Background

Insufficient detoxification caused by nitric oxide-related inhibition of cytochrome P450 may be important for metabolism of numerous drugs, including vecuronium. The present study investigated the pharmacodynamics and pharmacokinetics of vecuronium in rats with inflammatory liver dysfunction.

Methods

Male Sprague-Dawley rats (n = 56) were randomly allocated into two groups: In the sepsis group, liver inflammation was established by injection of 56 mg/kg heat-killed Corynebacterium parvum; control rats received the solvent. At day 4, groups were subdivided according to treatment with the nitric oxide synthase inhibitor N(G)-monomethyl-L-arginine (250 mg/kg) or placebo. The aminopyrine breath test was performed to assess cytochrome P450 activity. Rats were anesthetized with propofol and mechanically ventilated. Duration of action of vecuronium (1.2 mg/kg) was measured by evoked mechanomyography (stimulation of the sciatic nerve, contraction of the gastrocnemius muscle). In seven rats of each subgroup a 50% neuromuscular blockade was established by a continuous vecuronium infusion. Vecuronium plasma levels were measured and plasma clearance of vecuronium was calculated. Nitric oxide synthesis was assessed by measuring nitrite/nitrate serum levels.

Results

In sepsis/placebo rats, vecuronium-induced neuromuscular blockade was prolonged (144% of contro/placebo), vecuronium plasma levels at 50% neuromuscular blockade were increased (122% of control/placebo), and plasma clearance was decreased (68% of control/placebo). N(G)-monomethyl-L-arginine therapy in rats with sepsis improved cytochrome P450 activity and plasma clearance of vecuronium, shortened duration of action of vecuronium, but did not alter the elevated vecuronium plasma levels.

Conclusions

A systemic inflammatory response syndrome with liver dysfunction results in decreased sensitivity to and a decreased elimination of vecuronium. Modulation of nitric oxide synthesis may be a strategy that can be used in the future to improve xenobiotic metabolism in sepsis.

INSUFFICIENT detoxification of numerous drugs is a serious problem in the management of hepatic failure associated with a systemic inflammatory response syndrome. 1Some steps in the pathogenesis of this process have been characterized, demonstrating an important role for nitric oxide (NO) biosynthesis.

Exogenously applied NO 2as well as endogenously produced NO 3suppresses activity of cytochrome P450 enzymes, which are key enzymes of xenobiosis. In vivo  stimulation of inducible NO synthase (iNOS) results in a downregulation of cytochrome P450 activity. 4In a previous in vivo  experiment we demonstrated the effect of iNOS induction on the cytochrome P450–dependent detoxification using the aminopyrine breath test. Inhibition of iNOS 4 days after administration of the inflammatory stimulus resulted in a restitution of cytochrome P450 activity. 5However, the clinical relevance of these findings remained unclear.

Pittet et al.  6demonstrated a close correlation between the aminopyrine breath test and the speed of neuromuscular recovery following vecuronium administration in pigs after liver transplantation. If a similar effect occurs during inflammatory hepatic dysfunction, iNOS inhibition might prove to be a viable therapeutic option to restore P450 function.

Thus, the aim of this study was to investigate the effect of iNOS induction and its inhibition on drug metabolism by means of the pharmacokinetics and pharmacodynamics of vecuronium, using a rat model of inflammatory liver dysfunction.

Animal Model 

Fifty-six male Sprague–Dawley rats (Charles River GmbH, Kisslegg, Germany), 250–350 g, were allowed to accommodate to standard conditions with free access to chow and water for 14 days. The rats were randomly assigned to two groups according to the experimental model of granulomatous liver inflammation and its therapy with N  G-monomethyl-L-arginine (NMA). 5In rats of one group (n = 28), 56 mg/kg heat-killed Corynebacterium parvum  (Universal Biologicals, London, UK) was intravenously injected in total volume of 0.5 ml saline. Rats of the control group (n = 28) received the equivalent volume of saline. Four days later, each group was further divided into subgroups (n = 14) receiving two intravenous injections of either 125 mg/kg NMA (Clinalpha AG, Läufelfingen, Switzerland) in 0.5 ml saline or placebo (0.5 ml saline) 12 and 9 h before the measurements. This design resulted in four groups of rats: rats without infection not receiving NMA therapy (control/placebo: n = 14), rats without infection receiving NMA therapy (control/NMA: n = 14), infected rats not receiving NMA therapy (sepsis/placebo: n = 14), and infected rats receiving NMA therapy (sepsis/NMA: n = 14). To exclude variances in liver function caused by differences in food intake, rats were fasted beginning 12 h before the measurements.

Aminopyrine Breath Test 

Three hours after the second NMA or placebo injection, cytochrome P450 enzyme activity was assessed using the aminopyrine breath test based on the formation of [14C]carbon dioxide from [14C]dimethyl-aminopyrine via  cytochrome P450–dependent N -demethylation. 5The aminopyrine breath test was started by intravenous injection of [14C]dimethyl-aminopyrine (Amersham International, Braunschweig, Germany) with a specific activity of 57 MBq/kg in 0.5 ml saline. The animals were then placed into a gas-tight tubing with a continuous airflow of 0.7 l/min. The outflow was conducted through a solution of 0.5 M hyaminehydrochloride–alcohol (1:1; v/v; Zinsser Analytic, Frankfurt/Main, Germany) to bind expired CO2. In the animal model used in this experiment, the maximum [14C]dimethyl-aminopyrine turnover was found to be between 20 and 30 min after injection. 5Therefore, a vial containing hyamine was placed into the outflow of the gas tubing for this period, and the amount of exhaled [14C]carbon dioxide was determined by the use of a β-counter. The values were expressed as percentages of the totally applied radioactivity, making the test independent from total body CO2production.

Anesthesia and Vital Parameters 

Anesthesia was induced by inhalation of 4% isoflurane in 100% oxygen and maintained with 2% isoflurane in oxygen via  head mask in spontaneously breathing animals. Following tracheotomy, mechanical ventilation was continued with oxygen in air (inspired oxygen fraction = 0.4) and adjusted to maintain an end-tidal carbon dioxide partial pressure between 32 mmHg and 36 mmHg. After cannulation of the left external jugular vein, anesthesia was switched to a continuous infusion of propofol (20–40 mg · kg−1· h−1) and maintained according to cardiovascular signs of inadequate anesthesia. The left carotid artery was cannulated to measure mean arterial pressure and perform blood gas analyses.

After tracheotomy and cannulation, all animals were allowed to equilibrate over a period of 15 min, after which baseline recordings of heart rate, mean arterial pressure, rectal temperature, and base excess were performed. Afterward, hypovolemia, if present, was treated with hydroxyethyl starch (6%; 450/0.7) to establish a central venous pressure between 6 and 8 mmHg. The amount of necessary hydroxyethyl starch was documented. Following hydroxyethyl starch administration, mean arterial pressure ranged between 80 and 90 mmHg and was stable in each animal during the neuromuscular function tests.

Prior to the neuromuscular experiments ventilation was adjusted to maintain an arterial carbon dioxide partial pressure between 36 and 40 mmHg. Base excess was corrected with 1 mM sodium bicarbonate to values between 3 mM and −3 mM. Arterial oxygen partial pressure and concentration of ionized calcium were kept within normal limits, and rectal temperature was controlled between 36.5 and 37.5°C with warming blankets.

Neuromuscular Function 

The sciatic nerve of the immobilized leg was exposed at its exit from the lumbosacral plexus. The ankle joint was fixated and a force transducer connected to the foot at a right angle. Stimulation of the sciatic nerve was done by train-of-four stimuli every 12 s. The contraction of the gastrocnemius muscle was measured by evoked mechanomyography (Myograph, Biometer, Copenhagen, Denmark). Supramaximal stimulus and control twitch height (T0) were established. Baseline mechanomyographic response was stabilized over a period of 10 min before injection of a fourfold ED95of vecuronium (1.2 mg/kg; n = 14). In a preliminary investigation we had evaluated the ED95of vecuronium in five nonseptic and untreated rats using the same preparation. After the bolus injection, neuromuscular transmission was allowed to recover to baseline values. The intervals between injection of vecuronium and the recovery of the first twitch (T1) to 25%(duration 25%[seconds]) and 75%(duration 75%[seconds]) were measured. The recovery interval was calculated by: recovery interval [seconds]= duration 75%− duration 25%. In the first seven animals of each group the in vivo  experiment was terminated at this step. In the second seven animals of each group a continuous infusion of vecuronium was adjusted to achieve a constant T1/T0of 50%. Following 10 min of stable T1/T0= 50%, steady state conditions were assumed. The required infusion rates were documented, and 2 ml of heparinized blood was withdrawn for vecuronium plasma level determination. The withdrawn blood was centrifuged (3,500 rpm, 10 min, 4°C), the supernatant removed, and the equipotent amount of phosphate buffer added to the plasma. The samples were immediately frozen at −70°C. Following this, animals were killed by exsanguination. Plasma was separated by centrifugation and immediately stored at −70°C.

Chemical Analyses 

Nitric oxide synthesis was assessed by measuring levels of serum nitrite and nitrate, which are the stable products of NO oxidation. Plasma samples were deproteinized with 0.5 M NaOH and 10% ZnSO4. Nitrate was then converted to nitrite using high-performance liquid chromatography on a cadmium column. Nitrite concentrations were determined spectrophotometrically at 540 nm using a method based on the Griess reaction. 7The activity of the glutamate pyruvate transaminase (GPT) was measured with an automatic procedure at the Institute of Clinical Chemistry, Technische Universität München. Vecuronium plasma levels were analyzed by a high-performance liquid chromatography–tandem mass spectrometry at the Institute of Clinical Chemistry, University Hospital, Zürich, Switzerland. Plasma clearance of vecuronium during steady state conditions was calculated by the equation: clearance = infusion rate/plasma level.

Statistical Analyses 

Data are given as means ± SD. Statistical analyses were performed using factorial analysis of variance. Post hoc  analysis was performed in four of six possible comparisons by the Dunnett t  test, according to the objectives: autonomic effect of NMA (control/placebo vs.  control/NMA), effect of infection (control/placebo vs.  sepsis/placebo), effect of NMA therapy in infected rats compared with control rats (sepsis/NMA vs.  control/placebo), and effect of NMA therapy in infected rats compared with sick rats (sepsis/NMA vs.  sepsis/placebo). Differences were considered significant at P < 0.05.

Regressions were calculated between the aminopyrine breath test and the recovery interval of vecuronium as well as between the aminopyrine breath test and the vecuronium clearance during steady state conditions at 50% neuromuscular blockade. Correlations were calculated by linear least-squares regressions.

Systemic Signs 

To analyze systemic symptoms of inflammation, body weights of the animals were compared at day 4, prior to the first NMA or placebo injection. All control rats (control/placebo and control/NMA) gained weight (8 ± 5 g); all rats with corynebacterium parvum injections (sepsis/placebo and sepsis/NMA) showed a significant loss of body weight (−5 ± 8 g).

Vital parameters were evaluated after induction of anesthesia and cannulation of venous and arterial vessels. No differences between groups were found regarding mean arterial pressure (75–130 mmHg), heart rate (295–390 beats/min), base excess (−5.1–3.1 mM), or rectal temperature (34.1–37.9°C). The amount of hydroxyethyl starch necessary to increase the central venous pressure to 6–8 mmHg was significantly higher in both groups injected with Corynebacterium parvum  (sepsis/placebo and sepsis/NMA: 2.1 ± 1.4 ml) compared with control animals (control/placebo and control/NMA: 5.1 ± 1.1 ml).

Quantification of Cytochrome P450 Activity, NO Production, and GPT 

The aminopyrine turnover in sepsis/placebo rats was suppressed to 41% of that of control/placebo rats. Treatment with NMA significantly improved the aminopyrine turnover in the sepsis/NMA group (69% of control/placebo). Sepsis/placebo rats had extremely elevated nitrite/nitrate serum levels compared with control/placebo rats. Sepsis/NMA rats had significantly lower nitrite/nitrate serum concentrations compared with sepsis/placebo rats. Injection of Corynebacterium parvum  was followed by elevated GPT levels. Treatment with NMA did not significantly affect GPT levels in either the control group or animals injected with Corynebacterium parvum  (table 1).

Table 1. Nitrite Plus Nitrate Serum Concentrations, Aminopyrine Breath Test, and Liver Enzyme Release 

Table 1. Nitrite Plus Nitrate Serum Concentrations, Aminopyrine Breath Test, and Liver Enzyme Release 
Table 1. Nitrite Plus Nitrate Serum Concentrations, Aminopyrine Breath Test, and Liver Enzyme Release 

Vecuronium-induced Meeting Abstracts 

Vecuronium-induced neuromuscular blockade (as measured by duration 25%, duration 75%, and recovery interval) was prolonged in sepsis/placebo rats compared with control/placebo rats. Sepsis/NMA rats had duration 25%, duration 75% and recovery interval comparable to control/placebo rats (table 2).

Table 2. Neuromuscular Block Induced by 1.2 mg/kg Vecuronium (4 × ED95) 

Table 2. Neuromuscular Block Induced by 1.2 mg/kg Vecuronium (4 × ED95) 
Table 2. Neuromuscular Block Induced by 1.2 mg/kg Vecuronium (4 × ED95) 

Sepsis/placebo rats had significantly higher plasma levels of vecuronium compared with control/placebo rats. This significance was maintained between control/placebo rats with sepsis/NMA rats (table 3). Infusion rate and plasma clearance were significantly decreased in sepsis/placebo rats compared with control/placebo rats. Treatment of the sepsis group with NMA resulted in an increased infusion rate and plasma clearance compared with the sepsis/placebo group (table 3).

Table 3. Vecuronium Plasma Levels, Infusion Rate, and Plasma Clearance of Vecuronium during Steady-state Conditions at a 50% Meeting Abstracts 

Table 3. Vecuronium Plasma Levels, Infusion Rate, and Plasma Clearance of Vecuronium during Steady-state Conditions at a 50% Meeting Abstracts 
Table 3. Vecuronium Plasma Levels, Infusion Rate, and Plasma Clearance of Vecuronium during Steady-state Conditions at a 50% Meeting Abstracts 

Comparison of the regression analysis of the vecuronium plasma clearance and recovery index with the aminopyrine breath test showed that the vecuronium plasma clearance correlated better with the aminopyrine breath test (r = 0.707) than the recovery index (r = 0.447)(figs. 1 and 2).

Fig. 1. Regression between aminopyrine breath test (ABT) and recovery interval (R.I.) after administration of 1.2 mg/kg vecuronium. 

Fig. 1. Regression between aminopyrine breath test (ABT) and recovery interval (R.I.) after administration of 1.2 mg/kg vecuronium. 

Close modal

Fig. 2. Regression between aminopyrine breath test (ABT) and plasma clearance (ClSS) of vecuronium during steady state conditions at a 50% neuromuscular blockade. 

Fig. 2. Regression between aminopyrine breath test (ABT) and plasma clearance (ClSS) of vecuronium during steady state conditions at a 50% neuromuscular blockade. 

Close modal

Tremendous progress has been made in understanding the molecular basis of inflammatory response, demonstrating an important role of NO as a mediator during septic conditions. Besides many other effects, NO inhibits the activity of the iron-containing cytochrome P450. 8It leads to an oxidation of the iron, which is then released from the protein and becomes inactive. 4,8Metabolism of drugs depending on cytochrome P450 could therefore be affected by septic conditions.

In the rodent model used in the present study, injection of heat-killed Corynebacterium parvum  causes an induction of the iNOS expression with maximal nitrite/nitrate plasma levels between day 3 and day 6 after injection. 9,10The NO-induced suppression of the cytochrome P450 activity could be demonstrated with the aminopyrine breath test, reflecting the influence of NO synthesis on the hepatic detoxification processes. 5 

Evidence suggests a role of liver function in the pharmacodynamics of vecuronium. 11In patients with liver cirrhosis, a reduced plasma clearance has been reported. 12In addition, reduced infusion rates for the maintenance of a steady state neuromuscular blockade in the anhepatic phase during orthotopic liver transplantation have been described. 13 

In our model of a chronic inflammatory liver dysfunction, the duration of a vecuronium-induced neuromuscular blockade was prolonged, and the plasma clearance for vecuronium was decreased. These findings may be pertinent for clinical practice, because our data prove not only that the pharmacokinetics or metabolism of a diagnostic agent such as aminopyrine are affected by the activity of cytochrome P450, but also that the pharmacodynamic effect of a commonly used drug in clinical practice is altered. An impaired hepatic microsomal metabolism of ethylmorphine and midazolam during sepsis associated with increased NO synthesis has already been described, although without the possibility of monitoring the clinical effects. 14This dilemma was overcome in our study by using the neuromuscular blocking agent vecuronium. Its clinical effects can be easily measured by neuromuscular monitoring.

Because restitution of cytochrome P450 activity with NMA improves xenobiosis, NMA therapy was expected to shorten the vecuronium induced neuromuscular blockade. However, the normalized duration times and recovery intervals after NMA treatment in rats injected with Corynebacterium parvum  indicate a normalized drug metabolism, although restitution of cytochrome P450 activity was incomplete. Furthermore, the aminopyrine turnover and the recovery intervals following an injection of 1.2 mg/kg vecuronium did not correlate very well (r = 0.447) in this inflammation model with impaired liver function. In contrast, Pittet et al.  6found a strong correlation (r = 0.843) between aminopyrine turnover and the recovery interval of a vecuronium induced neuromuscular blockade after hepatic autotransplantation in pigs. Therefore, an additional mechanism has to be postulated.

Thus we determined plasma clearance and plasma levels of vecuronium during a 50% neuromuscular blockade in the second seven rats of each group. Plasma clearance of vecuronium during these steady state conditions correlated much better with the aminopyrine turnover (r = 0.707) than the recovery intervals. This different finding results from the higher vecuronium plasma levels necessary for 50% neuromuscular blockade, indicating a resistance to vecuronium.

Corresponding results have been reported by Tomera and Martyn, 15showing a threefold to fivefold rightward shift in the dose–response curves of d -tubocurarine 2 weeks after a three-times repeated intraperitoneal injected dose of Escherichia coli  lipopolysaccharide. They suggested an upregulation of perijunctional immature acetylcholine receptors in sepsis, similar to changes seen in burn injury or muscle disuse atrophy, to be the reason for those results. As additional acetylcholine receptors cause a relative resistance to nondepolarizing neuromuscular blocking drugs, 16shorter duration of action should be observed in septic rats. Thus, the prolonged duration of the vecuronium-induced neuromuscular blockade demonstrates that the impaired detoxification of vecuronium is the prevailing effect in our model.

The prolongation of the vecuronium-induced neuromuscular blockade and the reduced clearance of vecuronium could also be caused by reduced liver perfusion, which is considered to be one of the major factors in terminating the effect of vecuronium. 17However, Because NO improves liver perfusion, it is more conceivable that the negative effect of NO on metabolic pathways, via  inhibition of cytochrome P450 activity, may be counteracted by an improved perfusion. In addition, the effective improvement of the aminopyrine turnover with NMA did not indicate an impaired liver perfusion. Furthermore, GPT activity did not differ between NMA- and placebo-treated animals, regardless of whether they received a Corynebacterium parvum  injection.

Hepatocellular vecuronium uptake of rats is approximately 10 times higher than that of humans, 18which leads to a relatively shorter duration of action of vecuronium in rats given vercuronium on an ED95-equivalent base. However, because this study investigated groups with different activities of cytochrome P450, the fact that cytochrome P450 in hepatic microsomes of rats and humans has closely related properties 19is more relevant for transferability of our results. In addition, the expression of iNOS in human hepatocytes after stimulation with proinflammatory mediators and its effect on human cytochrome P450 is comparable to the situation in rats. 3 

In conclusion, inflammatory liver dysfunction resulted in a decreased sensitivity to and a decreased metabolism of vecuronium. The resistance to vecuronium may be caused by an upregulation of acetylcholine receptors in sepsis. NMA improved vecuronium metabolism as well as aminopyrine turnover in infected rats, indicating a restored drug metabolism. Thus, modulation of NO synthase may be a way to improve xenobiosis in sepsis.

The authors thank Dr. Katharina Rentsch (Institute of Clinical Chemistry, University Hospital Zürich, Switzerland) for measurement of vecuronium plasma levels, and our colleagues from the Institute of Clinical Chemistry, Technische Universität München, for measuring GPT levels.

1.
Marzi I: Bedeutung der Leber für den Verlauf des SIRS nach Schock, Trauma und bei Sepsis. Anästhesiol Intensivmed Notfallmed Schmerzther 1994; 29: 41–6
2.
Wink DA, Osawa Y, Darbyshire JF, Jones CR, Eshenaur SC, Nims RW: Inhibition of cytochromes P450 by nitric oxide and a nitric oxide-releasing agent. Arch Biochem Biophys 1993; 300: 115–23
3.
Stadler J, Trockfeld J, Schmalix W, Brill T, Siewert J, Greim H, Doehmer J: Inhibition of cytochromes P4501A by nitric oxide. Proc Natl Acad Sci 1994; 91: 3559–63
4.
Khatsenko OG, Gross SS, Rifkind AB, Vane JR: Nitric oxide is a mediator of the decrease in cytochrome P450-dependent metabolism caused by immunostimulants. Proc Natl Acad Sci 1993; 90: 11147–51
5.
Veihelmann A, Brill T, Blobner M, Scheller I, Mayer B, Prölls M, Himpel S, Stadler J: Inhibition of nitric oxide synthases improves detoxication in inflammatory liver dysfunction in vivo. Am J Physiol 1997; 273: G530–G6
6.
Pittet JF, Morel DR, Mentha G, Schopfer C, Belenger J, Benakis A, Tassonyi E: Vecuronium neuromuscular blockade reflects liver function during hepatic autotransplantation in pigs. A NESTHESIOLOGY 1994; 81: 168–75
7.
Archer S: Measurement of nitric oxide in biological models. Faseb J 1993; 7: 349–60
8.
Palmer RM: The discovery of nitric oxide in the vessel wall: A unifying concept in the pathogenesis of sepsis. Arch Surg 1993; 128: 396–401
9.
Farquhar D, Benvenuto JA, Kuttesch N, Li Loo T: Inhibition of hepatic drug metabolism in the rat after corynebacterium parvum treatment. Biochem Pharmacol 1983; 32: 1275–80
10.
Billiar TR, Curran RD, Harbrecht BG, Stuehr DJ, Demetris AJ, Simmons RL: Modulation of nitrogen oxide synthesis in vivo: NG-monomethyl- L -arginine inhibits endotoxin-induced nitrate/nitrate biosynthesis while promoting hepatic damage. J Leukoc Biol 1990; 48: 565–9
11.
Hunter JM, Parker CJ, Bell CF, Jones RS, Utting JE: The use of different doses of vecuronium in patients with liver dysfunction. Br J Anaesth 1985; 57: 758–64
12.
Lebrault C, Berger JL, D'Hollander AA, Gomeni R, Henzel D, Duvaldestin P: Pharmacokinetics and pharmacodynamics of vecuronium (ORG NC 45) in patients with cirrhosis. A NESTHESIOLOGY 1985; 62: 601–5
13.
O'Kelly B, Jayais P, Veroli P, Lhuissier C, Ecoffey C: Dose requirements of vecuronium, pancuronium, and atracurium during orthotopic liver transplantation. Anesth Analg 1991; 73: 794–8
14.
Müller CM, Scierka A, Stiller RL, Kim YM, Cook DR, Lancaster JR, Jr., Buffington CW, Watkins WD: Nitric oxide mediates hepatic cytochrome P450 dysfunction induced by endotoxin. A NESTHESIOLOGY 1996; 84: 1435–42
15.
Tomera JF, Martyn JJ: Intrapleural endotoxin but not protein malnutrition shifts d-tubocurarine dose-response curves in mouse gastrocnemius muscle. J Pharmacol Exp Ther 1989; 250: 216–20
16.
Martyn JAJ, White DA, Gronert GA, Jaffe RS, Ward JM: Up-and-down regulation of sceletal muscle acetylcholine receptors. A NESTHESIOLOGY 1992; 76: 822–43
17.
Mol WE, Rombout F, Paanakker JE, Oosting R, Scaf AH, Meijer DK: Pharmacokinetics of steroidal muscle relaxants in isolated perfused rat liver. Biochem Pharmacol 1992; 44: 1453–9
18.
Sandker GW, Weert B, Olinga P, Wolters H, Slooff MJ, Meijer DK, Groothuis GM: Characterization of transport in isolated human hepatocytes: A study with the bile acid taurocholic acid, the uncharged ouabain and the organic cations vecuronium and rocuronium. Biochem Pharmacol 1994; 47: 2193–200
19.
Imaoka S, Enomoto K, Oda Y, Asada A, Fujimori M, Shimada T, Fujita S, Guengerich FP, Funae Y: Lidocaine metabolism by human cytochrome P-450s purified from hepatic microsomes: Comparison of those with rat hepatic cytochrome P-450s. J Pharmacol Exp Ther 1990; 255: 1385–91