Eur. J. Immunol. 1991. 21: 2177-2184
Carlos Munozo+, Benoit Misseta, Catherine Fittingo, Jean-Pierre BICriot*, Jean-Carlet* and Jean-Marc CavaillonO Unit6 d’Immuno-Allergieo, Institut Pasteur and Intensive Care Unit*, H8pital Saint Joseph, Pans
Plasma and monocyte-associated cytokines
Dissociation between plasma and monocyte-associated cytokines during sepsis We report our investigations of circulating interleukin (IL) l p , IL 6 and tumor necrosis factor (TNF)-a, as well as cell-associated IL l a , IL l p and TNF-a in plasma and monocytes of 21 patients with sepsis syndrome and 6 patients with non-septic shock. Longitudinal studies reveal that (a) the most frequent detectable plasma cytokines were TNF-a and IL6, (b) the presence and the kinetics of circulating cytokines were independent of one other, (c) detectable levels of cytokines could be found for a long period of time, and (d) significantly higher levels of IL 6 were found for non-surviving patients. Because of the in vivo half-life of cytokines and of the existence of numerous specific high-affinity receptors, it is quite probable that detectable plasma cytokines represent the excess of produced mediators which have not been trapped by the target cells. TNF-a (410 k 65 pg/106monocytes) and IL 1fi(153 It: 60 pg/106monocytes) were frequently found associated to monocyte lysates (88% and 50%, respectively). Despite the fact that IL la is the most abundant cytokine found associated to monocytes following in vitro activation, IL la was rarely found in monocytes of intensive care unit patients (29%). No correlation was found to exist between the levels of plasma cytokines and cell-associated cytokines. Some patients had plasma TNF-a or IL 1p in the absence of the corresponding monocyte-associated cytokine. This observation suggests that cells other than monocytes can participate in the production of circulating cytokines. At the end of the longitudinal study (day 14 k 2), only 2/12 surviving patients still had plasma TNF-a,whereas 8/12 had monocyte-associated TNF-a.These results indicate that activation of monocytes still occurs in patients for whom no plasma cytokines can be detected. Thus, in addition to the measurement of plasma cytokine, measurement of cell-associated cytokine appears useful to assess cytokine production and monocyte activation in vivo.
1 Introduction Since the first observation in 1987 by Waage et al. [l] that TNF-a could be found in the plasma of patients with meningitis and septicemia, numerous authors have shown the presence of TNF-a, IFN-y, IL 1p and IL 6 in plasma or in the cerebrospinal fluid of patients with severe infections [2-121. However, controversial findings still exist concerning the levels of plasma cytokines, the proportion of patients with detectable cytokines and the correlation with the outcome. For example, Girardin et al. [2] found up to 1.7 ng/ml of IL 18 during Gram-negative septicemia, whereas levels found by Cannon et al. [3] did not exceed 270 pg/ml. Marks et al. [4] found circulatingTNF-a among only 36% of patients with septic shock, whereas Calandra et al. [5] found TNF-a in plasma of 79% of their patients. Several authors [2,5,7] noticed a correlation between high levels of circulating IL 1p and subsequent mortality; however, Cannon et al. [3] found that non-surviving patients
[I 95801 +
2177
Supported by a grant from the Fondation pour la Recherche MCdicale. Present address: Immunology Unit, INTA, Universidad de Chile, Casilla 15 138, Santiago 11, Chile.
Correspondence: Jean-Marc Cavaillon, Unit6 d’Immuno-Allergie , Institut Pasteur, 28 rue Dr. Roux, F-75724 Paris Cedex 15, France 0 VCH Verlagsgesellschaft mbH, D-6940 Weinheim, 1991
had lower levels of this cytokine. These discrepancies between different reports may reflect the use of different technical means of measuring circulating cytokines and the rarity of longitudinal studies. Indeed, measurement of circulating cytokines in infected patients is usually only made once, upon admission to the hospital. In the present investigation we have performed longitudinal studies among septic patients admitted into an intensive care unit (ICU) and have compared the levels of circulating IL l p, TNF-a and IL6. Experiments performed in vivo with animal models have established that cytokines have a rather short half-life [13-151. Furthermore, a great variety of cells have highaffinity surface receptors for these cytokines and, thus, are able to trap them very efficiently. As a consequence, it is quite probable that measurable circulating cytokines which can be detected during a given pathology represent the “tip of the iceberg” (i.e. the excess of production) and absence of detectable cytokines in plasma does not necessarily mean that they are not being produced. To address this question we have measured cell-associated cytokines in lysates of freshly isolated monocytes. Such an approach has been recently proved useful in demonstrating monocyte activation, as we reported during hemodialysis [16], or as it has been shown during cardiac bypass surgery [17]. Following monocyte activation in vitro, IL la appears to be the main cell-associated cytokine [18, 19J.Wehave, thus, studied the presence of IL la associated to monocytes as well as IL,10 and TNF-a.
+
0014-298Of91lO909-2177$3.50 .25/0
2178
C. Munoz, B. Misset, C. Fitting et al.
2 Materials and methods 2.1 Patients Twenty-seven patients admitted to the ICU were studied. Twenty-one fullfilled the criteria for "sepsis syndrome" as described by Bone et al. [20], including evidence of an infectious site and signs of inadequate organ perfusion. Of these, 12 met the criteria for septic shock [sepsis syndrome plus a systolic blood presure (BP) 40 mm Hg] and 7 met the criteria for adult respiratory distress syndrome (ARDS; bilateral alveolar infiltrates, pulmonary wedged pressure < 18 mm Hg and Pa02/FiO2 < 150 with positive expiratory pressure I 5 cm H20). The infectious conditions included pneumonia (n = 7), perinitis (n = 4), catheter-related sepsis (n = =3), pyelonephritis (n = 2), infectious aortic aneurism (n = l ) , aortic graft sepsis (n = l ) , Gram-negative septicemia due to granulocytopenia (n = 1), endocarditis (n = 1)and meningitis (n = l).The pathogenic agents were Gram-negative bacilli (GNB) in 6 cases, Gram-positive cocci (GPC) in 7 cases, association of GNB and GPC in 7 cases, and Mycobacteriurn tuberculosis in 1 case. Eleven patients were bacteremic. The 6 remaining patients were admitted because of cardiogenic shock (n = 3), hemorrhagic shock (n = 2) or hypovolemic shock (n = 1). The simplified acute physiological score (SAPS) [21] on admission was 12.7 k 4.2 (mean f SD) for the 21 septic patients and 12.7 f 4.3 for the 6 non-septic patients. Treatment of infected patients included antibiotics, morphine, benzodiazepine, heparin, cathecholamine, insulin and several occasional drugs. Eleven patients died in the ICU. Healthy controls (n = 8) were donors from the blood bank and research workers from our laboratory. The protocol had been approved by the ethical commitee of the Saint Joseph Hapit al. 2.2 Collection of blood samples Blood was collected within 2 days after admission of the patients in the ICU, thereafter daily for the 3 following days, and then weekly until discharge, death or day 21 if the patients were still in the ICU. Last measurements were performed on day 14 k 1.5 (range 5-21) for surviving patients and on day 10 f2.5 (range 1-21) for non-surviving patients. 2.3 Plasma Twenty milliliters of blood was drawn on EDTA. Blood was immediately placed on ice and processed within 2 h. Two hundred microliters of aprotinin (0.67 TUI/ml of blood; Sigma, St. Louis, MO) was added and tubes centrifuged at 400 x g, 4"C, for 10 min. Plasma was transferred into Eppendorf (Hamburg, FRG) tubes and re-centrifuged at 10000 x g, 4°C for 5 min. Plasma was then aliquoted and kept at -30°C. 2.4 Monocytes Cell pellets were resuspended in 25 ml RPMI 1640 medium (Gibco, Uxbridge, GB) and centrifuged over Ficoll (MSL,
Eur. J. Immunol. 1991. 21: 2177-2184 Eurobio, Paris, France) for 25 min at 150 x g at room temperature. After washing, PBMC were counted and nonspecific esterase (NSE) staining performed [22]. NSE+ cells (5 x 105)/0.5ml RPMI medium/well (24-well multidish plates, Falcon, Oxnard, CA) were allowed to adhere for 1h as previously described [23]. 2.5 Collection of cell-associated cytokine content Immediately after adherence, nonadherent cells were removed, adherent cells washed and 0.5 ml of fresh RPMI medium was added. Plates were then frozen as previously described [24]. Three cycles of freezing-thawing were performed before harvesting the cell lysate which was subsequently centrifuged and kept at -30 "C until cytokine contents were assessed. 2.6 IL6 bioassay IL 6 activity was determined using the specific 7TD1 IL6-dependent cell line [25], kindly provided by Dr. J.Van Snick (Ludwig Institute for Cancer Research, Brussels, Belgium). Cells were cultured at a density of 1200 cells/well (96-well multidish plates, Falcon) in 100 pl of RPMI medium supplemented with antibiotics, 2-ME (5 x lop5M) and 10% FCS, in the presence of either serial dilutions of cell SN or heat-inactivated (30 min, 56°C) plasma. After 4 days of culture at 37 "C, the proliferation was monitored by a staining method [26]. Briefly, 125 pg of tetrazolium salt (thiazolyl blue, M7T) were added to each well and after 1 to 2 h of incubation at 37"C, the test was stopped with 100 pl/well of extraction buffer (20% SDS, 50% dimethylformamide in H20, 2.5% 1N HCl, 2.5% of an 80% acetic acid solution, pH7.4). After an overnight incubation at 37 "C the absorbance was measured at 540 nm using an automated microELISA autoreader. One unit of IL6 corresponds t o half the maximum growth rate of the hybridoma cells. I L 6 activity detected in SN of LPSstimulated monocytes and in plasma was completely abolished by the addition of 10 pg rabbit polyclonal anti-human I L 6 antibodies (Genzyme, Boston, MA). 2.7 TNF-a RIA
The TNF-a RIA was based on the competitive inhibition assay for TNF-a developed by Van der Meer et al. [27],with minor modifications. In every assay, 10 TNF-a standards (RhBne Poulenc,Vitry/Seine, France) containing 0,40,80, 150, 300, 600, 1250, 2500, 5000 and 1OOOOpg/ml were included. All standards and cell SN or cell lysate samples were diluted in a BSA buffer.TNF-a in plasma was assessed in undiluted samples. On day 1, 100 p1 of a rabbit antiTNF-a antiserum (kindly provided by Catherine Rougeot, Institut Pasteur, Paris) diluted 1: 10000 in order to precipitate 35% of the radiolabeled TNF, was added to 100 yl of standards or samples.To determine the nonspecific binding, 100 p1 of BSA buffer was added to a tube instead of the sample. Subsequently, 300 p1 of BSA buffer was added to each tube. After vortexing, the tubes were incubated for 24 h at room temperature. On day 2, 100 pl of 1251-labeled TNF-a solution (30 pCi/pg = 111kBq/pg, New England Nuclear, Boston, MA) containing approximatively
Eur. J. Immunol. 1991. 21: 2177-2184
10000 cpm was added to each tube, and after vortexing left for a further incubation of 24 h at room temperature. On day 3, 500 pl of BSA buffer containing 6% PEG 8000 (Sigma), 1% horse anti-rabbit IgG (C. Rougeot, Institut Pasteur) and 0.1% normal rabbit serum were added. Tubes were vortexed, incubated for 1h at room temperature and then centrifuged at 1500 x g for 15 min at room temperature.Thereafter, the SN were discarded and the tubes kept inverted for 30 min and drained on absorbent paper. Tubes were counted in a gamma counter and the value for nonspecific binding subtracted. All standards and samples were expressed as a percentage of the standard containing noTNF-a (zero standard). The concentrations of TNF-a in pg/ml on the logarithmic X axis were plotted against the binding percentage on a logarithmic Y axis. The standard curve obtained was used to determine the TNF-a concentrations in the samples. The detection limit was 70 pg/ml (set binding of 95% of zero standard).
2.8 ILll3 RIA
To remove putative factors present in plasma that might interfere with IL 1p RIA, a chloroform extraction was performed as described by Cannon et al. [28]. One milliliter of chloroform was transferred into a 1.5 ml Eppendorf tube containing 500 pl of plasma. Each tube was mixed vigorously for 10 min and centrifuged at 10OOO x g for 10 min at 4 "C. This operation was repeated twice. The tube was left undisturbed for 20 rnin at room temperature. The aqueous phase was collected and stored at -30°C until assaying. A similar RIA protocol as that used for TNF-a measurement was used to determine the IL l p concentration. Standard IL 1p was obtained from RhGne Poulenc, rabbit anti-IL 1p antiserum diluted (1 : 100) was purchased from Genzyme and 1251-labeledIL l p (126-253 pCi/pg) from NEN. On day 3, 500 pl of BSA buffer containing 6% PEG 8000,1% of sheep anti-rabbit IgG (Sigma) and 0.05% normal rabbit serum were added to each tube. Determination of I L l p concentrations in plasma and samples were calculated as described above. The detection limit was 70 pg/ml.
2.9 I L l a ELISA IL la concentrations were determined by a specific ELISA method developed in our laboratory using two anti-IL la mAb which have been obtained in collaboration with J-C MaziC (Hydrolab, Institut Pasteur). Following immunization of BALB/c mice with human recombinant ILla (hrIL l a , kindly provided by RhGne Poulenc), spleen cells were fused with X63-Ag8.653 mouse myeloma cells [29]. Hybridoma cells were then tested for antibody production against I L l a by ELISA. Cells from positive wells were cloned by LD in microtiter plates. Ig subclass determination was performed using mAb isotyping from Amersham Int. (Amersham, GB). Hybridomas were expanded for antibody production by i.p. injection into pristane-primed mice. Two antibodies were selected: one mAb (28-9) is an IgG1, and the other (2-12) is an IgG2b. These do not recognize the same epitope on the I L l a molecule, as judged by competition ELISA and do not bind to I L l p . Both inhibit the biological activity of IL la in the thymocyte co-mitogenic assay.
Plasma and monocyte-associated cytokines
2179
On day 1, Luxlon ELISA microtiter plates were coated with 100 p1 of mouse IgGl anti-rhIL l a mAb (10 pg/ml in sodium carbonate buffer) and incubated 2 h at 37 "C. The plates were washed three times with 0.1% Tween-20PBS. Standards (0,10,30,100,300,1000,3000 and 10000 pg/ml rhILla) or cell SN or cell lysate samples, diluted in 1% BSA, 0.1% Tween-20/PBS,were added to coated wells and incubated overnight at 4°C. On day2, the plates were washed three times and 100 pl of the second mouse anti-rhIL la mAb (IgG2b; 1:2000 in BSA/Tween/PBS) added to each well. Plates were incubated for 3 h at 37 "C. After washing, 100 pl of peroxidase conjugate anti-mouse IgG2b (1: 1000; ICN; Irvine, CA) were added to each well and the platesleft for 1h a t 37 "C. After washing, enzymatic activity was detected with a phosphate citrate buffer containing 1mg/ml o-phenylenediamine dihydro-chloride (Sigma) and hydrogen peroxide (0.06%). The reaction was stopped with 50 pl of 3 N HCl and the absorbance read at 491 nm in a microplate reader (Titertek multiskan MC340, Flow Lab., Baar, Switzerland). The levels of IL la in the samples were calculated by reference to the standard curve. The detection limit of I L l a was 15 pg/ml. All cytokine measurements have been assessed using the international standards obtained from the National Institute for Biological Standards and Controls (Herfordshire).
2.10 Statistical analysis Statistical analysis was carried out using the paired or unpaired Student's t-test.
3 Results 3.1 Longitudinal study of plasma cytokines in septic patients Plasma of septic patients drawn at different intervals following their admission to the ICU contained significant levels of cytokines, whereas no circulating cytokines were detected in plasma of healthy donors. Eight different examples are given in Fig. 1. Such kinetics indicate independent variations of the different cytokines (i.e. IL l p, TNF-a and IL6): the presence and maximum levels for one given cytokine imply neither the presence nor the peak for another cytokine. The levels of circulating cytokines were usually higher during the course of ICU stay than at admission. For example, 3 out of the 21 septic patients who had no detectable plasma TNF-a initially, had significant levels of TNF-a later on. Maximum levels of circulating cytokines among septic patients are shown in Fig. 2. Circulating TNF-a and IL 6 were found in 86% and 81% of the septic patients, respectively. Importantly, only IL 6 plasma levels (initial as well as maximum) correlated with outcome. TNF-a was higher among septic non-surviving patients; however, initial or maximum levels of TNF-a among surviving and non-surviving septic patients were not statistically different. Initial and maximum levels of IL l p were lower, although not significantly, among patients with fatal outcome when compared to survivers. No IL la could be detected in any sera. No significant difference was noticed between patients with Gram-negative or Grampositive infections (data not shown).
2180
C. Munoz, B. Misset, C. Fitting et al.
Eur. J. Irnrnunol. 1991. 21: 2177-2184
AM1
TNFo
Figure 1. Longitudinal study of circulating TNF-a (grey bars), IL 1p (dark bars) and I L 6 (hatched bars) in eight septic patients (patients MOR, SAN and BOU did not survive). TNF-aand IL l p are expressed as pg/ml (left Z axis) and IL6 as unitshl (right Z axis). White bar represent values under detection limit.
3.2 Cell-associated cytokines in freshly isolated monocytes
Monocyte-associated -ILl a , IL l(3 and TNF-a levels at admission and during ICU stay (maximum levels) are shown in Table 1. Lower levels of IL 1@and TNF-a were observed at admission, as compared to latter observations. I L l a was very rarely found in cell lysates. IL1@ was
observed from time to time, with only 25% of the first determinations among septic patients being positive, and 50% of the septic patients having cell-associated IL 1@at any time during the follow up of their monocytes.The most frequent (88% of septic patients) and the most abundant (410 ? 65 pg/106 monocytes) cell-associated cytokine was TNF-a. Cell-associated IL la and IL l(3 were never detected in freshly isolated monocytes from healthy
Plasma and monocyte-associated cytokines
Eur. J. Immunol. 1991. 22: 2177-2184
2181
Table 1. Comparison between levels of monocyte-associated cytokines measured among septic patients ( n = 16) ad admission and maximum levels observed during the longitudinal study
A VO I KON MAR
SAN
Initial determinationa)
MOR
LET DOY PAR AM I JAN BOU
IL la
DUB
IL1 p
cou BIN GIL FER
TNF-a
HAL LER RAE ICA
EXj
=
nc/
Survival (mean f SEM A 4 3 ?.:25 pglml) Non survival (mean SEM 217 f 41 pg/ml)
0
NS")
0.046 O.(H)I
I
100
200
a) p g / W monocytes. b) Comparison between first and maximum levels (paired Student's t-test). c) Mean5SEM. d) Not significant. e) % positive patients and range.
300
P m l
B LET. ICA. MAR. KON.
donors, and most of these did not haveTNF-a associated to their monocytes (50 f 29 pg/106monocytes). Similar levels of cell-associated cytokines were found within cells of surviving or non-surviving patients (data not shown). No correlation was.found between the levels of cell-associated IL 16 and TNF-a in circulating monocytes and the levels of plasma cytokines. Some patients without detectable circulating TNF-a or IL 16 had the monocyte-associated corresponding cytokine and, inversely, some patients with circulating TNF-a or IL l p did not have any detectable monocyte-associated corresponding cytokine (see Fig. 2). Finally, at the end of the longitudinal study, most of the surviving patients no longer had plasmaTNF-a whereas most of them still had cell-associated TNF-a (Table 2). No significant difference was observed between the different type of infections.
cou.
LER. JAN. MOR. DEL. VOI. BOU.
AMI. DOY,
0
Carlos Munozo+, Benoit Misseta, Catherine Fittingo, Jean-Pierre BICriot*, Jean-Carlet* and Jean-Marc CavaillonO Unit6 d’Immuno-Allergieo, Institut Pasteur and Intensive Care Unit*, H8pital Saint Joseph, Pans
Plasma and monocyte-associated cytokines
Dissociation between plasma and monocyte-associated cytokines during sepsis We report our investigations of circulating interleukin (IL) l p , IL 6 and tumor necrosis factor (TNF)-a, as well as cell-associated IL l a , IL l p and TNF-a in plasma and monocytes of 21 patients with sepsis syndrome and 6 patients with non-septic shock. Longitudinal studies reveal that (a) the most frequent detectable plasma cytokines were TNF-a and IL6, (b) the presence and the kinetics of circulating cytokines were independent of one other, (c) detectable levels of cytokines could be found for a long period of time, and (d) significantly higher levels of IL 6 were found for non-surviving patients. Because of the in vivo half-life of cytokines and of the existence of numerous specific high-affinity receptors, it is quite probable that detectable plasma cytokines represent the excess of produced mediators which have not been trapped by the target cells. TNF-a (410 k 65 pg/106monocytes) and IL 1fi(153 It: 60 pg/106monocytes) were frequently found associated to monocyte lysates (88% and 50%, respectively). Despite the fact that IL la is the most abundant cytokine found associated to monocytes following in vitro activation, IL la was rarely found in monocytes of intensive care unit patients (29%). No correlation was found to exist between the levels of plasma cytokines and cell-associated cytokines. Some patients had plasma TNF-a or IL 1p in the absence of the corresponding monocyte-associated cytokine. This observation suggests that cells other than monocytes can participate in the production of circulating cytokines. At the end of the longitudinal study (day 14 k 2), only 2/12 surviving patients still had plasma TNF-a,whereas 8/12 had monocyte-associated TNF-a.These results indicate that activation of monocytes still occurs in patients for whom no plasma cytokines can be detected. Thus, in addition to the measurement of plasma cytokine, measurement of cell-associated cytokine appears useful to assess cytokine production and monocyte activation in vivo.
1 Introduction Since the first observation in 1987 by Waage et al. [l] that TNF-a could be found in the plasma of patients with meningitis and septicemia, numerous authors have shown the presence of TNF-a, IFN-y, IL 1p and IL 6 in plasma or in the cerebrospinal fluid of patients with severe infections [2-121. However, controversial findings still exist concerning the levels of plasma cytokines, the proportion of patients with detectable cytokines and the correlation with the outcome. For example, Girardin et al. [2] found up to 1.7 ng/ml of IL 18 during Gram-negative septicemia, whereas levels found by Cannon et al. [3] did not exceed 270 pg/ml. Marks et al. [4] found circulatingTNF-a among only 36% of patients with septic shock, whereas Calandra et al. [5] found TNF-a in plasma of 79% of their patients. Several authors [2,5,7] noticed a correlation between high levels of circulating IL 1p and subsequent mortality; however, Cannon et al. [3] found that non-surviving patients
[I 95801 +
2177
Supported by a grant from the Fondation pour la Recherche MCdicale. Present address: Immunology Unit, INTA, Universidad de Chile, Casilla 15 138, Santiago 11, Chile.
Correspondence: Jean-Marc Cavaillon, Unit6 d’Immuno-Allergie , Institut Pasteur, 28 rue Dr. Roux, F-75724 Paris Cedex 15, France 0 VCH Verlagsgesellschaft mbH, D-6940 Weinheim, 1991
had lower levels of this cytokine. These discrepancies between different reports may reflect the use of different technical means of measuring circulating cytokines and the rarity of longitudinal studies. Indeed, measurement of circulating cytokines in infected patients is usually only made once, upon admission to the hospital. In the present investigation we have performed longitudinal studies among septic patients admitted into an intensive care unit (ICU) and have compared the levels of circulating IL l p, TNF-a and IL6. Experiments performed in vivo with animal models have established that cytokines have a rather short half-life [13-151. Furthermore, a great variety of cells have highaffinity surface receptors for these cytokines and, thus, are able to trap them very efficiently. As a consequence, it is quite probable that measurable circulating cytokines which can be detected during a given pathology represent the “tip of the iceberg” (i.e. the excess of production) and absence of detectable cytokines in plasma does not necessarily mean that they are not being produced. To address this question we have measured cell-associated cytokines in lysates of freshly isolated monocytes. Such an approach has been recently proved useful in demonstrating monocyte activation, as we reported during hemodialysis [16], or as it has been shown during cardiac bypass surgery [17]. Following monocyte activation in vitro, IL la appears to be the main cell-associated cytokine [18, 19J.Wehave, thus, studied the presence of IL la associated to monocytes as well as IL,10 and TNF-a.
+
0014-298Of91lO909-2177$3.50 .25/0
2178
C. Munoz, B. Misset, C. Fitting et al.
2 Materials and methods 2.1 Patients Twenty-seven patients admitted to the ICU were studied. Twenty-one fullfilled the criteria for "sepsis syndrome" as described by Bone et al. [20], including evidence of an infectious site and signs of inadequate organ perfusion. Of these, 12 met the criteria for septic shock [sepsis syndrome plus a systolic blood presure (BP) 40 mm Hg] and 7 met the criteria for adult respiratory distress syndrome (ARDS; bilateral alveolar infiltrates, pulmonary wedged pressure < 18 mm Hg and Pa02/FiO2 < 150 with positive expiratory pressure I 5 cm H20). The infectious conditions included pneumonia (n = 7), perinitis (n = 4), catheter-related sepsis (n = =3), pyelonephritis (n = 2), infectious aortic aneurism (n = l ) , aortic graft sepsis (n = l ) , Gram-negative septicemia due to granulocytopenia (n = 1), endocarditis (n = 1)and meningitis (n = l).The pathogenic agents were Gram-negative bacilli (GNB) in 6 cases, Gram-positive cocci (GPC) in 7 cases, association of GNB and GPC in 7 cases, and Mycobacteriurn tuberculosis in 1 case. Eleven patients were bacteremic. The 6 remaining patients were admitted because of cardiogenic shock (n = 3), hemorrhagic shock (n = 2) or hypovolemic shock (n = 1). The simplified acute physiological score (SAPS) [21] on admission was 12.7 k 4.2 (mean f SD) for the 21 septic patients and 12.7 f 4.3 for the 6 non-septic patients. Treatment of infected patients included antibiotics, morphine, benzodiazepine, heparin, cathecholamine, insulin and several occasional drugs. Eleven patients died in the ICU. Healthy controls (n = 8) were donors from the blood bank and research workers from our laboratory. The protocol had been approved by the ethical commitee of the Saint Joseph Hapit al. 2.2 Collection of blood samples Blood was collected within 2 days after admission of the patients in the ICU, thereafter daily for the 3 following days, and then weekly until discharge, death or day 21 if the patients were still in the ICU. Last measurements were performed on day 14 k 1.5 (range 5-21) for surviving patients and on day 10 f2.5 (range 1-21) for non-surviving patients. 2.3 Plasma Twenty milliliters of blood was drawn on EDTA. Blood was immediately placed on ice and processed within 2 h. Two hundred microliters of aprotinin (0.67 TUI/ml of blood; Sigma, St. Louis, MO) was added and tubes centrifuged at 400 x g, 4"C, for 10 min. Plasma was transferred into Eppendorf (Hamburg, FRG) tubes and re-centrifuged at 10000 x g, 4°C for 5 min. Plasma was then aliquoted and kept at -30°C. 2.4 Monocytes Cell pellets were resuspended in 25 ml RPMI 1640 medium (Gibco, Uxbridge, GB) and centrifuged over Ficoll (MSL,
Eur. J. Immunol. 1991. 21: 2177-2184 Eurobio, Paris, France) for 25 min at 150 x g at room temperature. After washing, PBMC were counted and nonspecific esterase (NSE) staining performed [22]. NSE+ cells (5 x 105)/0.5ml RPMI medium/well (24-well multidish plates, Falcon, Oxnard, CA) were allowed to adhere for 1h as previously described [23]. 2.5 Collection of cell-associated cytokine content Immediately after adherence, nonadherent cells were removed, adherent cells washed and 0.5 ml of fresh RPMI medium was added. Plates were then frozen as previously described [24]. Three cycles of freezing-thawing were performed before harvesting the cell lysate which was subsequently centrifuged and kept at -30 "C until cytokine contents were assessed. 2.6 IL6 bioassay IL 6 activity was determined using the specific 7TD1 IL6-dependent cell line [25], kindly provided by Dr. J.Van Snick (Ludwig Institute for Cancer Research, Brussels, Belgium). Cells were cultured at a density of 1200 cells/well (96-well multidish plates, Falcon) in 100 pl of RPMI medium supplemented with antibiotics, 2-ME (5 x lop5M) and 10% FCS, in the presence of either serial dilutions of cell SN or heat-inactivated (30 min, 56°C) plasma. After 4 days of culture at 37 "C, the proliferation was monitored by a staining method [26]. Briefly, 125 pg of tetrazolium salt (thiazolyl blue, M7T) were added to each well and after 1 to 2 h of incubation at 37"C, the test was stopped with 100 pl/well of extraction buffer (20% SDS, 50% dimethylformamide in H20, 2.5% 1N HCl, 2.5% of an 80% acetic acid solution, pH7.4). After an overnight incubation at 37 "C the absorbance was measured at 540 nm using an automated microELISA autoreader. One unit of IL6 corresponds t o half the maximum growth rate of the hybridoma cells. I L 6 activity detected in SN of LPSstimulated monocytes and in plasma was completely abolished by the addition of 10 pg rabbit polyclonal anti-human I L 6 antibodies (Genzyme, Boston, MA). 2.7 TNF-a RIA
The TNF-a RIA was based on the competitive inhibition assay for TNF-a developed by Van der Meer et al. [27],with minor modifications. In every assay, 10 TNF-a standards (RhBne Poulenc,Vitry/Seine, France) containing 0,40,80, 150, 300, 600, 1250, 2500, 5000 and 1OOOOpg/ml were included. All standards and cell SN or cell lysate samples were diluted in a BSA buffer.TNF-a in plasma was assessed in undiluted samples. On day 1, 100 p1 of a rabbit antiTNF-a antiserum (kindly provided by Catherine Rougeot, Institut Pasteur, Paris) diluted 1: 10000 in order to precipitate 35% of the radiolabeled TNF, was added to 100 yl of standards or samples.To determine the nonspecific binding, 100 p1 of BSA buffer was added to a tube instead of the sample. Subsequently, 300 p1 of BSA buffer was added to each tube. After vortexing, the tubes were incubated for 24 h at room temperature. On day 2, 100 pl of 1251-labeled TNF-a solution (30 pCi/pg = 111kBq/pg, New England Nuclear, Boston, MA) containing approximatively
Eur. J. Immunol. 1991. 21: 2177-2184
10000 cpm was added to each tube, and after vortexing left for a further incubation of 24 h at room temperature. On day 3, 500 pl of BSA buffer containing 6% PEG 8000 (Sigma), 1% horse anti-rabbit IgG (C. Rougeot, Institut Pasteur) and 0.1% normal rabbit serum were added. Tubes were vortexed, incubated for 1h at room temperature and then centrifuged at 1500 x g for 15 min at room temperature.Thereafter, the SN were discarded and the tubes kept inverted for 30 min and drained on absorbent paper. Tubes were counted in a gamma counter and the value for nonspecific binding subtracted. All standards and samples were expressed as a percentage of the standard containing noTNF-a (zero standard). The concentrations of TNF-a in pg/ml on the logarithmic X axis were plotted against the binding percentage on a logarithmic Y axis. The standard curve obtained was used to determine the TNF-a concentrations in the samples. The detection limit was 70 pg/ml (set binding of 95% of zero standard).
2.8 ILll3 RIA
To remove putative factors present in plasma that might interfere with IL 1p RIA, a chloroform extraction was performed as described by Cannon et al. [28]. One milliliter of chloroform was transferred into a 1.5 ml Eppendorf tube containing 500 pl of plasma. Each tube was mixed vigorously for 10 min and centrifuged at 10OOO x g for 10 min at 4 "C. This operation was repeated twice. The tube was left undisturbed for 20 rnin at room temperature. The aqueous phase was collected and stored at -30°C until assaying. A similar RIA protocol as that used for TNF-a measurement was used to determine the IL l p concentration. Standard IL 1p was obtained from RhGne Poulenc, rabbit anti-IL 1p antiserum diluted (1 : 100) was purchased from Genzyme and 1251-labeledIL l p (126-253 pCi/pg) from NEN. On day 3, 500 pl of BSA buffer containing 6% PEG 8000,1% of sheep anti-rabbit IgG (Sigma) and 0.05% normal rabbit serum were added to each tube. Determination of I L l p concentrations in plasma and samples were calculated as described above. The detection limit was 70 pg/ml.
2.9 I L l a ELISA IL la concentrations were determined by a specific ELISA method developed in our laboratory using two anti-IL la mAb which have been obtained in collaboration with J-C MaziC (Hydrolab, Institut Pasteur). Following immunization of BALB/c mice with human recombinant ILla (hrIL l a , kindly provided by RhGne Poulenc), spleen cells were fused with X63-Ag8.653 mouse myeloma cells [29]. Hybridoma cells were then tested for antibody production against I L l a by ELISA. Cells from positive wells were cloned by LD in microtiter plates. Ig subclass determination was performed using mAb isotyping from Amersham Int. (Amersham, GB). Hybridomas were expanded for antibody production by i.p. injection into pristane-primed mice. Two antibodies were selected: one mAb (28-9) is an IgG1, and the other (2-12) is an IgG2b. These do not recognize the same epitope on the I L l a molecule, as judged by competition ELISA and do not bind to I L l p . Both inhibit the biological activity of IL la in the thymocyte co-mitogenic assay.
Plasma and monocyte-associated cytokines
2179
On day 1, Luxlon ELISA microtiter plates were coated with 100 p1 of mouse IgGl anti-rhIL l a mAb (10 pg/ml in sodium carbonate buffer) and incubated 2 h at 37 "C. The plates were washed three times with 0.1% Tween-20PBS. Standards (0,10,30,100,300,1000,3000 and 10000 pg/ml rhILla) or cell SN or cell lysate samples, diluted in 1% BSA, 0.1% Tween-20/PBS,were added to coated wells and incubated overnight at 4°C. On day2, the plates were washed three times and 100 pl of the second mouse anti-rhIL la mAb (IgG2b; 1:2000 in BSA/Tween/PBS) added to each well. Plates were incubated for 3 h at 37 "C. After washing, 100 pl of peroxidase conjugate anti-mouse IgG2b (1: 1000; ICN; Irvine, CA) were added to each well and the platesleft for 1h a t 37 "C. After washing, enzymatic activity was detected with a phosphate citrate buffer containing 1mg/ml o-phenylenediamine dihydro-chloride (Sigma) and hydrogen peroxide (0.06%). The reaction was stopped with 50 pl of 3 N HCl and the absorbance read at 491 nm in a microplate reader (Titertek multiskan MC340, Flow Lab., Baar, Switzerland). The levels of IL la in the samples were calculated by reference to the standard curve. The detection limit of I L l a was 15 pg/ml. All cytokine measurements have been assessed using the international standards obtained from the National Institute for Biological Standards and Controls (Herfordshire).
2.10 Statistical analysis Statistical analysis was carried out using the paired or unpaired Student's t-test.
3 Results 3.1 Longitudinal study of plasma cytokines in septic patients Plasma of septic patients drawn at different intervals following their admission to the ICU contained significant levels of cytokines, whereas no circulating cytokines were detected in plasma of healthy donors. Eight different examples are given in Fig. 1. Such kinetics indicate independent variations of the different cytokines (i.e. IL l p, TNF-a and IL6): the presence and maximum levels for one given cytokine imply neither the presence nor the peak for another cytokine. The levels of circulating cytokines were usually higher during the course of ICU stay than at admission. For example, 3 out of the 21 septic patients who had no detectable plasma TNF-a initially, had significant levels of TNF-a later on. Maximum levels of circulating cytokines among septic patients are shown in Fig. 2. Circulating TNF-a and IL 6 were found in 86% and 81% of the septic patients, respectively. Importantly, only IL 6 plasma levels (initial as well as maximum) correlated with outcome. TNF-a was higher among septic non-surviving patients; however, initial or maximum levels of TNF-a among surviving and non-surviving septic patients were not statistically different. Initial and maximum levels of IL l p were lower, although not significantly, among patients with fatal outcome when compared to survivers. No IL la could be detected in any sera. No significant difference was noticed between patients with Gram-negative or Grampositive infections (data not shown).
2180
C. Munoz, B. Misset, C. Fitting et al.
Eur. J. Irnrnunol. 1991. 21: 2177-2184
AM1
TNFo
Figure 1. Longitudinal study of circulating TNF-a (grey bars), IL 1p (dark bars) and I L 6 (hatched bars) in eight septic patients (patients MOR, SAN and BOU did not survive). TNF-aand IL l p are expressed as pg/ml (left Z axis) and IL6 as unitshl (right Z axis). White bar represent values under detection limit.
3.2 Cell-associated cytokines in freshly isolated monocytes
Monocyte-associated -ILl a , IL l(3 and TNF-a levels at admission and during ICU stay (maximum levels) are shown in Table 1. Lower levels of IL 1@and TNF-a were observed at admission, as compared to latter observations. I L l a was very rarely found in cell lysates. IL1@ was
observed from time to time, with only 25% of the first determinations among septic patients being positive, and 50% of the septic patients having cell-associated IL 1@at any time during the follow up of their monocytes.The most frequent (88% of septic patients) and the most abundant (410 ? 65 pg/106 monocytes) cell-associated cytokine was TNF-a. Cell-associated IL la and IL l(3 were never detected in freshly isolated monocytes from healthy
Plasma and monocyte-associated cytokines
Eur. J. Immunol. 1991. 22: 2177-2184
2181
Table 1. Comparison between levels of monocyte-associated cytokines measured among septic patients ( n = 16) ad admission and maximum levels observed during the longitudinal study
A VO I KON MAR
SAN
Initial determinationa)
MOR
LET DOY PAR AM I JAN BOU
IL la
DUB
IL1 p
cou BIN GIL FER
TNF-a
HAL LER RAE ICA
EXj
=
nc/
Survival (mean f SEM A 4 3 ?.:25 pglml) Non survival (mean SEM 217 f 41 pg/ml)
0
NS")
0.046 O.(H)I
I
100
200
a) p g / W monocytes. b) Comparison between first and maximum levels (paired Student's t-test). c) Mean5SEM. d) Not significant. e) % positive patients and range.
300
P m l
B LET. ICA. MAR. KON.
donors, and most of these did not haveTNF-a associated to their monocytes (50 f 29 pg/106monocytes). Similar levels of cell-associated cytokines were found within cells of surviving or non-surviving patients (data not shown). No correlation was.found between the levels of cell-associated IL 16 and TNF-a in circulating monocytes and the levels of plasma cytokines. Some patients without detectable circulating TNF-a or IL 16 had the monocyte-associated corresponding cytokine and, inversely, some patients with circulating TNF-a or IL l p did not have any detectable monocyte-associated corresponding cytokine (see Fig. 2). Finally, at the end of the longitudinal study, most of the surviving patients no longer had plasmaTNF-a whereas most of them still had cell-associated TNF-a (Table 2). No significant difference was observed between the different type of infections.
cou.
LER. JAN. MOR. DEL. VOI. BOU.
AMI. DOY,
0
