The bacterial endotoxins test (BET) is a sensitive assay for measuring endotoxin levels in solution and uses the limulus amebocyte lysate (LAL) coagulation reaction. We sought to identify the mechanisms through which certain substances interfere with the interaction between LAL and bacterial lipopolysaccharide (LPS).
Methods and Results
Endotoxin lipid A was inactivated by the addition of iron sulfate, which acted on endotoxin directly and strongly inhibited LAL coagulation activity. Size-exclusion, anion-exchange and reverse-phase liquid chromatography/mass spectrometry were used to examine changes in inactivated lipid A in terms of complex formation, negative charge status, hydrophobic interaction and structure. Furthermore, we verified the involvement of the lipid A phosphoryl group in the interference of iron sulfate with lipid A-factor C binding activity. Iron sulfate–inactivated lipid A was cleaved at its glycosidic bond, resulting in loss of hydrophobic interactions and disruption of lipid A complexes without alteration of negative charge status and lipid A-factor C interaction.
Lipid A cleavage was a direct result of interfering factors, including iron sulfate, which acted on endotoxin directly to disrupt lipid A complexes rather than interfering with LAL coagulation.
Significance and Impact of the Study
Our data provide new insights into the mechanisms of lipid A activity.
Lipopolysaccharide (LPS) is a bacterial endotoxin and a component of the bacterial cell wall (Nowotny 1969; Elin and Wolff 1976) and induce potent pathophysiological effects in mammals. LPS is an amphipathic molecule consisting of a polysaccharide chain and a fatty acid-containing lipid A component (Shands et al. 1967; Westphal 2009), which represents the active centre responsible for the endotoxic properties of LPS, such as lethality, fever, induction of cytokine production and coagulation of horseshoe crab haemolymph (Lüderitz et al. 1982). LPS bioactivities are thought to be greatly affected by the particle size or state of the LPS complex in aqueous media (Ribi et al. 1962, 1966; Roberson and Cromartie 1962; Milner et al. 1963; McIntire et al. 1967; Tarmina et al. 1968; Komuro et al. 1987). The lipid A component of LPS consists of two conjugated glucosamine residues attached to a variable number of long chain fatty acids and 1 phosphoryl group on each carbohydrate (Rietschel et al. 1984). This distinctive structure is necessary for the bioactivity of LPS, and omission of 1 phosphoryl group significantly reduces this bioactivity, with the monosaccharide structure being the least active (Rietschel et al. 1994).
The limulus amebocyte lysate (LAL) test was introduced by Levin and Bang (1968) and is based on an endotoxin-induced coagulation reaction in amebocyte lysates (Levin and Bang 1964). The bacterial endotoxins test (BET) is a simple, highly sensitive assay for determining endotoxin levels and is listed in the American, European and Japanese Pharmacopoeia for use in detecting endotoxin contamination in parenteral drugs (US Pharmacopeia 2012; European pharmacopoeia 2010; The Japanese Pharmacopoeia 2011). An endotoxin-induced coagulation reaction is triggered by binding of endotoxins to factor C present in LAL, and both the ionic bond contributed by the negative charge of the phosphoryl group or hydrophobic interactions with the fatty acid of lipid A are important for this coagulation.
Some substances interfere with the BET, preventing accurate determination of endotoxin levels present in a drug sample (Twohy et al. 1984; Ogawa et al. 1993). However, few studies have investigated the mechanisms through which such substances interfere with the BET due to the large size and microheterogeneity of LPS, especially in length and composition of its terminal glycan chains. We previously reported that some substances, including iron sulfate, strongly affect the LAL reaction and act on endotoxins directly, altering the state of endotoxin complexes in solution; such interference could be minimized by preparing sample solutions in saline, D-PBS, BES, or Tris buffers instead of water (Fujita et al. 2011).
We now report a qualitative and structural analysis of lipid A after inactivation by the addition of iron sulfate. We sought to elucidate the mechanisms through which iron sulfate interferes with lipid A function because this substance strongly and directly acts on endotoxins (Fujita et al. 2011). To this end, we analysed the negative charge, hydrophobic interactions, complex formation and structures of inactivated lipid A by size-exclusion high-performance liquid chromatography (HPLC), anion-exchange HPLC and reverse-phase liquid chromatography/mass spectrometry (LC/MS), respectively. Moreover, to investigate the involvement of the phosphoryl group of lipid A on the interfering action of iron sulfate, we compared the sensitivity of lipid A to iron sulfate with that of monophosphoryl lipid A (MPLA) prepared by acid hydrolysis of lipid A. Finally, we used an enzyme-linked immunosorbent assay (ELISA)-based binding assay to verify the iron sulfate–dependent binding activity of lipid A with factor C, a protease present in LAL that participates in the coagulation reaction and interacts with lipid A (Tan et al. 2000).
Materials and methods
Synthetic lipid A was purchased from the Peptide Institute Tokyo (Imoto et al. 1985, 1987). Purified lipid A from Escherichia coli Serotype R515 and MPLA from E. coli Serotype R515 were purchased from Enzo Life Sciences (Plymouth Meeting, PA, USA). Limulus ES-2 Test Wako was obtained from Wako Pure Chemical Industries (Osaka, Japan). Crude LAL was obtained from Charles River Laboratories International (Wilmington, MA, USA), and factor C was purified from crude LAL as previously described (Nakamura et al. 1985, 1986). Dextran sulfate gel was purchased from EY Laboratories (San Mateo, CA, USA). HiPrep 16/60 Sephacryl S-200 HR and HiTrap SP XL were purchased from GE Healthcare (Tokyo, Japan). Rabbit polyclonal antiserum was raised against factor C.
LAL coagulation activity was measured using the kinetic turbidimetric method; the Limulus ES-2 Test Wako (Kambayashi et al. 1991) was selected for this assay. Lipid A and MPLA sample solutions used for HPLC and LC/MS were diluted to 10 pg ml−1. The sample solutions were added to LAL, and mixture turbidities were measured in duplicate at 37 ± 1°C using a Toxinometer ET-2000 (Wako Pure Chemical Industries) (Oishi et al. 1988). The recovery of endotoxin added to sample solutions was calculated using a standard curve, which was made using endotoxin standard solutions at 20, 10 and 5 pg ml−1. As negative controls, the LAL activities of water, iron sulfate, ciprofloxacin and epirubicin solutions without lipid A or MPLA were measured.
Size-exclusion HPLC analysis
The size of the lipid A complex in solution was analysed with an HPLC system composed of Alliance 2695 and Alliance 2678 systems (Waters, Milford, MA, USA). A 5-μm 7·5 mm × 300 mm COSMOSIL Diol-300-II gel filtration column (Nacalai Tesque, Tokyo, Japan) was used as the analytical column. The mobile phase was water, and the UV detection wavelength was 205 nm. Synthetic lipid A, supplied in lyophilized form, was used and prepared at 0·1 mg ml−1. The flow rate was 0·5 ml min−1 for all samples, and the injection volume was 30 μl. Chromatographic data were recorded and processed using Millennium 32 software ver. 3.21 (Waters).
Anion-exchange HPLC analysis
The negative charge of purified lipid A was analysed on a TSKgel DEAE-5PW column (7·5 mm × 75 mm, 10 μm; TOSOH Corporation, Tokyo, Japan) using a flow rate of 0·7 ml min−1, an injection volume of 30 μl, and a UV detection wavelength of 246 nm. After running a 5 mmol l−1 ammonium acetate solution for 5 min, the column contents were eluted with a linear gradient of 5–70 mmol l−1 ammonium acetate in water over a period of 95 min. The column was re-equilibrated with 5 mmol l−1 ammonium acetate for 20 min prior to subsequent injections.
A Nexcera system (Shimadzu, Kyoto, Japan) was used for reverse phase. Lipid A was separated on a Waters X-bridge C8 column (3 μm, 3 mm × 150 mm i.d.; Waters). Mobile phase A was methanol/water/70% ethylamine (50/50/0·13), and mobile phase B was 2-propanol/water/70% ethylamine (93/7/0·13). The initial solvent, consisting of 90%/10% A/B, was maintained for 2 min, followed by a linear gradient to a final composition of 100% B after 20 min. A 10-min re-equilibration of the column with 90%/10% A/B was performed prior to subsequent injections. Twenty microlitres of synthetic lipid A solution was added to 50 μl of methanol/70%ethylamine (100/0·065), and the solution was neutralized with 30 μl of 1·25 mol l−1 ammonium bicarbonate aqueous solution supplemented with 20 mmol l−1 phosphoric acid. The sample solution was at a flow rate of 0·4 ml min−1 and an injection volume of 10 μl. Solvents were HPLC-grade and purchased from Wako Pure Chemical Industries.
The LC system described previously was coupled online to an LTQ Orbitrap XL mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA). The mass spectrometer was run in the negative-ion mode with the following instrument parameters: spray voltage of 2·5 kV, capillary temperature of 380°C, capillary voltage of −20 V and tube lens voltage of −50 V. Further structural analysis of lipid A was performed by MS/MS in an HCD mode.
ELISA-based lipid A binding assay
Clear polystyrene microplates (R&D Systems, Minneapolis, MN, USA) were first coated with 100 μl well−1 of 10 μmol l−1 lipid A or lipid A containing various concentrations of iron sulfate diluted in water. The plate was sealed and incubated overnight at 4°C. The wells were then washed five times with 200 μl wash solution (PBS containing 0·03% Tween-20). Blocking of unoccupied sites was achieved using PBS containing 1% BSA for 2 h at room temperature. Subsequently, the blocking solution was removed, and the wells were washed as described earlier. Varying concentrations of factor C were allowed to interact with bound lipid A at room temperature for 2 h. Bound factor C was detected by sequential incubation with antifactor C rabbit polyclonal antiserum (1 : 5000 dilution) and goat anti-rabbit antibody conjugated with horseradish peroxidase (HRP; 1 : 1000 dilution; Rockland Immunochemicals, Gilbertsville, PA, USA). Antibody incubations were carried out for 1 h at room temperature with washing between incubations as described previously. In the final step, 100 μl TMB peroxide substrate (R&D Systems) was added and reacted for 20 min before the addition of 50 μl of 2 mol l−1 sulfuric acid. The absorbance of the sample was determined at 450 nm with a reference wavelength of 540 nm using a microtitre plate reader. The values were correlated with the amount of lipid A bound and the amount of factor C present.
Lipid A complexes were disrupted by iron sulfate
To investigate the formation of lipid A complexes in solution, lipid A was analysed by size-exclusion HPLC, and LAL gelation activity was measured using the kinetic turbidimetric method to determine the LAL coagulation times (gelation times) when transmittance reached 94·9%. LAL coagulation times and gelation times both determine the polymerization capacity of the endotoxin. Lipid A suspensions of 5, 10 and 20 pg ml−1 produced gelation times of 115·3, 98·4 and 64·3 min, respectively (Table 1). However, lipid A solutions sonicated for 60 min had gelation times of 24·6, 19·9 and 16·3 min for 5-, 10- and 20-pg ml−1 suspensions, respectively (Table 1). The results of size-exclusion HPLC analysis of sonicated or nonsonicated lipid A dissolved at 0·1 mg ml−1 are shown in Fig. 1. Although the nonsonicated lipid A suspension eluted at 12 min as a single peak (peak B), the sonicated lipid A suspension produced a second peak at 8 min (peak A) in addition to peak B. We confirmed that the gelation time for 10 pg ml−1 purified lipid A was 20·8 min, and peaks were observed at 8 and 12 min, which was the same as for the sonicated synthetic lipid A suspension (data not shown). These results suggested that sonicated synthetic lipid A and purified lipid A produced the same LAL coagulation activity and complex sizes.
Table 1. Limulus amebocyte lysate coagulation activity of sonicated lipid A
Lipid A (pg ml−1)
Gelation time (min)
Next, to investigate a correlation between complexation and LAL coagulation activity of lipid A, sonicated synthetic lipid A solutions (0·1 mg ml−1) containing 0·125, 0·25, or 0·5 mmol l−1 of iron sulfate were also analysed by size-exclusion HPLC. We found that iron sulfate inhibited the coagulation activity of lipid A and increased gelation time in a concentration-dependent manner. Peak A corresponding to lipid A disappeared, coinciding with a reduction in LAL coagulation activity, and other peaks were observed in the presence of 0·25 and 0·5 mmol l−1 iron sulfate (Fig. 2). This finding demonstrated that lipid A complexation was important for LAL coagulation activity and could be inhibited by iron sulfate.
The negative charge of iron sulfate–inactivated lipid A was maintained
The LAL coagulation activity of lipid A has been reported to be reduced by dephosphorylation (Rietschel et al. 1994). To investigate whether dephosphorylation of lipid A was induced by iron sulfate, the negative charge status of lipid A was analysed by anion-exchange HPLC using a TSK gel DEAE-5PW column. Lipid A was eluted by a linear gradient of 5–70 mmol l−1 ammonium acetate and was resolved into two major peaks at 44 and 54 min and three minor peaks at 72, 75 and 92 min (Fig. 3). In this analysis, lipid A was resuspended in water, and the presence of ammonium acetate in the mobile phase caused the formation of lipid A complexes. Thus, the appearance of five peaks suggested that the lipid A solution contained complexes with differential negative charge states (Fig. 3).
Compared with lipid A alone, lipid A containing 0·5 mmol l−1 iron sulfate produced the same five peaks and the same retention times and peak areas as those seen in the absence of iron sulfate (Fig. 3). Despite the inhibition of LAL coagulation activity by iron sulfate, the retention of lipid A on the DEAE column was maintained, suggesting that the negative charge of lipid A was not affected by iron sulfate.
Lipid A was cleaved at the glycosidic bond by interfering factors
The negative charge of lipid A was not altered by iron sulfate; thus, to examine the effects of iron sulfate on the other components of lipid A, that is, glucosamine and fatty acids, we performed reverse-phase LC/MS. Lipid A was treated with 0·5 mmol l−1 iron sulfate and separated using a C8-bonded silica column under conditions in which the pH was >11. Intact lipid A was observed at m/z 1796·21 (Intact). When lipid A was treated with 0·5 mmol l−1 iron sulfate, the intact ion peak was decreased, and peaks with m/z 243·20 (Fragment 244), 453·40 (Fragment 454), 710·43 (Fragment 711), and 1102·79 (Fragment 1103) were increased; in particular, peak areas with m/z 710·43 and 1102·79 were increased to about 100-fold that of normal lipid A (Fig. 4). These peaks (m/z 710·43 and 1102·79) were identified as degradation products, where lipid A was cleaved at interglucosamine residues, while peaks with m/z 243·20 and 453·40 corresponded to fatty acids cleaved at the ester linkage between the lipid A glucosamine and fatty acid (Fig. 5). These results were confirmed by MS/MS (data not shown). Furthermore, we examined the cleavage of lipid A by ciprofloxacin and epirubicin as other interfering factors. Although the cleavage efficiency was low as compared to that of iron sulfate, peak areas with m/z 710·43 and 1102·79 were increased several fold and the LAL coagulation activity of lipid A was decreased following the addition of ciprofloxacin or epirubicin (Fig. 4). These results indicated that lipid A was cleaved at the glycosidic bond not only by iron sulfate but also by ciprofloxacin and epirubicin.
Inhibition of LAL coagulation activity and cleavage at the glycosidic bond of lipid A were attenuated by dephosphorylation
Next, we investigated the participation of the phosphoryl group of lipid A in the cleavage of the glycosidic bond by iron sulfate. To determine whether the LAL coagulation activity of MPLA, a dephosphorylated form of lipid A, was affected by iron sulfate, we performed an LAL test and determined the gelation time of each reaction. The gelation times of 5, 10 and 20 pg ml−1 MPLA were 42·6, 32·3 and 25·9 min, respectively, and the activity ratio of MPLA relative to lipid A was 23·1% (Table 2), which was lower than that previously reported (Takayama et al. 1984; Johnson et al. 1987). Although the endotoxin recovery rate of lipid A containing 0·5 mmol l−1 iron sulfate was only 0·12% that of lipid A in the absence of iron sulfate, the recovery rate for MPLA was 39% (Fig. 4b). Furthermore, peaks with m/z 1130·82 (Fragment 1131) and 630·45 (Fragment 631) corresponding to MPLA fragments cleaved at the glycosidic bond by iron sulfate were barely increased (Fig. 4a). This finding suggested that the sensitivity of MPLA to iron sulfate was lower than that of lipid A and that the phosphoryl group participated in the interfering action of iron sulfate.
Table 2. Limulus amebocyte lysate coagulation activity of MPLA
MPLA (pg ml−1)
Gelation time (min)
Activity ratio (lipid A = 100%) (%)
MPLA, monophosphoryl lipid A.
Lipid A containing iron sulfate retained factor C binding activity
Factor C is a LPS-sensitive, intracellular serine protease zymogen that consists of 1011 amino acid residues and has a calculated molecular mass of about 123 kDa (Tokunaga et al. 1991). The ability of lipid A to bind factor C in the presence of iron sulfate was investigated by an ELISA-based lipid A binding assay (Fig. 6). We purified factor C from crude LAL and verified a major band at 123 kDa by SDS-PAGE (Fig. 6a). Next, we detected factor C binding to lipid A immobilized on plates using antifactor C rabbit polyclonal antiserum. Although the LAL coagulation activity of lipid A containing 0·125, 0·25 or 0·5 mmol l−1 iron sulfate declined significantly, there was no reduction in the absorbance derived from factor C binding to lipid A, indicating that lipid A containing iron sulfate retained factor C binding activity and was not affected by the interfering action of iron sulfate.
In the present study, we sought to determine the mechanism through which iron sulfate interferes with lipid A activity and to develop BET methods to suppress this interference. Iron sulfate acts strongly and directly on endotoxins and is contained in various items and environments, such as injection needles; thus, this substance is one of the most influential interfering factors for the detection of endotoxin. We clearly demonstrated that iron sulfate-induced cleavage of lipid A at the glycosidic bond and hydrophilic/hydrophobic imbalance, which in turn reduced LAL coagulation activity. Our study also suggested that, because access of interfering factors to the lipid A phosphoryl group and the resulting cleavage of lipid A was a key event in interference, prevention of these actions would be highly effective in suppressing this interference event.
Principle factors interfering with the BET include suboptimal pH conditions, aggregation or adsorption of endotoxins, unsuitable cation concentrations, enzyme or protein modification, β-glucan and samples containing endotoxin (Cooper 1990; Duner 1993; Cooper et al. 1997; Williams 2007). Furthermore, we reported that some substances, such as metal ions, antibiotics, and anticancer drugs, directly affect endotoxins and produce a strong interfering action (Fujita et al. 2011). Despite these previous reports concerning interference with the LAL coagulation activity of LPS, the detailed mechanisms of these interference events remained unclear.
In this study, direct evidence for the interfering action of iron sulfate on lipid A was provided by the following experimental observations: (i) reverse-phase LC/MS showed cleavage at the glycosidic bond of lipid A containing iron sulfate, and such cleaved fragments were eluted faster than intact lipid A due to lack the capacity to participate in hydrophobic interactions (data not shown); and (ii) an increase in the peak area of corresponding lipid A complexes was observed at the same time as the onset of LAL coagulation, and this peak could be disrupted by iron sulfate, suggesting that the lipid A structure, its hydrophobic interactions, and formation of complexes were important factors for lipid A activity. In addition, lipid A containing 0·5 mmol l−1 iron sulfate, which has no coagulation activity, accounted for about 32% of intact lipid A in reverse-phase LC/MS, and this finding suggested that the activity of intact lipid A was suppressed by the fragments of lipid A cleaved at the glycosidic bond.
In contrast, the negative charge of lipid A was not affected by the presence of iron sulfate. The phosphoryl group and the negative charge it carries are important for lipid A bioactivity. Indeed, MPLA, which lacks this phosphoryl group, had weaker activity than lipid A. In LAL tests with lipid A and MPLA, lipid A had higher sensitivity to iron sulfate than did MPLA, supporting the role of the phosphoryl group in iron sulfate sensitivity. However, anion-exchange HPLC analysis indicated that the negative charge of lipid A in the presence of iron sulfate was unaltered in spite of the disappearance of peak A corresponding lipid A in size-exclusion HPLC; these findings indicated that the negative charge was sustained in lipid A containing iron sulfate. Therefore, the observed disruption of lipid A complex and interference with LAL coagulation activity were likely not due to changes in the lipid A negative charge status. Furthermore, reverse-phase LC/MS demonstrated that MPLA was infrequently cleaved at the glycosidic bond by iron sulfate. From these results, we hypothesize that the lipid A phosphoryl group has an important role in LAL coagulation activity and is involved in the interference event. For example, the Fe2+ ion may be attracted to lipid A due to its negative charge, and the glycosidic bond of lipid A will then be cleaved by the attracted Fe2+ ion; thus, iron sulfate would not be required to act on the phosphoryl group directly such as dephosphorylation. In addition, the interfering action of iron sulfate in the BET was minimized by using buffers (Fujita et al. 2011), and the attraction of iron sulfate to lipid A could likely be similarly inhibited by the high ionic strength supplied by buffers.
Lipid A-factor C binding activity was previously analysed by ELISA and surface plasmon resonance (Tan et al. 2000; Ariki et al. 2004), and these studies demonstrated that this binding event was necessary to trigger the gelation reaction. In the ELISA-based binding assay performed here, absorbance corresponding to factor C binding to lipid A was observed and was consistent with previous studies. However, iron sulfate–inactivated lipid A also had a similar degree of factor C binding. These results suggested that, in the presence of iron sulfate, lipid A retained its factor C binding activity, despite the inactivation of LAL coagulation, implying that lipid A-factor C binding alone was insufficient to trigger the LAL coagulation reaction. Moreover, the formation of the lipid A complex is thought to be necessary for some additional actions on factor C.
The most important factor for LAL coagulation activity is likely the formation of lipid A complexes resulting from its amphipathic characteristics, because disruption of lipid A complexes coincided with the reduction in LAL coagulation activity. In the results of size-exclusion HPLC, two peaks were observed for sonicated lipid A; however, only peak A was produced by the sonication of lipid A and disappeared by the addition of iron sulfate in a concentration-dependent manner. In addition, such production and disappearance of peak A coincided with a change in the LAL coagulation activity of lipid A. Therefore, we speculate that it is important for LAL coagulation activity of lipid A to form a particular size of complex corresponding with peak A. LPS forms a micelle complex in solution because of these amphipathic properties, and the amphipathicity of lipid A allows it to form micelle-like complexes in solution that are maintained by a balance between hydrophilic moieties, including the glucosamine residue, phosphoryl group, and hydrophobic subunits (i.e., fatty acids). In addition, a previous study reported that lipid IVA, a bioactive precursor of lipid A, also forms micelles or vesicles (Hofer et al. 1991). In this study, cleavage at the glycosidic bond was observed for inactivated lipid A, and the disappearance of the relevant peak in size-exclusion HPLC suggested that the amphipathic complex was disrupted by a hydrophilic/hydrophobic imbalance induced by cleavage at this glycosidic bond. Therefore, we speculate that cleavage at this glycosidic bond may be the result of direct action of iron sulfate on lipid A, generating the resulting lipid A lacking the ability to form hydrophobic interactions and thereby disrupting complexes (Fig. 7).
In conclusion, the results of the present study indicate that inactivation of the LAL coagulation reaction was induced by the disruption of lipid A complexes resulting from cleavage at the glycosidic bond and subsequent hydrophilic/hydrophobic imbalance. We used ciprofloxacin and epirubicin as other interfering factors in reverse-phase LC/MS, and cleavage at the glycosidic bond was also observed, similar to that found with iron sulfate. However, despite strong suppression of LAL coagulation activity by the addition of ciprofloxacin and epirubicin, the peak representing lipid A fragments cleaved at the glycosidic bond were only increased several fold over that of normal lipid A. Cleavage at the glycosidic bond may be an important event for interference, but ciprofloxacin and epirubicin may exert other effects on lipid A. Moreover, other substances that also act directly on lipid A, such as aluminium, copper, quinolone antibacterial agents and anticancer agents (Fujita et al. 2011), would be predicted to cleave lipid A in a manner similar to that of iron sulfate or ciprofloxacin and epirubicin, because such substances produce an irreversible interfering effect on lipid A that cannot be decreased through filtration or rinsing of LPS-substance mixed solutions. Thus, our data provide important insights into the mechanisms of lipid A complex formation and activity and have implications in the optimization and further development of the BET.
The authors thank Professor K. Yagi (University of Osaka) for his helpful advice during the preparation of this manuscript and Dr. M. Tsuchiya (Charles River Laboratories International) for providing crude LAL.