Complement-mediated lipopolysaccharide release and outer membrane damage in Escherichia coli J5: requirement for C9

Authors


Dr A. P. Moran, Laboratory of Molecular Biochemistry, Department of Microbiology, National University of Ireland, Galway, Ireland. E-mail: anthony.moran@nuigalway.ie

Summary

Lipopolysaccharides (LPS) are major antigenic components of the outer membrane of Gram-negative bacteria and can stimulate activation of the complement system. Such activation leads to formation of the complement membrane attack complex (MAC) on the cell walls, LPS release and, in serum-sensitive strains, to cell death. In this study, Escherichia coli J5 strains, which incorporate exogenous galactose exclusively into LPS, were used to generate target strains with different LPS chemotypes, and the LPS of the strains was labelled with tritium (3H-LPS). The ability of normal human serum (NHS) and human complement-deficient sera to release LPS was subsequently monitored. NHS-induced release of 64–95·7% of 3H-LPS within 30 min; overall, no significant difference was observed between release of LPS from E. coli J5 strains with different LPS chemotypes. In functional assays, maximum LPS release had occurred by 30 min and before maximum bacterial killing. Electron microscopy revealed NHS-induced outer-membrane disruption in the form of blebs at 15 min; at this time-point the inner membrane remained intact. Background LPS release and no bactericidal activity were detected in heat-inactivated serum or human sera deficient in C6, C7 or C8. The C9-deficient (C9D) serum had low bactericidal activity and failed to induce LPS release; however, addition of purified human C9 reconstituted its ability to release LPS. This study demonstrated the need for functional C9 molecules for LPS-releasing activities in serum-sensitive E. coli J5 strains.

Introduction

Despite intensive research, Gram-negative sepsis remains a major cause of human morbidity and mortality.1,2 Lipopolysaccharides (LPS), also termed endotoxins, are major components of the outer membrane of Gram-negative bacteria and are a significant contributory factor in the pathogenesis of Gram-negative septicaemia,3 particularly meningococcal septicaemia.4,5 LPS structures can interact with complex host response systems,6,7 including cellular and humoral components of the immune system,8,9 and may induce the production of a diverse spectrum of immunoregulatory cytokines, which in turn mediate the pathophysiological responses characteristic of septic shock.10

LPS can stimulate activation of the complement system, both after binding by antibodies and through direct interaction with complement proteins.8,9 The activation of complement on bacterial cell surfaces leads to the formation of the membrane attack complex (MAC) by the terminal complement components. The MAC comprises complement components C5b through C8, and up to nine or more C9 molecules.11 The C9 molecules interact to form a cylinder that is inserted into a target membrane. The outcome of complement interaction on target cell membranes includes LPS release and cell death, although the latter depends largely on the properties of the target cells.

The role of C9 in functional complement activity has been debated. Some researchers reported C9-deficient (C9D) serum to be cytotoxic for serum sensitive, but not serum resistant, Escherichia coli.12,13 Nevertheless, others have shown C5b−9 formation and subsequent C9 polymerization to be necessary for efficient bacterial killing and inhibition of inner membrane activity.14,15 However, another group concluded that multimeric C9 within C5b−9 is not an absolute requirement for inner membrane damage and cell death.16 More recently, Wang et al.17 reported that the bactericidal activity of complement is dependent upon C9 as it is the only terminal complement protein that, when shocked into the cytoplasm, elicits any toxic effect. Hereditary deficiency of C9 is relatively common in Japan18 and, although associated with an increased risk of infection with Neisseria meningitidis, the risk is lower than that in Japanese C7-deficient (C7D) individuals19 and in C5-deficient (C5D), C6-deficient (C6D), C7D or C8-deficient (C8D) individuals elsewhere.20

Complement has been clearly implicated in the release of LPS from Gram-negative bacteria;21 moreover it has been reported that infusion of fresh-frozen plasma to a C6D patient suffering from meningococcal sepsis restored plasma endotoxin-releasing activity.22 An intact complement system has also been shown to be necessary for the effective clearance of LPS in mice.23 However, Tesh et al.21 reported that C9 was not required for LPS release from E. coli J5. The importance of target cell properties in the outcome of their interaction with complement led us to re-examine this question using different E. coli J5 strains with known differences in LPS structure.

Chemically, high-molecular-weight smooth-form LPS comprises lipid A, core oligosaccharide and an O-specific polysaccharide chain,24 whereas low-molecular-weight rough-form LPS lacks an O-specific chain. E. coli J5, a galactose epimerase-negative mutant derived from E. coli 0111-B4, can incorporate exogenous galactose exclusively into the LPS component of the bacterial outer membrane.25 In the absence of galactose, most E. coli J5 strains produce rough-form LPS, whereas in the presence of galactose a smooth-form LPS, apparently identical to that of the parent strain, is produced.26 Antigenic studies using monoclonal antibodies (mAbs) reactive with LPS of Rc (incomplete core) or R3 (complete core) core chemotypes, have revealed differences between the LPS of E. coli J5 strains.27

The involvement of the terminal complement proteins in LPS release and clearance can affect the outcome of Gram-negative infections and may be influenced both by the LPS structure and by the immune competence of the host. This study examined serum-induced LPS release, in particular the role of C9 in the release of LPS, and assessed whether antigenic differences in LPS structure affected complement-mediated LPS release from E. coli J5.

Materials and methods

Bacterial strains

E. coli J5(U), J5(UK) and J5(2877) were obtained from Dr B. J. Appelmelk (Vrije Universiteit, Amsterdam, the Netherlands). LPS antigenic differences between these strains have been determined serologically in a previous study27 and were reconfirmed in this work using immunoblotting techniques and mAbs reactive with LPS of a given chemotype. When grown in the presence of galactose, E. coli J5(U) strongly expressed the more truncated Rc core-type LPS, whereas E. coli J5(UK) predominantly expressed complete R3 core type; however, neither E. coli J5(U) nor E. coli J5(UK) expressed an O-specific chain. In the presence of galactose, E. coli J5(2877) strongly expressed both incomplete Rc and complete R3 core types and an O-specific polysaccharide chain, but in the absence of galactose the O-specific chain was not detected. Thus, E. coli J5(2877) behaves most like the original J5 strain described previously25 and consequently was the strain used predominantly in our investigations. Stock cultures of the E. coli J5 strains were maintained at −20° in 20% (v/v) glycerol in nutrient broth (Difco, Detriot, MI). The core region and phosphorylated oligosaccharides from the LPS of one of the E. coli J5 strains have been chemically and structurally analysed28,29 and, although the authors do not specify the strain, information on its source suggests that this work was performed using E. coli J5(U).27,28

Sera

Normal human serum (NHS) was obtained from healthy volunteers. Heat-inactivated serum (HIS) was prepared by incubation at 56° for 30 min. Sera from patients deficient in complement components C6 and C7 (C6D and C7D), were obtained from patients reported previously who were diagnosed following presentation with recurrent meningococcal infections.30,31 The C8D serum was obtained from a subject diagnosed during a study of complement deficiency in Ireland.32 The C9D serum was a gift from Dr M. J. Hobart (De Montfort University, Leicester, UK):33,34 it contains 1·2% of the antigenic C9 levels found in NHS, but it is not functionally active and the terminal complement complex (TCC) cannot be formed.34

Growth and radiolabelling conditions

The basic salts medium35 used to radiolabel the E. coli J5 with tritiated galactose (3H-galactose) was modified to include 0·005% (w/v) unlabelled galactose.36 A colony of one of the E. coli J5 strains was inoculated into 10 ml of basic salts medium and grown at 37° for 8 hr with gentle shaking. The bacteria were centrifuged at 1000 g for 10 min and resuspended in basic salts medium supplemented with 4 µCi/mmol of 3H-galactose (Amersham International, Amersham, Bucks., UK). The LPS was radiolabelled during growth of the bacteria (37° for 16 hr) to logarithmic phase. The bacteria were then washed twice in phosphate-buffered saline (PBS) to remove unincorporated galactose and resuspended to a volume of 10 ml in PBS. Total counts of 3H-galactose incorporated into the LPS were determined by liquid scintillation counting using Optiphase Hisafe III scintillation fluid (Fisher Chemicals, Loughborough Leics., UK) and an automatic quench compensation program on the 1600TR Liquid Scintillation Analyzer (Packard Instruments, Meriden, CT).

LPS-release assay

Two-millilitre aliquots of the 3H-galactose-labelled E. coli J5 suspended in PBS were centrifuged at 2000 g for 10 min and the supernatants discarded. The pellets were resuspended in either 160 µl of basic salts medium (controls)36 or 160 µl of basic salts medium containing 25% or 50% (v/v) serum and subsequently incubated for up to 3 hr at 37°. At 30-min intervals, samples were centrifuged, the supernatant collected and the radioactivity in the supernatant determined by liquid scintillation counting. The amount of 3H-galactose incorporated into the pellet of control tubes, which contained no serum, was determined by collecting the pellet on 0·2-µm membrane filters (Costar, Badhoevedorp, the Netherlands) and determining the radioactivity by liquid scintillation counting. The total 3H incorporated into the LPS (3H-LPS) of the E. coli J5 strain was defined as the combined radioactivity in the supernatant and in the pellet of control samples. The amount of serum-released 3H-LPS in the supernatant was expressed as a percentage of the total detectable 3H incorporated into the LPS of E. coli J5. To test for the formation of LPS aggregates by released LPS, 4% (v/v) triethylamine (Sigma Chemical Co., St. Louis, MO), which solubilizes and disrupts LPS aggregates,37 was added to culture tubes of E. coli J5(2877) at each time-point, prior to centrifugation. The resulting supernatant was collected and the radioactivity determined as described above.

C9-reconstitution assay

The purified C9 protein38 was provided by Prof. B. P. Morgan (Welsh National School of Medicine, Cardiff, Wales, UK) and retains good functional activity (B. P. Morgan, personal communication). The concentration of the purified C9 and serum C9 concentrations were determined using an enzyme-linked immunosorbent assay (ELISA).34,39 NHS contains ≈ 60 µg/ml of C9, and the C9 concentration of the C9D serum had been determined previously to be 1·2% of the C9 concentration of NHS.34 The C9D serum was reconstituted up to normal C9 concentrations using the purified C9 and used in LPS-release assays, as described above.

Serum bactericidal assay

Unlabelled E. coli J5 were prepared and centrifuged as described for the LPS-release assays. The pellets were resuspended in 200 µl of basic salts medium containing 50% (v/v) serum and incubated for 3 hr at 37°. At 0, 0·5 (NHS only), 1·0, 2·0 and 3·0 hr time-points, serial 10-fold dilutions in sterile PBS were plated onto tryptic soy agar (Difco) supplemented with 0·25% (w/v) glucose and 0·25% (w/v) galactose, and incubated for 16 hr at 37°. Bacterial viability was assessed by determining the number of colony-forming units (CFU).

Transmission electron microscopy

Unlabelled E. coli J5 suspended in basic salts medium (control) or basic salts medium containing 25% (v/v) serum, were prepared as described above for the LPS-release assays. Tubes were incubated at 37° for 15 and 30 min, centrifuged and the pellets fixed in 3% (v/v) glutaraldehyde (Sigma) for 1 hr then postfixed in 1% (w/v) osmium tetraoxide (Sigma) for 1 hr. The pellets were then dehydrated in a graded ethanol series. After addition of a 50 : 50 solution of ethanol in Spurr's resin (Agar Scientific Ltd, Essex, UK) for 1 hr, the pellets were resuspended in 100% Spurr's resin for 20 hr. Subsequently, the pellet was embedded in 100% Spurr's resin and polymerized for 3 days at 60°. Ultra-thin sections were cut on a Reichert-Jung Ultra Microtome (C. Reichert Optische Werke AG, Vienna, Austria) at 70–90 nm and mounted on copper grids. Sections were stained with uranyl acetate (Leica UK Ltd, Bucks., UK) for 30 min at 40° and then with lead citrate (Leica UK Ltd) for 5 min at room temperature, and subsequently viewed on a Hitachi H7000 Transmission Electron Microscope (Tokyo, Japan).

Statistical analysis

The overall effect of the different sera on LPS release from the E. coli J5 strains at each time-point was analysed by analysis of variance (anova). Where a treatment was found to be significant, differences between the strains were analysed by Scheffe's s-test and the Dunnett two-tailed test, both Post-Hoc tests.

Results

Incorporation of 3H-galactose into different E. coli J5 strains

E. coli J5(2877), which expressed an O-specific chain and both incomplete and complete core types, was the E. coli J5 strain most effective at incorporating 3H-galactose into its LPS. This reflects the availability of additional galactose sites on the complete core and O-specific polysaccharide chain. E. coli J5(UK), which reacted weakly with anti-Rc antibodies and strongly expressed a complete R3-type LPS, incorporated ≈ 55% of the 3H-galactose that was detected in E. coli J5(2877), whereas the most truncated Rc core LPS-type of E. coli J5(U) incorporated only 5·5%. These differences reflect the LPS antigenic variations between the strains and, consequently, the E. coli strains that incorporated the radiolabel most efficiently – J5(2877) and J5(UK) – were used in preference to J5(U). The expression of the amount of 3H-LPS released as a percentage of the total uptake of 3H-galactose into the LPS of each strain allowed a more direct comparison of the effect of serum on LPS release.

LPS release by NHS

The total radioactivity detected did not differ in either the presence or absence of serum and thus serum was considered not to quench the detection of 3H-galactose. Incubation of 3H-labelled E. coli J5(2877) or J5(UK) with NHS resulted in a significant, rapid release of LPS into the supernatant within 30 min (P < 0·01) (Fig. 1). At this time-point, substantial levels of 3H-LPS, ranging from 64 to 95·7% of the total radiolabel, were detected in the supernatants, indicating that both smooth-form LPS and complete R3 LPS types are very susceptible to NHS. Maximum release of 3H-LPS from E. coli J5(2877) and E. coli J5(UK) had occurred within 30 min, and after 1 hr the percentage of 3H-LPS released had apparently decreased. The differences between NHS-induced LPS release from E. coli J5(2877) and E. coli J5(UK) were not statistically significant (P > 0·05) at any time-point examined (0, 0·5, 1·0, 1·5, 2·0 and 3·0 hr). Although 3H-labelling of E. coli J5(U) LPS was less efficient, scintillation counts suggested similar levels of 3H-LPS release from the incomplete Rc-type LPS of E. coli J5(U) in the presence of NHS. No statistically significant difference was observed between LPS release from E. coli J5(2877) in the presence of 25% (v/v) or 50% (v/v) NHS (P > 0·05).

Figure 1.

Percentage of tritium-labelled lipopolysaccharide (3H-LPS) released into the supernatant from Escherichia coli J5(2877) and J5(UK) during 1 hr of incubation in 50% (v/v) normal human serum (NHS). Data points represent the means of four determinations and error bars indicate standard deviations. The release of 3H-LPS by NHS following 0·5 and 1 hr of incubation at 37° was statistically significant (P < 0·01).

As demonstrated in Fig. 1, and in further experiments with E. coli J5(2877) and E. coli J5(UK), there were unexpected reductions in the amounts of radioactivity detected in the supernatants after maximum levels of 3H-LPS were detected at 30 min. To test if these changes were caused by the formation of LPS aggregates by released LPS, 4% (v/v) triethylamine (Sigma) was added to culture tubes of E. coli J5(2877) and, in the presence of triethylamine, fluctuations in the amounts of radioactivity detected were greatly reduced (Fig. 2). At the 1·5 and 3·0 hr time-points there was significantly more released 3H-LPS detected in the presence of triethylamine than in its absence (P < 0·05), indicating that the serum-released LPS formed aggregates that precipitated and were not detectable in the supernatant.

Figure 2.

Percentage of tritium-labelled lipopolysaccharide (3H-LPS) released into the supernatant from Escherichia coli J5(2877) incubated with 50% (v/v) normal human serum (NHS) or with 50% (v/v) NHS supplemented with 4% (v/v) triethylamine, plotted against duration of incubation. Data points represent the means of four determinations, error bars show the standard deviations and the arrows indicate time-points where there was significantly more 3H-LPS detected in the supernatant in the presence of triethylamine than in its absence (P < 0·05).

LPS release by HIS and complement deficient sera

As shown in Fig. 3, incubation of E. coli J5(2877) with HIS, or with the C6D or the C9D sera, was associated with minimal LPS release, indicating that LPS release is dependent on an intact complement system. The same was true for the C7D and the C8D sera (data not shown).

Figure 3.

Release of tritium-labelled lipopolysaccharide (3H-LPS) from Escherichia coli J5(2877) by normal human serum (NHS) and human complement-deficient sera over a 1-hr incubation period. The figure shows the percentage of 3H-LPS detected in the supernatant during incubation of E. coli J5(2877) with 50% (v/v) NHS, heat-inactivated serum (HIS), C6-deficient (C6D) serum or C9-deficient (C9D) serum. LPS release by C7-deficient and C8-deficient sera (data not shown) was similar to LPS release by C6D serum. Data points represent the means of four determinations and error bars indicate standard deviations.

C9-reconstitution assay

The concentration of C9 in the purified C9 preparation was determined by ELISA to be 0·3 mg/ml. Reconstitution of the C9D serum with purified human C9 to 60 µg/ml of C9 (normal C9 concentration) had a significant effect on the LPS-releasing activity from E. coli J5(2877) (P < 0·0015) (Fig. 4). Incubation of E. coli J5(2877) with reconstituted C9D serum resulted in a 2·4-fold and 2·9-fold increase in the amounts of 3H-LPS detected in the supernatant after 0·5 and 1·0 hr at 37°, respectively.

Figure 4.

The percentage of tritium-labelled lipopolysaccharide (3H-LPS) detected in the supernatant during incubation of Escherichia coli J5(2877) with 50% (v/v) normal human serum (NHS), 50% (v/v) C9-deficient (C9D) serum or 50% (v/v) C9D serum reconstituted to 60 µg/ml of C9, is shown. Data points represent the means of four determinations and error bars indicate standard deviations.

Serum bactericidal assays

Incubation with NHS had a marked effect on the viability of all three E. coli J5 strains. In the presence of NHS the numbers of CFU of E. coli J5(2877) were reduced by 75% after 30 min and by > 99·9% after 1 hr at 37° (Fig. 5), and NHS had a similar effect on the viability of E. coli J5(U) and E. coli J5(UK) (data not shown). Parallel LPS release and bactericidal assays in the presence of NHS suggested that maximum LPS release preceded maximum cell death, as up to 95% release had occurred within 30 min (Fig. 1). After the same 30-min incubation period, 25% of the bacteria were still viable (Fig. 5). Incubation of E. coli J5(2877) with HIS (Fig. 5), C6D, C7D, or C8D sera (data not shown) had no bactericidal effect and the bacteria continued to grow. However, the C9D serum reduced the numbers of CFU by 73% after 1 hr and by 95·3% after 3 hr (Fig. 5).

Figure 5.

Bactericidal effects on Escherichia coli J5(2877) of normal human serum (NHS) at 0-, 0·5-, 1-, 2- and 3-hr time-points; and of heat-inactivated serum (HIS) and the C9-deficient (C9D) serum at 0-, 1-, 2- and 3-hr time-points. The figure depicts the percentage viability of the bacteria at each time-point. The results shown are the means of three separate experiments.

Transmission electron microscopy

Complement-mediated damage to the outer membrane of E. coli J5(2877) was observed by transmission electron microscopy following incubation with NHS. E. coli J5(2877) that were not exposed to NHS remained fully intact and no blebs were visible (Fig. 6a). Incubation with NHS for 15 min caused outer-membrane disruption and led to the formation of blebs or membrane-bound vesicles (Fig. 6c); however, the inner membrane remained intact (Fig. 6b). The diameters of the blebs ranged from 0·9 to 1·0 µm, whereas those of the whole cells ranged from 7·0 to 8·5 µm. After 30 min of incubation, NHS caused widespread bacteriolysis.

Figure 6.

Transmission electron micrographs. Complement-mediated outer membrane damage was directly observed following incubation of Escherichia coli J5(2877) with normal human serum (NHS). Thin sections of the pellet were examined by transmission electron microscopy. (a) Control cells were not exposed to NHS, the bacterial inner and outer membranes remained fully intact and no blebs were visible (magnification × 50 000). (b) Severe disruption of the outer membrane was observed following exposure of the bacteria to NHS for 15 min (indicated by →), but the inner membrane remained undisturbed (magnification × 40 000). (c) Blebs or membrane-bound vesicles (indicated by →) were also visible (magnification × 50 000). Bars, 100 nm.

Discussion

The results of the present study show that NHS caused a rapid dissociation of LPS from the outer membranes of the E. coli J5 strains and no major difference was observed between LPS release from the E. coli J5 strains expressing smooth-form LPS, or Rc or R3-type rough-form LPS, although the type of core expressed and the presence/absence of the O-specific chain did influence the uptake of 3H-galactose. The study also demonstrated a significant decrease in the viability of E. coli J5 (of > 99·9%) after 1 hr at 37° in the presence of 50% (v/v) NHS. Moreover, the results of LPS release and bactericidal assays performed in parallel suggested that the maximum LPS release had occurred before maximum bacterial killing was reached; thus some cells that had released LPS at 30 min were still viable at this time-point and further exposure to serum was required to achieve optimal killing. This suggests that LPS release was not inevitably followed by cell death. Visual examination of electron micrographs confirmed that at 15 min the bacterial inner membrane remained intact, whereas NHS-induced outer membrane disruption was clearly visible in the form of blebs. Considering that the inner membrane is composed of phospholipids, and that LPS forms the outer leaflet of the outer membrane,40 the blebs would contain LPS and therefore LPS release can precede cell death. The toxic effects exerted by LPS incorporated into blebs and by aggregated LPS may have important pathological effects in vivo.

After maximum LPS release at 30 min (Fig. 2), fluctuations in the amounts of radioactivity being released were observed, but the addition of triethylamine markedly reduced these fluctuations, suggesting that LPS aggregation occurred in the supernatants after LPS release. LPS aggregate formation in vivo can influence its stimulatory potency and may have an important effect on subsequent host responses.41,42 However, the exact relationship between the LPS aggregation state and its ability to induce cellular responses remains unclear.43

Repeated investigations of LPS release by NHS consistently showed a high initial release of 64–95·7% of 3H-LPS after 30 min. Tesh and co-workers21,44 found lower levels of LPS release during the logarithmic phase of growth and reported that maximal LPS release was 30% of the total radiolabel incorporated into the LPS molecules. They reported even less LPS release from bacteria selected for serum resistance and from bacteria in the stationary phase of growth. The differences detected in the levels of LPS released during the logarithmic phase of growth are probably attributable to technical factors and the definition of serum-released LPS. Results of the present study were expressed as a percentage of radioactivity detected in the supernatant following centrifugation at 2000 g, and were calculated by comparison to the total detectable radioactivity incorporated into the washed bacteria. Tesh and co-workers define serum-released LPS as the radiolabelled LPS fraction that remains soluble after 5 min of centrifugation at 9000 g.21,44 Nevertheless, in agreement with the data presented here, they found no difference in the percentage of 3H-LPS released during the logarithmic phase of growth from organisms expressing or not expressing an O-side chain. Properties of the strains used may affect experimental results; however, we have no evidence that the E. coli J5 strain used by Tesh and co-workers differed from E. coli J5(2877). The higher levels of NHS-induced LPS release described here do correlate with another report of > 80% release of 3H-LPS from E. coli J5 by serum from a C6-deficient patient with meningococcal septicaemia whose complement function had been restored by infusion of fresh-frozen plasma.22

The lack of significant LPS release or bactericidal activity presented here, in the presence of HIS or C6–C8-deficient sera, confirms the vital role of complement in the host defence against Gram-negative bacteria. The use of human C9D sera in this study is particularly informative. The C9D serum did have some bactericidal effect but induced no LPS release. Thus, even though formation of the incomplete C5b−8 MAC eventually leads to the death of serum-sensitive cells, incomplete MAC formation does not induce LPS release. The importance of C9 in complement-mediated LPS release was confirmed in the C9-reconstitution assay presented here, where the addition of purified human C9 to the C9D serum was found to significantly increase the LPS-releasing activity (Fig. 4). The reason(s) that the amount of LPS released by the reconstituted C9D serum remained less than that released by NHS, could be attributed to aggregation of the purified C9 or to the consumption of C9 by pre-existing C5b−8 complexes in the C9D serum, as has been shown for C7D sera.45 In addition, very low levels of C6 and C7 have been shown to be efficient in enabling TCC formation,39,46 but, considering that the C9D serum contains antigenic C9, it is unlikely that inhibitory antibodies are an explanation.

In addition to the debate, outlined in the introduction, on the role of C9 in bactericidal function, there is some controversy as to the role of C9 in LPS release. Tesh and co-workers reported that a C9D serum lacking bactericidal activity was able to release 3H-LPS in amounts that were quantitatively similar to amounts released by NHS.21 Results of the present study did not demonstrate LPS-releasing activity in the C9D serum. However, it is possible the discrepancies arose because Tesh et al. used a serum sample from a subject who had unrecognized C9 subtotal deficiency with a functional activity similar to that described previously for an Irish family.34 We report here that the addition of purified C9 to serum totally deficient in C9 activity was required for reconstitution of LPS-releasing activity, demonstrating that fully functional C9 molecules are required for the optimal release of LPS. Just as the exact mechanism of C9 action in bactericidal activity remains unclear, our study has shown the requirement for, but not the exact mechanism of, C9 in LPS release. Considering that bacterial killing is dependent upon perturbation of the inner membrane,47 and that LPS release involves the outer membrane and may precede cell death, we believe that these two mechanisms may be different.

The role of serum-released LPS in endotoxaemia is not yet fully understood, but circulating LPS plays a vital role in the outcome of many Gram-negative infections.1,3,9 However, the presence of LPS in the circulation does not necessarily mean that it has been released by the host immune system, and complement activation can be protective against endotoxic shock.23 Therefore, the precise role of complement proteins, in particular C9, in interactions with LPS during Gram-negative infections are an important issue requiring further research.

Acknowledgments

This work was supported by the Irish Health Research Board, the Meningitis Research Foundation (Bristol, UK) and by the BIOMED BMH4-CT96-1005 programme. We are grateful to Dr B. J. Appelmelk for providing the E. coli J5 strains, to Dr M. J. Hobart for the C9D serum and to Professor B. P. Morgan for the purified human C9. We also thank N. O'Donoghue for assistance with the transmission electron microscopy, Dr J. W. Patching for use of the radioactive laboratory and Dr A. Hynes for assistance with the statistical analysis.

Ancillary