Chronic and Intradialytic Effects of High-Flux Hemodialysis on Tumor Necrosis Factor-a Production: Relationship to Endotoxins J.L. Mege, MD, PhD, M.V. Sanguedolce, PhD. R. Purgus, MD, B. Moulin, MD. P. Bongrand, MD, PhD, C. Capo, PhD, and M. Olmer, MD • Tumor necrosis factor-a (TNFa) likely plays a role in hemodialysis-associated complications. As TNFa is mainly produced by monocytes in response to endotoxins, we studied its production and the presence of circulating endotoxins in patients dialyzed on polyacrylonitrile (PAN) membrane. Spontaneous production of TNFa was observed in patients before the dialysis session and increased during the session. Endotoxins were present in serum from patients chronically dialyzed with PAN and increased during hemodialysis session. In addition, intradialytic decrease In CD14 antigen expression on circulating monocytes, which could be caused by endotoxins, was found. The continuous presence of low amounts of circulating endotoxins between sessions may explain the chronic increase in TNFa secretion, while high amounts of circulating endotoxlns may account for intradlalytic oversecretion of TNFa and downmodulation of CD14. We suggest that endotoxin-free dialysates should be a prerequisite for the use of high-flux membranes.
© 1992 by the National Kidney Foundation, Inc. INDEX WORDS: Tumor necrosis factor-a; endotoxins; monocytes; polyacrylonitrile; CD14 antigen.
H
EMODIALYSIS is associated with several side effects, induding reversible neutropenia and monocytopenia, increased susceptibility to infections, amyloidosis, carpal-tunnel syndrome, and osteopenia.I- 3 Although the activation of the complement system may partially account for deleterious effects of hemodialysis, the involvement of other components such as cytokines is now largely suspected. 4,5 It is established that the overproduction of cytokines may favor ,82-microglobulin (,82m) production and ,82 amyloidosis. 6 ,7 Different studies have reported an abnormal production of interleukin (IL)-l and more recently tumor necrosis factor-a (TNFa) in chronic dialysis patients.5,8,9 Moreover, we have recently observed sustained and chronic monocyte production of TNFa, IL-l,8, and IL6, which was independent of complement activation.1O But the factor(s) responsible for the enhanced cytokine production and the influence of the type of membrane is not well understood. 11-l3 Indeed, high-flux dialysis membranes are increasingly used because they achieve a significant clearance of ,82m and seem to prevent amyloidFrom the Laboratoire d 'Immun%gie, Hopita/ de SainteMarguerite, Marseille; and the Service de Nephr%gie et d'Hemodia/yse, Hopita/ de /a Conception, Marseille, France. Received March 6, 1992; accepted in revised/orm June 30, 1992. Correspondence to: Dr. Jean-Louis Mege, Laboratoire d 'Immunologie, Hopita/ de Sainte-Marguerite, 270 Bd de Sainte-Marguerite, 13277 Marseille Cedex 9, France © 1992 by the National Kidney Foundation, Inc. 0272-6386/92/2005-0006$3.00/0 482
osis. 7 Nevertheless, since their introduction, several deleterious effects of these high-flux membranes, such as anaphylactoid reactions l 4-16 or destructive arthropathy,17 have been described. In addition, an enhanced production of IL-l and TNFa was found in some predialysis patients dialyzed with high-permeability membranes. 5 ,9,18 As these membranes do not activate the complement pathway, it is likely that another mechanism related to a possible transmembrane passage of endotoxins may account for the stimulation of cytokine synthesis. 18 Our study was undertaken to evaluate TNFa production by monocytes and the presence of circulating endotoxins in patients dialyzed with polyacrylonitrile (PAN), a high-permeability membrane. We found an oversecretion of TNFa in these patients, which seems to be related to the presence of circulating endotoxins. MATERIALS AND METHODS
Patients Ten patients consisting of six men and four women (mean age, 55 ± 7 years; range, 38 to 67) were selected. They were clinically asymptomatic with normal clinical and chest x-ray examination. They regularly attended the hospital dialysis unit three times weekly for a 4-hour dialysis session. The mean duration of hemodialysis was 50 months (range, 12 to 114). The underlying diseases consisted of chronic interstitial nephritis (n = 3), renal vascular disease (n = 3). chronic glomerulonephritis (n = 2), and unknown etiology (n = 2). All patients were dialyzed with PAN hollow-fiber hemodialyzerfilter (FILTRAL, 12 Hospal, Lyon, France, AN69 HF; surface area, 1.15 m 2; ultrafiltration coefficient. 32 mL/h . mm Hg). Patients were dialyzed with bicarbonate bath for at least 3 months before any biological investigation. Ten healthy subjects (five men and five women; mean age, 62 ± 6 years; range. 35 to 70) recruited among volunteer
American Journal of Kidney Diseases, Vol XX, No 5 (November), 1992: pp 482-488
TNFa, ENDOTOXINS AND HEMODIALYSIS
blood donors served as controls. They were selected as free from any infection, inflammatory syndrome, or known renal disease. Seven patients (four men and three women, mean age, S9 ± 4 years; range, 4S to 71) dialyzed with Cuprophan hollow-fiber dialyzer (Allegro, Organon Teknica, Fresnes, France; surface area, 1.4 m 2; ultrafiltration coefficient, S mL/ h . mm Hg) were included in the study of circulating endotoxins as controls. The mean duration of hemodialysis was 47 months (range, 12 to 99). Five of the patients were first dialyzed on Cuprophan and then treated with PAN. Two patients were only dialyzed on Cuprophan. Patients were dialyzed with each type of membrane for 3 months before the study.
Cells Blood was drawn in heparinized tubes. Mononuclear cells were separated by Ficoll gradient centrifugation (MSL, Eurobio, Les Ulis, France), as previously described,19 and suspended in RPMI 1640 supplemented with 10% fetal calf serum, antibiotics, and L-g1utamine. Platelet contamination was limited by using low-speed centrifugation, and the percentage of polymorphonuclear neutrophils contaminating mononuclear cells was less than S%. One aliquot of mononuclear cells was saved to measure the ability of endotoxins to elicit TNFa secretion. Then, cells at 2 X 106/mL were incubated in a l-mL volume for 2 hours at 37°C in flat-bottom 24-well culture plates (Nunc, Intermed, France). Nonadherent cells were removed by washing. Adherent cells consisted of approximately 9S% monocytes as assessed by nonspecific esterase staining and were incubated for 18 hours at 37°C in supplemented RPMI 1640.20 Supernatants were collected and stored at -70°C before TNFa determination. The spontaneous secretion of cytokine by monocytes reflects in vivo stimulation of monocytes. 1O We assessed the population of monocytes by microscopic counting of adherent cells (S X 105 cells/well) and measurement of protein concentration by using the Bradford method. No significant variation in protein content was observed between the different assays.
Immunoassay for TNFa Determination TNFa was measured by radioimmunometric assay (IRMA, Medgenix, Eria Diagnostics, Pasteur, France) in serum and monocyte supernatants from controls and dialyzed patients. This assay is based on an oligoclonal system including several monoclonal antibodies directed against distinct epitopes of TNFa and an iodinated antibody that triggers the immunological reaction. The IRMA was sensitive up to S pg/mL TNFa.
Measurement of Endotoxins Endotoxins in water and dialysate baths were determined by using Limulus Amebocyte Lysate (LAL) assay (Whittaker Bioproducts, Walkersville, MD, USA). Briefly, SO ~L of sample or standard was dispensed in microtiter plates at 37°C followed by SO ~L of LAL. Then, substrate solution was added for S minutes and the reaction was stopped with 2S% acetic acid. Changes in absorbance were measured at 40S nm with a Titertek multiscan autoreader. The concentration of endotoxin in a sample was calculated from the absorbance values of solutions containing known amounts of endotoxin standard. Endotoxins in serum from controls and dialyzed patients were measured by modified LAL assay and polymyxin B in-
483 hibition of TNFa secretion by mononuclear cells. First, a slight modification of LAL procedure was introduced as described elsewhere. 21 .22 Serum was first diluted 1:20 with pyrogen-free saline and heated to 100°C for 10 minutes to reduce the effects of nonspecific inhibitors. 22 Lipopolysaccharide (LPS; OS S BS Escherichia coli, Sigma, St Louis, MO) was diluted in control serum to assess the putative interference of serum with LAL. Second, it has been shown that some LAL-reactive materials are not related to endotoxins, may consist of cellulose derivatives, and thus may be unable to elicit cytokine secretion. 23 We used a complementary procedure that measures the ability of serum containing endotoxins to stimulate TNFa secretion in mononuclear cells. 23 The inhibition of serum-induced TNFa secretion by polymyxin B, known to block the effects of endotoxins, confirms the presence of endotoxin materials. Autologous serum from controls and dialyzed patients was incubated with polymyxin B (Sigma) at S ~gfmL or diluent for S minutes before assay. Then, treated sera were added to 2 X 106 mononuclear cells in RPMI 1640 to obtain 10% serum concentration and incubated for 18 hours at 37°C, The supernatants were collected and assayed for the presence of TNFa. Control experiments consisted of LPS (200 ngfmL), which was added to mononuclear cells in RPM I 1640 supplemented with autologous serum and incubated for 18 hours at 37°C in the presence or not of polymyxin B (S ~gfmL).
Surface Antigen Expression Mononuclear cells at 2 X 106/mL were incubated for 30 minutes at 4°C in RPMI 1640 containing 1% bovine serum albumin with different FlTC-tagged monoclonal antibodies or control immunoglobulins. IOMI (CDllb, CR3) and lOP 49d (CDw49d, VLA4) were IgG 1 antibodies (Immunotech, Marseille, France), and MY4 (CDI4) was IgG2b antibody (Coulter). Then, cells were washed and fixed with 1% formaldehyde before being subjected to flow cytometry. Fluorescence staining was assessed using an Epics Profile (Coulter, Coultronics, France) equipped with a IS-mW argon laser emitting at 488 nm. Ten thousand cells were scored each time and the fluorescence results were expressed as percentage of labeled cells and mean fluorescence.
Data Analysis Results are given as the mean ± SEM and were compared using Student's t test. Differences were considered significant at P < O.OS.
RESULTS
Effect of Hemodialysis on Circulating TNFa and TNFa Production by Monocytes Circulating TNFa was slightly but not significantly increased in patients chronically dialyzed on PAN membrane (16 ± 5 pgfmL), as compared with controls (10 ± 5 pg/mL). No variation in TNFa levels was observed during the hemodialysis session (Fig 1). Spontaneous production of TNFa by monocytes was significantly (P < 0.005) enhanced in monocytes from patients chronically treated with
484
MEGE ET AL 1200
SO
1000
40
.§ Cl C.
LL
Z I-
30
E 800 C, c. 600
20
Z I-
LL
10
400 200 0
0
C
0
60 30 Time (min)
C
240
0
60 30 Time (min)
240
Fig 1. Effect of hemodialysis on circulating TNFa. Sera from 10 controls (C) and 10 patients dialyzed on PAN were assayed for the presence of TNFa by using an immunoradiometric assay. Samples were collected before the dialysis session (0) and after 30, 60, and 240 minutes. Results in pg/mL are the means ± SEM.
Fig 2. Effect of hemodialysis on TNFa production by monocytes. Monocytes from 10 controls (C) and 10 dialyzed patients (before dialysis session [0], and after 30, 60, and 240 minutes) were incubated in supplemented RPM I 1640 for 18 hours at 37°C. TNFa was assayed in supernatants by immunoradiometric assay. Results in pg/mL are means ± SEM.
PAN membrane (522 ± 75 pg/mL) as compared with monocyte from controls (190 ± 53 pg/mL) (Fig 2). During the first 30 minutes of the dialysis session, TNFa production remained similar to predialytic values. TNFa production leveled off after 60 minutes and reached a maximum value at the end of the dialysis session (847 ± 94 pg/ mL), a significant (P < 0.05) increase as compared with predialytic values (Fig 2).
with serum from PAN patients, but endotoxin levels were significantly higher than control values (P < 0.01). During the dialysis session, endotoxin levels started to increase after 30 minutes, reached a maximum value at 60 minutes (89.8 ± 10.3 ng/mL), and then declined to a value similar to predialytic levels. Conversely, in patients dialyzed on Cuprophan, endotoxin levels did not significantly vary during the session (Fig 3). The assay determining ifTNFa production by mononuclear cells was inhibited by polymyxin
Measurement of Circulating Endotoxins in Dialyzed Patients
Endotoxins in serum were assessed by two complementary methods, LAL assay and polymyxin B-inhibitable TNFa secretion by mononuclear cells. The former assay detects LAL-reactive materials, the latter is a sensitive test to evaluate the activity of endotoxins. 23 As serum may contain inhibitory compounds that may interfere with LAL assay, serum samples were diluted 1:20 and boiled before being tested. Control serum contained residual amounts of endotoxin (1.8 ± 0.5 ng/mL), which is equivalent to the sensitivity of the assay (data not shown). As shown in Fig 3, before the dialysis session, patients dialyzed with PAN exhibited higher levels of serum endotoxins (41.3 ± 3.1 ng/mL) than did controls (P < 0.005). Serum from patients chronically dialyzed on Cuprophan, known to be a low-flux membrane, contained low amounts of endotoxins (7.9 ± 1.8 ng/mL) as compared
100
E C, c:
(/)
80 60
c: )(
0
0
40
"C
c: w
20 0
o
60 30 Time (min)
240
Fig 3. Endotoxin levels in serum. Sera from 10 patients dialyzed with PAN (0 ) and seven patients dialyzed with Cuprophan (CUP [ ]) were diluted 1:20, boiled, and assayed for the presence of endotoxin by LAL assay. Samples were collected before the dialysis session (0), and after 30, 60, and 240 minutes of dialysis. Results in ng/mL are means ± SEM.
485
TNFa, ENDOTOXINS AND HEMODIALYSIS
B confirmed the LAL results. First, 5 ~g/mL polymyxin B markedly inhibited TNFa secretion induced by exogenous LPS at 200 ng/mL (Table 1). Second, sera from patients dialyzed on PAN were incubated with mononuclear cells in the presence or absence of polymyxin B. Before the dialysis session, patient serum displayed a significant (P < 0.005) TNFa-inducing activity as compared with controls, which was sensitive to polymyxin B. After 30 and 60 minutes of dialysis, TNFa production elicited by patient serum was similar to that observed before the dialysis session. At the end of the dialysis session, although TNFainducing activity, sensitive to polymyxin B, was significantly decreased as compared with intradialytic values, some TNFa-inducing activity, insensitive to polymyxin B, was still present (Table 1).
50
E C, c
rn
40 30
c
)(
0
0
20
"0 C
w
10 0 0
30 60 Time (min)
240
Fig 4. Endotoxin levels in dialysates. Endotoxins were assessed in bicarbonate dialysate baths by LAL assay before the dialysis session and after 30, 60, and 240 minutes. Results in ng/mL represent the mean of 10 values ± SEM.
Endotoxins in Dialysates
Effect of Hemodialysis on Monocyte Membrane Antigens
Contaminant endotoxins were assessed in water and dialysate baths by LAL assay. The levels of endotoxins in purified water were 2.0 ± 0.3 ng/mL. As shown in Fig 4, before the dialysis session, bicarbonate dialysates contained significantly more endotoxins than purified water (22.7 ± 7.5 ng/mL, P < 0.02). During the session, the level of endotoxins in dialysates decreased steadily to reach a minimal value after 240 minutes (10.7 ± 4.2 ng/mL). The composition of dialysate (acetate versus bicarbonate) did not modify the degree of endotoxin contamination (data not shown).
Endotoxins may bind some membrane antigens expressed on monocytes, which could modulate their expression. The expression of CD 14 antigen, which is recognized as one of the binding sites for endotoxins, is shown in Table 2. The percentage of monocytes expressing CD 14 was similar in patients before the dialysis session and in controls. This percentage did not vary during the first 60 minutes of the session, but significantly (P < 0.02) decreased at the end of the session. The decrease involved both the number of labeled cells and the intensity of the fluorescence, thus suggesting that CD 14 was partially down-
Table 1. Effect of Serum on TN Fa Production by Mononuclear Cells Dialysis Time (min)
PBPB+
Control
LPS
o
30
60
240
332 ± 50 405 ± 70
1,606 ± 64 592 ± 51
929 ± 79* 512 ± 81
869 ± 129* 471 ± 191
1,109 ± 181* 449 ± 130
591 ± 36*:j:§ 660 ± 24t
NOTE. Control mononuclear cells (2 x 106 /mL) were suspended in RPMI 1640 containing 10% serum from controls and patients dialyzed after 0,30,60, and 240 minutes of dialysis on PAN. Then, mononuclear cells were incubated for 18 hours at 37°C in the presence (PB+) or the absence (PB-) of polymyxin B (5ILg/mL). In some experiments, LPS (200 ng/mL) was added to mononuclear cells in RPMI 1640 supplemented with 10% normal serum as described. Supernatants were removed and TNFa was measured by IRMA. Results are expressed in pg/mL and represent the mean ± SE of five distinct controls and patients at different dialysis times. * p < 0.005, tP < 0.01 corresponded to the comparison between controls and patients at different dialysis times. :j: P < 0.005 corresponded to the comparison between patients before and after dialysis session. § P < 0.02 corresponded to the comparison between patients after the 60- and 240-minute dialysis sessions.
486
MEGE ET AL Table 2. Effect of Hemodialysis on CD14 Antigen Expression Dialysis Time (min)
Percentage Mean
Control
o
30
60
240
10.7±1.5 55.6 ± 2.1
9.8 ± 1.0 48.0 ± 2.5
10.8 ± 1.2 50.5 ± 4.2
11.2±1.7 50.3 ± 4.1
6.3 ± 0.6' 39.5 ± 2.1t
NOTE. Mononuclear cells (2 X 106 /mL) were incubated with FITC-Iabeled 10M2 monoclonal antibodies directed against CD14 for 30 minutes at 4°C. After washing, they were subjected to flow cytometry. Results are expressed as the percentage of positive cells and the mean fluorescence intensity. They represent the mean ± SEM of 10 patients and 10 controls. 'P < 0.02. tP < 0.005.
modulated in monocytes during PAN hemodialysis. White blood cell count remained constant during the dialysis session and the expression of other monocyte membrane antigens such as CDllb/CDI8 and VLA4 was not affected (data not shown). DISCUSSION
High-flux membranes, such as PAN or polysulfone, improve the treatment of chronic renal failure, but their high permeability to compounds such as endotoxins may limit their use. As endotoxins stimulate in vitro TNFa production by monocytes, we studied TNFa secretion and circulating endotoxins in patients dialyzed with PAN membrane. Monocytes from these patients exhibited a marked spontaneous secretion of TNFa. This was evident before the dialysis session and corresponded to a chronic effect of hemodialysis, thus strengthening the results of Chollet-Martin et al obtained with polysulfone membranes. 9 We also observed an increase in TNFa secretion during the dialysis session as compared with predialytic values. This intradialytic effect of hemodialysis was highest at the end of the session. The in vitro oversecretion of TNFa was not associated with an increase in circulating TNFa, thus confirming previous findings obtained with PAN and polysulfone. 8 ,9,24 The in vivo half-life of cytokines, the existence of numerous specific high-affinity receptors, and the relatively poor sensitivity of immunoassays for cytokines could account for the lack of detection of TNFa in serum. 25 ,26 The main mechanisms thought to playa role in the oversecretion of TNFa by monocytes are the activation of the complement and the presence of circulating endotoxins. The first mecha-
nism is unlikely, since PAN is a poor activator of complement pathway.27 We found no modification of hemolytic activity of complement and its components such as C3, C4, and B during the hemodialysis session, regardless of the time of sampling (data not shown). The transient activation of complement cannot account for the chronic effect of hemodialysis on TNFa secretion we observed. In addition, we have previously shown that TNFa production was not increased in patients chronically dialyzed with Cuprophan, which is known to stimulate complement pathway.iO The second mechanism is based on the presence of endotoxins in the bloodstream, which might account for the overproduction of TNFa, but it supposes that endotoxins contained in dialysates have crossed the hemodialysis membrane. We found that bicarbonate dialysate baths contained significant amounts of endotoxins, which steadily decreased during the dialysis session. Numerous reports favor the hypothesis of a transmembrane passage of endotoxins. To a greater extent than low-flux membranes, highflux membranes offer the possibility of backtransport of dialysate contaminants into bloodstream. 28 ,29 Serum from patients treated with high-flux membranes contained more antibodies against endotoxins than those from patients treated with low-flux membranes. 3o IL-l release was more increased in patients dialyzed with contaminated dialysates than in patients treated with dialysates devoid of endotoxins. 31 IL-l secretion was not elicited in an ex vivo system consisting of pyrogen-free dialysate and polysulfone membrane. 32 The addition of labeled LPS to the dialysate compartment (separated from the blood compartment by PAN or polysulfone mem-
487
TNFa, ENDOTOXINS AND HEMODIALYSIS
branes) resulted in the recovery of a component able to induce IL-1 in the blood compartment. 18 Our results appear to support the role of a transmembrane passage of endotoxins through PAN membranes in TNFa secretion. Indeed, circulating endotoxins were found in patients before the PAN dialysis session. They markedly increased after 60 minutes of dialysis and decreased thereafter. Since we cannot exclude that some circulating LAL-reactive material not related to endotoxins may be present during dialysis session,33 we used a complementary method based on polymyxin B-inhibitable cytokine secretion. Indeed, serum from patients dialyzed with PAN displayed significant ability to induce TNFa secretion by mononuclear cells in a polymyxinsensitive manner. Nonendotoxinic material was probably present chronically and during the dialysis session, since serum from PAN patients was able to induce TNFa secretion in the absence of polymyxin B, as compared with healthy controls. Moreover, serum from patients dialyzed with Cuprophan, a low-permeable membrane, contained only residual amounts of endotoxins, suggesting that some endotoxins may cross Cuprophan membranes as previously observed with bacterial LPSs during in vitro dialysis. 18 The presence of endotoxins in both serum and dialysates suggests transmembrane passage. While the main fraction of LPS is present in dialysates as high-molecular weight aggregates,33 low-molecular weight species of LPS are able to cross hemodialysis membranes and to elicit cytokine secretion. 18,34 However, it has been proposed that only a small fraction of endotoxins contained in dialysates would be able to cross PAN membranes, while the main fraction of endotoxins would be adsorbed on these membranes. That hypothesis is strengthened by the high ability of high-flux membranes to bind endotoxins. 18•35 Indeed, it has been shown that the main fraction of exogenous LPS infused in the dialysate compartment of a closed system was recovered on membrane. 18 We hypothesize that large amounts of endotoxins are susceptible to bind PAN membranes and a fraction of these endotoxins can cross PAN membranes to reach circulating flow. In the bloodstream, endotoxins interact with LPS-binding proteins and the complex binds receptor structures such as CD 14 antigen to elicit cytokine synthesis. 36 Recently, it has been shown
that high concentrations ofLPS downmodulated the expression ofCD14 antigen on monocytes in vitro through a shedding process that required several hours. 37 We observed a marked decrease in CD 14 antigen expression on monocytes at the end ofthe dialysis session, when TNFa production was the highest. We suggest that high intradialytic levels of circulating endotoxins resulted in delayed downmodulation of CD 14 antigen. The soluble CD14 molecules may bind endotoxins in excess, thus accounting for their decrease at the end of the session. The chronic presence of circulating endotoxins able to elicit sustained TNFa production gives an insight into the chronic effect of high flux hemodialysis on cytokine synthesis. The use of highflux membranes and sterile dialysates would prevent deleterious effects oflong-term hemodialysis.
REFERENCES 1. Burton BT, Krueger NK, Bryan FA: National registry of long term dialysis patients. JAMA 218:718-722,1971 2. Lewis SL, Van Epps DE: Neutrophil and monocyte alterations in chronic dialysis patients. Am J Kidney Dis 5:381395, 1987 3. Drueke TB, Zingraff J, Noel LH, et al: Amyloidosis and dialysis: Pathophysiological aspects. Contrib Nephrol 62:6066, 1988 4. Jacobs AA, Ward RA, Wellhausen SR, et al: Polymorphonuclear leukocyte function during hemodialysis: Relationship to complement activation. Nephron 52: 119-124, 1989 5. Haeffner-Cavaillon N, Fisher E, Bacle F, et al: Complement activation and induction of interleukin-1 production during hemodialysis. Contrib Nephrol 62:86-98, 1988 6. Zaoui PM, Stone WJ, Hakim RM: Effects of dialysis membranes on betarmicroglobulin production and cellular expression. Kidney Int 38:962-968, 1990 7. O'Donoghue DJ, Plant WD, Winney RJ: Dialysis amyloidosis. Lancet 2:821, 1991 8. Herbelin A, Nguyen AT, Zingraff J, et al: Influence of uremia and hemodialysis on circulating interleukin-1 and tumor necrosis factor a. Kidney Int 37:116-125, 1990 9. Chollet-Martin S, Stamatakis G, Bailly S, et al: Induction of tumour necrosis factor-alpha during haemodialysis. Influence of the membrane type. Clin Exp Immunol 83:329-333, 1990 10. Mege JL, Olmer M, Purgus R, et al: Haemodialysis membranes modulate chronically the production of TNFa, IL1~ and IL6. Nephrol Dial Transplant 6:868-875, 1991 11. Amato M, Cozzolino F, Bergesio F, et al: In vitro interleukin-1 production by different dialysis membranes. Nephrol Dial Transplant 3:432-434, 1988 12. Shaldon S, Dinarello CA, Elie M, et al: Interleukin-l and bioincompatibility in hemodialysis. Contrib Nephrol 62: 128-131, 1988
488 13. Bingel M, Lonneman G, Koch KM, et al: Plasma interleukin-I activity during hemodialysis: the influence of dialysis membranes. Nephron 50:273-276, 1988 14. Verresen L, Waer M, Van Renterghem Y, et al: Angiotensin-converting-enzyme inhibitors and anaphylactoid reactions to high-flux membrane dialysis. Lancet 2: 1360-1362, 1990 15. Jadoul M, Struyven J, Stagier A, et al: Angiotensinconverting-enzyme inhibitors and anaphylactoid reactions to high-flux membrane dialysis. Lancet 337: 112, 1991 16. Alvarez-Lara MA, Martin-Malo A, Espinosa M, et al: ACE inhibitors and anaphylactoid reactions to high flux membrane dialysis. Lancet 337:370, 1991 17. Kessler M, Netter P, Maheut H, et al: Highly permeable and biocompatible membranes and prevalence of dialysis-associated arthropathy. Lancet 337: 1092-1093, 1991 18. Laude-Sharp M, Caroff M, Simard L, et al: Induction ofIL-1 during hemodialysis: Transmembrane passage of intact endotoxins (LPS). Kidney Int 38: 1089-1094, 1990 19. Grob JJ, Mege JL, Capo C, et al: The role ofTNF in Sneddon-Wilkinson subcorneal pustulosis: A model of neutrophil priming in vivo. J Am Acad Dermatol 25:944-947, 1991 20. Sanguedolce V, Capo C, Bongrand P, et al: Zymosanstimulated tumor necrosis factor alpha production by human monocytes: Downmodulation by phorbol ester. J Immunol 148:2229-2236, 1992 21. Wakebayashi G, Gelfand JA, Jung WK, et al: Staphylococcus epidermidis induces complement activation, tumor necrosis factor and interleukin-I, a shock-like state and tissue injury in rabbits without endotoxinemia. J Clin Invest 87: 1925-1935, 1991 22. Cohen J, McConnel JS: Observation on the measurement and evaluation of endotoxemia by a quantitative limulus lysate microassay. J Infect Dis 150:916-924, 1984 23. Schindler R, Dinarello CA: A method for removing interleukin-I and tumor necrosis factor-inducing substances from bacterial cultures by ultrafiltration with polysulfone. J Immunol Methods 116:159-165, 1989 24. Ghysen J, De Plaen JF, Van Ypersele de Strihou C: The effect of membrane characteristics on tumour necrosis
MEGE ET AL
factor Kinetics during haemodialysis. Nephrol Dial Transplant 5:270-274, 1990 25. Whicher J, Ingham E: Cytokine measurements in body fluids. Eur Cytokine Net 1:239-243, 1990 26. Munoz C, Misset B, Fitting C, et al: Dissociation between plasma and monocyte-associated cytokines during sepsis. Eur J Immunol 21:2177-2184, 1991 27. Baurmeister U, Vienken J, Ansorge W, et al: Cellulosic versus synthetic membranes: a reasonable comparison. Artif Organs 13:52-57, 1989 28. Baurmeister V, Travers M, Vienken J, et al: Dialysate contamination and back filtration may limit the use of high flux dialysis membranes. ASIAO Trans 35:519-522, 1989 29. Klinkmann H, Falkenhagen D, Smollich BP: Investigation of the permeability of highly permeable polysulfone membranes for pyrogens. Contrib NephroI46:174-183, 1985 30. Yamagami S, Adachi T, Sugimura T, et al: Detection of endotoxin antibodies against bacteria derived from the dialysate in hemodialysis patients. Nephrol Dial Transplant 4: 489-496, 1989 31. Port FK, Van De Kherkhove KM, Kunkel SL, et al: The role of dialysate in the stimulation of interleukin-I production during clinical hemodialysis. Am J Kidney Dis 10: 118-122, 1987 32. Ward RA, Schmidt B, Blumenstein M, et a1: Evaluation of phagocytic cell function in an ex vivo model of hemodialysis. Kidney Int 37:776-782, 1990 33. Pearson FC, Dubczak J, Weary M, et al: Determination of endotoxin levels and their impact on interleukin-I generation in continuous ambulatory peritoneal dialysis and hemodialysis. Blood Purif 6:207-212, 1988 34. Haeffner-Cavaillon N, Cavaillon JM, Ciancioni C, et al: In vivo induction of interleukin-I during hemodialysis. Kidney Int 35:1212-1218, 1989 35. Bommer J, Becker KP, Urbaschek R, et a1: No evidence for endotoxin transfer across high flux polysulfone membranes. Clin Nephrol 27:278-282, 1987 36. Maliszewski CR: CDI4 and immune response to lipopolysaccharide. Science 252: 1321-1322, 1991 37. Bazil V, Strominger JL: Shedding as a mechanism of downmodulation ofCDI4 on stimulated human monocytes. J Immunol 147:1567-1574, 1991
© 1992 by the National Kidney Foundation, Inc. INDEX WORDS: Tumor necrosis factor-a; endotoxins; monocytes; polyacrylonitrile; CD14 antigen.
H
EMODIALYSIS is associated with several side effects, induding reversible neutropenia and monocytopenia, increased susceptibility to infections, amyloidosis, carpal-tunnel syndrome, and osteopenia.I- 3 Although the activation of the complement system may partially account for deleterious effects of hemodialysis, the involvement of other components such as cytokines is now largely suspected. 4,5 It is established that the overproduction of cytokines may favor ,82-microglobulin (,82m) production and ,82 amyloidosis. 6 ,7 Different studies have reported an abnormal production of interleukin (IL)-l and more recently tumor necrosis factor-a (TNFa) in chronic dialysis patients.5,8,9 Moreover, we have recently observed sustained and chronic monocyte production of TNFa, IL-l,8, and IL6, which was independent of complement activation.1O But the factor(s) responsible for the enhanced cytokine production and the influence of the type of membrane is not well understood. 11-l3 Indeed, high-flux dialysis membranes are increasingly used because they achieve a significant clearance of ,82m and seem to prevent amyloidFrom the Laboratoire d 'Immun%gie, Hopita/ de SainteMarguerite, Marseille; and the Service de Nephr%gie et d'Hemodia/yse, Hopita/ de /a Conception, Marseille, France. Received March 6, 1992; accepted in revised/orm June 30, 1992. Correspondence to: Dr. Jean-Louis Mege, Laboratoire d 'Immunologie, Hopita/ de Sainte-Marguerite, 270 Bd de Sainte-Marguerite, 13277 Marseille Cedex 9, France © 1992 by the National Kidney Foundation, Inc. 0272-6386/92/2005-0006$3.00/0 482
osis. 7 Nevertheless, since their introduction, several deleterious effects of these high-flux membranes, such as anaphylactoid reactions l 4-16 or destructive arthropathy,17 have been described. In addition, an enhanced production of IL-l and TNFa was found in some predialysis patients dialyzed with high-permeability membranes. 5 ,9,18 As these membranes do not activate the complement pathway, it is likely that another mechanism related to a possible transmembrane passage of endotoxins may account for the stimulation of cytokine synthesis. 18 Our study was undertaken to evaluate TNFa production by monocytes and the presence of circulating endotoxins in patients dialyzed with polyacrylonitrile (PAN), a high-permeability membrane. We found an oversecretion of TNFa in these patients, which seems to be related to the presence of circulating endotoxins. MATERIALS AND METHODS
Patients Ten patients consisting of six men and four women (mean age, 55 ± 7 years; range, 38 to 67) were selected. They were clinically asymptomatic with normal clinical and chest x-ray examination. They regularly attended the hospital dialysis unit three times weekly for a 4-hour dialysis session. The mean duration of hemodialysis was 50 months (range, 12 to 114). The underlying diseases consisted of chronic interstitial nephritis (n = 3), renal vascular disease (n = 3). chronic glomerulonephritis (n = 2), and unknown etiology (n = 2). All patients were dialyzed with PAN hollow-fiber hemodialyzerfilter (FILTRAL, 12 Hospal, Lyon, France, AN69 HF; surface area, 1.15 m 2; ultrafiltration coefficient. 32 mL/h . mm Hg). Patients were dialyzed with bicarbonate bath for at least 3 months before any biological investigation. Ten healthy subjects (five men and five women; mean age, 62 ± 6 years; range. 35 to 70) recruited among volunteer
American Journal of Kidney Diseases, Vol XX, No 5 (November), 1992: pp 482-488
TNFa, ENDOTOXINS AND HEMODIALYSIS
blood donors served as controls. They were selected as free from any infection, inflammatory syndrome, or known renal disease. Seven patients (four men and three women, mean age, S9 ± 4 years; range, 4S to 71) dialyzed with Cuprophan hollow-fiber dialyzer (Allegro, Organon Teknica, Fresnes, France; surface area, 1.4 m 2; ultrafiltration coefficient, S mL/ h . mm Hg) were included in the study of circulating endotoxins as controls. The mean duration of hemodialysis was 47 months (range, 12 to 99). Five of the patients were first dialyzed on Cuprophan and then treated with PAN. Two patients were only dialyzed on Cuprophan. Patients were dialyzed with each type of membrane for 3 months before the study.
Cells Blood was drawn in heparinized tubes. Mononuclear cells were separated by Ficoll gradient centrifugation (MSL, Eurobio, Les Ulis, France), as previously described,19 and suspended in RPMI 1640 supplemented with 10% fetal calf serum, antibiotics, and L-g1utamine. Platelet contamination was limited by using low-speed centrifugation, and the percentage of polymorphonuclear neutrophils contaminating mononuclear cells was less than S%. One aliquot of mononuclear cells was saved to measure the ability of endotoxins to elicit TNFa secretion. Then, cells at 2 X 106/mL were incubated in a l-mL volume for 2 hours at 37°C in flat-bottom 24-well culture plates (Nunc, Intermed, France). Nonadherent cells were removed by washing. Adherent cells consisted of approximately 9S% monocytes as assessed by nonspecific esterase staining and were incubated for 18 hours at 37°C in supplemented RPMI 1640.20 Supernatants were collected and stored at -70°C before TNFa determination. The spontaneous secretion of cytokine by monocytes reflects in vivo stimulation of monocytes. 1O We assessed the population of monocytes by microscopic counting of adherent cells (S X 105 cells/well) and measurement of protein concentration by using the Bradford method. No significant variation in protein content was observed between the different assays.
Immunoassay for TNFa Determination TNFa was measured by radioimmunometric assay (IRMA, Medgenix, Eria Diagnostics, Pasteur, France) in serum and monocyte supernatants from controls and dialyzed patients. This assay is based on an oligoclonal system including several monoclonal antibodies directed against distinct epitopes of TNFa and an iodinated antibody that triggers the immunological reaction. The IRMA was sensitive up to S pg/mL TNFa.
Measurement of Endotoxins Endotoxins in water and dialysate baths were determined by using Limulus Amebocyte Lysate (LAL) assay (Whittaker Bioproducts, Walkersville, MD, USA). Briefly, SO ~L of sample or standard was dispensed in microtiter plates at 37°C followed by SO ~L of LAL. Then, substrate solution was added for S minutes and the reaction was stopped with 2S% acetic acid. Changes in absorbance were measured at 40S nm with a Titertek multiscan autoreader. The concentration of endotoxin in a sample was calculated from the absorbance values of solutions containing known amounts of endotoxin standard. Endotoxins in serum from controls and dialyzed patients were measured by modified LAL assay and polymyxin B in-
483 hibition of TNFa secretion by mononuclear cells. First, a slight modification of LAL procedure was introduced as described elsewhere. 21 .22 Serum was first diluted 1:20 with pyrogen-free saline and heated to 100°C for 10 minutes to reduce the effects of nonspecific inhibitors. 22 Lipopolysaccharide (LPS; OS S BS Escherichia coli, Sigma, St Louis, MO) was diluted in control serum to assess the putative interference of serum with LAL. Second, it has been shown that some LAL-reactive materials are not related to endotoxins, may consist of cellulose derivatives, and thus may be unable to elicit cytokine secretion. 23 We used a complementary procedure that measures the ability of serum containing endotoxins to stimulate TNFa secretion in mononuclear cells. 23 The inhibition of serum-induced TNFa secretion by polymyxin B, known to block the effects of endotoxins, confirms the presence of endotoxin materials. Autologous serum from controls and dialyzed patients was incubated with polymyxin B (Sigma) at S ~gfmL or diluent for S minutes before assay. Then, treated sera were added to 2 X 106 mononuclear cells in RPMI 1640 to obtain 10% serum concentration and incubated for 18 hours at 37°C, The supernatants were collected and assayed for the presence of TNFa. Control experiments consisted of LPS (200 ngfmL), which was added to mononuclear cells in RPM I 1640 supplemented with autologous serum and incubated for 18 hours at 37°C in the presence or not of polymyxin B (S ~gfmL).
Surface Antigen Expression Mononuclear cells at 2 X 106/mL were incubated for 30 minutes at 4°C in RPMI 1640 containing 1% bovine serum albumin with different FlTC-tagged monoclonal antibodies or control immunoglobulins. IOMI (CDllb, CR3) and lOP 49d (CDw49d, VLA4) were IgG 1 antibodies (Immunotech, Marseille, France), and MY4 (CDI4) was IgG2b antibody (Coulter). Then, cells were washed and fixed with 1% formaldehyde before being subjected to flow cytometry. Fluorescence staining was assessed using an Epics Profile (Coulter, Coultronics, France) equipped with a IS-mW argon laser emitting at 488 nm. Ten thousand cells were scored each time and the fluorescence results were expressed as percentage of labeled cells and mean fluorescence.
Data Analysis Results are given as the mean ± SEM and were compared using Student's t test. Differences were considered significant at P < O.OS.
RESULTS
Effect of Hemodialysis on Circulating TNFa and TNFa Production by Monocytes Circulating TNFa was slightly but not significantly increased in patients chronically dialyzed on PAN membrane (16 ± 5 pgfmL), as compared with controls (10 ± 5 pg/mL). No variation in TNFa levels was observed during the hemodialysis session (Fig 1). Spontaneous production of TNFa by monocytes was significantly (P < 0.005) enhanced in monocytes from patients chronically treated with
484
MEGE ET AL 1200
SO
1000
40
.§ Cl C.
LL
Z I-
30
E 800 C, c. 600
20
Z I-
LL
10
400 200 0
0
C
0
60 30 Time (min)
C
240
0
60 30 Time (min)
240
Fig 1. Effect of hemodialysis on circulating TNFa. Sera from 10 controls (C) and 10 patients dialyzed on PAN were assayed for the presence of TNFa by using an immunoradiometric assay. Samples were collected before the dialysis session (0) and after 30, 60, and 240 minutes. Results in pg/mL are the means ± SEM.
Fig 2. Effect of hemodialysis on TNFa production by monocytes. Monocytes from 10 controls (C) and 10 dialyzed patients (before dialysis session [0], and after 30, 60, and 240 minutes) were incubated in supplemented RPM I 1640 for 18 hours at 37°C. TNFa was assayed in supernatants by immunoradiometric assay. Results in pg/mL are means ± SEM.
PAN membrane (522 ± 75 pg/mL) as compared with monocyte from controls (190 ± 53 pg/mL) (Fig 2). During the first 30 minutes of the dialysis session, TNFa production remained similar to predialytic values. TNFa production leveled off after 60 minutes and reached a maximum value at the end of the dialysis session (847 ± 94 pg/ mL), a significant (P < 0.05) increase as compared with predialytic values (Fig 2).
with serum from PAN patients, but endotoxin levels were significantly higher than control values (P < 0.01). During the dialysis session, endotoxin levels started to increase after 30 minutes, reached a maximum value at 60 minutes (89.8 ± 10.3 ng/mL), and then declined to a value similar to predialytic levels. Conversely, in patients dialyzed on Cuprophan, endotoxin levels did not significantly vary during the session (Fig 3). The assay determining ifTNFa production by mononuclear cells was inhibited by polymyxin
Measurement of Circulating Endotoxins in Dialyzed Patients
Endotoxins in serum were assessed by two complementary methods, LAL assay and polymyxin B-inhibitable TNFa secretion by mononuclear cells. The former assay detects LAL-reactive materials, the latter is a sensitive test to evaluate the activity of endotoxins. 23 As serum may contain inhibitory compounds that may interfere with LAL assay, serum samples were diluted 1:20 and boiled before being tested. Control serum contained residual amounts of endotoxin (1.8 ± 0.5 ng/mL), which is equivalent to the sensitivity of the assay (data not shown). As shown in Fig 3, before the dialysis session, patients dialyzed with PAN exhibited higher levels of serum endotoxins (41.3 ± 3.1 ng/mL) than did controls (P < 0.005). Serum from patients chronically dialyzed on Cuprophan, known to be a low-flux membrane, contained low amounts of endotoxins (7.9 ± 1.8 ng/mL) as compared
100
E C, c:
(/)
80 60
c: )(
0
0
40
"C
c: w
20 0
o
60 30 Time (min)
240
Fig 3. Endotoxin levels in serum. Sera from 10 patients dialyzed with PAN (0 ) and seven patients dialyzed with Cuprophan (CUP [ ]) were diluted 1:20, boiled, and assayed for the presence of endotoxin by LAL assay. Samples were collected before the dialysis session (0), and after 30, 60, and 240 minutes of dialysis. Results in ng/mL are means ± SEM.
485
TNFa, ENDOTOXINS AND HEMODIALYSIS
B confirmed the LAL results. First, 5 ~g/mL polymyxin B markedly inhibited TNFa secretion induced by exogenous LPS at 200 ng/mL (Table 1). Second, sera from patients dialyzed on PAN were incubated with mononuclear cells in the presence or absence of polymyxin B. Before the dialysis session, patient serum displayed a significant (P < 0.005) TNFa-inducing activity as compared with controls, which was sensitive to polymyxin B. After 30 and 60 minutes of dialysis, TNFa production elicited by patient serum was similar to that observed before the dialysis session. At the end of the dialysis session, although TNFainducing activity, sensitive to polymyxin B, was significantly decreased as compared with intradialytic values, some TNFa-inducing activity, insensitive to polymyxin B, was still present (Table 1).
50
E C, c
rn
40 30
c
)(
0
0
20
"0 C
w
10 0 0
30 60 Time (min)
240
Fig 4. Endotoxin levels in dialysates. Endotoxins were assessed in bicarbonate dialysate baths by LAL assay before the dialysis session and after 30, 60, and 240 minutes. Results in ng/mL represent the mean of 10 values ± SEM.
Endotoxins in Dialysates
Effect of Hemodialysis on Monocyte Membrane Antigens
Contaminant endotoxins were assessed in water and dialysate baths by LAL assay. The levels of endotoxins in purified water were 2.0 ± 0.3 ng/mL. As shown in Fig 4, before the dialysis session, bicarbonate dialysates contained significantly more endotoxins than purified water (22.7 ± 7.5 ng/mL, P < 0.02). During the session, the level of endotoxins in dialysates decreased steadily to reach a minimal value after 240 minutes (10.7 ± 4.2 ng/mL). The composition of dialysate (acetate versus bicarbonate) did not modify the degree of endotoxin contamination (data not shown).
Endotoxins may bind some membrane antigens expressed on monocytes, which could modulate their expression. The expression of CD 14 antigen, which is recognized as one of the binding sites for endotoxins, is shown in Table 2. The percentage of monocytes expressing CD 14 was similar in patients before the dialysis session and in controls. This percentage did not vary during the first 60 minutes of the session, but significantly (P < 0.02) decreased at the end of the session. The decrease involved both the number of labeled cells and the intensity of the fluorescence, thus suggesting that CD 14 was partially down-
Table 1. Effect of Serum on TN Fa Production by Mononuclear Cells Dialysis Time (min)
PBPB+
Control
LPS
o
30
60
240
332 ± 50 405 ± 70
1,606 ± 64 592 ± 51
929 ± 79* 512 ± 81
869 ± 129* 471 ± 191
1,109 ± 181* 449 ± 130
591 ± 36*:j:§ 660 ± 24t
NOTE. Control mononuclear cells (2 x 106 /mL) were suspended in RPMI 1640 containing 10% serum from controls and patients dialyzed after 0,30,60, and 240 minutes of dialysis on PAN. Then, mononuclear cells were incubated for 18 hours at 37°C in the presence (PB+) or the absence (PB-) of polymyxin B (5ILg/mL). In some experiments, LPS (200 ng/mL) was added to mononuclear cells in RPMI 1640 supplemented with 10% normal serum as described. Supernatants were removed and TNFa was measured by IRMA. Results are expressed in pg/mL and represent the mean ± SE of five distinct controls and patients at different dialysis times. * p < 0.005, tP < 0.01 corresponded to the comparison between controls and patients at different dialysis times. :j: P < 0.005 corresponded to the comparison between patients before and after dialysis session. § P < 0.02 corresponded to the comparison between patients after the 60- and 240-minute dialysis sessions.
486
MEGE ET AL Table 2. Effect of Hemodialysis on CD14 Antigen Expression Dialysis Time (min)
Percentage Mean
Control
o
30
60
240
10.7±1.5 55.6 ± 2.1
9.8 ± 1.0 48.0 ± 2.5
10.8 ± 1.2 50.5 ± 4.2
11.2±1.7 50.3 ± 4.1
6.3 ± 0.6' 39.5 ± 2.1t
NOTE. Mononuclear cells (2 X 106 /mL) were incubated with FITC-Iabeled 10M2 monoclonal antibodies directed against CD14 for 30 minutes at 4°C. After washing, they were subjected to flow cytometry. Results are expressed as the percentage of positive cells and the mean fluorescence intensity. They represent the mean ± SEM of 10 patients and 10 controls. 'P < 0.02. tP < 0.005.
modulated in monocytes during PAN hemodialysis. White blood cell count remained constant during the dialysis session and the expression of other monocyte membrane antigens such as CDllb/CDI8 and VLA4 was not affected (data not shown). DISCUSSION
High-flux membranes, such as PAN or polysulfone, improve the treatment of chronic renal failure, but their high permeability to compounds such as endotoxins may limit their use. As endotoxins stimulate in vitro TNFa production by monocytes, we studied TNFa secretion and circulating endotoxins in patients dialyzed with PAN membrane. Monocytes from these patients exhibited a marked spontaneous secretion of TNFa. This was evident before the dialysis session and corresponded to a chronic effect of hemodialysis, thus strengthening the results of Chollet-Martin et al obtained with polysulfone membranes. 9 We also observed an increase in TNFa secretion during the dialysis session as compared with predialytic values. This intradialytic effect of hemodialysis was highest at the end of the session. The in vitro oversecretion of TNFa was not associated with an increase in circulating TNFa, thus confirming previous findings obtained with PAN and polysulfone. 8 ,9,24 The in vivo half-life of cytokines, the existence of numerous specific high-affinity receptors, and the relatively poor sensitivity of immunoassays for cytokines could account for the lack of detection of TNFa in serum. 25 ,26 The main mechanisms thought to playa role in the oversecretion of TNFa by monocytes are the activation of the complement and the presence of circulating endotoxins. The first mecha-
nism is unlikely, since PAN is a poor activator of complement pathway.27 We found no modification of hemolytic activity of complement and its components such as C3, C4, and B during the hemodialysis session, regardless of the time of sampling (data not shown). The transient activation of complement cannot account for the chronic effect of hemodialysis on TNFa secretion we observed. In addition, we have previously shown that TNFa production was not increased in patients chronically dialyzed with Cuprophan, which is known to stimulate complement pathway.iO The second mechanism is based on the presence of endotoxins in the bloodstream, which might account for the overproduction of TNFa, but it supposes that endotoxins contained in dialysates have crossed the hemodialysis membrane. We found that bicarbonate dialysate baths contained significant amounts of endotoxins, which steadily decreased during the dialysis session. Numerous reports favor the hypothesis of a transmembrane passage of endotoxins. To a greater extent than low-flux membranes, highflux membranes offer the possibility of backtransport of dialysate contaminants into bloodstream. 28 ,29 Serum from patients treated with high-flux membranes contained more antibodies against endotoxins than those from patients treated with low-flux membranes. 3o IL-l release was more increased in patients dialyzed with contaminated dialysates than in patients treated with dialysates devoid of endotoxins. 31 IL-l secretion was not elicited in an ex vivo system consisting of pyrogen-free dialysate and polysulfone membrane. 32 The addition of labeled LPS to the dialysate compartment (separated from the blood compartment by PAN or polysulfone mem-
487
TNFa, ENDOTOXINS AND HEMODIALYSIS
branes) resulted in the recovery of a component able to induce IL-1 in the blood compartment. 18 Our results appear to support the role of a transmembrane passage of endotoxins through PAN membranes in TNFa secretion. Indeed, circulating endotoxins were found in patients before the PAN dialysis session. They markedly increased after 60 minutes of dialysis and decreased thereafter. Since we cannot exclude that some circulating LAL-reactive material not related to endotoxins may be present during dialysis session,33 we used a complementary method based on polymyxin B-inhibitable cytokine secretion. Indeed, serum from patients dialyzed with PAN displayed significant ability to induce TNFa secretion by mononuclear cells in a polymyxinsensitive manner. Nonendotoxinic material was probably present chronically and during the dialysis session, since serum from PAN patients was able to induce TNFa secretion in the absence of polymyxin B, as compared with healthy controls. Moreover, serum from patients dialyzed with Cuprophan, a low-permeable membrane, contained only residual amounts of endotoxins, suggesting that some endotoxins may cross Cuprophan membranes as previously observed with bacterial LPSs during in vitro dialysis. 18 The presence of endotoxins in both serum and dialysates suggests transmembrane passage. While the main fraction of LPS is present in dialysates as high-molecular weight aggregates,33 low-molecular weight species of LPS are able to cross hemodialysis membranes and to elicit cytokine secretion. 18,34 However, it has been proposed that only a small fraction of endotoxins contained in dialysates would be able to cross PAN membranes, while the main fraction of endotoxins would be adsorbed on these membranes. That hypothesis is strengthened by the high ability of high-flux membranes to bind endotoxins. 18•35 Indeed, it has been shown that the main fraction of exogenous LPS infused in the dialysate compartment of a closed system was recovered on membrane. 18 We hypothesize that large amounts of endotoxins are susceptible to bind PAN membranes and a fraction of these endotoxins can cross PAN membranes to reach circulating flow. In the bloodstream, endotoxins interact with LPS-binding proteins and the complex binds receptor structures such as CD 14 antigen to elicit cytokine synthesis. 36 Recently, it has been shown
that high concentrations ofLPS downmodulated the expression ofCD14 antigen on monocytes in vitro through a shedding process that required several hours. 37 We observed a marked decrease in CD 14 antigen expression on monocytes at the end ofthe dialysis session, when TNFa production was the highest. We suggest that high intradialytic levels of circulating endotoxins resulted in delayed downmodulation of CD 14 antigen. The soluble CD14 molecules may bind endotoxins in excess, thus accounting for their decrease at the end of the session. The chronic presence of circulating endotoxins able to elicit sustained TNFa production gives an insight into the chronic effect of high flux hemodialysis on cytokine synthesis. The use of highflux membranes and sterile dialysates would prevent deleterious effects oflong-term hemodialysis.
REFERENCES 1. Burton BT, Krueger NK, Bryan FA: National registry of long term dialysis patients. JAMA 218:718-722,1971 2. Lewis SL, Van Epps DE: Neutrophil and monocyte alterations in chronic dialysis patients. Am J Kidney Dis 5:381395, 1987 3. Drueke TB, Zingraff J, Noel LH, et al: Amyloidosis and dialysis: Pathophysiological aspects. Contrib Nephrol 62:6066, 1988 4. Jacobs AA, Ward RA, Wellhausen SR, et al: Polymorphonuclear leukocyte function during hemodialysis: Relationship to complement activation. Nephron 52: 119-124, 1989 5. Haeffner-Cavaillon N, Fisher E, Bacle F, et al: Complement activation and induction of interleukin-1 production during hemodialysis. Contrib Nephrol 62:86-98, 1988 6. Zaoui PM, Stone WJ, Hakim RM: Effects of dialysis membranes on betarmicroglobulin production and cellular expression. Kidney Int 38:962-968, 1990 7. O'Donoghue DJ, Plant WD, Winney RJ: Dialysis amyloidosis. Lancet 2:821, 1991 8. Herbelin A, Nguyen AT, Zingraff J, et al: Influence of uremia and hemodialysis on circulating interleukin-1 and tumor necrosis factor a. Kidney Int 37:116-125, 1990 9. Chollet-Martin S, Stamatakis G, Bailly S, et al: Induction of tumour necrosis factor-alpha during haemodialysis. Influence of the membrane type. Clin Exp Immunol 83:329-333, 1990 10. Mege JL, Olmer M, Purgus R, et al: Haemodialysis membranes modulate chronically the production of TNFa, IL1~ and IL6. Nephrol Dial Transplant 6:868-875, 1991 11. Amato M, Cozzolino F, Bergesio F, et al: In vitro interleukin-1 production by different dialysis membranes. Nephrol Dial Transplant 3:432-434, 1988 12. Shaldon S, Dinarello CA, Elie M, et al: Interleukin-l and bioincompatibility in hemodialysis. Contrib Nephrol 62: 128-131, 1988
488 13. Bingel M, Lonneman G, Koch KM, et al: Plasma interleukin-I activity during hemodialysis: the influence of dialysis membranes. Nephron 50:273-276, 1988 14. Verresen L, Waer M, Van Renterghem Y, et al: Angiotensin-converting-enzyme inhibitors and anaphylactoid reactions to high-flux membrane dialysis. Lancet 2: 1360-1362, 1990 15. Jadoul M, Struyven J, Stagier A, et al: Angiotensinconverting-enzyme inhibitors and anaphylactoid reactions to high-flux membrane dialysis. Lancet 337: 112, 1991 16. Alvarez-Lara MA, Martin-Malo A, Espinosa M, et al: ACE inhibitors and anaphylactoid reactions to high flux membrane dialysis. Lancet 337:370, 1991 17. Kessler M, Netter P, Maheut H, et al: Highly permeable and biocompatible membranes and prevalence of dialysis-associated arthropathy. Lancet 337: 1092-1093, 1991 18. Laude-Sharp M, Caroff M, Simard L, et al: Induction ofIL-1 during hemodialysis: Transmembrane passage of intact endotoxins (LPS). Kidney Int 38: 1089-1094, 1990 19. Grob JJ, Mege JL, Capo C, et al: The role ofTNF in Sneddon-Wilkinson subcorneal pustulosis: A model of neutrophil priming in vivo. J Am Acad Dermatol 25:944-947, 1991 20. Sanguedolce V, Capo C, Bongrand P, et al: Zymosanstimulated tumor necrosis factor alpha production by human monocytes: Downmodulation by phorbol ester. J Immunol 148:2229-2236, 1992 21. Wakebayashi G, Gelfand JA, Jung WK, et al: Staphylococcus epidermidis induces complement activation, tumor necrosis factor and interleukin-I, a shock-like state and tissue injury in rabbits without endotoxinemia. J Clin Invest 87: 1925-1935, 1991 22. Cohen J, McConnel JS: Observation on the measurement and evaluation of endotoxemia by a quantitative limulus lysate microassay. J Infect Dis 150:916-924, 1984 23. Schindler R, Dinarello CA: A method for removing interleukin-I and tumor necrosis factor-inducing substances from bacterial cultures by ultrafiltration with polysulfone. J Immunol Methods 116:159-165, 1989 24. Ghysen J, De Plaen JF, Van Ypersele de Strihou C: The effect of membrane characteristics on tumour necrosis
MEGE ET AL
factor Kinetics during haemodialysis. Nephrol Dial Transplant 5:270-274, 1990 25. Whicher J, Ingham E: Cytokine measurements in body fluids. Eur Cytokine Net 1:239-243, 1990 26. Munoz C, Misset B, Fitting C, et al: Dissociation between plasma and monocyte-associated cytokines during sepsis. Eur J Immunol 21:2177-2184, 1991 27. Baurmeister U, Vienken J, Ansorge W, et al: Cellulosic versus synthetic membranes: a reasonable comparison. Artif Organs 13:52-57, 1989 28. Baurmeister V, Travers M, Vienken J, et al: Dialysate contamination and back filtration may limit the use of high flux dialysis membranes. ASIAO Trans 35:519-522, 1989 29. Klinkmann H, Falkenhagen D, Smollich BP: Investigation of the permeability of highly permeable polysulfone membranes for pyrogens. Contrib NephroI46:174-183, 1985 30. Yamagami S, Adachi T, Sugimura T, et al: Detection of endotoxin antibodies against bacteria derived from the dialysate in hemodialysis patients. Nephrol Dial Transplant 4: 489-496, 1989 31. Port FK, Van De Kherkhove KM, Kunkel SL, et al: The role of dialysate in the stimulation of interleukin-I production during clinical hemodialysis. Am J Kidney Dis 10: 118-122, 1987 32. Ward RA, Schmidt B, Blumenstein M, et a1: Evaluation of phagocytic cell function in an ex vivo model of hemodialysis. Kidney Int 37:776-782, 1990 33. Pearson FC, Dubczak J, Weary M, et al: Determination of endotoxin levels and their impact on interleukin-I generation in continuous ambulatory peritoneal dialysis and hemodialysis. Blood Purif 6:207-212, 1988 34. Haeffner-Cavaillon N, Cavaillon JM, Ciancioni C, et al: In vivo induction of interleukin-I during hemodialysis. Kidney Int 35:1212-1218, 1989 35. Bommer J, Becker KP, Urbaschek R, et a1: No evidence for endotoxin transfer across high flux polysulfone membranes. Clin Nephrol 27:278-282, 1987 36. Maliszewski CR: CDI4 and immune response to lipopolysaccharide. Science 252: 1321-1322, 1991 37. Bazil V, Strominger JL: Shedding as a mechanism of downmodulation ofCDI4 on stimulated human monocytes. J Immunol 147:1567-1574, 1991
