Intact rough- and smooth-form lipopolysaccharides from Escherichia coli separated by preparative gel electrophoresis exhibit differential biologic activity in human macrophages

Authors

  • Elder Pupo,

    1. Division of Medical and Biochemical Microbiology, Department of Molecular Infection Biology, Research Center Borstel, Leibniz-Center for Medicine and Biosciences, Borstel, Germany
    Current affiliation:
    1. Vaccinology Unit, National Institute for Public Health and the Environment, 3720 BA Bilthoven, The Netherlands
    Search for more papers by this author
  • Buko Lindner,

    1. Division of Immunochemistry, Department of Molecular Infection Biology, Research Center Borstel, Leibniz-Center for Medicine and Biosciences, Borstel, Germany
    Search for more papers by this author
  • Helmut Brade,

    1. Division of Medical and Biochemical Microbiology, Department of Molecular Infection Biology, Research Center Borstel, Leibniz-Center for Medicine and Biosciences, Borstel, Germany
    Search for more papers by this author
  • Andra B. Schromm

    Corresponding author
    1. Division of Immunobiophysics, Department of Molecular Infection Biology, Research Center Borstel, Leibniz-Center for Medicine and Biosciences, Borstel, Germany
    • Division of Medical and Biochemical Microbiology, Department of Molecular Infection Biology, Research Center Borstel, Leibniz-Center for Medicine and Biosciences, Borstel, Germany
    Search for more papers by this author

Correspondence

A. B. Schromm, Division of Immunobiophysics, Research Center Borstel, Leibniz-Center for Medicine and Biosciences, Parkallee 10, 23845 Borstel, Germany

Fax: +49 4537 1886320

Tel: +49 4537 1882960

E-mail: aschromm@fz-borstel.de

Abstract

We established a new preparative separation procedure, based on DOC/PAGE, to isolate intact lipopolysaccharide (LPS) fractions from natural LPS preparations of Escherichia coli. Analysis of the chemical integrity of LPS fractions by MS showed that no significant chemical modifications were introduced by the procedure. Contamination with toll-like receptor 2 (TLR2)-reactive cell-wall components present in the natural LPS mixture was effectively removed by the procedure, as determined by the absence of reactivity of the purified fractions in a HEK293-TLR2 cell line. Biologic analysis of LPS fractions derived from E. coli O111 in human macrophages demonstrated that the rough (R), semirough (SR) and smooth (S) LPS fractions were highly active at inducing tumor necrosis factor-alpha (TNF-α) in the presence of human serum; however, on a weight basis the R-LPS and SR-LPS fractions were more active, by a factor of 10–100, than was the S-LPS fraction. Under serum-free conditions, the natural LPS mixture, as well as the R-LPS and SR-LPS fractions, showed dose-dependent activation of macrophages, although the response was attenuated by about 10- to 100-fold. In contrast, the S-LPS fraction failed to induce TNF-α. Remarkably, the dose–response of the natural LPS mixture resembled that of the R-LPS and SR-LPS fractions, supporting that short-chain (R and SR) forms of LPS dominate the innate immune response of human macrophages to LPS in vitro. Biologic activity to the S-LPS fraction under serum-free conditions could be restored by the addition of recombinant lipopolysaccharide-binding protein (LBP). In contrast, soluble cluster of differentiation antigen 14 was not able to confer activity of the S-LPS fraction, indicating a crucial role of LBP in the recognition of S-LPS by human macrophages.

Abbreviations
C14:0(3-OH)

3-hydroxy-tetradecanoic acid

CD14

cluster of differentiation antigen 14

dideoxyHex

dideoxyhexose

DOC/PAGE

sodium deoxycholate–polyacrylamide gel electrophoresis

ESI FT-ICR MS

electrospray ionization-Fourier transform ion cyclotron mass spectrometry

Hep

heptose

Hex

hexose

HexN

hexosamine

HexNAc

N-acetylhexosamine

IL

interleukin

Kdo

3-deoxy-d-manno-oct-2-ulosonic acid

LBP

lipopolysaccharide-binding protein

LPS

lipopolysaccharide

MD-2

myeloid differentiation protein 2

MNC

mononuclear cells

Myd88

myeloid differentiation primary response gene (88)

NHS

normal human serum

R

rough

Ra-LPS

rough mutants with a complete core oligosaccharide

Re-LPS

deep-rough mutants with the shortest core

S

smooth

SR

semirough

TEA

triethylamine

TLR

toll-like receptor

TNF-α

tumor necrosis factor-alpha

Introduction

Lipopolysaccharides (LPS; endotoxins) from Gram-negative bacteria are responsible for many of the pathophysiological effects of bacterial infections. LPS represent a highly active stimulus of the innate immune system, providing an immediate immune response, that can be protective but, in the event of an overshooting systemic immune alert, also detrimental, leading to sepsis and septic shock [1]. The chemical architecture of the LPS molecule is strongly associated with its biologic activity. Lipid A anchors the molecule in the bacterial membrane and is the minimal active component required to induce the immune response [2]. Lipid A is composed of a 1,4′-bisphosphorylated β(1→6)-linked d-glucosamine (GlcN) disaccharide backbone asymmetrically acylated by amide- or ester-bound (R)-3-hydroxymyristoyl residues. In mammalian assay systems, maximal endotoxicity is induced by the hexa-acylated and bisphosphorylated enterobacterial-type lipid A structure [3, 4].

Natural LPS from the bacterial cell wall is composed of a mixture of different molecules that can vary in the length and composition of their carbohydrate chain as well as in the composition of their lipid A. In addition to lipid A, LPS contains a core oligosaccharide and, in Enterobacteria and many other genera, the O side-chain, which is made up of repeating oligosaccharide units. Many clinically relevant Gram-negative bacteria express this type of LPS, which is denoted as smooth (S). In contrast, rough (R) mutant strains express a core oligosaccharide of varying length but lack an O polysaccharide. A number of clinical and laboratory strains express core oligosaccharides with different degrees of completion: deep-rough mutants with the shortest core (Re-LPS) to rough mutants with a complete core oligosaccharide (Ra-LPS).

LPS belong to the most powerful classes of immunostimulators. They are recognized by the combined action of several pattern-recognition receptors. The complex of toll-like receptor (TLR) and the myeloid differentiation protein 2 (MD-2) constitute the central cellular signaling system that mediates the recognition of lipid A [1]. The TLR4/MD-2 receptor system is strongly enhanced by the action of the lipopolysaccharide-binding protein (LBP) and the membrane-bound or soluble form of the cluster of differentiation antigen 14 (CD14). Although cell activation by LPS is based on the interaction of the pattern recognition receptors with lipid A, there is indication in the literature demonstrating an influence of the oligosaccharide portion on the molecular pathway of cell activation. Thus, the requirement of receptor proteins, such as CD14, has been found to be different for Re-LPS and S-LPS, suggesting differential recognition pathways for smooth and rough mutant chemotype LPS [5, 6].

Knowledge on the principles of biologic immune recognition are crucial for the development of treatment strategies for endotoxin-related pathology in diseases such as sepsis and septic shock, which are the second leading cause of death in noncoronary intensive care units [7]. A central problem in elucidating the mechanisms of recognition of individual molecular species in LPS preparations is the fact that these preparations derived from the bacterial cell wall represent natural mixtures of a variety of chemically different LPS molecules ranging from Re LPS and LPS with varying core length to LPS molecules with high numbers of O-chain oligosaccharide repeats. In addition to these variations in carbohydrate composition, the molecules may be differently substituted with acyl chains. Therefore, most of our current knowledge on the contribution of individual molecular species is based on the investigation of LPS from mutants lacking specific enzymes involved in the synthesis of LPS core oligosaccharides. We have established a new methodology for the preparative separation of different chemotypes from wild-type LPS, which enables the investigation of the chemotypes of natural LPS mixtures individually (i.e. the definition of their chemical composition and the investigation of their specific biologic recognition pathways). In this report we present data on the isolation and analysis of R-fractions from the E. coli Re chemotype (strain F515) and of R-LPS and S-LPS fractions from the wild-type E. coli strain O111 and describe their ability to activate human macrophages.

Results

Isolation of intact LPS species by sodium deoxycholate/PAGE

We have demonstrated recently that the combination of SDS- or DOC/PAGE with reverse staining and passive elution in 5% triethylamine (TEA) is effective for the isolation of electrophoretically homogenous fractions from rough- and smooth-type LPS mixtures in quantities up to several hundreds of micrograms [8]. The LPS fractions obtained are amenable to structural analysis by MALDI-TOF MS or electrospray ionization (ESI)-MS/MS when they are chemically derivatized before MS to produce O-deacylated LPS or dephosphorylated and permethylated oligosaccharides [9, 10]. In addition, it was verified that the isolation method did not introduce chemical modifications in the LPS part-structures analyzed [9, 10]. Nevertheless, a potential shortcoming of using 5% TEA for the elution of LPS from gels for biologic studies was that, depending on the time of incubation, the temperature and the concentration, TEA solutions could variably promote the release of alkali-labile ester-linked fatty acids from LPS [11-13]. For this reason, we first addressed the question of to what extent this isolation procedure was capable of preserving the chemical integrity of these alkaline-labile LPS components by analyzing the intact LPS using ESI Fourier transform ion cyclotron (FT-ICR) MS, and if elution conditions could be changed to minimize any deleterious effect on the alkali-labile ester-linked fatty acids residues without significantly compromising the recovery of LPS from gels.

To investigate this, the Re-type LPS mixture from E. coli F515 was used. This preparation was mainly composed of three molecular species carrying six, five or four fatty acids, of 2237.34, 2027.14 and 1800.95 Da, respectively, as determined by ESI FT-ICR MS (Fig. 1A). Also, an LPS species with four fatty acids and lacking one 3-deoxy-d-manno-oct-2-ulosonic acid (Kdo) unit (1580.89 Da; Fig. 1A) and other minor LPS species were detected, which are not further described. This LPS mixture migrated as a single band on the 15% DOC polyacrylamide gel. A total of 500 μg of LPS was applied to preparative DOC/PAGE and the LPS band displayed by reverse staining was recovered from the gel by passive elution in 5% TEA [8]. When the recovered LPS species were analyzed using ESI FT-ICR MS, the same main LPS species, as described for the natural LPS mixture (Fig. 1B), were detected. However, notable changes in the peak intensities are obvious, including, for instance, increases in the signals of penta-acyl Re-LPS (2011.15 Da), triacyl Re-LPS (1574.76 Da) and triacyl LPS species lacking one Kdo unit (1354.70 Da) and decreases in the signals of hexa-acyl (2237.34 Da) and penta-acyl (2027.14 Da) LPS species. This change in the proportion of LPS species can be explained by the loss of a 3-hydroxy-tetradecanoic acid [C14:0(3-OH), Δm = −226.19 Da].

Figure 1.

Charge-deconvoluted negative ion ESI FT-ICR mass spectra of the Re-LPS of Escherichia coli F515. (A) Natural LPS mixture; (B) LPS recovered in 5% TEA from reverse-stained DOC-polyacrylamide gels. Signals labeled within boxes correspond to LPS species generated from the cleavage of 3-hydroxy-tetradecanoic fatty acid [C14:0(3-OH)]. TEA was removed from LPS samples by lyophilization (previously reported procedure [8-10]). Note that LA(5*) and LA(5) carry the same number of fatty acids but have a different fatty-acid composition. Compared with LA(6), LA(5*) contains one less C14:0(3-OH) while LA(5) contains one less n-tetradecanoic fatty acid (C14:0). LA(n), diphosphoryl n-acyl lipid A (n = 3, 4, 5 or 6); P, phosphate; C12:0, n-dodecanoic acid.

In order to approach this problem, the effect of incubation time in 5% TEA at 20–23 °C on the extent of LPS deacylation was studied. Immediately after each incubation period the pH of the solution was brought to pH 7 by the addition of 10% acetic acid. As expected, the proportion of partially deacylated LPS species considerably increased with a prolonged incubation time, as determined by ESI FT-ICR MS (Fig. 2). However, when the LPS species were incubated for only 5–10 min, the relative proportion of partially deacylated LPS was negligible and comparable with that of the LPS species that were not exposed to this basic solution (natural LPS mixture; Fig. 2). Furthermore, incubation of other chemotypes of LPS (e.g. the Ra-LPS of E. coli K-12 W3100 and the S-LPS of E. coli O111) in 5% TEA for 5 min at 20–23 °C did not produce any significant modification of the lipid A from these LPS preparations (data not shown).

Figure 2.

Relative intensity ratio of penta-acyl Re-LPS species (Mr = 2011.15 Da) to that of hexaacyl Re-LPS species (Mr = 2237.34 Da) upon incubation in 5% TEA at 20–23 °C. As a control, the ratio is given of LPS suspended in water only (natural LPS mixture). An increasing cleavage of one hydroxytetradecanoic fatty acid from the hexa-acyl Re-LPS species of Escherichia coli F515 starts after 20 min of incubation.

To test, in more detail, the applicability of these incubation conditions a total of 500 μg of the Re-LPS mixture from E. coli F515 was subjected to the DOC/PAGE-based isolation. For LPS elution and subsequent preparation steps, a modified extraction method was applied consisting of the passive elution of LPS by incubating the LPS-containing gel microparticles for only 5 min in 5% TEA followed by neutralization with 10% acetic acid. The samples were then dialyzed, freeze-dried and desalted by precipitation with ethanol/acetone.

Indeed, the MS spectrum of the Re-LPS extracted according to the new procedure comprised similar peaks and peak intensities to that of the natural LPS mixture (Fig. 3). Mass peaks originating from partially deacylated LPS species that were notably increased in the spectra of the LPS recovered using the previous isolation procedure (2011.15, 1574.76 and 1354.70 Da; Fig. 1B), were of only minimal intensity. Thus, proof was obtained that by using the newly adopted procedure for LPS elution and subsequent preparation it is possible to isolate LPS species by DOC/PAGE without significantly altering LPS integrity. Consequently, the newly elaborated method was selected for use in following DOC/PAGE preparative experiments.

Figure 3.

Negative-ion charge-deconvoluted spectrum of the Re-LPS of Escherichia coli F515 isolated using the newly adopted procedure. The spectra of the natural LPS mixture (A), used as a reference, and that of the LPS recovered from reverse-stained DOC-polyacrylamide gels (B) are shown. For LPS elution, LPS bands were incubated for 5 min in 5% TEA and the solution was neutralized afterwards with acetic acid (new procedure). Note that LA(5*) and LA(5) carry the same number of fatty acids but have a different fatty acid composition. LA (5*) contains one less C14:0(3-OH), whilst LA(5) contains one less n-tetradecanoic fatty acid (C14:0) compared with LA(6). LA(n), diphosphoryl n-acyl lipid A (n = 3, 4, 5 or 6).

Fractionation and MS analysis of the S-LPS mixture of E. coli O111

The S-LPS mixture of E. coli O111 was fractionated by the established DOC/PAGE-based method into three fractions consisting of R-LPS, semirough (SR)-LPS and high-molecular-mass smooth S-LPS. To do this, 12, 16% polyacrylamide gels, each loaded with 3.33 mg of LPS, were run simultaneously. As a result, quantities of 3.11, 1.11 and 1.15 mg of the R-LPS (band 1), SR-LPS (band 2) and S-LPS (smooth-fraction) fractions, respectively, were obtained (Table 1). Analysis of the isolated LPS fractions by re-electrophoresis did not show any noticeable cross-contamination (Fig. 4).

Table 1. Quantities of the LPS fractions of Escherichia coli O111 recovered from DOC-polyacrylamide gels
LPS fractionLPS chemotypeQuantity (mg)a
  1. a

    The quantities of LPS were obtained from a total of 40 mg of starting LPS mixture. Values represent the mean ± standard deviation of three determinations.

Band 1R-form: complete core oligosaccharide3.11 ± 0.02
Band 2SR-form: complete core oligosaccharide plus one O-chain repeat1.11 ± 0.05
Smooth-fractionS-form: complete core oligosaccharide plus 10–18 O-chain repeats1.15 ± 0.06
Figure 4.

DOC/PAGE analysis of isolated LPS fractions from Escherichia coli O111. Lanes: 1, S-LPS fraction; 2, natural LPS mixture, used as a reference; 3, R-LPS fraction; and 4, SR-LPS fraction. The gel was stained with silver.

Analysis of the lipid A of the natural LPS mixture of E. coli O111 by unspecific fragmentation in the collision cell showed that this LPS preparation was almost exclusively composed of fully acylated hexa-acyl diphosphoryl lipid A species (1797.22 Da; Fig. 5A). Only minor signals corresponding to penta-acyl and tetra-acyl diphosphoryl lipid A species were observed (1569.02 and 1360.83 Da, respectively; Fig. 5A). Also, some heterogeneity in the composition of the main hexa-acyl lipid A species was present as a result of the addition of phosphate (1877.20 Da; Fig. 5A) or phosphoethanolamine (1920.25 Da; Fig. 5A), the loss of phosphate (1717.26 Da; Fig. 5A) or the presence of fatty acyl chains one or two carbons shorter (1783.22 and 1769.19 Da; Fig. 5A).

Figure 5.

Analysis of the lipid-A moiety of LPS fractions from Escherichia coli O111 recovered from DOC-polyacrylamide gels by unspecific fragmentation in the collision cell. Negative-ion charge-deconvoluted spectra in the mass range from 1200 to 2000 Da covering lipid-A fragments of the natural LPS mixture (A); the R-LPS fraction (B) and the SR-LPS fraction (C) isolated from DOC-polyacrylamide gels. No lipid-A species of Mr > 2000 Da were observed.

The isolated R-LPS and SR-LPS fractions were analyzed under the same unspecific fragmentation conditions. The spectra closely resembled the lipid A structures found in the natural LPS mixture (Fig. 5A). It is of note that the mass spectra of the isolated fractions were dominated by the same signals of fully acylated hexa-acyl diphosphoryl lipid A species (Fig. 5B,C). Minor signals of penta-acyl and tetra-acyl diphosphoryl lipid A species were also present (Fig. 5B,C). These results confirmed that the isolation method preserved the chemical integrity of the lipid-A region of LPS. MS analysis of the R-LPS and SR-LPS fractions of E. coli O111 obtained from two additional fractionation experiments showed mass spectra similar to those presented in Fig. 5B,C, thus indicating that these results were reproducible (data not shown). Intact high-molecular-mass peaks of the heterogeneous S-LPS, with high numbers of O-chain repeats could not be detected using mass spectrometric techniques.

The LPS of E. coli are known to contain one of five distinct core types, namely R1, R2, R3, R4 and K-12 [14]. E. coli O111, in particular, has previously shown genetic sequences which account for the biosynthesis of an R3 core-type [15]. In accordance with this was the sugar-composition analysis of the rough fraction of LPS isolated from E. coli O111, whereby hexosamine (HexN)-heptose (Hep) and HexN-hexose (Hex) disaccharides were detected. The structure of the O-antigen-repeating unit of this LPS has previously been determined [16].

As shown in Fig. 6, high-quality ESI FT-ICR MS spectra of the intact LPS species of the isolated R-LPS and SR-LPS fractions could be obtained. In the spectra of the isolated R-LPS (Fig. 6A) the molecular mass of the main signal (3905.85 Da) is consistent with LPS species containing an R3 core-type oligosaccharide with the composition Hex4·HexNAc·HexN·Hep3·P·Kdo2 linked to a hexa-acyl diphosphoryl lipid A (Mr LPScalc = 3905.85 Da). The spectra of the isolated SR-LPS fraction contained a main signal of 4693.20 Da (Fig. 6B), which is in agreement with hexa-acylated diphosphorylated LPS species that contain an R3 core oligosaccharide structure of the same composition as that of the major species of the R-LPS fraction plus one (dideoxyHex2·Hex2·HexNAc) O-chain repeating unit (Mr LPScalc = 4693.16 Da) [16-18]. Note that the molecular ions present in the spectra of the R-type LPS fraction (Fig. 6A) were not present in the spectra of the SR–LPS fraction (Fig. 6B), and vice versa. Therefore, the highly efficient separation of the LPS fractions, as previously observed by DOC/PAGE analysis (Fig. 4), was confirmed.

Figure 6.

Charge-deconvoluted negative-ion ESI FT-ICR mass spectra of intact LPS fractions of Escherichia coli O111 were recovered from DOC-polyacrylamide gels. Signals correspond to complete LPS molecular ions from the R-LPS fraction (A) and the SR-LPS fraction (B). The molecular mass of the most prominent signals of LPS containing four, five or six fatty acids are depicted. For simplicity, other molecular ions are not annotated.

In order to estimate the molecular mass of the S-fraction of LPS, we performed analytical SDS/PAGE of this fraction using the natural LPS mixture as the standard. The S-fraction was determined to contain between 10 and 18 repeating units, from which a molecular mass range of 11.8–18.0 kDa could be calculated (data not shown). Based on these data, a mean difference in molecular mass of a factor of 3.8 between the R-LPS fraction (band 1, 3905.85 Da, Fig. 6A) and the S-LPS fraction could be estimated.

Biologic analysis of LPS fractions isolated by DOC/PAGE

DOC/PAGE efficiently removes TLR2 activity from LPS preparations

Established LPS purification procedures provide LPS preparations that can contain significant amounts of contaminants from the bacterial cell wall. Thus, a major problem is the efficient removal of TLR2 ligands, such as lipoproteins and lipopeptides, which are constituents of the outer membrane of Gram-negative bacteria and comprise a whole class of microbial molecules engaging pattern-recognition receptors. These contaminants exhibit biologic activities on a variety of host cells and their contribution to cell activation by engagement of the TLR2 receptor system can be a major problem in biologic test systems when signaling pathways are investigated. Phenol/water extraction in the presence of TEA and DOC has been demonstrated to be an efficient procedure for the removal of TLR2-activating molecules from LPS preparations [19, 20]. However, this method is not intended to reduce the heterogeneity of LPS preparations in terms of LPS molecular species. To address the purity of our purification procedure with respect to TLR2 contamination, we analyzed the natural LPS mixture from E. coli O111 and the R-LPS (band 1), SR-LPS (band 2) and S-LPS fractions derived using the DOC/PAGE procedure, as well as the natural mixture and the SR-LPS fraction (band 2) derived from E. coli O12, for their TLR2 activity. An HEK293 cell line, which stably expresses human TLR2, was stimulated and the chemokine interleukin (IL)-8 was determined in the supernatant as a well-established TLR2-dependent activation response of these cells. The natural LPS mixtures of E. coli O111 and E. coli O12 exhibited a significant dose-dependent activation of HEK-293 TLR2 cells. In contrast, neither of the isolated fractions investigated induced TLR2 activation up to LPS concentrations of 1 μg·mL−1 (Fig. 7). Thus, the procedure described here provides highly purified LPS species without any contaminating TLR2 ligands.

Figure 7.

TLR2 activity of LPS preparations before and after preparative gel fractionation. HEK293 cells stably expressing TLR2 were stimulated with LPS from the Escherichia coli O111 natural preparation and fractions derived thereof (Band 1, R-LPS fraction; Band 2, SR-LPS fraction; Smooth-fraction, S-LPS fraction) or with LPS from the E. coli O12 natural preparation and a fraction isolated thereof (Band 2, SR-LPS fraction) at the indicated concentrations or with 100 ng·mL−1 of the synthetic lipopeptide Pam3CSK4 (LP). Unstimulated cells were included as negative controls (X). After 24 h, supernatants were collected for determination of the chemokine IL-8. Data shown are mean and SD of triplicate samples of one experiment representative of at least three independent experiments.

Activation of human macrophages by the LPS of E. coli O111 and fractions isolated thereof

Role of the antigenic property of O glycosylation in biologic activity

The O glycosylation of S-LPS activates adaptive immunity through its antigenic character (referred to as O-antigen). Natural LPS antibodies in human serum can interfere with the biologic responses to LPS. Therefore, special attention has been paid to the exclusion of effects induced by LPS antibodies in human serum. The human AB serum used in these experiments was depleted of LPS-reactive antibodies by LPS affinity-chromatography. Immunoblot analysis showed efficient removal of anti-LPS reactivity from normal human serum (NHS), as seen by the disappearance of the typical ladder staining by LPS antibodies in normal human AB serum (Fig. 8A, lanes a, b) compared with the same serum after antibody depletion (Fig. 8A, lanes c, d).

Figure 8.

LPS antibodies in serum do not inhibit TNF-α production of LPS-stimulated human macrophages. (A) Western blots of human AB serum before and after absorption of O111 antibodies. O111 LPS (8 μg per lane) was separated by SDS/PAGE on a 13% resolving gel, blotted onto nitrocellulose and stained with normal human AB serum diluted 1 : 10 (lane a) or 1 : 50 (lane b), or with normal human AB serum after removal of O111-specific antibodies by affinity chromatography, as described in the Materials and Methods. The dilutions were 1 : 5 (lane c) or 1 : 10 (lane d). (B) TNF-α production from human macrophages following stimulation with the R-LPS fraction (band 1, squares) and the S-LPS fraction (circles) from Escherichia coli O111. Cells were seeded at 2 × 105 cells per well into 96-well plates in RPMI-1640 in the presence of 4% normal human AB-type serum (NHS; filled symbols) or 4% serum depleted of antibodies against LPS from E. coli O111 (Ab-depleted human serum; open symbols) and stimulated with LPS at the indicated concentrations or left unstimulated as controls. TNF-α was determined in cell-free supernatants after 4 h of stimulation (the mean and SD values are from duplicate determinations). The data are representative of four independent experiments with cells from different donors.

The isolated and chemically characterized fractions derived from the LPS mixture were then analyzed with respect to their biologic activity in human macrophages. As a central mediator of the pro-inflammatory innate immune response of human mononuclear cells (MNCs) to endotoxin and a major contributing factor in the development of the septic shock syndrome, the production of the cytokine tumor necrosis factor-alpha (TNF-α) was determined in the supernatant of stimulated macrophages. Human macrophages stimulated with R-LPS or S-LPS fractions (Fig. 8B) in the presence of NHS or antibody-depleted serum showed very similar dose–dependent responses, suggesting that LPS antibodies do not have a major impact on the biologic reactivity observed in these experiments. However, to exclude any potential effects of antibodies interfering with the biologic recognition process, all further experiments were performed with antibody-depleted human serum. Note that the S-LPS fractions were about one to two orders of magnitude less active (converting to a 2.6– to 26-fold lower activity when considering the differences in molecular weight) than the R-LPS fractions, which induced maximal TNF-α production at 1 ng·mL−1. However, at concentrations above 10 ng·mL−1 also the S-LPS fraction induced maximal cell activation.

S-LPS and R-LPS fractions have differential requirements for serum to induce macrophage activation

The isolated LPS fractions derived from E. coli O111 were then analyzed in detail and compared with the biologic activity of the natural S-LPS mixture. Stimulation experiments demonstrated a highly similar dose-dependent release of TNF-α in response to the natural S-LPS mixture as well as to the isolated R-LPS and SR-LPS fractions derived thereof when human macrophages were stimulated in vitro in the presence of 4% human serum (Fig. 9, solid lines). In contrast and consistent with the abovementioned results (Fig. 8), the S-LPS fraction was about 1–2 orders of magnitude less active (translating to a 2.6- to 26-fold lower activity on a molar basis), but induced maximal activation at concentrations above 10 ng·mL−1. These results support the conclusions that (a) the fractionation procedure did not confer negative effects on the fractions interfering with the activation of the macrophages because the R-LPS and SR-LPS fractions showed dose-dependent activity similar to that of the natural mixture and (b) the molecular species in the R-LPS and SR-LPS fractions have considerably higher abilities to induce TNF-α compared with the S-LPS fraction.

Figure 9.

TNF-α production from human macrophages by isolated LPS fractions in the presence and absence of human serum. Human macrophages were seeded at 2 × 105 cells per well into 96-well plates in RPMI-1640 containing 4% antibody-depleted human serum (solid lines) or in serum-free RPMI-1640 (dashed lines) and stimulated with LPS of Escherichia coli O111 or isolated LPS fractions (band 1 and band 2 = R-LPS and SR-LPS fractions, respectively; Smooth-fraction = highly-glycosylated fraction) at the indicated concentrations. Control cells were left unstimulated in the presence of serum (−) or in serum-free medium (x). TNF-α was determined in cell-free supernatants after 4 h stimulation (mean and SD are from duplicate determinations). The data are representative of three independent experiments with cells from different donors.

Serum proteins, such as LBP and soluble CD14, are important contributors conferring a sensitive recognition of LPS to mononuclear phagocytes. Notably, the results of the biologic activity of LPS fractions were different when experiments were performed in the absence of serum. Consistent with the data observed in the presence of serum, the R-LPS and SR-LPS fractions exhibited a dose-dependent activity that was very similar to that of the natural, S-LPS mixture under the same conditions (Fig. 9, dashed lines). Compared with their activity in the presence of serum, cell stimulation was reduced by two orders of magnitude for the natural LPS mixture and for both R-LPS and SR-LPS fractions. When human macrophages were stimulated under serum-free culture conditions, a strong decrease of activity was observed for the smooth LPS fraction, which did not induce TNF-α production, up to a concentration of 10 μg·mL−1. In some experiments, partial cell activation was observed only at a concentration of 100 ng·mL−1 of the S-LPS fraction under serum-free conditions, whereas also in these cases higher concentrations did not induce cell activation (data not shown). In conclusion, the biologic response of human macrophages to the isolated S-LPS fraction is strongly dependent on the presence of serum and requires serum factors for the activation, whereas recognition of the R-LPS and SR-LPS fractions clearly occurs independently of serum factors. The biologic activity of the natural LPS mixture was slightly reduced when compared with the activity of R-LPS and SR-LPS fractions under serum-free conditions, which could be interpreted in a way that the smooth fraction of LPS might affect the biologic recognition of the latter.

We thus investigated whether the presence of the S-LPS fraction would alter the immunologic response to the R-LPS fraction. Human macrophages were pre-incubated in serum-free culture medium with different concentrations (1 ng·mL−1 to 10 μg·mL−1) of the S-LPS fraction for 10 min and were subsequently stimulated with 1 ng·mL−1 of the R-LPS fraction under serum-free conditions. The production of TNF-α was reduced by about 20–25% in the presence of the S-LPS fraction at a ratio of 1 : 1 to a ratio of 1 : 100 (R-fraction: S-fraction, w/w). Thus, in this concentration range the S-LPS did interfere, but only slightly, with cell activation by R-LPS. On the other hand, higher concentrations of the S-LPS fraction, of 1 : 1000 up to 1 : 10 000, almost completely inhibited the activation of macrophages by the R-LPS fraction (Fig. 10). Although statistical analysis of independent experiments with cells from different donors indicated that the inhibition of the R-LPS fraction was already significant at a ratio of 1 : 10 (one-tailed t-test of paired samples; P ≤ 0.05), only when present in large excess did the S-LPS fraction completely prevent the activation of macrophages by the R-LPS fraction.

Figure 10.

S-LPS can inhibit activity of R-LPS. Human macrophages were seeded at 2 × 105 cells per well into 96-well plates in RPMI-1640 medium without serum. Cells were pre-incubated with the indicated concentrations of the S-LPS fraction from Escherichia coli O111 or control medium for 10 min at 37 °C and subsequently stimulated with 1 ng·mL−1 of the R-LPS fraction (band 1) from E. coli O111. TNF-α was determined in cell-free supernatants after 4 h of stimulation (results are expressed as mean and SD from duplicate determinations). The data are representative of three independent experiments with cells from different donors. n.d., not detectable.

To address the question of which components in serum are required for the activity of the S-LPS fraction on human macrophages, we performed stimulation experiments under serum-free conditions in the presence of recombinant human LBP and soluble CD14. These experiments clearly demonstrate that solely LBP was able to restore the activation of human macrophages to the S-LPS fraction, whereas soluble CD14 did not confer activation to the S-LPS fraction (Fig. 11).

Figure 11.

LBP imparts activation of human macrophages by S-LPS under serum-free conditions. Human macrophages were seeded at 2 × 105 cells per well into 96-well plates in RPMI-1640 without serum. Cells were pre-incubated with human recombinant LBP or human recombinant soluble CD14 (sCD14) at the indicated concentrations for 15 min at 37 °C or with control buffer (C) and subsequently stimulated with 1 ng·mL−1 of the S-LPS fraction from Escherichia coli O111. The concentration of TNF-α was determined in cell-free supernatants after 4 h of stimulation (results are expressed as mean and SD of four independent experiments with cells from different donors; *P ≤ 0.05). n.d., not detectable.

Discussion

Bacterial LPS is among the most sensitive of stimuli of the innate immune response, and LPS derived from Enterobacteriaceae are widely used to investigate immune responses and signaling pathways. Recent advances in elucidating the mechanisms of LPS recognition have revealed that activation of the TLR4/MD-2 receptor system can employ different co-receptors and activate different signaling pathways. Specific differences have recently been connected to the degree of glycosylation of LPS as it appears in LPS from wild-type strains with a high content of sugars as compared to LPS from rough mutant strains with only a minor content of sugars. However, investigation of the precise mechanisms and routes of activation of specific chemical subsets of LPS has so far been hampered by the fact that both S-LPS and R-LPS from bacteria represent natural mixtures of LPS variants. The optimized preparative protocol described in this study provides clearly separated chemotype subsets from natural LPS preparations and is thus a new technique for generating samples suitable for the elucidation of chemotype-specific activation pathways. As evidenced by analysis of the LPS fractions in acrylamide gel (Fig. 4) and proven by MS (Fig. 6), there is no indication of any cross-contamination of the individual chemotypes. The LPS fractions obtained exhibited an acylation pattern comparable with that observed in the original LPS mixture (Fig. 5). The purification procedure also removed TLR2 activity (Fig. 7) from the LPS preparations and therefore the procedure also provides a cleaning tool to remove TLR2-activating contaminants, such as bacterial lipopeptides, which might otherwise influence the signaling pathways under investigation. Compared with other purification protocols, our method can be used with very small amounts of LPS. As demonstrated for different strains of E. coli O111 and O12 (data not shown) in this study and for Salmonella abortus equi [21], this protocol presents a suitable technique to provide LPS fractions with a defined degree of glycosylation from the Enterobacteriaceae. Compared with other methods previously used for preparative separation of S-LPS mixtures, based on centrifugal partition chromatography [22], gel-filtration chromatography [23] and DOC/PAGE in cylindrical or slab gels [24-26], the present procedure has the advantages of providing a higher resolution through precise detection and fractionation of separated LPS species, effective removal of salts and detergents associated with the purification and is suitable for the analysis of SDS/PAGE- or DOC/PAGE-derived samples by MS.

The most important result of the biologic characterization of LPS fractions isolated from wild-type E. coli is the fact that the R and SR forms of LPS had a broader capacity to activate human macrophages in vitro compared with the isolated S form of LPS, which required the presence of serum to induce the induction of TNF-α as a central mediator of the pro-inflammatory response to bacterial infection. Macrophages represent a large pool of innate immune cells that reside in almost all human tissues and are major players in the first line of defense against Gram-negative infections. Macrophages express the GPI-anchored protein, CD14 (a co-receptor of the TLR4/MD-2 system, which enables sensitive recognition of LPS), on their surface [27]. Serum proteins, such as the LBP and soluble CD14, have been shown to provide transport for LPS from aggregates to the cellular receptor molecules, thereby lowering the recognition threshold toward a more sensitive response [28, 29]. In accordance with this, the response of human macrophages was about one order of magnitude more sensitive in the presence, than in the absence, of human serum. However even under serum-free conditions, TNF-α production was observed with 1–100 ng·mL−1 of LPS mixture derived from wild-type E. coli (Fig. 9). Strikingly, this was not the case for the isolated S-LPS fraction, which did not induce cell activation at a concentration of up to 10 μg·mL−1 in the absence of serum (Fig. 9), clearly demonstrating that the highly glycosylated LPS fraction requires different receptors or molecular pathways to induce cell activation. Huber et al. also employed a fractionation strategy and found that an R-form fraction isolated from S. abortus equi S-LPS exhibited much higher biologic activity compared with the parental LPS in inducing IL-6 from murine macrophages and mast cells, as well as mitogenic responses in murine splenocytes, assigning R-LPS a key role in biologic activation [6]. The biologic activities of R-LPS and lipid-A structures depend on the aggregate structure and physicochemical characteristics [3, 4, 30, 31]. Aggregates have been proposed to be the biologically active unit [32], whereas several other studies have suggested that the sequential transport of LPS from aggregates to receptor proteins is important for cell activation [29, 33]. Independently of the route of activation, the physico-chemical properties of S-LPS (such as hydrophobicity, aggregate structure and stability) are likely to be different from those of R-LPS and could explain the observed differences in the biologic activity and requirement for serum proteins.

Basic differences in the receptor requirement of R-LPS and S-LPS have recently been described. The mutant mouse strain ‘Heedless’ has been shown to be deficient in recognition of S- LPS, whereas R-LPS retains full activity in cells from these mice. The mutation was identified to encode a premature stop-codon in the gene of CD14, leading to a CD14-negative phenotype [5]. Peritoneal macrophages from these mice did not respond to S-LPS, but showed normal activation to R-LPS. In this study, a similar phenotype was also demonstrated for peritoneal macrophages from CD14−/− mice. Another study, which compared the response of macrophages from CD14 wild-type mice and CD14−/− mice, also suggested that CD14 is critically involved in the recognition of S-LPS and discriminates between LPS structures with a different degree of glycosylation [34]. A role for CD14 in the recognition of S-LPS has also been demonstrated for other cell types such as mast cells, which do not express GPI-anchored CD14 and thus represent a CD14-negative immune cell. Murine mast cells were found to respond normally to R-LPS but did not respond to S-LPS in the absence or at physiological serum concentrations of CD14 [6]. The production of IL-6 by murine macrophages in response to S-LPS was found to be dependent on the expression of CD14 and was enhanced by LBP [6]. Also, dendritic cells and natural killer cells appear to mount differential responses to R-LPS and to S-LPS [35]. In most of these investigations, LPS preparations from different wild-type or rough-mutant bacteria were employed to elucidate chemotype-dependent cell activation. Our data provide, for the first time, clear evidence for a chemotype-specific immune recognition in human immune cells and supports that human phagocytes employ differential mechanisms of recognition for S, R and SR forms of LPS. CD14 expression on human macrophages is not sufficient for the production of TNF-α, but requires the presence of LBP. LBP was able to confer cell activation to S-LPS fractions under serum-free conditions, whereas soluble CD14 was not (Fig. 11). Of note, the TLR4/MD-2 receptor complex activates two independent signaling pathways - the myeloid differentiation primary response gene (88) (MyD88)-dependent and -independent pathways – via recruitment of the adaptors TIR domain-containing adapter-inducing interferon-β and translocating chain-associating membrane protein. Whereas our studies have so far been focused on the MyD88-dependent pathway leading to nuclear factor-kappaB-dependent production of TNF-α, different requirements might be considered for the MyD88-independent pathway, especially because this pathway is activated in the endosomal compartment. The observation that the human macrophages investigated in our study did not respond to the isolated S-LPS fraction under serum-free conditions clearly demonstrates that investigation of the isolated fractions can give valuable additional information on the pathways of LPS recognition.

To explain the strong differences in the biologic activity among R, SR and S forms of LPS, it was speculated that S-LPS contains a less acylated fraction in LPS preparations and therefore might express lower biologic activity. Unfortunately, it is technically not possible to analyze the acylation pattern in the intact S form of LPS without chemical degradation of the molecule to generate free lipid A. However, the MS analysis of lipid A derived by unspecific fragmentation of the natural LPS mixture in the collision cell did not indicate the presence of major amounts of lower acylated species (Fig. 5). Also, our biologic data are not in agreement with a lower acylation pattern of S-LPS because the S-LPS fraction showed strong biologic activity in the presence of serum (Figs 8 and 9). Human myeloid cells show a peculiar species-specific immune response that is dependent on the acylation pattern and the molecular conformation of lipid A. Thus, only fully hexa-acylated lipid A expresses strong biologic activity on these cells, whereas lower acylated penta- and tetra-acylated lipid-A structures do not activate myeloid cells but express antagonistic activity. In line with this would be our observation that the S-LPS fraction was able to inhibit activation of the R-LPS fraction (Fig. 10). However, in contrast to known antagonists such as tetra-acyl lipid-A compound 406, lipid A from Rhodopseudomonas sphaeroides or Pseudomonas gingivalis, or the synthetic penta-acylated lipid A-like compound, E5564, which inhibit cell activation by biologically active LPS at low molar excess or even at an equimolar ratio [36-39], a high excess of more than 1000-fold on a weight basis (or of more than 263-fold on a molar basis) was required for efficient inhibition of cell activation by the S-LPS fraction. Considering, in addition, that the abovementioned antagonists have not been reported to depend on the presence of serum proteins to display their antagonistic activity, it can be concluded that the inhibitory effect observed for the S-LPS fraction is probably based on a different mode of action. The amount of S-LPS contained in the wild-type LPS mixture of E. coli O111 used in this study can be estimated from the DOC gels of the natural mixture (Fig. 4, lane 2) and the quantities of LPS fractions isolated (Table 1). The proportion of S-LPS to R-LPS and SR-LPS in the unfractionated natural mixture is approximately 1 : 4 in terms of weight, as calculated from the quantities of recovered LPS fractions (R = band 1, SR = band 2 and S = smooth fraction). The calculated quantities of the S-LPS fraction would, according to the in-vitro inhibition experiments (Fig. 10), not be sufficient to exhibit strong inhibitory effects. Supporting this, the biologic activity observed for the isolated R-LPS and SR-LPS fractions (R-LPS fraction = band 1; and SR-LPS fraction = band 2) showed a very similar, but slightly more sensitive, dose–response compared with the natural LPS mixture under serum-free conditions (Fig. 9), indicating the presence of major amounts of R-LPS and SR-LPS in the mixture. Thus, the S-LPS preparation of E. coli O111 contains significant amounts of R-LPS and SR-LPS, which strongly dominate the biologic response.

In summary, this study demonstrates that human macrophages clearly differentiate between LPS chemotypes, with S-LPS having a strict requirement for LBP as a cofactor in serum for the activation of macrophages in vitro, and thereby enhances our knowledge from murine experimental systems to the human host. This finding may also have important implications for our understanding of the complex physiological mechanism orchestrating the inflammatory response in vivo. As shown here, immune recognition of the isolated S form of LPS by human macrophages was dependent on the presence of serum and can therefore be understood to be restricted in vivo to the activation of immune responses in the bloodstream. In contrast, R and SR forms of LPS represent the LPS structure identified by the mononuclear phagocytic system independently of serum and can therefore express activity under all physiological conditions – in the bloodstream as well under serum-free conditions in tissue and organs. Thus, our study strongly supports that R- and SR-forms of LPS are master stimuli of immune activation.

Materials and Methods

Bacterial LPS

The LPS of the deep-rough E. coli F515 mutant (Re chemotype) and wild-type E. coli O111 and O12 were isolated from cell pastes using the phenol/chloroform/light petroleum method [40]. Electrophoresis, staining and elution reagents were from Merck (Darmstadt, Germany), Bio-Rad (Richmond, CA, USA) or Sigma (St Louis, MO, USA). Distilled and deionized water (18.2 MΩ cm) was obtained using a Milli-Q water purification system (Millipore, Schwalbach, Germany) and used in all experiments.

Isolation of LPS species by DOC/PAGE

LPS species were isolated in four steps: preparative DOC/PAGE [41]; reverse staining; zinc chelation; and passive elution. The elution of LPS from gels was performed according to previously reported [8-10] or newly adopted procedures, as described below.

Preparative DOC/PAGE, reverse staining and zinc chelation

LPS was prepared for electrophoresis and separated using a glycine DOC/PAGE system, as described previously [41]. Up to 12 preparative (single-well) gels were accommodated in PROTEAN II xi cells (Bio-Rad) and run in parallel in a single separation experiment. The gel dimensions were 16 cm × 16 cm × 0.1 cm. The acrylamide concentrations of the stacking and separating gels were 4% and either 15% or 16%, respectively. The sample mixture was loaded into the gel wells and run at 15 mA per slab gel at 20–23 °C. After electrophoresis, the gel-resolved LPS patterns were detected by reverse staining with zinc-imidazole [42, 43]. Following detection, the LPS bands were separately excised and incubated (twice, for 10 min each incubation) under agitation in 100 mm EDTA, pH 7, to chelate zinc ions and then washed (three times, for 10 min each wash) thoroughly with water to remove the chelating solution.

Elution of LPS from gels

The single band of the LPS from E. coli F515 was isolated from polyacrylamide gels, as described previously [8-10]. Briefly, a 1-mL gel extruder was assembled by inserting two disks of 32-μm and 125-μm metal sieves (Carl Schroeter, Hamburg, Germany) into the bottom of a 1-mL polypropylene syringe. Gel slices were transferred to the syringe and extruded through the sieves by pressing the plunger. The resulting gel microparticles were collected in 50-mL tubes. To passively elute LPS molecules, 20 mL of 5% (v/v) TEA was added to the tubes and the gel slurries were incubated, under vortexing, for 10 min at 20–23 °C. The tubes were centrifuged for 2 min at 1200 g and the overlaying solution was collected. This elution step was repeated once. Samples containing the overlaying solution from these two elution steps were combined and lyophilized.

Newly adopted procedure for eluting LPS from polyacrylamide gels

The recovery of the gel-separated LPS bands of E. coli F515, E. coli O111 or E. coli O12 involved three main steps: gel extrusion, passive elution and desalting of LPS. The extrusion of LPS bands from a single polyacrylamide gel was carried out using a 1-mL gel extruder, as described above. When LPS bands had to be combined from several preparative gels run simultaneously, a 5-mL gel extruder was employed [43]. For the elution of LPS, three volumes of a 5% TEA solution in water were added to the extruded gel and the gel slurry was agitated for 5 min at 20–23 °C. Immediately afterwards, a volume of 10% acetic acid, equal to one-fifth of the volume of the 5% TEA solution added initially, was poured into the mixture whilst stirring. The pH was monitored and a few more drops of 10% acetic acid were added until a pH of 7 was reached. The gel slurry was further stirred in the neutral solution for 5 min at 20–23 °C. Subsequently, the mixtures were centrifuged for 5 min at 1200 g and the overlaying solution, which contains the eluted LPS, was collected using a pipette. To remove any remaining gel microparticles, LPS samples were filtered through syringe filters (5 μm pore size; Whatman GmbH, Dassel, Germany). LPS samples were desalted by transfer to dialysis bags (12–16 kDa cut-off), dialyzed at 4 °C against water and lyophilized. To remove any remaining salt, the samples were suspended in 20–100 μL of water and subjected to three consecutive precipitations with ethanol/acetone (9 : 1, v/v). The ratio of the volume of the sample to that of ethanol/acetone was 1 : 10. The final LPS precipitates were suspended in 0.2 mL of water, transferred to 1-mL glass vials and lyophilized. The yield of LPS was determined on a microbalance.

DOC/PAGE analysis of isolated LPS

LPS fractions were separated in a glycine DOC/PAGE system, as described above. Aliquots of 10–20 μL were loaded into gel wells and run at 15 mA per slab gel. Gel-separated LPS fractions were visualized by silver staining [44]. To avoid interference of DOC in the technique, the gels were incubated, before staining, in 30% acetonitrile (twice for 10 min) and washed with distilled water (three times for 10 min each wash) under shaking.

Mass spectrometry

ESI FT-ICR MS was performed in negative ion mode using an APEX Qe (Bruker Daltonics, Billerica, MA, USA), equipped with a 7 Tesla actively shielded magnet and a dual ESI/MALDI ion source. Samples at a concentration of ~ 10 ng·μL−1 were sprayed at a flow rate of 2 μL·min−1. The capillary entrance voltage was set to 3.8 kV, and the dry gas temperature was set to 200 °C. Under normal soft ionization conditions no significant fragmentation of LPS is generated. For unspecific fragmentation the collision voltage of the quadrupole interface was set from 5 to 30 V. Under these conditions the labile linkage between the lipid A and the core oligosaccharide is cleaved, as previously described [45]. If not otherwise stated the mass spectra were charge deconvoluted using the Bruker Data Analysis software. The mass numbers given refer to the monoisotopic masses of the neutral molecules. The mass scale was calibrated externally with rough LPS of known structure. The mass accuracy in the broad band modus was better than 5 p.p.m.

Mass spectrometric study of the time dependence of LPS deacylation in 5% TEA

Forty micrograms of the Re-LPS mixture from E. coli F515 was suspended in 1 mL of 5% TEA and incubated for 5, 10, 20, 30, 60 or 120 min at 20–23 °C under agitation. Then, the pH of the solution was brought to 7 by the addition of 10% acetic acid. After this neutralization step, the samples were kept at the same temperature, under agitation, to complete a total incubation time of 120 min. The samples were concentrated by freeze-drying, suspended in 20 μL of water and analyzed using ESI FT-ICR MS. As a control, one sample was suspended in water and analyzed directly using MS.

Monosaccharide composition analysis of the rough-type LPS fraction of E. coli O111

The sugar components of the rough-type LPS fraction from E. coli O111, isolated by preparative DOC/PAGE as described under newly adopted procedure for eluting LPS from polyacrylamide gels, were identified as peracetylated methyl glycosides by GC/MS. Briefly, 60 μg of LPS was subjected to methanolysis (0.5 m HCl in anhydrous methanol) at 85 °C for 20 min and subsequently peracetylated with pyridine and acetic anhydride (1 : 1, v/v) at 85 °C for 15 min. After evaporating the solvents, the sample was dissolved in chloroform and analyzed, by GC/MS, on an Agilent Technologies 5975 Gas Chromatograph with inert XL Mass Selective Detector, using a Hewlett-Packard fused silica capillary column (30 m × 0.25 mm I.D. × 0.25 μm) with helium as the carrier gas.

Stimulation of TLR2 receptor in HEK293 cells

The HEK293 cell line, stably expressing human TLR2 (HEK-TLR2), has been described previously [46]. Cells were cultivated in Dulbecco's modified Eagle's medium (DMEM; Biochrom AG, Berlin, Germany) containing 10% fetal bovine serum (Linaris, Bettingen am Main, Germany), 0.5 U·mL−1 of penicillin, 0.5 mg·mL−1 of streptomycin (Biochrom AG) and 0.4 mg·mL−1 of G418, and grown at 37 °C under an atmosphere of 5% CO2. For cell-stimulation experiments, cells were seeded at a density of 2.5 × 105 cells·mL−1 in 96-well plates in 200 μL of DMEM and stimulated with LPS preparations at the indicated concentrations for 24 h. Cell-free supernatants were collected and human IL-8 was determined using IL-8 cytoset from Biosource (Solingen, Germany).

Differentiation of human macrophages

MNCs were isolated from human whole blood drawn from healthy volunteers using procedures approved by the Ethics Committee of the University of Lübeck. Heparinized blood was fractionated using the Hypaque–Ficoll gradient method. MNCs were cultivated for 7 days at 37 °C, in a 5% CO2 atmosphere in Teflon bags in RPMI-1640 containing 100 U·mL−1 of penicillin, 100 μg·mL−1 of streptomycin, 200 mm l-glutamine, 4% heat-inactivated NHS and 2 ng·mL−1 of human macrophage colony-stimulating factor (R&D Systems, Wiesbaden, Germany) to differentiate monocytes to macrophages.

NHS and antibody-depleted serum

Serum was obtained from the blood of healthy donors of serogroup AB and heat inactivated by incubation for 30 min at 56 °C. The heat-inactivated serum was aliquoted and stored at −20 °C. In experiments, this serum is referred to as NHS.

Removal of O111-specific antibodies from NHS by affinity chromatography

LPS of E. coli O111 was deacylated as previously reported [17] and conjugated to EAH–Sepharose (30 mg of ligand per 2 g of EAH–Sepharose) by the glutardialdehyde method. NHS (5 mL) was centrifuged (11 000 g in a Minifuge) and circulated three times (0.25 mL·min−1) over a column with the affinity support. This serum is referred to as antibody-depleted human serum. Efficacy of the absorption was ascertained by western blotting using a monoclonal antibody against O111 LPS. The LPS of E. coli O111 (8 μg per lane) was separated by SDS/PAGE on a 13% gel, blotted onto nitrocellulose and stained with NHS diluted 1 : 10 or 1 : 50 or with NHS after removal of O111-specific antibodies at a dilution of 1 : 5 or 1 : 10. Alkaline phosphatase-conjugated goat anti-human IgG/A/M (diluted 1 : 4000; Dianova, Hamburg, Germany) was used as the second antibody and developed with 5–bromo-4-chloro-3-indoylphosphate and p-toluidine p-nitroblue tetrazolium chloride (Bio-Rad, München, Germany) as substrates, according to the supplier's instructions. After 15 min the reaction was stopped by the addition of distilled water.

Stimulation of human macrophages

Lyophilized LPS samples were suspended in pyrogen-free water (Aqua B. Braun; Braun, Melsungen, Germany) at 1 mg·mL−1 stock solution, by sonication for 30 min and temperature cycled twice between 4 and 56 °C. Samples were stored for at least 12 h at 4 °C before use in biologic experiments.

Macrophages were harvested at day 7 of differentiation, washed twice in RPMI and seeded at 1 × 105 cells per well in a 96-well dish in 200 μL of RPMI-1640 containing 100 U·mL−1 of penicillin, 100 μg·mL−1 of streptomycin, 200 mm l-glutamine and 4% heat-inactivated NHS or antibody-depleted serum. For stimulation, LPS was serially diluted in serum-free RPMI-1640 and added to the cells at the indicated final concentrations. In inhibition experiments, cells were pre-incubated with the S-LPS fraction for 10 min at 37 °C at the indicated concentrations and then stimulated with 1 ng·mL−1 of the R-LPS fraction. To investigate the role of LBP (kindly provided by XOMA Corp. Berkeley, CA, USA) and soluble CD14 (Biometec, Greifswald, Germany), recombinant human proteins were added to the macrophages 15 min before stimulation with LPS. After 4 h of stimulation, cell-free supernatants were collected and analyzed for TNF-α using an ELISA for human TNF-α (BD Biosciences, Heidelberg, Germany). Samples were analyzed in duplicate and the results were reported as mean and SD.

Acknowledgements

The authors thank Sabrina Groth, Brigitte Kunz, Gudrun Lehwark-Yvetot, Christine Schneider, Veronika Susott and Hermann Moll for technical assistance. This work was supported by the Deutsche Forschungsgemeinschaft DFG grant SCHR 621/3-1 to A.B. Schromm. E. Pupo was a recipient of an Alexander von Humboldt Research Fellowship.

Ancillary