The relationship between age and production of tumour necrosis factor-α in healthy volunteers and patients with chronic heart failure

International Journal of Cardiology 90 (2003) 197–204 www.elsevier.com / locate / ijcard

The relationship between age and production of tumour necrosis factor-a in healthy volunteers and patients with chronic heart failure a,c , a,d a a Stephan von Haehling *, Sabine Genth-Zotz , Rakesh Sharma , Aidan P. Bolger , a b a a,c Wolfram Doehner , Peter J. Barnes , Andrew J.S. Coats , Stefan D. Anker b

a Department of Clinical Cardiology, National Heart and Lung Institute, London SW3 6 LY, UK Department of Thoracic Medicine, National Heart and Lung Institute, Imperial College, London, UK c Franz-Volhard-Klinik ( Charite´ , Campus Berlin-Buch), Berlin, Germany d Department of Medicine II, Johannes Gutenberg-University, Mainz, Germany

Received 8 October 2002; accepted 25 October 2002

Abstract Background: Ageing is associated with an altered immune response. Elevated plasma levels of tumour necrosis factor-a (TNF-a) are present in patients with advanced chronic heart failure (CHF). However, the relationship between age and the immune response in CHF is unknown. Methods: We investigated the relationship between age and the TNF-a generating capacity of lipopolysaccharide (LPS) stimulated peripheral blood mononuclear cells (PBMC) in nine healthy control subjects (mean age 51.663.6 years, age range 39–75 years) and 22 stable patients with CHF (mean age 68.361.5 years, age range 52–78 years, NYHA class 3.060.2). We also tested the TNF-a generating capacity of all control subjects and 18 CHF patients in whole blood cultures. Results: Subjects were subgrouped according to baseline TNF-a secretion in PBMC cultures into low- and high-responders, with the latter producing TNF-a even without LPS stimulation. High-responders produced more TNF-a than low-responders at all LPS doses (0.001–10 ng / ml, P,0.0001, repeated measures ANOVA), and high-responders were significantly older than low-responders (controls: 65.869.2 vs. 47.562.5 years; patients: 71.961.9 vs. 65.961.9 years, both P,0.05). Age correlated with TNF-a production in both patients and controls. This effect was independent of NYHA class. Conclusions: LPS-responsiveness appears to relate to age in both healthy controls and CHF patients. When assessing the immune status of CHF patients, age should therefore be considered an important confounding factor. In whole blood these findings could only be confirmed at the highest LPS concentration used, thus suggesting that certain factors in the blood may be able to abolish LPS activity at lower concentrations.  2003 Published by Elsevier Ireland Ltd. Keywords: Ageing; Endotoxin; Heart failure; Immune system; Tumour necrosis factor-a

Introduction Ageing is known to be associated with an altered immune response with the elderly being more susceptible to infections than younger adults [1]. The *Corresponding author. Tel.: 144-20-7351-8718; fax: 144-20-73518733. E-mail address: [email protected] (S. von Haehling).

reasons for this and the precise relationship between ageing and cellular immune responses are poorly understood. Several studies have sought to investigate the influence of age on the production of tumour necrosis factor-a (TNF-a) in ex vivo models using one of the key stimulators for its secretion: bacterial lipopolysaccharide (LPS, endotoxin). Most studies to date using cell culture techniques with either diluted whole blood [2] or peripheral blood mononuclear

0167-5273 / 03 / $ – see front matter  2003 Published by Elsevier Ireland Ltd. doi:10.1016 / S0167-5273(02)00566-1

198

S. von Haehling et al. / International Journal of Cardiology 90 (2003) 197–204

cells (PBMC) [3–5] have applied only one dose of LPS [2,5], which was typically far higher than the plasma levels found in vivo [2,3,5]. Furthermore, the results from these studies are somewhat conflicting: Riancho et al. failed to demonstrate age-related differences in the secretion of TNF-a in PBMC [5], whereas Fagiolo and co-workers described significant increases in TNF-a production in a similar model [4]. On the other hand, Bruunsgard et al. reported a decreased production of TNF-a in elderly control subjects compared to young adults after LPS-stimulation in a whole blood model [2]. Inflammatory cytokine activation has been documented in patients with advanced chronic heart failure (CHF), and this is thought possibly to play an important role in the pathogenesis of this condition [6,7]. Recent reports have demonstrated that raised levels of TNF-a correlate with prognosis in CHF [8,9]. Although the aetiology of cytokine release in CHF patients is not known, LPS has been proposed as being a potential trigger [10]. In CHF, TNF-a levels relate to age in male patients and postmenopausal female patients [8], however, the influence of age on cellular immune responses in CHF is not known. Thus, we investigated the relationship between age and the production of cytokines by PBMC and whole blood in response to LPS both in healthy subjects and in CHF patients. This study also examined a range of LPS concentrations which are more likely to be physiologically relevant. Finally, in order to detect a possible relationship between plasma LPS levels and the responsiveness of monocytes to this substance, we measured the plasma LPS in both healthy subjects and CHF patients.

Materials and methods Study population We investigated the TNF-a generating capacity of PBMCs from nine healthy control subjects (age 51.663.6 years, mean6S.E.M., range 39–75 years, six female) and 22 patients with stable CHF (68.361.5 years, range 52–78 years, NYHA class 3.060.2, three female). Additionally, whole blood from all control subjects and 18 CHF patients (age 68.161.6 years, range 52–78 years, NYHA class 3.260.2, three female) was tested for its TNF-a

generating capacity. The patients were recruited consecutively from the Royal Brompton Hospital heart failure clinic. The diagnosis of heart failure was based on symptoms, clinical signs, and documented left ventricular impairment (left ventricular ejection fraction ,40%). All patients were stable on medication for at least four weeks (except adjustments for diuretics in NYHA class IV patients). Medication consisted of ACE inhibitors (95% of all patients), diuretics (100%) and b-blockers (59%) in varying combinations. No patient was taking non-steroidal anti-inflammatory drugs or any steroid hormones. Subjects with clinical signs of infection, rheumatoid arthritis, renal failure (creatinine for all patients ,180 mmol / l) or cancer were excluded. The local ethics committee approved the study, and all subjects gave written informed consent. PBMC preparation, culture and lipopolysaccharide treatment PBMC were isolated using Ficoll Paque (Amersham Pharmacia Biotech AB, Uppsala, Sweden) gradient centrifugation from venous citrated blood taken after at least 15 min supine rest in the morning. Cells were washed twice in Hanks’ balanced salt solution (Sigma–Aldrich, Irvine, UK) and then resuspended at a concentration of 1310 6 / ml in RPMI 1640 (Life Technologies Ltd., Paisley, UK) supplemented with 10% foetal calf serum (FCS), 2 mM L-glutamine, 100 U / ml Penicillin and 100 mg / ml Streptomycin (all from Sigma–Aldrich). After seeding 1 ml aliquots in 12-well plates the cells were allowed to rest for 24 h in a humidified atmosphere (37 8C, 5% CO 2 ). Subsequently, Escherichia coliderived LPS (serotype 0111:B4, Sigma–Aldrich) was added to achieve nine different final concentrations (0.001, 0.01, 0.03, 0.06, 0.1, 0.3, 0.6, 1 and 10 ng / ml). The addition of Hanks’ balanced salt solution alone served as a control. Following 24 h incubation under the above conditions, the supernatants were harvested and frozen immediately at 280 8C for later analysis. Whole blood culture and lipopolysaccharide treatment After whole blood collection, 1 ml aliquots were placed in 1.5 ml Eppendorf tubes. Prior to LPS-

S. von Haehling et al. / International Journal of Cardiology 90 (2003) 197–204

stimulation, the samples were allowed to rest for 24 h in a humidified atmosphere. Subsequent addition of LPS yielded nine different final LPS concentrations as described above. Application of Hanks’ balanced salt solution alone again served as a control. Supernatants were harvested after another 24 h incubation period under the above conditions and frozen immediately (280 8C) for later analysis. Detection of TNF-a The detection of TNF-a was performed using standard ELISA kits according to the manufacturer’s instructions (R&D Systems, Minneapolis, USA). The detection limit was 15 pg / ml. All samples were thawed only once for immediate analysis. Plasma LPS measurement Endotoxin-free tubes (Endo Tube ET, Chromogenix AB, Sweden) were used to collect blood for LPS assessment. Plasma LPS levels were determined using the Limulus Amebocyte Lysate assay (QCL-1000 test kit, Bio Whittaker Inc, Walkerswill, USA) with the lower limit of detection being 0.05 EU / ml. At concentrations of 0.35 EU / ml and 0.82 EU / ml the within-assay coefficients of variation were 9.9 and 9.6% and the between-assay coefficients of variation were 16.8 and 13.3%, respectively [11]. Statistical analysis Statistical analysis was performed using StatView 5.0 software (SAS Institute Inc., Cary, USA). Results are reported as mean6S.E.M. TNF-a levels were square root-transformed before statistical analysis to

199

achieve a normal distribution. Unpaired Student’s t-test, ANOVA with Fisher’s post hoc test, ANOVA for repeated measures, simple and logistic regression analysis were used as appropriate. All P-values ,0.05 were considered significant.

Results After incubation without LPS stimulation, two controls and nine CHF patients had detectable levels of TNF-a in PBMC supernatant (211662 and 4366101 pg / ml, respectively). This group constituted the high-responders, and the remainder the lowresponsers. The plasma LPS levels did not correspond with high- or low-responder status, either in the control or in the CHF patients group (all P.0.2, Tables 1 and 2). There was a significant increase in TNF-a release from PBMC in response to LPS in both the low- and high-responder groups (both P, 0.0001, repeated measures ANOVA). We found a significant positive correlation between TNF-a production and age in CHF patients for all LPS doses between 0 and 0.6 ng / ml (all P,0.05) with the exception of 0.001 and 0.3 ng / ml LPS (P50.1 and 0.05, respectively). In the control subjects, there was also a significant correlation between TNF-a production and age for all LPS doses between 0 and 0.1 ng / ml (all P,0.05) with the exception of 0.06 ng / ml (P50.10, Fig. 1). In both groups, high-responders were significantly older than low-responders (Fig. 2, all P,0.05). This was also true when combining the control and CHF patient groups (P,0.05). In the control group the mean age of low-responders was 47.562.5 years, that of high-responders 65.869.2 years (P50.023), re-

Table 1 Baseline data for patients and control subjects

NYHA class Age (year) Hemoglobin (g / dl) Hematocrit (%) White blood cells (310 9 / l) Sodium (mmol / l) Creatinine (mmol / l) Uric acid (mmol / l) CRP (mg / l) Plasma LPS (EU / ml) NS, non-significant with P.0.20.

Control subjects

CHF patients

P-value

51.663.6 13.660.5 39.961.2 7.760.7 137.860.5 75.265.1 313.0631.4 11.662.8 0.41760.19

3.060.2 68.361.5 12.660.4 38.561.0 7.960.4 136.560.8 126.767.4 451.3629.4 12.661.9 0.46260.21

,0.05 0.179 NS NS NS ,0.05 ,0.05 NS NS

200

S. von Haehling et al. / International Journal of Cardiology 90 (2003) 197–204

Table 2 Baseline data for CHF patients according to LPS-responsiveness in mononuclear cell cultures (PBMC)

Number NYHA Age (yr) Hemoglobin (g / dl) Hematocrit (%) White Blood Cells (x 10 9 / l) Sodium (mmol / l) Creatinine (mmol / l) Uric Acid (mmol / l) CRP (mg / l) Plasma LPS (EU / ml)

Low responders

High responders

P-value

13 3.460.3 65.961.9 12.860.5 39.461.2 8.060.4 136.463.2 132.669.8 418.0641.7 10.862.0 0.45860.030

9 2.660.2 71.961.9 12.160.6 35.861.3 7.660.9 136.661.3 118.6611.5 499.3636.1 15.964.0 0.46960.031

,0.05 NS 0.121 NS NS NS 0.081 NS NS

NS, non-significant with P.0.20.

spectively. Similarly, in the CHF patient group the low-responders were younger (65.961.9 years) than the high-responders (71.961.9 years, P50.040). In logistic regression we found that higher age independently of the clinical status of the study subjects (controls or CHF patients, NYHA class) related to the presence of high-responder status (P,0.05, Table 3). The mean TNF-a production at all LPS doses in

both controls and CHF patients was higher in highresponders than in low-responders (Fig. 3). Interestingly, these results did not reach significance at the highest LPS dose used. In control subjects, significance was not reached for 0.3 and 1 ng / ml LPS, although there was a trend towards significance at 0.3 ng / ml LPS (P50.064). Applying our above definition of responder status

Fig. 1. The relationship between TNF-a release from LPS-stimulated mononuclear cells (PBMC) and age in 22 patients with chronic heart failure and nine healthy control subjects. * P,0.05, (*) P50.051.

S. von Haehling et al. / International Journal of Cardiology 90 (2003) 197–204

201

Fig. 2. The mean age of control subjects and CHF patients subgrouped according to LPS-responsiveness of mononuclear cells (PBMC) into high- (dashed bars) and low-responders (solid bars). * P,0.05. Table 3 Logistic regression model analysing the likelihood of being high- vs. low-responder depending on age or clinical status

x2

P-value

Relative risk for the likelihood of high-responder status

Model 1 Age .70 yes / no CHF yes / no

5.89 0.05

0.015 NS

9.33 –

Model 2 Age .70 yes / no NYHA class II NYHA class III NYHA class IV

5.10 2.15 0.60 0.43

0.024 0.143 NS NS

15.61 – – –

Model 3 Age (years) CHF patients

5.77 1.44

0.016 NS

1.23 –

Model 4 Age (years) NYHA class II NYHA class III NYHA class IV

4.39 0.2 2.2 2.1

0.036 NS 0.137 0.147

1.25 – – –

Comment: In models with NYHA class, control subjects were set as NYHA class 0. NS5non-significant with P.0.20.

202

S. von Haehling et al. / International Journal of Cardiology 90 (2003) 197–204

Fig. 3. TNF-a production by LPS-stimulated PBMC in control subjects and CHF patients subgrouped according to low- and high-responders. Unpaired t-tests between high and low responders at each LPS-concentration: * P,0.05, ** P,0.01, *** P,0.001.

to whole blood cultures, two control subjects and four CHF patients fulfilled the criteria of high-responders. There was a significant increase in TNF-a release from whole blood samples in response to LPS in both the low- and high-responder groups (both P,0.0001, repeated measures ANOVA). Interestingly, in whole blood we could not confirm the hypothesis that highresponders are older than low-responders, neither in the control (P50.16), nor in the CHF group (P5 0.68). No correlation between age and TNF-a production could be detected with the exception of the highest LPS concentration used, i.e. 10 ng / ml. Whilst in healthy control subjects, at this LPS concentration, a positive correlation between these parameters showed only a trend towards significance (r50.663, P50.052), the relationship between age and TNF-a production in CHF patients was negatively correlated (r520.561, P50.015).

Discussion The results of our study show that healthy control subjects and CHF patients subdivided into high- and low-responders according to PBMC TNF-a production at baseline differ significantly in age. High responders are significantly older than low responders, and high-responders secrete more TNF-a at all LPS doses. Since our definition uses the production of TNF-a at baseline in PBMC, i.e. unstimulated by LPS, our results can be interpreted as an unascertained status of subclinical activation of TNF-a production. Moreover, the spontaneous production of TNF-a in high-responders may be pathophysiologically relevant particularly in CHF patients in the presence of increased LPS activity, e.g. in acute oedematous decompensation. Other authors have described conflicting responses

S. von Haehling et al. / International Journal of Cardiology 90 (2003) 197–204

to LPS in PBMC and whole blood in healthy aged humans compared to young volunteers. Riancho et al. failed to demonstrate age-related differences in the secretion of TNF-a in PBMC using 20 mg / ml LPS [5]. This concentration far exceeds the doses used in this study as well as the LPS levels found in vivo in the setting of CHF or sepsis [12]. We found that even at a relatively modest 10 ng / ml the differences in TNF-a production between high-responders and lowresponders and the effect of age were lost (see Figs. 1 and 3). Lower concentrations of stimultion may better be able to discriminate these differences [13]. Fagiolo and collegues, for example, found significant increases in TNF-a production in aged healthy compared to young healthy volunteers using a similar stimulation model with 1 mg / ml phytohemagglutinin [4]. Bruunsgard et al., however, reported a decreased production of TNF-a in elderly control subjects compared to young adults after applying 1 mg / ml LPS in a whole blood model [2]. We could confirm this finding inasfar as only the highest concentration of LPS, i.e. 10 ng / ml, in our whole blood model showed a significant relationship between age and TNF-a generating capacity. However, our results differed between controls and CHF patients, and only in CHF patients a negative correlation was found. Delpedro et al. found a decrease in TNF-a production by separated monocytes in aged compared to young adults for a dose-range from 0.6 ng / ml to 10 mg / ml, thus presenting the only available dose–response data in this respect [3]. Delpedro and co-workers made use of a bioassay for the detection of TNF-a (rather than ELISA). Bioassays are dependent on the presence of serum, which has to be seen as a potential source of confounding cytokines [14]. Another important consideration is that in most cases LPS plasma levels are given as LPS bioactivity, i.e. Endotoxin units (EU) / ml. Thus, the comparison of cell culture results and plasma levels is anything but simple. Yokota et al. provided a new assay for the measurement of LPS in the plasma [12]. Using E. coli 0111:B4 as a reference, they found plasma concentrations below 10 pg / ml in healthy controls. In patients suffering from severe peritonitis they detected plasma LPS-levels of up to 500 pg / ml. The plasma concentrations in normal subjects are usually referred to as being below 0.50 EU / ml [11], which

203

was confirmed by our data. In acute decompensated CHF higher levels of LPS at a mean value of 0.74 EU / ml are found [11]. Thus, we used a broad range of LPS concentrations between 0.001 and 10 ng / ml assuming the LPS concentrations found in vivo to be between 0.01 and 0.6 ng / ml. There are important pathophysiological changes associated with ageing, particularly relating to cardiovascular abnormalities, like arterial stiffening [15] and myocardial adaptations specifically found in the elderly [16]. Alterations in the immune response are less obvious: It has been reported that the total number of T lymphocytes in the peripheral blood are reduced in aged humans [17], and a dysbalance between the T-helper (T H ) 1 and T H 2 subsets with a predominant production of T H 2 cytokines has been described [18]. Even the B cell population is known to be diminished in the elderly, although both the IgG and the IgA fraction are usually found to be increased [19]. Interestingly, the total number of monocytes in the blood is not altered [18]. These cells and their descendants, i.e. macrophages, represent the most important source of proinflammatory cytokines and are highly positive for the lipopolysaccharide (LPS) receptor CD14. However, it should be noted that the available data only indicate cell numbers in the blood and not in the body as a whole. No information are yet obtainable on total cell numbers, e.g. in lymphoid organs. A body of evidence has accumulated in the past few years that LPS might be the causative stimulus for the enhanced secretion of TNF-a in CHF [10]. The close relationship between LPS and TNF-a in CHF was demonstrated by high levels of soluble CD14, a marker of LPS–cell interaction and shedding from the cell membrane, in these patients [10]. The precise cause for enhanced TNF-a levels still remains uncertain, although TNF-a seems to play a major role in disease progression [20]. The highest level of TNF-a found in CHF are seen in patients with cardiac cachexia [21]. It has therefore been speculated that especially these patients might benefit from immunosuppressive therapy [22]. The results of the present study provide insights into activation mechanisms of the immune system at LPS-concentrations found in vivo. Bacterial LPS is probably the most powerful trigger for the secretion

204

S. von Haehling et al. / International Journal of Cardiology 90 (2003) 197–204

of TNF-a both in vivo and in vitro. Our study provides evidence, that an increased production of TNF-a may at least in part be due to an aged immune response. Ageing may partly explain the higher levels of TNF-a found in CHF patients with severe disease but other possibilities, such as LPS entering the circulation through the oedematous gut wall, may also be important [10]. Experiments with PBMCs provide insights into cellular mechanisms and the situation found in tissues and in lymphoid organs. Interestingly, no relationship between age and LPS-responsiveness was found for controls and CHF patients in whole blood at concentrations assumed to be found in vivo. Other factors such as lipoproteins and LPS-binding protein may have an important immunoregulatory role in the blood, particularly on LPS bioactivity [23]. However, these factors may be unable to abolish the effects of LPS at higher concentrations, although these concentrations are less likely to be found in vivo.

References [1] Castle SC. Impact of age-related immune dysfunction on risk of infections. Z Gerontol Geriat 2000;33:341–9. [2] Bruunsgaard H, Pedersen AN, Schroll M, Skinhoj P, Pedersen BK. Impaired production of proinflammatory cytokines in response to lipopolysaccharide (LPS) stimulation in elderly humans. Clin Exp Immunol 1999;118:235–41. [3] Delpedro AD, Barjavel MJ, Mamdouh Z, Faure S, Bakouche O. Signal transduction in LPS-activated aged and young monocytes. J Interferon Cytokine Res 1998;18:429–37. [4] Fagiolo U, Cossarizza A, Scala E, Fanales-Belasio E, Ortolani C, Cozzi E et al. Increased cytokine production in mononuclear cells of healthy elderly people. Eur J Immunol 1993;23:2375–8. [5] Riancho JA, Zarrabeitia MT, Amado JA, Olmos JM, GonzalezMacias J. Age-related differences in cytokine secretion. Gerontology 1994;40:8–12. [6] Bolger AP, Anker SD. Tumour necrosis factor in chronic heart failure: a peripheral view on pathogenesis, clinical manifestations and therapeutic implications. Drugs 2000;60:1245–57. [7] Levine B, Kalman J, Mayer L, Fillit HM, Packer M. Elevated circulating levels of tumor necrosis factor in severe chronic heart failure. N Engl J Med 1990;323:236–41.

[8] Deswal A, Petersen NJ, Feldman AM, Young JB, White BG, Mann DL. Cytokines and cytokine receptors in advanced heart failure: an analysis of the cytokine database from the vesnarinone trial (VEST). Circulation 2001;103:2055–9. [9] Rauchhaus M, Doehner W, Francis DP, Davos C, Kemp M, Liebenthal C et al. Plasma cytokine parameters and mortality in patients with chronic heart failure. Circulation 2000;102:3060–7. [10] Anker SD, Egerer KR, Volk HD, Kox WJ, Poole-Wilson PA, Coats AJS. Elevated soluble CD14 receptors and altered cytokines in chronic heart failure. Am J Cardiol 1997;79:1426–30. [11] Niebauer J, Volk HD, Kemp M, Dominguez M, Schumann RR, Rauchhaus M et al. Endotoxin and immune activation in chronic heart failure: a prospective cohort study. Lancet 1999;353:1838–42. [12] Yokota M, Kambayashi J, Tanaka T, Tsujinaka T, Sakon M, Mori T. A simple turbidimetric time assay of the endotoxin in plasma. J Biochem Biophys Methods 1989;18:97–104. [13] Genth-Zotz S, von Hoehling S, Bolger AP, Kalra PR, Wensel R, Coats AJ, Anker SD. Pathophysio logic quantities of endotoxininduced tmor necrosis factor-alpha release in whole blood from patients with chronic heart failure. Am J Cardiol 2002;90:1226–30. [14] Rauchhaus M, Haehling von Lanzenauer S, Coats AJS. Nitric oxide synthase in CHF (letter). Eur Heart J 2000;21:856–7. [15] Kendall MJ. Hypertension in the elderly. Basic Res Cardiol 1998;93(Suppl 2):43–6. [16] Lakatta EG. Myocardial adaptations in advanced age. Basic Res Cardiol 1993;88(Suppl 2):125–33. [17] Lehtonen L, Eskola J, Vainio O, Lehtonen A. Changes in lymphocyte subsets and immune competence in very advanced age. J Gerontol 1990;45:M108–12. [18] Rink L, Cakman I, Kirchner H. Altered cytokine production in the elderly. Mech Ageing Dev 1998;102:199–209. [19] Paganelli R, Quinti I, Fagiolo U, Cossarizza A, Ortolani C, Guerra E et al. Changes in circulating B cells and immunoglobulin classes and subclasses in a healthy aged population. Clin Exp Immunol 1992;90:351–4. [20] Torre-Amione G, Kapadia S, Benedict C, Oral H, Young JB, Mann DL. Proinflammatory cytokine levels in patients with depressed left ventricular ejection fraction: a report from the studies of left ventricular dysfunction (SOLVD). J Am Coll Cardiol 1996;27:1201–6. [21] Anker SD, Chua TP, Ponikowski P, Harrington D, Swan JW, Kox WJ et al. Hormonal changes and catabolic / anabolic imbalance in chronic heart failure and their importance for cardiac cachexia. Circulation 1997;96:526–34. [22] von Haehling S, Genth-Zotz S, Anker SD, Volk HD. Cachexia: a therapeutic approach beyond cytokine antagonism. Int J Cardiol 2002;85:173–83. [23] Rauchhaus M, Coats AJS, Anker SD. The endotoxin–lipoprotein hypothesis. Lancet 2000;356:930–3.