The liver contains the largest pool of cytokine-producing mac- rophages in the body and may therefore play an important role in the development and outcome of ...
First publ. in: The Journal of Pharmacology and Experimental Therapeutics 276 (1996), 3, pp. 968-976
Tumor Necrosis Factor Production in the Perfused Mouse Liver and Its Pharmacological Modulation by Methylxanthines 1,2 MARC EL LEIST, SIGRID AUER-BARTH and ALBRECHT WENDEL Biochemical Pharmacology, Faculty of Biology, University of Konstanz, Konstanz, Germany
ABSTRACT
The liver contains the largest pool of cytokine-producing macrophages in the body and may therefore play an important role in the development and outcome of systemic inflammatory response syndromes. Therefore, we investigated the tumor necrosis factor-a (TNF) releasing capacity of the in situ perfused mouse liver and its modulation by methylxanthines, i.e., by a class of well-established inflammatory cytokine-suppressing drugs. We have shown that pretreatment of mice with either lipopolysaccharide or TNF elicited a dose-dependent TNF release into the perfusate which was inhibited by in vivo pretreatment of mice with pentoxifylline or A-802715 [1-(5-hydroxy-5methyl)hexyl-3-methyl-7 -propylxanthin]. Infusion of these
methylxanthines into livers from mice pretreated with lipopolysaccharide or TNF also inhibited TNF release in an immediate and reversible way even after TNF production had been initiated. The inhibitory effect of methylxanthines was prevented by pretreatment of mice with the adenylate cyclase inhibitor dideoxyadenosine, suggesting upregulation of the cyclic adenosine monophosphate system as a possible mechanism of action of these drugs_ Our findings demonstrate that the liver is a potent cytokine producer and identify it as one of the target organs of methylxanthines or other phosphodiesterase inhibitors in murine models of shock and inflammatory liver failure.
TNF is an important immunomodulator, the excessive production of which has been associated with lethality in LPS shock (Beutler et al., 1985), cerebral malaria (McGuire et al., 1994; Grau et al., 1987) and other systemic inflammatory response syndromes (Tracey and Cerami, 1993; Ozmen et al., 1994). It has been identified as a necessary (Lehmann et al., 1987) and terminal (Tiegs et al., 1989) mediator of hepatic apoptosis and subsequent fulminant liver failure in different mu shock models. These pathological actions of TNF are observed both in nonsensitized mice challenged with the T-cell mitogen concanavalin A (Gantner et al., 1995b) and in GalN-sensitized mice stimulated with the polyclonal T-cell activators anti-CD3, staphylococcal enterotoxin B (Gantner et al., 1995a) or the macrophage stimulator LPS (Leist et al., 1995a). Under various experimental conditions, TNF may be released from a variety of cells, including hepatocytes (Gonza-lez-Amaro et al., 1994), T-cells (Cuturi et al., 1987), neutrophils (Cassatella, 1995) and macrophages/monocytes (Tracey and Cerami, 1993). The most potent producers are macro-
phages_ This cell population was suggested to be responsible for systemic LPS effects (Rosenstreich and Vogel, 1980) and was subsequently shown to be both necessary and sufficient to cause liver damage in GalN-sensitized mice after LPS injection (Chojkier and Fierer, 1985; Freudenberg et al., 1986)_ Although Kupffer cells have been found to be involved in the development of LPS-induced hepatocytotoxicity in vivo and in vitro (Shiratori et al., 1988; Hartung and Wendel, 1991; Shibayama et al., 1993), a study using the CAT-reporter gene assay was unable to demonstrate hepatic TNF synthesis (Giroir et al., 1992). On the other hand, TNF production in vitro (Reinstein et al., 1994; Hansen et al., 1994) or mRNA accumulation in vitro (Hansen et al., 1994) or in vivo (Nanji et al., 1994; Shirwan et al., 1994) have been shown repeatedly in rodent Kupffer cells (Kark et al., 1988; Decker et al., 1989; Decker, 1990). Neither the regulation of LPS-induced TNF release from macrophages nor that of autocrine stimulation by TNF itself are understood completely. Released bioactivity of TNF is determined by various processes like mRNA induction, an alteration of translational efficiency or mRNA stability, and finally by proteolytic cleavage of the 26 kDa membraneanchored TNF precursor to yield the 17 kDa soluble TNF (Mohler et al., 1994). LPS-induced TNF release from macrophages in vitro or mice in vivo is inhibited completely by
Received for publication June 29, 1995. 1 This study was part of the M.D. thesis of S. A.-B. 2 This paper is dedicated to Professor Ernst Mutscheler, Frankfurt, on the occasion of his 65th birthday.
ABBREVIATIONS: TNF, tumor necrosis factor-a; LPS, lipopolysaccharide; mu, murine; GaiN, D-galactosamine; cAMP, cyclic AMP; PDE,
phosphodiesterase; POF, pentoxifylline; hu, human; BST, bromosulphothalein; LDH, lactate dehydrogenase; cGMP, cyclic GMP; DDA, 2'3'dideoxyadenosine.
968
Konstanzer Online-Publikations-System (KOPS) URN: http://nbn-resolving.de/urn:nbn:de:bsz:352-opus-82547 URL: http://kops.ub.uni-konstanz.de/volltexte/2009/8254/
969
1996
pretreatment with low doses of LPS, which leads to a refractory state of the producing cell (Mengozzi and Ghezzi, 1993). Inhibition of TNF production, predominantly on the transcriptional level (Strieter et al., 1988; Kunkel et al., 1988; Han et al., 1990), has been observed after treatment of isolated macrophages or mice with cAMP elevating drugs (Renz et al., 1988) such as adenosine (Parmely et al., 1993; Reinstein et al., 1994), prostaglandin E (Kark et al., 1988; Endres et al., 1991), prostacyclin analogs (Crutchley et al., 1994; Sinha et al., 1995) or PDE inhibitors (Fischer et al., 1993; Reinstein et al., 1994). The effect of the PDE inhibitors was increased synergistically by coadministration of cAMP-raising agonists (Crutchley et al., 1994; Sinha et al., 1995). In some experimental systems, cAMP-increasing treatment was still effective after the initiation of TNF production by LPS (Schonharting and Schade, 1989; Prabhakar et al., 1994). Among the methylxanthines, POF has been especially well examined. This drug was shown to inhibit TNF release in mice and in humans in vivo (Zabel et al., 1989; Dezube et al., 1993), in vitro (Endres et al., 1991; Schandene et al., 1992; Semmler et al., 1993) and ex vivo (Neuner et al., 1994). This inhibitory capacity has been closely associated with the protective action of this drug and its congeners (Schade, 1989; Niehorster et al., 1992; Bahrami et al., 1992) against organ damage and lethality in systemic inflammatory response syndromes (SchOnharting and Schade, 1989; Harada et al., 1989; Coccia et al., 1989; Alegre et al., 1991; Kremsner et al., 1991). The present study was designed to investigate the capacity of the liver to release bioactive TNF when stimulated with LPS or TNF in vivo under similar conditions to those used in mu models of inflammatory hepatic failure. In addition, in these two models of ex vivo-stimulated TNF release, we studied the pharmacological mode of action of two different methylxanthines that have been demonstrated to prevent liver failure in mice in vivo. The data presented suggest that the liver is not only the target organ ofTNF toxicity, but possibly also an important and pharmacologically accessible source of this mediator.
Materials and Methods Materials Salmonella abortus equi LPS was purchased from Sigma Chemical Co. (Deisenhofen, Germany). TNF was a generous gift from Dr. G. R. Adolph, Bender & Co. (Vienna, Austria) and had an endotoxin content of < 10 pg/lLg according to a limulus assay (Coatest Endotoxin, Chromogenix AB, Molndal, Sweden). Cell culture media were from Biochrom (Berlin, Germany). GaIN was from Roth Chemicals (Karlsruhe, Germany). TNF antiserum was prepared in our laboratories by S. Jilg as described previously (Leist et al., 1995b). POF and A-802715 [1-(5-hydroxy-5-methyl)hexyl-3-methyl-7-propylxanthine] were a gift from Dr. M. Schtinharting, Hoechst AG (Wiesbaden, Germany). All other reagents not further specified were from Sigma Chemical Co. (Deisenhofen, Germany) or Merck (Darmstadt, Germany) Animals and treatment schedules. Male NMRI mice (30 g) were from the Hannover'sches Zentralinstitut (Hannover, Germany). They were kept for at least 1 week at 22°C, 55% humidity and a 12-hr day/night rhythm with free access to food and water. All animals received hu care in compliance with legal requirements (Regierungsbezirk; Freiburg, Germany). Before the start of the liver perfusion, animals were injected i.v. (200 ILl) or i.p. (300 ILl) with
different inflammogens or pharmacological modulators. Ten minutes before the start of the perfusion, mice were anesthesized with 100 mg/kg of phenobarbitone (Sanofi, Munich, FRG) and fixed to a temperature-controlled (37°C) operation table. Liver perfusion. Mouse livers were perfused in situ in a nonrecirculating system essentially as described previously (Hartung et al., 1991). After laparotomy, the gall bladder was incised and the portal vein cannulated with a 20-gauge Teflon cannula. The diaphragm was severed, the anterior vena cava ligated and the perfusate was collected from a tube connected via a Teflon cannula to the posterior vena cava. This surgical procedure took about 5 min and any period of ischemia was strictly avoided. Perfusate flow was adjusted to 3 ml/min/g of liver based on an experimentally determined average relative liver weight of 4.75 ± 0.34% of b.wt. The buffer was a modified Krebs-Henseleit buffer (millimolar: Na+, 144; K+, 5.9; Ca++, 1.2; Mg++, 1.2; Cl-, 130; HCO 18; SOl-, 1.2; and phosphate, 1.2) supplemented with 2.1 mM lactate, 0.3 mM pyruvate and sometimes 1.5 ILM BST. After gassing with 5% CO 2 -95% O2 , the pH was 7.4 and oxygen concentration 0.95 mM at 37°C. Oxygen concentration and pH were measured continuously in the effluent perfusate. Samples for biochemical determinations were taken by a fraction collector (Pharmacia, Freiburg, Germany) at 2.5-min intervals. Additions to the basic buffer were made by a special infusion pump (Braun-Melsungen AG, Melsungen, Germany) connected to a T-junction just before the entry into the liver. Viability parameters. Viability and function of the liver was determined by various methods. Routinely, LDH release into the buffer was measured on an automatic Eppendorff ACP 5040 enzyme analyzer (Netheler und Hinz, Hamburg, Germany) according to Bergmeyer (1984). BST clearance was determined by measuring the residual BST concentration in the effluent buffer at 578 nm after alkalization with NaOH. The effluent pH of 7.23, the oxygen consumption of 2.7 ILmol/g/min and the perfusion pressure of 9 cm of H 2 0 remained constant over a period of 180 min in control perfusions
a,
(n = 9).
TNF determination. Samples for TNF determinations were frozen immediately at -20°C until analysis in the WEHI 164 clone 13 bioassay (Espevik and Nissen-Meyer, 1986) This assay was performed using recombinant muTNF as a standard and the specificity for muTNF was checked routinely by coincubation with a neutralizing anti-muTNF antibody. The detection limit was 10 pg/m!. The assay was sensitive to both huTNF and mTNF. The activity of muTNF was blocked completely by anti-muTNF antibodies. These did not interfere with the bioactivity of huTNF. Therefore, the amount of huTNF in samples containing both huTNF or muTNF could be determined by running parallel assays in the presence or absence of anti-muTNF antibody. The cumulative amount of TNF released into the perfusate was determined by integration of the TNF concentration in the buffer over time using a spreadsheet calculation software (QPRO, Borland, MA). TNF release over time was usually constant, i.e., cumulative TNF increased linearly under different experimental conditions. Therefore, with the exception of figures 4 and 5, in which TNF release was not constant over time, we show, in the other diagrams, the release ofTNF over the first 70 min. PDE inhibition. PDE activity was determined with some modifications (Bauer and Schwabe, 1980) as described by Thompson and Appleman (1979). PDE I [Ca++ (1 mM)/calmodulin (100 nM)-dependent] from bovine brain was kindly provided by Dr. Gietzen (Ulm, Germany). PDE 11 [cGMP-stimulated (5 !LM)] was purified from rat heart (Schudt et al., 1991b). PDE III (cGMP-inhibited) and PDE V (cGMP-specific) were assayed in homogenates of human platelets as described by Schudt et al. (1991a). PDE IV (cAMP-specific) was tested in the cytosol ofhu polymorphonuclear cells in the presence of the PDE Ill-specific inhibitor motapizone (1 ILM) as described by Schudt et al. (1991a). IC 50 values were calculated from concentration-inhibition curves by nonlinear regression analysis using the GraphPad software (Graphpad Inc., Sorrento Valley, CA).
970 Statistics. Experiments were performed in triplicate. Perfusions belonging to one series of experiments were performed within 1 day or on subsequent days. Control perfusions were repeated for every individual set of experiments, because variations of the absolute inducible TNF release were noticed between different sets of experiments. Concentration-response curves were fitted by nonlinear regression analysis using the inplot 4.0 software (Graphpad Inc.). When appropriate, the Student's t test or Welch test (for inhomogeneous variances) were used for significance analysis
Results Hepatotoxicity of pretreatment schedules. First of all we checked the basic properties of our perfusion system, in particular whether the pretreatment protocols chosen affected hepatocellular integrity. The LDH release from untreated mouse livers during the first 90 min of perfusion was about 150 mU/g of liver per hr (table 1). A similar LDH release was observed after pretreatment with any single substance used in this study, except GaIN. This latter compound caused an overall LDH release of about 300 mU/g of liver per hr when given alone or in combination with either LPS or TNF 60 min before the start of the perfusion. As damage controls, we used 5 mM hydrogen peroxide or 60 min noflow ischemia, both of which caused an LDH releases of 2,000 to 25,000 mU/g ofliver per hr. Infusion of 1 mM Triton X-lOO released 65,000 mU of LDH per g of liver per hr into the perfusate. From these data we conclude that our ex vivo treatment schedule using lactate-pyruvate supplemented buffer and livers from fed mice caused only negligible toxicity and was therefore better suited for the following pharmacological studies than a previously used perfusion system without buffer supplements and livers from starved mice (Wang and Wendel, 1990). Concordant results on hepatic integrity were obtained when BST clearance was determined as a measure of liver function (table 1). GaIN or combinations of LPS or TNF with GaIN caused a reduction of the hepatic extraction of BST from 92 ± 2 to 80 ± 4% after 70 min. All other substances had no significant effect. Such a slightly reduced BST clearance capacity of the liver may.be associated with a reduced clearance of TNF by the liver. To check this we infused 75 ng/ml ofTNF (a concentration that may be TABLE 1 Viability and ex vivo TNF release in perfused mouse livers from mice pretreated in vivo Treatmenta
LDH release b % of tota/liver LDH170 min
Saline GaiN (700 mg/kg) LPS (30 f.Lg/kg) GalN/LPS huTNF (30 f.Lg/kg) GalN/huTNF muTNF (30 f.Lg/kg) GalN/muTNF
0.055 0.103 0.064 0.082 0.073 0.113 0.076 0.111
± 0.007 ± 0.007
± 0.007
± 0.008 ± 0.011 ± 0.009
± 0.009
± 0.006
BST clearance c
%
92 ± 1 79 ± 2 89 ± 3 81 ± 5 N.D.e ND. 93 ± 4 79 ± 2
TNF release nglg liverlh n.d.d
n.d. 15 ± 26 ± 11 ± 22 ± 48 ± 151 ±
2 4 3 1 8 39
a Mice were injected 60 min before the start of the perfusion. Data are means ::t S.E.M. from triplicate determinations. b Total LDH content of liver homogenate was 275 ::t 20 U/9 of liver. LDH release from perfused livers was determined at 5-min intervals and the total amount was obtained by integration over 70 min. e BST clearance was measured 70 min after the start of the perfusion as the ratio of BST concentration before and after the liver. d n.d., not detectable. e N.D., not determined.
found in vivo) for 10 min into livers from mice pretreated with GaIN or solvent control. In both groups, the effluent TNF concentration was < 10 pg/ml. This finding suggests that the liver has a very high TNF -inactivating capacity that is not influenced by GaIN treatment. The role of GaIN in TNF release from the perfused mu liver. Because the liver contains the greatest number of macrophages of any organ in the body and may therefore account for a major part of circulating cytokines, we examined whether the release of the inflammatory cytokine TNF from perfused mouse livers can be determined quantitatively in a blood-free perfusion system. Stimulation ofTNF release during the first 90 min was not observed in vitro after infusion ofLPS or TNF. However, we did observe a spontaneous TNF release starting after about 90 min. When mice were treated in vivo with the inflammogens LPS or TNF 60 min before liver perfusion, a constant and consistent dose-dependent release of TNF into the perfusate was detected. Cotreatment of mice with GaIN, a sensitizer commonly used in mu models of sepsis and inflammatory liver failure, caused an approximately 1.5-fold increase in TNF release with any stimulus chosen (table 1). Because GaIN alone did not cause any significant TNF release and because we focused on the role of liver macrophages in experimental models of hepatic inflammation, we routinely used GaIN-treated mice for all subsequent experiments. Ex vivo induction of TNF release from the perfused mouse liver. Livers from mice pretreated with 2: 10 JLg/kg of LPS released TNF into the perfusate (fig. 1). TNF release was dose-dependent, was observed over a similar dose range to that seen in vivo and, at any given dose, was relatively constant over the first 70 min of perfusion. The amounts of TNF released were comparable to those observed in vivo after injection of LPS in concentrations that cause liver failure (Galanos et al., 1979). In our studies, a maximum dose of 60 JLg/kg of LPS was not exceeded, because this was acutely toxic to GaIN-sensitized mice in vivo. Next we examined whether TNF, as one ofthe main proinflammatory cytokines, might act as an autocrine stimulator 300 ~GaINlLPS
~ ~
.g: .:.
~ GalNlhuTNF
~ GalNlmuTNF
LL
z
I-
"E
'0
GI
1Il
.lI!
f!
o~~----~--~~~~~~~~~
0.0
1.0
dose of LPS or TNF [1Ig/k9 body weight]
Fig. 1. TNF release from the GaiN-sensitized perfused mouse liver elicited by the inflammatory stimuli LPS and TNF ex vivo. Mice were injected 60 min before the start of the perfusion with a combination of 700 mg/kg of GaiN (Lp.) and different doses of either of the three inflammatory stimuli: LPS (Lp.), muTNF (Lv.) or huTNF (Lv.). Subsequently, livers were perfused and TNF was determined in the perfusate at 5-min intervals for 70 min. Data are means ± S.D. from three independent perfusions.
971 of its own production (Philip and Epstein, 1986) in the liver. Pretreatment of mice with 2: 10 /Lg/kg of muTNF in vivo caused a dose-dependent release ofTNF from perfused livers ex vivo (fig. 1). TNF-induced TNF release was relatively constant during the perfusion time, a finding that cannot be explained only by a washout of previously injected TNF. TNF release due to pretreatment with muTNF may be significantly higher than that observed with huTNF, inasmuch as both types of TNF receptors are stimulated by muTNF, but not by huTNF (Leist et al., 1995b). In order to show that the released TNF was actually produced by the liver, we examined whether the injection ofhuTNF would cause the release ofmuTNF. By using an antibody able to selectively neutralize muTNF, the bioactivity of muTNF and huTNF can be determined in parallel in the WEHI 164 bioassay. By using this approach, we found that no huTNF, injected in vivo before the start of the perfusion, was washed out from the perfused liver ex vivo, and that any TNF bioactivity released into the perfusate was due to muTNF. From these experiments, we conclude that either LPS or TNF itself may induce an inflammatory response in the liver resulting in a dosedependent TNF production and release from the isolated liver. In order to exclude the possibility that TNF release into the perfusate was the result of an unspecific stress response due to a liver damage caused by the in vivo pretreatment, we examined whether physiological regulation· similar to that described as tolerance induction (Mengozzi and Ghezzi, 1993; Porat et al., 1993) also could be observed in the isolated organ. Therefore, we examined whether low-dose LPS pretreatment also would induce tolerance in the ex vivo perfusion system. As shown in table 2, GaINILPS- or GalNITNFinduced TNF release, respectively, was completely or strongly reduced under such conditions. These findings suggest that TNF release from the liver is more than a stress response, but rather is dependent on the activation of the Kupffer cells.
Ex vivo effects of methyIxanthines on hepatic TNF release. In order to examine whether methylxanthines had a direct effect on TNF production by the liver, we tested and compared the effects of the established TNF-suppressing drug POF with a more recent congener, the investigational drug A-802715 [1-(5-hydroxy-5-methyl)hexyl-3-methyl-7propylxanthin]. Neither compound caused any significant TNF release when injected into control mice. When mice were pretreated with these compounds before the injection of LPS, a dose-dependent reduction of the TNF release from subsequ.ently perfused livers was observed. The IC so for the in vivo pretreatment was 73 mg/kg for POF or 87 mg/kg for A-802715, respectively. TNF release was completely inhibited in livers from animals pretreated with 200 mg/kg of either compound (fig. 2A). Essentially, similar results were obtained when TNF was used as a stimulus. The IC so values for the inhibition of TNF release from the perfused liver (after in vivo stimulation with TNF) by in vivo pretreatment with methylxanthines were 79 mg/kg for POF and 96 mg/kg for A-802715. Inhibition of TNF-induced TNF release was 80% at a methylxanthine pretreatment dose of 200 mg/kg (fig. 2B). Thus, in vivo pretreatment of mice with methylxanthines 60 min before challenge with LPS or TNF prevented
e'E 0
u
(J)
ll..
80
1000b
Additional Modulation a
PDE 11
PDE III
PDE IV
PDE V
80 130
/L NF 215 210
47 44
85 16D
a Enzymatic activity of isolated isoenzymes and its inhibition by methylxanthines was measured as described under "Materials and Methods." b 30% inhibition at 1000 /LM.
functions, but also all major organs. Kupffer cells (3-9 X lOG/liver) by far outnumber the circulating monocytes (1-1.5 X 106/mouse), i.e., the chiefTNF producers of the blood, and may therefore contribute significantly to systemic TNF concentrations found in experimental models of sepsis. Recent observations actually demonstrate that hepatic TNF production, elicited by hepatic ischemia, is responsible for the development of subsequent pulmonary injury in rats (Colletti et al., 1990). Further indirect evidence for the pathogenic role of hepatic TNF comes from data showing that endotoxin neutralization or the elimination or inactivation of macrophages with agents such as silica, praseodymium nitrate or GdCl3 reduces or prevents hepatic injury elicited by different stimuli (Grful and Liehr, 1976; Nolan and Leibowitz, 1978; Camara et al., 1983; Tiegs et al., 1992; Adachi et al., 1994; Czaja
i
20
,
30
,
40
i
50
i
60
,
70
i
80
Fig. 5. Role of cAMP in the rapid onset of the inhibition of TNF release by methylxanthines: influence of the infusion of DDA with or without dibutyryl-cAMP (db-cAMP). Mice were injected Lp. with 700 mg/kg of GaiN and 30 p,g/kg of LPS 60 min before the start of the perfusion. Thirty minutes before LPS, mice were injected with saline or 200 mg/kg of POF or A-802715 (A-80). Fifteen minutes before the methylxanthine treatment, a group of mice were injected with 1.6 mg/kg of DDA as indicated. Additionally, some livers that received this treatment were infused with 0.1 mM db-cAMP after the initial 20 min of perfusion. Data = means::!: S.D., n = 3.
--GalNffNF - - + A80 (20-50 min) ...... + A80 (0-30 min) ... saline
~ 120 U. 90
30
,
10
time [min]
..:-240
-a- + POF (20-50 min)
i
40
30
time[h]
270
60
E ::s
i
i
10
B
120 U. 90
f!:
i
0
80
~ .2:
0
i
70
time [min]
~180
-0- +A80 + DDA +db-cAMP ... saline
.~ 10
12::s
i'240 :5:en 210
""-+A80+DDA -0- +POF +DDA +db-cAMP
I-
.~ 10
270
..... GalNlLPS .... +POF+DDA
.m u.ze 20
f!:
A
--GalNlLPS - + A80 (20-50 min) - + A80 (0-30 min) -+-saline
I40
..!l!
E ::s ()
60
50
--- + POF (0-30 min) ... saline
974
Treatment Saline GalN/LPS GalN/LPS GalN/LPS GalN/LPS
Pretreatment
Saline a
Saline 0 Saline 100::!: 15 200 mg/kg 0 Pentoxifyllined d 200 mg/kg of A-802715 0 150 mg/kg of db-cAMp d 10 ::!: 4
160/Lg/kg DDAb
16 mg/kg DDAa
0 110 19
N.D.c N.D. 59
17 N.D.
49 N.D.
DDA was given 45 min before LPS. Data are given in percentage of cumulative TNF release compared to mice treated i.p. with 700 mg of GaiN plus 30 /Lg of LPS only 60 min before perfusion (GaIN/LPS). TNF release from perfused mouse livers was measured for 70 min at 5-min intervals to calculate the cumulative amount released. Data are means ± S.D. from triplicate experiments. Data without S.D. are derived from a single experiment. C N.D., not determined. d Methylxanthines or db-cAMP were injected i.p. 30 min before LPS. a
b
et al., 1994; Laskin et al., 1995; Barriault et al., 1995; Ishiyama et al., 1995). Because TNF has been identified as one of the Kupffer cell products capable of inducing fulminant hepatic injury (Lehmann et al., 1987; Tiegs et al., 1989), it may be concluded that Kupffer cell blockade protects the liver primarily by inhibition of hepatic TNF release. We showed that the inducible TNF release observed in our liver perfusion system was subject to regulation by endoge-
nous and exogenous mediators. We therefore believe that neither direct hepatotoxic effects of the stimulators, surgical stress nor the 1- to 3-sec period of noflow ischemia when the perfusion cannula in the portal vein was connected to the perfusion pump account for LPS-stimulated TNF release. The ex vivo tolerance induction in the perfused liver by lowdose LPS is an example of regulation by endogenous mediators. Our findings from these experiments suggest that mu macrophages in situ may be regulated by mechanisms similar to those described in hu monocytes (Porat et al., 1993) or other macrophage populations (Mengozzi and Ghezzi, 1993). Knowing this, it was important to make sure that pharmacological effects of methylxanthines were not due to LPS contamination. Measurement ofLPS in a chromogenic limuIus assay showed that the endotoxin contamination of the methylxanthines was extremely low. The best evidence for pharmacological actions of the methylxanthines unrelated to a possible tolerance induction is, however, the fact that the TNF-reducing effect also was observed when they were given in vivo after the LPS stimulus or even ex vivo, i.e., 1 hr after the initiating LPS injection. Both POF and A-802715 [1-(5-hydroxy-5-methyl)hexyl-3methyl-7-propylxanthinl have been shown to suppress LPSinduced TNF release in vitro (Sinha et al., 1995). Very recently, A-802715 was found to be the most potent among different methylxanthines, being about twice as effective as POF and about 10 times more effective than theophylline (Sinha et al., 1995). Therefore, we thought it may be interesting to test the effects of this novel compound on TNF release from the perfused mouse liver in comparison to the well-characterized TNF-suppressing drug POF. Both substances were found to be of similar efficacy, when mice were pretreated in vivo or when drugs were given 1 hr after the TNF releasing challenge in an ex vivo design directly in the perfusate. This finding held true for two different stimuli, i.e., LPS or TNF. The inhibition of LPS-induced TNF release was observed at about 10-fold lower concentrations of meth-
ylxanthines in the perfusion buffer than was required for the inhibition of TNF-induced TNF release. With the in vivo pretreatment protocol, such a potency difference toward different stimuli was not observed. Another difference between the in vivo and the ex vivo protocol is the reversibility of the inhibition effect by simple washout. Ex vivo, the methylxanthines inhibited TNF release only during the period of time when they were present in the perfusion buffer. When mice had been pretreated with the drugs in vivo, however, the inhibition of TNF release persisted throughout the entire subsequent perfusion period. Several explanations are possible for these observations: 1) they may be explained pharmacokinetically, i.e., by different distribution or by the action of different metabolites; 2) methylxanthines may inhibit TNF release on multiple levels (e.g., transcriptional, translational or post-translational) and different levels of cAMP may be required to inhibit TNF release after different stimuli; and 3) direct cytotoxicity to macrophages may be a further cause of methylxanthine-inhibited TNF release. This latter possibility can, however, be ruled out in our system, inasmuch as inhibited TNF release could be restored by DDA, and because inhibition was reversible after omission of methylxanthine from the perfusion buffer. Further known methylxanthine actions include interaction with adenosine receptors (Parmely et al., 1993). However, in our system we provide experimental evidence that raised cAMP is causally involved in the regulation ofTNF release by methylxanthines. Both substances were shown to inhibit various PDE families and have been shown by others to increase cAMP in macrophages (Semmler et al., 1993). Confirming these results for mu Kupffer cells in situ was not within the scope of this study and measurements of total cellular cAMP may be flawed by the fact that average cAMP increase does not correlate with the inhibition ofTNF release (Sinha et al., 1995) and highly localized concentration changes of even flux changes of cAMP may be sufficient to cause effects such as inhibition ofTNF release (Pryzwansky et al., 1981; Barsony and Marx, 1990; Nicholson et al., 1991). However, we showed that raising cAMP with alternative effectors such as iloprost or dibutyryl-cAMP has similar pharmacological effects as the treatment with methylxanthines. In addition, we checked the causal involvement of cAMP in the suppression of TNF in this system by using DDA. This adenylase inhibitor is a functional antagonist of methylxanthines with respect to their cAMP-raising actions. Indeed we found that DDA antagonized the methylxanthine-induced inhibition of TNF release from the liver. Therefore, we conclude that TNF regulation by POF or A-802715 in mouse liver is caused by an increase of cAMP in the effector cell, i.e.,. the Kupffer cell. Such a suppression of TNF release within the liver itself may explain the prevention of fulminant LPS-induced liver failure by various cAMP-raising substances such as forskolin (Fischer et al., 1993), PDE-inhibitors (Schtinharting and Schade, 1989) or prostacyclin analogs (Wendel et al., 1987) observed in experimental animals. Our study provides evidence that the perfused mouse liver is a system that allows the investigation of the regulation of TNF release within the largest resident macrophage pool in the body. This means that TNF release from Kupffer cells within their natural cellular environment may be examined under conditions of maintained regulation by endogenous or exogenous mediators. As demonstrated by two examples,
969
1996
pretreatment with low doses of LPS, which leads to a refractory state of the producing cell (Mengozzi and Ghezzi, 1993). Inhibition of TNF production, predominantly on the transcriptional level (Strieter et al., 1988; Kunkel et al., 1988; Han et al., 1990), has been observed after treatment of isolated macrophages or mice with cAMP elevating drugs (Renz et al., 1988) such as adenosine (Parmely et al., 1993; Reinstein et al., 1994), prostaglandin E (Kark et al., 1988; Endres et al., 1991), prostacyclin analogs (Crutchley et al., 1994; Sinha et al., 1995) or PDE inhibitors (Fischer et al., 1993; Reinstein et al., 1994). The effect of the PDE inhibitors was increased synergistically by coadministration of cAMP-raising agonists (Crutchley et al., 1994; Sinha et al., 1995). In some experimental systems, cAMP-increasing treatment was still effective after the initiation of TNF production by LPS (Schonharting and Schade, 1989; Prabhakar et al., 1994). Among the methylxanthines, POF has been especially well examined. This drug was shown to inhibit TNF release in mice and in humans in vivo (Zabel et al., 1989; Dezube et al., 1993), in vitro (Endres et al., 1991; Schandene et al., 1992; Semmler et al., 1993) and ex vivo (Neuner et al., 1994). This inhibitory capacity has been closely associated with the protective action of this drug and its congeners (Schade, 1989; Niehorster et al., 1992; Bahrami et al., 1992) against organ damage and lethality in systemic inflammatory response syndromes (SchOnharting and Schade, 1989; Harada et al., 1989; Coccia et al., 1989; Alegre et al., 1991; Kremsner et al., 1991). The present study was designed to investigate the capacity of the liver to release bioactive TNF when stimulated with LPS or TNF in vivo under similar conditions to those used in mu models of inflammatory hepatic failure. In addition, in these two models of ex vivo-stimulated TNF release, we studied the pharmacological mode of action of two different methylxanthines that have been demonstrated to prevent liver failure in mice in vivo. The data presented suggest that the liver is not only the target organ ofTNF toxicity, but possibly also an important and pharmacologically accessible source of this mediator.
Materials and Methods Materials Salmonella abortus equi LPS was purchased from Sigma Chemical Co. (Deisenhofen, Germany). TNF was a generous gift from Dr. G. R. Adolph, Bender & Co. (Vienna, Austria) and had an endotoxin content of < 10 pg/lLg according to a limulus assay (Coatest Endotoxin, Chromogenix AB, Molndal, Sweden). Cell culture media were from Biochrom (Berlin, Germany). GaIN was from Roth Chemicals (Karlsruhe, Germany). TNF antiserum was prepared in our laboratories by S. Jilg as described previously (Leist et al., 1995b). POF and A-802715 [1-(5-hydroxy-5-methyl)hexyl-3-methyl-7-propylxanthine] were a gift from Dr. M. Schtinharting, Hoechst AG (Wiesbaden, Germany). All other reagents not further specified were from Sigma Chemical Co. (Deisenhofen, Germany) or Merck (Darmstadt, Germany) Animals and treatment schedules. Male NMRI mice (30 g) were from the Hannover'sches Zentralinstitut (Hannover, Germany). They were kept for at least 1 week at 22°C, 55% humidity and a 12-hr day/night rhythm with free access to food and water. All animals received hu care in compliance with legal requirements (Regierungsbezirk; Freiburg, Germany). Before the start of the liver perfusion, animals were injected i.v. (200 ILl) or i.p. (300 ILl) with
different inflammogens or pharmacological modulators. Ten minutes before the start of the perfusion, mice were anesthesized with 100 mg/kg of phenobarbitone (Sanofi, Munich, FRG) and fixed to a temperature-controlled (37°C) operation table. Liver perfusion. Mouse livers were perfused in situ in a nonrecirculating system essentially as described previously (Hartung et al., 1991). After laparotomy, the gall bladder was incised and the portal vein cannulated with a 20-gauge Teflon cannula. The diaphragm was severed, the anterior vena cava ligated and the perfusate was collected from a tube connected via a Teflon cannula to the posterior vena cava. This surgical procedure took about 5 min and any period of ischemia was strictly avoided. Perfusate flow was adjusted to 3 ml/min/g of liver based on an experimentally determined average relative liver weight of 4.75 ± 0.34% of b.wt. The buffer was a modified Krebs-Henseleit buffer (millimolar: Na+, 144; K+, 5.9; Ca++, 1.2; Mg++, 1.2; Cl-, 130; HCO 18; SOl-, 1.2; and phosphate, 1.2) supplemented with 2.1 mM lactate, 0.3 mM pyruvate and sometimes 1.5 ILM BST. After gassing with 5% CO 2 -95% O2 , the pH was 7.4 and oxygen concentration 0.95 mM at 37°C. Oxygen concentration and pH were measured continuously in the effluent perfusate. Samples for biochemical determinations were taken by a fraction collector (Pharmacia, Freiburg, Germany) at 2.5-min intervals. Additions to the basic buffer were made by a special infusion pump (Braun-Melsungen AG, Melsungen, Germany) connected to a T-junction just before the entry into the liver. Viability parameters. Viability and function of the liver was determined by various methods. Routinely, LDH release into the buffer was measured on an automatic Eppendorff ACP 5040 enzyme analyzer (Netheler und Hinz, Hamburg, Germany) according to Bergmeyer (1984). BST clearance was determined by measuring the residual BST concentration in the effluent buffer at 578 nm after alkalization with NaOH. The effluent pH of 7.23, the oxygen consumption of 2.7 ILmol/g/min and the perfusion pressure of 9 cm of H 2 0 remained constant over a period of 180 min in control perfusions
a,
(n = 9).
TNF determination. Samples for TNF determinations were frozen immediately at -20°C until analysis in the WEHI 164 clone 13 bioassay (Espevik and Nissen-Meyer, 1986) This assay was performed using recombinant muTNF as a standard and the specificity for muTNF was checked routinely by coincubation with a neutralizing anti-muTNF antibody. The detection limit was 10 pg/m!. The assay was sensitive to both huTNF and mTNF. The activity of muTNF was blocked completely by anti-muTNF antibodies. These did not interfere with the bioactivity of huTNF. Therefore, the amount of huTNF in samples containing both huTNF or muTNF could be determined by running parallel assays in the presence or absence of anti-muTNF antibody. The cumulative amount of TNF released into the perfusate was determined by integration of the TNF concentration in the buffer over time using a spreadsheet calculation software (QPRO, Borland, MA). TNF release over time was usually constant, i.e., cumulative TNF increased linearly under different experimental conditions. Therefore, with the exception of figures 4 and 5, in which TNF release was not constant over time, we show, in the other diagrams, the release ofTNF over the first 70 min. PDE inhibition. PDE activity was determined with some modifications (Bauer and Schwabe, 1980) as described by Thompson and Appleman (1979). PDE I [Ca++ (1 mM)/calmodulin (100 nM)-dependent] from bovine brain was kindly provided by Dr. Gietzen (Ulm, Germany). PDE 11 [cGMP-stimulated (5 !LM)] was purified from rat heart (Schudt et al., 1991b). PDE III (cGMP-inhibited) and PDE V (cGMP-specific) were assayed in homogenates of human platelets as described by Schudt et al. (1991a). PDE IV (cAMP-specific) was tested in the cytosol ofhu polymorphonuclear cells in the presence of the PDE Ill-specific inhibitor motapizone (1 ILM) as described by Schudt et al. (1991a). IC 50 values were calculated from concentration-inhibition curves by nonlinear regression analysis using the GraphPad software (Graphpad Inc., Sorrento Valley, CA).
970 Statistics. Experiments were performed in triplicate. Perfusions belonging to one series of experiments were performed within 1 day or on subsequent days. Control perfusions were repeated for every individual set of experiments, because variations of the absolute inducible TNF release were noticed between different sets of experiments. Concentration-response curves were fitted by nonlinear regression analysis using the inplot 4.0 software (Graphpad Inc.). When appropriate, the Student's t test or Welch test (for inhomogeneous variances) were used for significance analysis
Results Hepatotoxicity of pretreatment schedules. First of all we checked the basic properties of our perfusion system, in particular whether the pretreatment protocols chosen affected hepatocellular integrity. The LDH release from untreated mouse livers during the first 90 min of perfusion was about 150 mU/g of liver per hr (table 1). A similar LDH release was observed after pretreatment with any single substance used in this study, except GaIN. This latter compound caused an overall LDH release of about 300 mU/g of liver per hr when given alone or in combination with either LPS or TNF 60 min before the start of the perfusion. As damage controls, we used 5 mM hydrogen peroxide or 60 min noflow ischemia, both of which caused an LDH releases of 2,000 to 25,000 mU/g ofliver per hr. Infusion of 1 mM Triton X-lOO released 65,000 mU of LDH per g of liver per hr into the perfusate. From these data we conclude that our ex vivo treatment schedule using lactate-pyruvate supplemented buffer and livers from fed mice caused only negligible toxicity and was therefore better suited for the following pharmacological studies than a previously used perfusion system without buffer supplements and livers from starved mice (Wang and Wendel, 1990). Concordant results on hepatic integrity were obtained when BST clearance was determined as a measure of liver function (table 1). GaIN or combinations of LPS or TNF with GaIN caused a reduction of the hepatic extraction of BST from 92 ± 2 to 80 ± 4% after 70 min. All other substances had no significant effect. Such a slightly reduced BST clearance capacity of the liver may.be associated with a reduced clearance of TNF by the liver. To check this we infused 75 ng/ml ofTNF (a concentration that may be TABLE 1 Viability and ex vivo TNF release in perfused mouse livers from mice pretreated in vivo Treatmenta
LDH release b % of tota/liver LDH170 min
Saline GaiN (700 mg/kg) LPS (30 f.Lg/kg) GalN/LPS huTNF (30 f.Lg/kg) GalN/huTNF muTNF (30 f.Lg/kg) GalN/muTNF
0.055 0.103 0.064 0.082 0.073 0.113 0.076 0.111
± 0.007 ± 0.007
± 0.007
± 0.008 ± 0.011 ± 0.009
± 0.009
± 0.006
BST clearance c
%
92 ± 1 79 ± 2 89 ± 3 81 ± 5 N.D.e ND. 93 ± 4 79 ± 2
TNF release nglg liverlh n.d.d
n.d. 15 ± 26 ± 11 ± 22 ± 48 ± 151 ±
2 4 3 1 8 39
a Mice were injected 60 min before the start of the perfusion. Data are means ::t S.E.M. from triplicate determinations. b Total LDH content of liver homogenate was 275 ::t 20 U/9 of liver. LDH release from perfused livers was determined at 5-min intervals and the total amount was obtained by integration over 70 min. e BST clearance was measured 70 min after the start of the perfusion as the ratio of BST concentration before and after the liver. d n.d., not detectable. e N.D., not determined.
found in vivo) for 10 min into livers from mice pretreated with GaIN or solvent control. In both groups, the effluent TNF concentration was < 10 pg/ml. This finding suggests that the liver has a very high TNF -inactivating capacity that is not influenced by GaIN treatment. The role of GaIN in TNF release from the perfused mu liver. Because the liver contains the greatest number of macrophages of any organ in the body and may therefore account for a major part of circulating cytokines, we examined whether the release of the inflammatory cytokine TNF from perfused mouse livers can be determined quantitatively in a blood-free perfusion system. Stimulation ofTNF release during the first 90 min was not observed in vitro after infusion ofLPS or TNF. However, we did observe a spontaneous TNF release starting after about 90 min. When mice were treated in vivo with the inflammogens LPS or TNF 60 min before liver perfusion, a constant and consistent dose-dependent release of TNF into the perfusate was detected. Cotreatment of mice with GaIN, a sensitizer commonly used in mu models of sepsis and inflammatory liver failure, caused an approximately 1.5-fold increase in TNF release with any stimulus chosen (table 1). Because GaIN alone did not cause any significant TNF release and because we focused on the role of liver macrophages in experimental models of hepatic inflammation, we routinely used GaIN-treated mice for all subsequent experiments. Ex vivo induction of TNF release from the perfused mouse liver. Livers from mice pretreated with 2: 10 JLg/kg of LPS released TNF into the perfusate (fig. 1). TNF release was dose-dependent, was observed over a similar dose range to that seen in vivo and, at any given dose, was relatively constant over the first 70 min of perfusion. The amounts of TNF released were comparable to those observed in vivo after injection of LPS in concentrations that cause liver failure (Galanos et al., 1979). In our studies, a maximum dose of 60 JLg/kg of LPS was not exceeded, because this was acutely toxic to GaIN-sensitized mice in vivo. Next we examined whether TNF, as one ofthe main proinflammatory cytokines, might act as an autocrine stimulator 300 ~GaINlLPS
~ ~
.g: .:.
~ GalNlhuTNF
~ GalNlmuTNF
LL
z
I-
"E
'0
GI
1Il
.lI!
f!
o~~----~--~~~~~~~~~
0.0
1.0
dose of LPS or TNF [1Ig/k9 body weight]
Fig. 1. TNF release from the GaiN-sensitized perfused mouse liver elicited by the inflammatory stimuli LPS and TNF ex vivo. Mice were injected 60 min before the start of the perfusion with a combination of 700 mg/kg of GaiN (Lp.) and different doses of either of the three inflammatory stimuli: LPS (Lp.), muTNF (Lv.) or huTNF (Lv.). Subsequently, livers were perfused and TNF was determined in the perfusate at 5-min intervals for 70 min. Data are means ± S.D. from three independent perfusions.
971 of its own production (Philip and Epstein, 1986) in the liver. Pretreatment of mice with 2: 10 /Lg/kg of muTNF in vivo caused a dose-dependent release ofTNF from perfused livers ex vivo (fig. 1). TNF-induced TNF release was relatively constant during the perfusion time, a finding that cannot be explained only by a washout of previously injected TNF. TNF release due to pretreatment with muTNF may be significantly higher than that observed with huTNF, inasmuch as both types of TNF receptors are stimulated by muTNF, but not by huTNF (Leist et al., 1995b). In order to show that the released TNF was actually produced by the liver, we examined whether the injection ofhuTNF would cause the release ofmuTNF. By using an antibody able to selectively neutralize muTNF, the bioactivity of muTNF and huTNF can be determined in parallel in the WEHI 164 bioassay. By using this approach, we found that no huTNF, injected in vivo before the start of the perfusion, was washed out from the perfused liver ex vivo, and that any TNF bioactivity released into the perfusate was due to muTNF. From these experiments, we conclude that either LPS or TNF itself may induce an inflammatory response in the liver resulting in a dosedependent TNF production and release from the isolated liver. In order to exclude the possibility that TNF release into the perfusate was the result of an unspecific stress response due to a liver damage caused by the in vivo pretreatment, we examined whether physiological regulation· similar to that described as tolerance induction (Mengozzi and Ghezzi, 1993; Porat et al., 1993) also could be observed in the isolated organ. Therefore, we examined whether low-dose LPS pretreatment also would induce tolerance in the ex vivo perfusion system. As shown in table 2, GaINILPS- or GalNITNFinduced TNF release, respectively, was completely or strongly reduced under such conditions. These findings suggest that TNF release from the liver is more than a stress response, but rather is dependent on the activation of the Kupffer cells.
Ex vivo effects of methyIxanthines on hepatic TNF release. In order to examine whether methylxanthines had a direct effect on TNF production by the liver, we tested and compared the effects of the established TNF-suppressing drug POF with a more recent congener, the investigational drug A-802715 [1-(5-hydroxy-5-methyl)hexyl-3-methyl-7propylxanthin]. Neither compound caused any significant TNF release when injected into control mice. When mice were pretreated with these compounds before the injection of LPS, a dose-dependent reduction of the TNF release from subsequ.ently perfused livers was observed. The IC so for the in vivo pretreatment was 73 mg/kg for POF or 87 mg/kg for A-802715, respectively. TNF release was completely inhibited in livers from animals pretreated with 200 mg/kg of either compound (fig. 2A). Essentially, similar results were obtained when TNF was used as a stimulus. The IC so values for the inhibition of TNF release from the perfused liver (after in vivo stimulation with TNF) by in vivo pretreatment with methylxanthines were 79 mg/kg for POF and 96 mg/kg for A-802715. Inhibition of TNF-induced TNF release was 80% at a methylxanthine pretreatment dose of 200 mg/kg (fig. 2B). Thus, in vivo pretreatment of mice with methylxanthines 60 min before challenge with LPS or TNF prevented
e'E 0
u
(J)
ll..
80
1000b
Additional Modulation a
PDE 11
PDE III
PDE IV
PDE V
80 130
/L NF 215 210
47 44
85 16D
a Enzymatic activity of isolated isoenzymes and its inhibition by methylxanthines was measured as described under "Materials and Methods." b 30% inhibition at 1000 /LM.
functions, but also all major organs. Kupffer cells (3-9 X lOG/liver) by far outnumber the circulating monocytes (1-1.5 X 106/mouse), i.e., the chiefTNF producers of the blood, and may therefore contribute significantly to systemic TNF concentrations found in experimental models of sepsis. Recent observations actually demonstrate that hepatic TNF production, elicited by hepatic ischemia, is responsible for the development of subsequent pulmonary injury in rats (Colletti et al., 1990). Further indirect evidence for the pathogenic role of hepatic TNF comes from data showing that endotoxin neutralization or the elimination or inactivation of macrophages with agents such as silica, praseodymium nitrate or GdCl3 reduces or prevents hepatic injury elicited by different stimuli (Grful and Liehr, 1976; Nolan and Leibowitz, 1978; Camara et al., 1983; Tiegs et al., 1992; Adachi et al., 1994; Czaja
i
20
,
30
,
40
i
50
i
60
,
70
i
80
Fig. 5. Role of cAMP in the rapid onset of the inhibition of TNF release by methylxanthines: influence of the infusion of DDA with or without dibutyryl-cAMP (db-cAMP). Mice were injected Lp. with 700 mg/kg of GaiN and 30 p,g/kg of LPS 60 min before the start of the perfusion. Thirty minutes before LPS, mice were injected with saline or 200 mg/kg of POF or A-802715 (A-80). Fifteen minutes before the methylxanthine treatment, a group of mice were injected with 1.6 mg/kg of DDA as indicated. Additionally, some livers that received this treatment were infused with 0.1 mM db-cAMP after the initial 20 min of perfusion. Data = means::!: S.D., n = 3.
--GalNffNF - - + A80 (20-50 min) ...... + A80 (0-30 min) ... saline
~ 120 U. 90
30
,
10
time [min]
..:-240
-a- + POF (20-50 min)
i
40
30
time[h]
270
60
E ::s
i
i
10
B
120 U. 90
f!:
i
0
80
~ .2:
0
i
70
time [min]
~180
-0- +A80 + DDA +db-cAMP ... saline
.~ 10
12::s
i'240 :5:en 210
""-+A80+DDA -0- +POF +DDA +db-cAMP
I-
.~ 10
270
..... GalNlLPS .... +POF+DDA
.m u.ze 20
f!:
A
--GalNlLPS - + A80 (20-50 min) - + A80 (0-30 min) -+-saline
I40
..!l!
E ::s ()
60
50
--- + POF (0-30 min) ... saline
974
Treatment Saline GalN/LPS GalN/LPS GalN/LPS GalN/LPS
Pretreatment
Saline a
Saline 0 Saline 100::!: 15 200 mg/kg 0 Pentoxifyllined d 200 mg/kg of A-802715 0 150 mg/kg of db-cAMp d 10 ::!: 4
160/Lg/kg DDAb
16 mg/kg DDAa
0 110 19
N.D.c N.D. 59
17 N.D.
49 N.D.
DDA was given 45 min before LPS. Data are given in percentage of cumulative TNF release compared to mice treated i.p. with 700 mg of GaiN plus 30 /Lg of LPS only 60 min before perfusion (GaIN/LPS). TNF release from perfused mouse livers was measured for 70 min at 5-min intervals to calculate the cumulative amount released. Data are means ± S.D. from triplicate experiments. Data without S.D. are derived from a single experiment. C N.D., not determined. d Methylxanthines or db-cAMP were injected i.p. 30 min before LPS. a
b
et al., 1994; Laskin et al., 1995; Barriault et al., 1995; Ishiyama et al., 1995). Because TNF has been identified as one of the Kupffer cell products capable of inducing fulminant hepatic injury (Lehmann et al., 1987; Tiegs et al., 1989), it may be concluded that Kupffer cell blockade protects the liver primarily by inhibition of hepatic TNF release. We showed that the inducible TNF release observed in our liver perfusion system was subject to regulation by endoge-
nous and exogenous mediators. We therefore believe that neither direct hepatotoxic effects of the stimulators, surgical stress nor the 1- to 3-sec period of noflow ischemia when the perfusion cannula in the portal vein was connected to the perfusion pump account for LPS-stimulated TNF release. The ex vivo tolerance induction in the perfused liver by lowdose LPS is an example of regulation by endogenous mediators. Our findings from these experiments suggest that mu macrophages in situ may be regulated by mechanisms similar to those described in hu monocytes (Porat et al., 1993) or other macrophage populations (Mengozzi and Ghezzi, 1993). Knowing this, it was important to make sure that pharmacological effects of methylxanthines were not due to LPS contamination. Measurement ofLPS in a chromogenic limuIus assay showed that the endotoxin contamination of the methylxanthines was extremely low. The best evidence for pharmacological actions of the methylxanthines unrelated to a possible tolerance induction is, however, the fact that the TNF-reducing effect also was observed when they were given in vivo after the LPS stimulus or even ex vivo, i.e., 1 hr after the initiating LPS injection. Both POF and A-802715 [1-(5-hydroxy-5-methyl)hexyl-3methyl-7-propylxanthinl have been shown to suppress LPSinduced TNF release in vitro (Sinha et al., 1995). Very recently, A-802715 was found to be the most potent among different methylxanthines, being about twice as effective as POF and about 10 times more effective than theophylline (Sinha et al., 1995). Therefore, we thought it may be interesting to test the effects of this novel compound on TNF release from the perfused mouse liver in comparison to the well-characterized TNF-suppressing drug POF. Both substances were found to be of similar efficacy, when mice were pretreated in vivo or when drugs were given 1 hr after the TNF releasing challenge in an ex vivo design directly in the perfusate. This finding held true for two different stimuli, i.e., LPS or TNF. The inhibition of LPS-induced TNF release was observed at about 10-fold lower concentrations of meth-
ylxanthines in the perfusion buffer than was required for the inhibition of TNF-induced TNF release. With the in vivo pretreatment protocol, such a potency difference toward different stimuli was not observed. Another difference between the in vivo and the ex vivo protocol is the reversibility of the inhibition effect by simple washout. Ex vivo, the methylxanthines inhibited TNF release only during the period of time when they were present in the perfusion buffer. When mice had been pretreated with the drugs in vivo, however, the inhibition of TNF release persisted throughout the entire subsequent perfusion period. Several explanations are possible for these observations: 1) they may be explained pharmacokinetically, i.e., by different distribution or by the action of different metabolites; 2) methylxanthines may inhibit TNF release on multiple levels (e.g., transcriptional, translational or post-translational) and different levels of cAMP may be required to inhibit TNF release after different stimuli; and 3) direct cytotoxicity to macrophages may be a further cause of methylxanthine-inhibited TNF release. This latter possibility can, however, be ruled out in our system, inasmuch as inhibited TNF release could be restored by DDA, and because inhibition was reversible after omission of methylxanthine from the perfusion buffer. Further known methylxanthine actions include interaction with adenosine receptors (Parmely et al., 1993). However, in our system we provide experimental evidence that raised cAMP is causally involved in the regulation ofTNF release by methylxanthines. Both substances were shown to inhibit various PDE families and have been shown by others to increase cAMP in macrophages (Semmler et al., 1993). Confirming these results for mu Kupffer cells in situ was not within the scope of this study and measurements of total cellular cAMP may be flawed by the fact that average cAMP increase does not correlate with the inhibition ofTNF release (Sinha et al., 1995) and highly localized concentration changes of even flux changes of cAMP may be sufficient to cause effects such as inhibition ofTNF release (Pryzwansky et al., 1981; Barsony and Marx, 1990; Nicholson et al., 1991). However, we showed that raising cAMP with alternative effectors such as iloprost or dibutyryl-cAMP has similar pharmacological effects as the treatment with methylxanthines. In addition, we checked the causal involvement of cAMP in the suppression of TNF in this system by using DDA. This adenylase inhibitor is a functional antagonist of methylxanthines with respect to their cAMP-raising actions. Indeed we found that DDA antagonized the methylxanthine-induced inhibition of TNF release from the liver. Therefore, we conclude that TNF regulation by POF or A-802715 in mouse liver is caused by an increase of cAMP in the effector cell, i.e.,. the Kupffer cell. Such a suppression of TNF release within the liver itself may explain the prevention of fulminant LPS-induced liver failure by various cAMP-raising substances such as forskolin (Fischer et al., 1993), PDE-inhibitors (Schtinharting and Schade, 1989) or prostacyclin analogs (Wendel et al., 1987) observed in experimental animals. Our study provides evidence that the perfused mouse liver is a system that allows the investigation of the regulation of TNF release within the largest resident macrophage pool in the body. This means that TNF release from Kupffer cells within their natural cellular environment may be examined under conditions of maintained regulation by endogenous or exogenous mediators. As demonstrated by two examples,