis-CM-406 Hexaacyl lipid A (506) CM - 506 Bis-CM-506 R 1 P CH COOH P CH COOH 2 2 CH COOH 2 R 2 OH OH C-OH C-OH C-OH 12 12 12 R 3 OH OH C-OH C-OH C-OH 14 14 14 R 4 P CH COOH P P 2 CH COOH 2 OH O O O NH O . O . 14 14 O O HO NH O O O OH . 14 14 O OH . R 3 R 4 R 2 R 1 Fig. 1. Chemical structures of synthetic tetraacyl and hexaacyl lipid A analogs. P, phosphate groups; R 1 –R 4 , side groups as indicated. 26 Characterization of carboxymethyl lipid A U. Seydel et al. 328 FEBS Journal 272 (2005) 327–340 ª 2004 FEBS membrane plane of the synthetic compounds were investigated as potentially important determinants of bioactivity [13]. In Fig. 2, the temperature dependence of the peak position of the symmetric stretching vibrational band v s (CH 2 ) for the different compounds reveals a con- siderable influence of the phosphate substitution in the case of the tetraacyl compounds. The phase transition temperature T c of compound 406 around 25 °C is shifted for compound Bis-CM-406 to 46 °C, concomitantly the wavenumber values in the gel phase decrease from around 2851 cm )1 to lower than 2850 cm )1 , indicating a higher state of order (lower fluidity). In contrast, the T c values of the hexaacyl compounds are very similar and lie slightly above 50 °C. The wavenumber values in the gel phase decrease from 2850 cm )1 for compound 506 to 2849.5 for CM-506 and to 2849.0 cm )1 for Bis-CM- 506. The comparison of the different compounds at the biologically relevant temperature 37 °C (vertical line) shows the same sequence. Infrared spectra in the wavenumber range 1800– 1500 cm )1 of the amide and ester vibrational bands are compared exemplarily for 406 and Bis-CM-406 (Fig. 3). The peak positions of the ester band contours are around 1740–1742 cm )1 and 1728–1731 cm )1 , respectively (obtained from the second derivative), and thus similar for both. The bandwidth, however, for Bis-CM-406 is smaller, indicating less mobility of the ester groups than those of phosphorylated 406. Importantly, also the amide I band, predominantly resulting from C¼O stretching of the amide group, is located at lower wavenumbers and is sharper, indica- ting stronger water and ⁄ or cation binding and, with that, higher order. Furthermore, the shoulder at 1600– 1607 cm )1 for Bis-CM-406 should correspond to the antisymmetric stretching of the negatively charged carboxylate group [v as (COO – )], showing the molecule in the charged state at neutral pH. To determine the orientation of the diglucosamine group with respect to the membrane plane, infrared dichroic measurements on an attennuated total reflectance (ATR) plate were performed with hydra- ted lipid multilayers. The dichroic ratios, R, were 0 10203040506070 2849 2850 2851 2852 2853 2854 406 Bis-CM-406 506 CM-506 Bis-CM-506 Wavenumber (cm –1 ) Temperature (C°) Fig. 2. Gel to liquid crystalline phase behav- ior of the hydrocarbon chains of the various synthetic lipid A analogs. The peak position of the symmetric stretching vibration of the methylene groups v s (CH 2 ) is plotted vs. temperature. In the gel phase it is located around 2850 cm )1 , in the liquid crystalline phase around 2852.5 to 2853.0 cm )1 . Fig. 3. Infrared spectra in the range 1800–1500 cm )1 for com- pounds 406 and Bis-CM-406, exhibiting the ester carbonyl band around 1730 cm )1 , the amide I band (predominantly C¼O stretch- ing) in the range 1620–1660 cm )1 , and amide II band (predomin- antly N–H bending) around 1550 cm )1 . The band around 1602 cm )1 for compound Bis-CM-406 corresponds to the antisymmetric stretching vibration of the the negatively charged carboxylate group v as (COO – ). U. Seydel et al. Characterization of carboxymethyl lipid A FEBS Journal 272 (2005) 327–340 ª 2004 FEBS 329 measured for the diglucosamine ring vibrational bands at 1170 and 1045 cm )1 , allowing calculation of the inclination angle of the diglucosamine ring plane with respect to the membrane plane [14]. In Table 1, the data are summarized showing high R values for the tetraacyl, corresponding to small inclination angles (5–20°), and much smaller R values for the hexaacyl compounds corresponding to high inclina- tion angles (47–48°). Supramolecular aggregate structures The aggregate structure has been described to be an essential determinant for the ability of endotoxins to induce cytokines in immune cells [5]. We have applied small-angle X-ray diffraction using synchrotron radiation to elucidate the aggregate structure of Bis- CM-406 and Bis-CM-506 (Fig. 4). For the tetraacyl compound (Fig. 4A), in the temperature range 20– 80 °C only one sharp diffraction peak at 4.72 nm is observed, which can be assigned to the periodicity of a multilamellar stack. The diffraction pattern of the hexaacyl compound (Fig. 4B) is more complex. At 40 °C, the main diffraction peak at 4.53 nm may be assumed to result from a multilamellar aggregate. The further reflections at 2.46, 1.94, and 1.16 nm, however, belong to a different type of aggregate structure. Thus, a superposition of a lamellar and a cubic phase is sug- gested, because the relations 2.46 nmÆv5 ¼ 5.50 nm and 1.94 nmÆv8 ¼ 5.50 nm hold. At 80 °C, the three observable reflections are clearly indicative of an inverted hexagonal H II structure, because 4.61 nm ¼ 2.67 nmÆv3 and 4.61 nm ¼ 2.31 nmÆv4. This phase is already observable at 60 °C (data not shown). Respect- ive data for the phosphorylated compounds show only (multi)lamellar structures for 406 [6], and a higher ten- dency of compound 506 towards a cubic structure (data not shown). Incorporation into phospholipid cell membranes As a prerequisite for agonistic as well as antagonistic activity, the intercalation of endotoxins into target cell membranes corresponding to the composition of the macrophage membrane mediated by the LBP has been described [15]. The results for the synthetic compounds are presen- ted in Fig. 5, showing most pronounced intercalation of the tetraacylated compound 406, and significantly lower efficiencies for the other compounds. Binding of monoclonal antibodies The reactivities of several lipid A monoclonal antibod- ies with compounds CM-506 and Bis-CM-506 were compared to compound 506 using an ELISA. Selected antibodies recognize variations in the hydrophilic backbone as follows (Fig. 6A): the monoclonal anti- bodies A6 and 8A1 recognize the bisphosphorylated backbone, mAb S1 binds to the 4¢-monophosphory- lated backbone and mAb A43 reacts with phosphoryl- ated as well as with phosphate-free compounds [16]. As one example, the data of the phosphate-dependent mAb A6 and of the phosphate-independent mAb A43 are shown in Fig. 6B, displaying antibody binding curves determined by checkerboard titration. The results show that both mAb’s bind with high affinity to compound 506. mAb A43 binds with similarly high affinities to compounds 506, CM-506, and Bis-CM- 506, whereas mAb A6 binds with high affinity only to compound 506. Binding to compound CM-506 was observed only at much higher antibody concentrations and no binding was observed to Bis-CM-506. mAb 8A1 gave similar results as mAb A6 (data not shown). mAb S1, which recognizes the 4¢-monophosphorylated backbone, reacted not only with compound 506 but also with compound CM-506 although with somewhat lower affinity. No binding of mAb S1 to compound Bis-CM-506 was observed. Limulus amebocyte lysate (LAL) The ability of the synthetic compounds to activate the clotting cascade of the horseshoe crab Limulus Table 1. Dichroic ratio R, order parameter S, and inclination angle h between diglucosamine ring plane and membrane plane for the synthetic lipid A analogs. The R values were evaluated from the ratio of the band intensities of the 90° and 0° polarized infrared spectra of the diglucosamine ring vibrations at 1045 and 1170 cm )1 . The error of h results from calculating the Gaussian error propagation by using the functional relation between R, S and h [14]. Compound Parameter Dichroic ratio R Order parameter S Inclination angle h (°) 406 1.33 0.55 20.4 ± 14.0 0.46 16.5 ± 18.2 0.33 7.1 ± 28.3 Bis-CM-406 1.60 0.69 15.0 ± 3.8 0.56 9.0 ± 7.1 0.49 4.9 ± 12.0 506 0.98 0.96 46.8 ± 1.3 CM-506 1.05 0.88 47.6 ± 1.6 Bis-CM-506 1.01 0.93 47.5 ± 1.2 Characterization of carboxymethyl lipid A U. Seydel et al. 330 FEBS Journal 272 (2005) 327–340 ª 2004 FEBS polyphemus gave highest values for compound 406, followed by 506 and CM-506 (Table 2). Considering that the diglucosamine 4¢-phosphate backbone is the recognition structure of the LAL assay [17,18], it seems surprising that compounds Bis-CM406 and Bis- CM-506 also showed high reactivity down to 10 ngÆmL )1 . Cytokine-inducing capacity in macrophages As a characteristic endotoxic reaction, the induction of tumor necrosis factor a (TNFa) production in human macrophages by the synthetic compounds was determined. Concomitantly, the influence of the specific MaxiK channel blocker paxilline on TNFa production was monitored. Data are shown for com- pounds 506 and Bis-CM-506 (Fig. 7) revealing cyto- kine-inducing capacity down to 1 ngÆmL )1 for the former, whereas the activity of the latter is one order of magnitude lower. The activity of compound CM-506 is nearly the same as for compound 506 (data not shown). For both compounds inhibi- tion due to the addition of paxilline can clearly be observed, in particular at the lower lipid log I log I 0.2 0.4 0.6 0.8 Bis-CM-406 4.72 nm 20 °C 40 80 Bis-CM-506 1.16 nm 2.32 nm 2.46 nm 4.61 nm 1.94 nm 2.67 nm 4.53 nm 40 °C 80 s (nm -1 ) A B Fig. 4. Synchrotron radiation small-angle X-ray diffraction patterns for compounds Bis-CM-406 (A) and Bis-CM-506 (B) at high water content (90%) and different temperatures. 50 100 150 200 250 300 0.5 1.0 1.5 2.0 2.5 +LBP PL liposomes Buffer CM-506 Bis-CM-506 506 Bis-CM-406 406 I D /I A Time (s) Fig. 5. LBP-mediated intercalation of the synthetic lipid A analogs into phospholipid liposomes corresponding to the composition of the macrophage membrane, derived from the increase of the ratio of the donor fluor- escence intensity, I D , to that of the accep- tor, I A . At 50 s, the lipid A analogs were added to the liposomes, and at 100 s LBP was added. U. Seydel et al. Characterization of carboxymethyl lipid A FEBS Journal 272 (2005) 327–340 ª 2004 FEBS 331 concentrations, which is, however, significantly higher for the CM- as compared to the phosphate-contaning compound. This observation allows one to conclude that channel blocking leads to inhibition of signal transduction. Again, similar results are found for CM-506. GlcN II GlcN I P P O O N N 14 1414 141214 OH OH S1 A43 A B A6, 8A1 Fig. 6. (A) Schematic representation of spe- cificities of various monoclonal antibodies against the lipid A backbone. The recogni- tion structures of the mAbs A43, S1, and A6 ⁄ 8A1 are GlcNII, diglucosamine-4¢-phos- phate, and the entire backbone, respect- ively. (B) Binding curves of monoclonal antibodies A6 (left column) and A43 (right column) to compounds 506 (top), CM-506 (middle), and Bis-CM-506 (bottom). ELISA plates were coated with 400 (d), 200 (m), 100 (j), 50 (r), 25 (s), 12.5 (n), 6.3 (h) and 3.1 (e) ng of compound per mL. Antibody concentrations are indicated. Values are the mean of quadruplicates with confidence values not exceeding 10%. Characterization of carboxymethyl lipid A U. Seydel et al. 332 FEBS Journal 272 (2005) 327–340 ª 2004 FEBS Antagonistic activity Compound 406 is a well known effective antagonistic agent against agonistically active LPS and lipid A [19]. The antagonistic action of compound Bis-CM-406 was compared to that of 406 by the addition of these com- pounds to mononuclear cells under serum-free condi- tions 15 min prior to the addition of deep rough mutant LPS in defined [antagonist] ⁄ [LPS] molar ratios. Figure 8 shows a strong cytokine-inhibiting activity of compound 406, which is also expressed, but to signifi- cantly lesser extent by compound Bis-CM-406. HEK cell system Similar to compound 506, compound Bis-CM-506 acti- vates HEK293 cells via TLR4 ⁄ MD2, but not via TLR2 (Fig. 9). Thus, the change in the nature of charges does not cause changes in the affinity to recep- tors decisive for cell activation. Reactivity with Kdo transferases To assess the lipid A analogs CM-506 and Bis-CM-506 in comparison to compound 506 as acceptors for 3-de- oxy-d-manno-oct-2-ulosonic acid (Kdo) transferases, they were submitted to in vitro enzyme assays, and the reaction products were detected with mAb A20, react- ing with a terminal Kdo residue (Fig. 10). Kdo trans- ferases from Haemophilus influenzae (monofunctional; lane A) [20], E. coli (bifunctional, lane B) [21], Table 2. Endotoxic activity in endotoxin unitsÆmL )1 (EUÆmL )1 )in the chromogenic Limulus amebocyte lysate test for the synthetic lipid A analogs at various concentrations. The data are representa- tive of three independent sets of measurements. In the test, 14 EUÆmL )1 corresponds to 1 ngÆmL )1 of LPS from Escherichia coli O55:B5. Lipid Concentration 1 lgÆmL )1 100 ngÆmL )1 10 ngÆmL )1 1 ngÆmL )1 100 pgÆmL )1 406 > 125 > 125 > 125 45.84 1.96 Bis-CM-406 > 125 61.5 11.3 1.76 0.89 506 > 125 76.1 15.6 3.16 CM-506 > 125 23.6 11.8 2.20 Bis-CM-506 > 125 82.8 12.5 4.5 1.43 Fig. 8. Antagonistic activity of compounds 406 and Bis-CM-406. Human mononuclear cells were incubated under serum-free condi- tions with the antagonistic compound, and after 15 min LPS from Salmonella minnesota R595 was added at the given molar ratios. The data result from one representative experiment. The mean and standard deviation are based on the data from the determination of TNFa in duplicate at two different dilutions. Fig. 7. Induction of TNFa production in human macrophages by compounds 506 (A) and Bis-CM-506 (B) in the absence (left-hand bars) and presence (right-hand bars) of the K + -channel blocker paxil- line at different lipid concentrations. The concentration of paxilline was 20 l M.The data result from one representative experiment. The mean and standard deviation are based on the data from the determination of TNFa in duplicate at two different dilutions. A repetition of the experiments yielded the same dependences except for the absolute amount of TNFa production which may vary significantly between different donors. U. Seydel et al. Characterization of carboxymethyl lipid A FEBS Journal 272 (2005) 327–340 ª 2004 FEBS 333 Burkholderia cepacia (bifunctional, lane C) [22], and Chlamydia psittaci 4 (tetrafunctional, lane D) [23] were tested for their ability to use the artificial lipid A ana- logs as substrates. All Kdo transferases gave a product when compound 506 served as acceptor, and the same were observed when compound CM-506 was offered as acceptor (data not shown). 5 However, compound Bis-CM-506 was transformed to a Kdo-containing product only when incubated with Kdo transferases from H. influenzae or E. coli, but not with enzymes from B. cepacia or C. psittaci. The use of monophos- phoryl compound 504 (having the 4¢-phosphate) and 505 (1-phosphate) showed that both phosphate resi- dues were necessary for binding of Kdo transferases from B. cepacia and C. psittaci , whereas those from E. coli and H. influenzae were reactive with each of the monophosphoryl analogs (data not shown). Discussion The CM derivatives of lipid A exhibit in many aspects comparable behavior to that of the phosphate-contain- ing compounds. The substitution of the 1-phosphate by CM has already been shown not to alter the basic cyto- kine-inducing capacity of compound 506 [14], and this is similarly true for the further substitution with the 4¢-CM group (Fig. 7B). There is, however, a significant decrease of the activity in particular for the lower con- centrations 10 and 1 ngÆmL )1 (Fig. 7), which could be confirmed in three independent experiments. The observed aggregate structures may provide an explan- ation for this difference (Fig. 4B and U Seydel, AB Schromm, L Brade, S Gronow, J Andra ¨ ,MMu ¨ ller, MHJ Koch, K Fukase, M Kataoka, M Hashimoto, S Kusumoto & K Brandenburg, unpublished data) 6 . Compound Bis-CM-506 adopts a mixed multilamellar ⁄ zcubic aggregate structure at 40 °C. According to previ- ous work, one main structural prerequisite for endo- toxic activity is the existence of a pure cubic or a mixed unilamellar ⁄ cubic aggregate structure of the lipid A part of LPS [24], whereas multilamellar structures have been shown to reflect inactive lipid A. Thus, the signifi- cant amount of a multilamellar portion within the aggregate may be responsible for the reduction of activ- ity as compared to natural lipid A or synthetic com- pound 506. Regarding the other prerequisites for endotoxin activity, no significant difference between the two compounds can be found. For example, the incli- nation angle of the diglucosamine plane with respect to the membrane plane is very similar (Table 1). Thus, the interpretation is confirmed that the driving force for an optimal packing of the acyl chains is their linkage and distribution to the two glucosamines, causing the observed high inclination independent of the type of charges [14]. Also, all compounds show an LBP- induced intercalation into phospholipid liposomes (Fig. 5), for which the presence of negative charges is a prerequisite. The presence of the infrared band around 1605 cm )1 (Fig. 3) clearly indicates that the carboxylate group is in the ionized state. Analogously to the agonistically active hexaacyl compounds is the antagonistic activity of the tetraacyl compounds: Bis-CM-406 is also antagonistic, although to a significantly lower degree than 406 (Fig. 8). Cor- responding to the data of the antagonistic activity 0 500 1000 1500 2000 no Compound 506 Compound Bis-CM-506 IL-1 Stimuli Control huTLR4/MD2 IL-8 concentration (pg/ml) TLR2/MD2 Fig. 9. Activation of HEK293 cells by compounds 506 and Bis-CM- 506 in dependence on TLR-expression. HEK293 cell were transi- ently transfected with a control plasmid (pcDNA3), cotransfected with expression plasmids for human TLR4 and human MD2 or for human TLR2 as described in Materials and methods. After 24 h, cells were stimulated with compounds 506, Bis-CM-506 (1 lgÆmL )1 ), or recombinant interleukin-1 (5 ngÆmL )1 ) for 24 h. Cell activation was determined by measuring the IL-8 concentration in an ELISA test. Transfections were performed in triplicate, and data given are mean and standard deviation of one experiment represen- tative of three. 506 CM-506 Bis-CM-506 406 A C D E B Fig. 10. In vitro assays of different Kdo transferases using com- pounds 506, CM-506, Bis-CM-506 and 406 as lipid acceptors. Reac- tion products of Kdo transferases from H. influenzae (A), E. coli (B), B. cepacia (C), C. psittaci (D) or controls without enzyme added (E) were detected with mAb A20, recognizing a terminal Kdo residue. Characterization of carboxymethyl lipid A U. Seydel et al. 334 FEBS Journal 272 (2005) 327–340 ª 2004 FEBS found for compound 406 [6], a pure multilamellar aggregate structure is observed for Bis-CM-406 (Fig. 4A), and the inclination angle of the diglucosa- mine ring plane with respect to the membrane plane is small (Table 1). Furthermore, LBP mediates the inter- calation of compound 406 into phospholipid mem- branes more effectively than compound Bis-CM-406 (Fig. 5). This observation might be connected with the much higher fluidity of compound 406 as compared to compound Bis-CM-406: for these compounds the behavior of the carboxymethyl derivatives is strongly deviating from that of compounds substituted with phosphate groups (Fig. 2). The replacement of phos- phate by carboxymethyl leads to an overall decrease of the wavenumbers of v s (CH 2 ) and concomitantly an increase in T c (Fig. 2), i.e., a considerable ordering of the hydrophobic moiety takes place. This observation may be understood in the light of a dramatically increased water ⁄ cation binding of the interface as deduced from the red shift of the amide I band (Fig. 3), which is caused by the presence of the carb- oxymethyl groups. These data are in accordance with the fact that the most potent antagonists described so far [25,26] have an extremely high fluidity. This holds for lipid A from Rhodobacter capsulatus and Rhodobacter sphaeroides as well as for lipid A from Chromobacterium violaceum ([5] and U Seydel, AB Schromm, L Brade, S Gronow, J Andra ¨ ,MMu ¨ ller, MHJ Koch, K Fukase, M Kata- oka, M Hashimoto, S Kusumoto & K Brandenburg, unpublished data) 7 . The results from the LAL test for the bioactivity (Table 2) are fundamentally different from those for the cytokine-inducing capacity in human cells, because in the former the tetra- and hexaacyl lipid A CM-analogs exhibit similar activities except for minor modifications, in that the biscarboxymethylated com- pounds exhibit a lower activity than the phosphate- containing compounds. This may be understood by considering previous findings that the main epitope in the Limulus test is the GlcNII-4¢-phosphate region in the lipid A backbone [17,18]. A substitution of the 4¢-phosphate by another charge does not lead to a complete inhibition, which implies that the Limulus test recognizes patterns rather than a particular charged group. The binding specificities of the monoclonal antibod- ies A6, 8A1, S1, and A43 are determined by the phos- phorylation pattern [16]. So far it has not be determined whether phosphate groups are specifically required or whether they may be replaced by other negatively charged groups. By synthesizing compounds CM-506 and Bis-CM-506, we are in the position to answer this question. The binding of mAb A43, which recognizes the nonreducing GlcN moiety in the back- bone of compound 506 is not influenced by the replacement of phosphate groups by CM residues. This is in agreement with the known specificity of mAb A43, which does not require the phosphate substitu- tion in either position. The binding of mAb A6 and 8A1 to compound CM-506 (for the latter, data not shown) is considerably reduced as compared to com- pound 506 (Fig. 6). This result is also not unexpected because previous data had already shown that binding of both mAbs to 4¢-monophosphorylated lipid A required 30-fold higher antigen concentrations than those needed with bisphosphorylated lipid A [16]. No binding of mAbs A6 and 8A1 was observed with Bis- CM-506 showing that the bisphosphorylated backbone is a prerequisite for high affinity binding. In summary, the data confirm that lipid A antibod- ies, except mAb A43, recognize a distinct phosphoryla- tion pattern of the lipid A backbone and confirm that phosphates are part of the epitopes recognized, which cannot be replaced simply by other negatively charged groups. This may be understood by different negative charge densities within the phosphate and the carboxy- methyl groups. Recently, it has been described that a blockade of the K + -channel, MaxiK, by the specific blocker paxil- line is connected to an inhibition of cytokine induc- tion in macrophages [11]. The test of the synthetic hexaacyl compounds showed that although Bis-CM- 506 is less effective than 506 in the cytokine assay, the action of the K + -channel (MaxiK) blocker paxil- line is reversed: our data indicate a higher efficiency in the case of the carboxylated compound; a complete blockade of the cytokine induction is observed already at 100 ngÆmL )1 for Bis-CM-506 (Fig. 7). Together with the data of the activation of HEK cells (Fig. 9), which indicate the involvement of the TLR4 ⁄ MD2 complex in cell activation, these data provide clear evidence that not a single molecular species but rather a complete receptor cluster [27] governs cell signalling. Interestingly, both lipid A analogs show different properties when used as acceptors for Kdo transfer- ases. Whereas the enzymes from H. influenzae and E. coli are not depending on either phosphate residue substituting the lipid A backbone and recognize both compounds CM-506 and Bis-CM-506 as acceptors, the respective enzymes from B. cepacia and C. psittaci are strongly depending on the phosphate group in position 4¢. Both Kdo transferases accept compound CM-506 but are not able to recognize compound Bis- CM-506 as a Kdo acceptor. Presently, the reason for U. Seydel et al. Characterization of carboxymethyl lipid A FEBS Journal 272 (2005) 327–340 ª 2004 FEBS 335 this distinctive behavior of Kdo transferases cannot be explained completely, in particular with respect to the similarity of the molecular conformations of compounds 506 and Bis-CMz-506 (Table 1). How- ever, together with the data of the monophosphoryl lipid A compounds a necessary prerequisite for the binding of Kdo transferases is the existence of a charge at position 4¢. That two of the transferases bind to both compounds 506 as well as Bis-CM-506 and the other two do not, may have to do with the acyl chain substitution of the acceptor, which again may play a decisive role for the molecular conforma- tion. It was shown that LPS from H. influenzae sim- ilar to that from E. coli is hexaacylated [21], whereas those from B. cepacia [23] and C. psittaci [28] are pentaacylated. Conclusions The mere presence of two negative charges but not their kind within the lipid A backbone is essential for the bioactivity of endotoxins, for the agonistic as well as the antagonistic activities. These findings confirm and extend our ‘conformational concept’ of endotoxicity, presuming for agonists a conical shape of the lipid A moiety with a high inclination angle of the acyl chains with respect to the membrane surface and for antago- nists a cylindrical shape with a low inclination angle. For other effects, however, for which particular binding epitopes are necessary, exchange of charges may largely change the bioactivities. This is valid for the recognition by monoclonal antibodies, the binding of Kdo trans- ferases, and the reactivity in the Limulus test. Materials and methods Synthesis The synthesis of hexaacyl lipid A compounds 506 [29,30], CM-506 [30], and tetraacyl lipid A 406 [31,32] has been pub- lished previously. Monophosphoryl compounds 504 (4¢-phosphate) and 505 (1-phosphate), used in some cases, were synthesized as described [29]. The synthesis of Bis-CM- 506 and Bis-CM-406 will be reported elsewhere (K Fukase, M Kataoka, M Hashimodo & S Kusumodo, unpublished data). 8 Briefly, the 3-position of 1-O-allyl 4,6-O-benzylidene- 2-deoxy-2-(2,2,2-trichloroethoxycarbonylamino)-a-d-gluco- pyranoside was acylated with (R)-3-benzyloxytetradecanoic acid. The 2-N-Troc (Troc ¼ 2,2,2-trichloroethoxycarbonyl) group was cleaved, and the 2-amino group was then acylated with (R)-3-benzyloxytetradecanoic acid. The allyl group was then oxidized with OsO 4 , and the resulting diol was oxida- tively cleaved with Pb(OAc) 4 to give 1-O-formylmethyl glycoside, which was further oxidized with NaClO 2 . The resulting carboxyl group was protected with benzyl group by using phenyldioazomethane. Deprotection of benzylidene gave the reducing-end GlcN moiety as a common glycosyl acceptor for the synthesis of Bis-CM-506 and Bis-CM-406. For the synthesis of the nonreducing-end GlcN moiety, the common intermediate 1-O-allyl 4,6-O-benzylidene-3- O-((R)-3-benzyloxytetradecanoyl)-2-deoxy-2-(2,2,2-trichloro- ethoxy carbonylamino)-a-d-glucopyranoside was treated under the conditions of regioselective reduction of benzylid- ene (BF 3 ÆOEt 2 9 and Et 3 SiH) to give the 6-O-benzyl-4-OH GlcN derivative. The benzyloxy carbonyl group was then introduced to the 4-O-position using Ag 2 O and ICH 2 COO- Bn. The 1-O-allyl group was then cleaved and transformed to 1-O-trichloroacetimidate, which was used as a glycosyl donor for the synthesis of Bis-CM-406. Glycosylation of the above glycosyl acceptor with the glycosyl donor gave the desired a(1-6) disaccharide. The 2¢-N-Troc group was cleaved, and the resulting amino group was acylated with (R)-3-benzyloxytetradecanoic acid to give the protected Bis- CM-406. The final catalytic hydrogenation gave the desired Bis-CM-406 (m ⁄ z ¼ 1360.0 [(M-H) – ]). For the synthesis of Bis-CM-506, the benzyl group of the benzyloxytetradecanoyl moiety in the above synthetic intermediate 1-O-allyl 6-O-benzyl-4-O-benzyloxycarbonyl- methyl-3-O-((R)-3-benzyloxytetradecanoyl)-2-deoxy-2-(2,2,2- trichloroethoxycarbonylamino)-a-d-glucopyra-noside was selectively cleaved using 2,3-dichloro-5,6-dicyanobenzoqui- none. The resulting hydroxy group of the fatty acid moi- ety was then acylated with tetradecanoic acid. After deprotection of the 1-O-allyl group and formation of 1-O- trichloroacetoimidate, glycosylation of the acceptor with 1-O-trichloroacetoimidate gave the disaccharide, which was then transformed to Bis-CM-506 in a similar manner (m ⁄ z ¼ 1753.1 [(M-H) – ]). The chemical structures of the synthesized compounds are plotted in Fig. 1. Lipids and reagents Lipopolysaccharide from the deep rough mutant Escheri- chia coli strain WBB01 (kindly provided by W Brabetz, Biomet, Dresden, Germany) 10 was extracted by the phe- nol ⁄ chloroform ⁄ petrol ether method [33] from bacteria grown at 37 °C, purified, and lyophilized. Paxilline was purchased from Sigma (Deisenhofen, Germany). LBP was a kind gift of RL Dedrick (XOMA Co., Berkeley, CA, USA). LBP was stored at ) 70 °Casa1mgÆmL )1 stock solution in 10 mm Hepes, pH 7.5, 150 mm NaCl, 0.002% (v ⁄ v) Tween 80, 0.1% F68. The lipids 3-sn-phosphatidylserine, egg 3-sn-phosphatidyl- choline, sphingomyelin from bovine brain, and 3-sn-phos- phatidylethanolamine from E. coli were from Avanti Polar Lipids (Alabaster, AL, USA). Characterization of carboxymethyl lipid A U. Seydel et al. 336 FEBS Journal 272 (2005) 327–340 ª 2004 FEBS [...]... read with a microplate reader (Dyna23 tech MR5000, Dynatech Laboratories, Chantilly, VA, USA) at a wavelength of 405 nm All tests were set up in quadruplicate Confidence values of the means were less than 10% Binding of Kdo transferases Characterization of carboxymethyl lipid A 3 4 5 Cell-free extracts containing Kdo transferase from E coli, 24 C psittaci or H in uenzae were prepared from Corynebacterium... synthetic lipid A analogs FTIR spectroscopy Characterization of carboxymethyl lipid A to a relationship between S, h and R as reported previously [14], was used X-ray diffraction X-ray diffraction experiments to define aggregate structures of the lipid A analogues were performed at the European Molecular Biology Laboratory (EMBL) outstation at the Hamburg synchrotron radiation facility HASYLAB using the... mononuclear protocol Transient transfections were carried out in triplicells, and incubated at 37 °C cate and data given are the mean and standard deviation Antagonism Binding of monoclonal antibodies Mononuclear cells (MNC) were isolated from peripheral For the analysis of the binding of antibodies to the backblood taken from healthy donors by the Hypaque–Ficoll bone of lipid A, monoclonal antibodies with... (2.5% casein in atants were collected after centrifugation of the culture NaCl ⁄ Pi) and then incubated for 1 h with mAb diluted in plates for 10 min at 400 g and stored at )20 °C until deterblocking buffer (50 lL) Plates were washed in NaCl ⁄ Pi mination of cytokine content (see above) and incubated for 1 h with peroxidase-conjugated goat anti-mouse IgG (heavy and light chain specific, (Dianova 22 Hamburg,... type AB at 37 °C and an approximate value of the order parameter, S, is known 6% CO2 On day 6 the cells were washed with NaCl ⁄ Pi, As a first approximation, the S value calculated according FEBS Journal 272 (2005) 327–340 ª 2004 FEBS 337 Characterization of carboxymethyl lipid A U Seydel et al (LAL) at 37 °C, using test kits of LAL Coamatic Chromodetached by trypsin ⁄ EDTA treatment and seeded at 1 ·... polysaccharide Infect Immun 55, 462–466 ¨ ¨ The use of a quantitative assay in endotoxin testing In 24 Brandenburg K, Richter W, Koch MHJ, Meyer HW & Detection of Bacterial Endotoxin with the Limulus AmeSeydel U (1998) Characterization of the nonlamellar bocyte Lysate Test (Watson SW, Levin J & Novitzky cubic and HII structures of lipid A from Salmonella enterica serovar Minnesota by X-ray diffraction and... Germany); diluted 1 : 1000 in blocking bufDetermination of endotoxin activity by the fer, 50 lL) After three washes in NaCl ⁄ Pi, the plates chromogenic Limulus test were washed in substrate buffer (0.1 m sodium citrate, pH 4.5) Substrate solution, which was prepared freshly, Endotoxin activity of the lipid A analogs was determined was composed of azinodi-3-ethylbenzthiazolinsulfonic acid by a quantitative... (LPS) corresponds to the lethal toxi39 Bode CE, Brabetz W & Brade H (1998) Cloning and city and is inhibited by nontoxic Rhodobacter capsulatus characterization of 3-deoxy-d-manno-oct-2-ulosonic acid LPS Infect Immun 58, 3743–3750 (Kdo) transferase genes (kdtA) from Acinetobacter bau26 Kirikae T, Schade FU, Kirikae F, Qureshi N, Takamannii and Acinetobacter haemolyticus Eur J Biochem yama K & Rietschel... evaluated at 1170 and (endotoxin < 0.01 endotoxin units (EU)ÆmL)1 in Limulus 1045 cm)1 The angle h represents the angle between the test; Biochrom, Berlin, Germany) containing 2 mm l-glutadiglucosamine backbone and the direction of the hydrocarmine, 100 UÆmL)1 penicillin, and 100 lgÆmL)1 streptomycin, bon chains, and can be calculated from the measured R, if and 4% heat-inactivated human serum type AB... For preparation of liposomes corresponding to the composition of the macrophage membrane, phosphatidylcholine, phosphatidylserine, phosphatidylethanolamine and sphingomyelin in a molar ratio of 1 : 0.4 : 0.7 : 0.5 [12] were solubilized in chloroform, the solvent was evaporated under a stream of nitrogen, the lipids were resuspended in the appropriate volume of buffer, and treated as described above for . high cytokine-inducing capacity and antagonistic activity, respectively. In other parameters such as acyl chain melting behaviour, antibody binding, activity. were carried out in tripli- cate and data given are the mean and standard deviation. Binding of monoclonal antibodies For the analysis of the binding of antibodies