Abstract
Binding, transport, and metabolism are factors that influence morphine (M) removal in the rat liver. For M and the morphine 3β-glucuronide metabolite (M3G), modest binding existed with 4% bovine serum albumin (unbound fractions of 0.89 ± 0.07 and 0.98 ± 0.09, respectively), and there was partitioning of M into red blood cells. Transport studies of M (<750 μM) showed similar, concentration-independent uptake clearances (CLs) of 1.5 ml min-1 g-1 among zonal and homogeneous, isolated rat hepatocytes. Transport of M3G, ascertained in multiple indicator dilution studies at various steady-state M3G concentrations (10-262 μM), uncovered a low and concentration-independent influx clearance (<10% of flow rate). The outflow dilution curve of [3H]M3G was superimposable onto that of [14C]sucrose, the extracellular reference, displaying similarity in transit times (23.5 and 22.2 s), negligible biliary excretion, and almost complete dose recovery from perfusate. In contrast, M3G occurred abundantly in both perfusate and bile in single-pass perfusion studies of the precursor, M, and revealed a biliary clearance of formed M3G that was 12.3-fold that of preformed M3G, suggesting a sinusoidal, diffusional barrier for M3G. With increasing concentrations of M (9-474 μM), clearance decreased, and metabolism and biliary excretion displayed concentration-dependent kinetics. Fitting of the data to a physiologically based liver model revealed that M removal mechanisms were saturable, with a Km,met of 52.2 μM and Vmax,met of 58.8 nmol min-1 g-1 for metabolism, and a Km,ex of 41.2 μM and Vmax,ex of 8.1 nmol min-1 g-1 for excretion. Sinusoidal transport was not rate-limiting for M removal.
Morphine (M) is one of the commonly used opioid analgesic drugs. It is almost completely absorbed, and a large fraction of the absorbed dose is removed by first-pass metabolism. M has also been extremely well studied over the past few decades in both animals (Walsh and Levine 1975) and humans (Drewe et al., 2000) where considerable efforts were expended on understanding the contribution of the intestine and liver in the first-pass glucuronidation of M. A survey of the literature pertinent to morphine disposition reveals a progression of concepts and in vitro experimental techniques, including cellular (Iwamoto et al., 1978; Dechelotte et al., 1993) and subcellular (Del Villar et al., 1977; Miners et al., 1988) fractions, perfused liver (Evans and Shanahan, 1995; Milne et al., 1997) and intestine (Doherty and Pang, 2000) preparations, and more recently, molecular biological studies on the cloning, expression, and regulation of the UDP-glucuronosyltransferases (Coffman et al., 1997; King et al., 1997). M has a well characterized metabolic profile, including formation of the pharmacologically important glucuronide conjugates as well as N-demethylation to normorphine (Klutch, 1974; Evans and Shanahan, 1995) and sulfoconjugation (Smith et al., 1973). M is metabolized by Ugt2b1 in the rat (King et al., 1997) and UGT2B7 in human (Coffman et al., 1997). Although morphine 6β-glucuronide (M6G) is suggested to be more effective than M as an analgesic (Osborne et al., 1990), recent studies showed that this conclusion was exaggerated, because M6G barely contributed to the central opioid effects after administration of analgesic doses of M (Sharke et al., 2003). Morphine 3β-glucuronide (M3G) is devoid of analgesic activity and significant antagonistic effects on morphine and M6G (Penson et al., 2000). The metabolism of M is species-dependent. In the rat, the natural (-)-morphine [(-)M] is glucuronidated only at the 3-position (Rane et al., 1985).
The transport of morphine and its glucuronides has been studied. Saturable transport of morphine was reported in the rat (Iwamoto et al., 1978). In addition, being lipophilic, M enters the cell by passive diffusion. M is suggested to be a substrate of multidrug-resistant protein 1 (MDR1), also known as P-glycoprotein (P-gp) (Xie et al., 1999). However, digoxin, another known MDR1 substrate, was not inhibited by M or its metabolites in vitro (Wandel et al., 2002). Others also failed to observe changes in M disposition in humans treated with valspodar (PSC833), a MDR1 modulator (Drewe et al., 2000). The glucuronide metabolites are hydrophilic organic anions and, as a class, do not enter the liver cells well by passive diffusion. Reports by O'Brien et al. (1996) postulated that M3G is too polar to traverse the sinusoidal membrane. Furthermore, M3G was not well excreted when it was given to the guinea pig liver, whereas much more was excreted when its precursor, morphine, was given (Milne et al., 1997). Studies with Mdr1a(-/-) knockout mice failed to show changes in M3G and M6G kinetics in the brain, ruling out the involvement of P-gp for transport of the glucuronides (Xie et al., 1999). By contrast, the multidrug resistance-associated protein (Mrp) 2 shows avid transport of glucuronide conjugates exemplified by acetaminophen glucuronide at the canalicular membrane (Xiong et al., 2000), whereas Mrp3 is associated with sinusoidal efflux (Xiong et al., 2002). It is likely that these transporters are also involved in morphine glucuronide excretion and sinusoidal efflux.
The aim of this study is to test the hypothesis that transport barriers for lipophilic drugs are not important in dictating metabolite formation and excretion; rather, enzymes and canalicular transporters are important. This view is presently tested with (-)-morphine that only forms M3G in the rat (Rane et al., 1985). However, no data exist to discern whether binding, transport, excretion, or metabolism is rate-controlling the overall removal of morphine in the intact liver. This study examined the in vitro uptake of (-)M in rat hepatocytes and appraised the interactions of uptake, metabolism, and excretion in the in situ perfused liver preparation. For the study of uptake, efflux, and sequestration of M3G, the multiple indicator dilution (MID) technique was used to ascertain both uptake and excretion over a wide concentration range. Single-pass studies were also conducted with varying concentrations of M (9-474 mM) to study saturable transport and metabolism. The importance of transport and metabolism with concentration under nonfirst-order conditions was appraised with a physiologically based pharmacokinetic model of the liver.
Materials and Methods
Materials
Unlabeled (-)M, M3G, and ethylmorphine were obtained from the National Institute on Drug Abuse (Rockville, MD). [N-Methyl-3H]morphine ([3H]M; specific activity 2.5 mCi/mmol) was purchased from New England Nuclear (Boston, MA). The radiochemical purity exceeded 96% as confirmed by high-performance liquid chromatography (HPLC). Collagenase A (Boehringer Mannheim, Laval, QC, Canada), digitonin (Sigma-Aldrich, St. Louis, MO), and silicone oils (Dow Corning Corp., Midland, MD; densities of 1.07 and 1.00 g/ml) were used for the preparation of isolated and zonal rat hepatocytes. Bovine serum albumin (BSA) fraction V, glucose, and phenolphthalein glucuronide were obtained from Sigma-Aldrich. Solvents were of HPLC grade (Caledon Labs, Mississauga, ON, Canada), and all other chemicals were of the highest grade available. Tritiated M3G was biosynthesized from [3H]M in saline recirculating through the in situ rat liver preparation at 40 ml/min. Following lyophilization of bile and perfusate, [3H]M3G was isolated and purified by HPLC (Beckman Ultrasphere ODS; 5 μm, 4.5 mm × 25 cm; flow rate of 1 ml/min) under isocratic conditions (95:5, 0.1% trifluoroacetic acid/methanol). The purity of [3H]M3G was >95% as determined by an HPLC method (see assay for unlabeled M3G).
Uptake Studies
The isolation of homogeneous (even) populations of rat hepatocytes was performed by collagenase perfusion, as described previously, and the preparation of periportal and perivenous rat hepatocytes was achieved with the digitonin procedure (Tirona et al., 1999). Marker enzymes alanine aminotransferase and glutamine synthase were then determined in the enriched population of periportal (PP) and perivenous (PV) hepatocytes (Tirona and Pang, 1999). Uptake studies were carried out at 37°C in a rotating water bath after admixture of hepatocyte suspensions (0.5 or 0.9 ml of 2 × 106 cell/ml) and M (0.1 ml containing varying concentrations of unlabeled and [3H]M) to result in final concentrations of 1 to 750 μM in an atmosphere of 95% oxygen and 5% carbon dioxide; [14C]sucrose was added as an extracellular marker to correct for volumes adhered onto the cells. The uptake of [3H]M into hepatocytes at various times (0.25, 0.5, 0.75, and 1 min) was studied with retrieval of the incubation mixture (100 μl) into 250-μl polyethylene tubes containing silicon oil (density of 1.02 g/ml) atop 50 μl of 3 N NaOH. This was followed immediately by centrifugation (9650g). The radioactivities in the resultant supernatant (25 μl) and in the tip containing the centrifuged cells were quantified by liquid scintillation counting (LSC) (Beckman liquid scintillation counter, model 5801; Beckman Canada, Mississauga, ON, Canada).
Liver Perfusion System
Male Sprague-Dawley rats (300-350 g; Charles River, St. Constant, QC, Canada), housed under artificial light on a 12-h light/dark cycle, were fed ad libitum and allowed free access to water in accordance with the approved protocols of the University of Toronto Animal Care Committee. Before surgery, animals were anesthetized with an intraperitoneal injection of 50 mg/kg sodium phenobarbital. The surgical procedure and perfusion apparatus were identical to those described previously (Geng et al., 1995). Perfusate consisted of 20% washed fresh bovine RBCs (a kind gift of Ryding Regency Meat Packers Ltd., Toronto, ON, Canada), 4% BSA, 5 mM glucose, and a complement of 20 amino acids in Krebs-Henseleit bicarbonate solution buffered to pH 7.4 and was oxygenated with carbogen (95% oxygen and 5% carbon dioxide; Matheson, Mississauga, ON, Canada). Perfusate entered the liver through the portal vein and exited the hepatic vein in a single-pass manner. Throughout the experiment, perfusate pH and pressure in the inlet perfusate entering the portal vein cannula were monitored; the pH was adjusted by altering the inflow of oxygen or carbogen.
Plasma Protein Binding and RBC Distribution Studies. The binding of both M and M3G to albumin in perfusate plasma was determined by equilibrium dialysis (molecular cut-off at 12,000 to 14,000 Da; Spectrapor; Spectrum Medical Industries, Inc., Los Angeles, CA) at 37°C over a time course of 6 h, a predetermined time whereby equilibrium was reached. Equilibrium dialysis was conducted at seven M3G plasma concentrations (ranging from tracer to 500 μM) containing [3H]M3G (30,000 dpm/ml) and at various M concentrations (ranging from tracer to 662 μM) with perfusate medium containing [3H]M (25,000 dpm/ml). Comparison of the volumes of the protein and buffer compartments before and after sampling revealed that little volume change had occurred; leakage of protein (Lowry et al., 1951) into the buffer side was also insignificant. The equilibrated concentrations of [3H]M3G or [3H]M in the plasma (protein) and buffer sides were determined by LSC. The unbound plasma fraction (fu) was estimated by taking the ratio of the concentrations in buffer/plasma sides at equilibrium.
The RBC binding/partitioning of M was determined at various concentrations (ranging from tracer to 589 μM) containing [3H]M (11,500 dpm/ml) and [14C]sucrose (18,300 dpm/ml). Red cell binding of [3H]M3G was determined at tracer concentrations (87,500 dpm/ml) only; [14C]sucrose (63,500 dpm/ml) was again used as the extracellular reference. The red cell entry of 3H-labeled M (with unlabeled M) or M3G was appraised by the perturbation in 3H-labeled substrate plasma concentration, relative to that of [14C]sucrose, upon addition of an equal volume of blank blood perfusate that consisted of double the RBC content (40%) (Liu et al., 2005). The resultant mixture (20% RBC in 4% albumin) was incubated at 37°C in a rotating water bath, and aliquots were removed at predetermined time intervals. The hematocrit (Hct) of each perfusate sample was determined, and the blood perfusate was centrifuged to provide plasma. The plasma contents of [3H]M3G or [3H]M were assayed by LSC, whereas unlabeled M was assayed by HPLC. Samples were removed from the incubation mixture between 0.25 and 120 min. The equation below (eq. 1) shows that the concentrations of drug (C) in blood (in subscript), RBC, and plasma (p) are influenced by the hematocrit, when the drug partitions into red blood cells. The CRBC/Cp concentration ratio is obtained upon rearrangement of eq. 1, and the unbound fraction in blood, fu,b, is given by the quotient of the unbound fraction in plasma (fu) and the blood to plasma concentration ratio,
In Situ Perfused Rat Liver.M3G studies. Liver perfusion with M3G (10-262 μM) was carried out at 12 ml/min for 60 min. Only one concentration was used for each perfusion study. An MID injection dose (0.23 ml) containing 51Cr-labeled RBCs (0.47 ± 0.15 μCi), 125I-labeled albumin (2.8 ± 0.84 μCi), [14C]sucrose (0.632 ± 0.748 μCi), [3H]M3G (0.89 ± 0.71 μCi), unlabeled M3G, and D2O (0.096 ± 0.013 ml) was introduced into the inflow system by an electronically controlled HPLC injection valve at 30 min, as described previously (Geng et al., 1995). Outflow samples were rapidly collected immediately at successive 1- (35 samples), 2- (35 samples), and 3-s (20 samples) intervals. Bile was collected at 5-min intervals for 60 min. Sham experiments (without liver) were conducted to characterize the dispersion due to the inflow and outflow catheters as well as binding to the tubing. The hematocrit of the blood perfusate was determined for each experiment by the use of a hematocrit centrifuge (model MB microhematocrit centrifuge; IEC, Fisher Scientific, Mississauga, ON, Canada).
Morphine studies. Livers were perfused with morphine (9-474 μM) and [3H]M (50,000 dpm/ml) at 10 ml/min in a single-pass manner. After an initial equilibration period of 20 min with blank perfusate, perfusion with M was initiated, and outflow samples were collected at various times (5, 15, 22.5, or 25 and at 5 min thereafter). At least three or four constant values at steady state were used for determination of the average output plasma concentration of morphine, COut, whereas the mean of two reservoir-perfusate samples was taken as the input plasma concentration, CIn. The specific activity of the sample was used to convert disintegrations per minute per milliliter to nanomoles per milliliter. The extraction ratio E was estimated as 1 - COut/CIn, and CL was estimated as the product of flow (Q) and E. Bile was collected at 5-, 10-, or 15-min intervals into preweighed vials. The difference in weight before and after the bile collection was taken to be the volume of bile (density of bile was assumed to be 1). The excretion rate was given as the product of the bile flow rate and the concentration in bile. All blood and bile samples were stored at -80°C before HPLC analysis.
Assays.M and M3G. A specific HPLC assay was used for the quantification of [3H]M and [3H]M3G by radioelution and UV detection at 280 nm of the internal standard ethylmorphine (Doherty and Pang, 2000). Whole-blood perfusate samples (500 μl) were spiked with 50 μl of ethylmorphine (100 μg/ml); the sample was deproteinized with 50 μl of perchloric acid (5 M), mixed, and centrifuged. Immediately thereafter, 200 μl of the supernatant was subjected to reverse-phase chromatographic separation using a C18 column (Beckman Ultrasphere ODS; 5 μm, 4.6 mm × 25 cm) with a mobile phase based on that used by Svensson et al. (1982), buffered to pH 3.1. The HPLC system comprised a Shimadzu pump LC6A, UV spectrophotometric detector SPD-6A, autoinjector SIL-6A, and a system controller SCL6A. A gradient system (from 10 to 40% acetonitrile) was delivered at a flow rate of 1 ml/min. Bile samples were analyzed for [3H]M and [3H]M3G by HPLC following the dilution of 10 μl of bile with 50 μl of internal standard and 40 μl of water. A calibration curve consisting of varying known amounts of [3H]M or [3H]M3G as standards was constructed in similar manner.
For the quantification of unlabeled M, blood perfusate samples were processed using a similar procedure as described above. Blood perfusate samples (250 μl) were spiked with 50 μl of ethylmorphine (100 μg/ml); the sample was deproteinized with 25 μl of perchloric acid (5 M), mixed, and centrifuged, and 100 μl of the supernatant was injected onto the HPLC system (UV detection 280 nm). Another gradient system (from 17 to 30% acetonitrile) was used for separation of M from impurities. Linear calibration curves (r2 > 0.995) were constructed for each run (1-350 μg/ml M).
For the quantification of unlabeled M3G in MID studies, the sample preparation procedure was based on that reported by Evans and Shanahan (1995). In brief, a mixture of 500 μl of perfusate plasma, 25 μl of internal standard (1 mg/ml phenolphthalein glucuronide aqueous solution), and 1.25 ml of bicarbonate buffer (0.5 M; pH 9.3) were applied to a C18 Sep-Pak cartridge (Waters Associates, Milford, MA), previously conditioned with 10 ml of methanol, 10 ml of deionized water, and 1 ml of 0.5 M bicarbonate buffer. The cartridges were rinsed successively with 1.25 ml of bicarbonate buffer (5 mM; pH 9.3) and 0.5 ml of water. The analytes were eluted with 1.5 ml of methanol and evaporated to dryness at 45°C under a gentle stream of nitrogen. Samples were reconstituted in 150 μl of mobile phase (20% acetonitrile/80% dihydrogen phosphate buffer, pH 2.1, with 1.3 μM sodium dodecyl sulfate), and 20 μl was injected on to the HPLC system. The chromatographic separation used an isocratic flow of mobile phase at a rate of 0.9 ml/min with UV detection (220 nm). The typical retention times for M3G and phenolphthalein glucuronide were 12 and 17 min, respectively. Linear calibration curves (r2 > 0.995) were obtained for M3G (2.5-400 μM).
Multiple indicator dilution. The 51Cr, 125I, 14C, and 3H radioactivities and D2O in the dose and in outflow perfusate samples were performed as described previously by counting and Fourier transform infrared spectrometry (Geng et al., 1995).
Modeling.MID for M3G. The outflow radioactivity for each indicator was expressed as a fraction of the radioactivity of injected mixture per milliliter of blood and expressed in relation to the outflow recovery. The plasma recovery of M3G was assigned as 1 - biliary excretion (percentage of dose). The fraction of the M3G plasma outflow that is returning from the hepatocytes can thus also be obtained from the extraction ratio and the outflow profile of the reference indicator. An adequate fit to the data was attained after modification of the barrier-limited model of Goresky and co-workers (Geng et al., 1995; Pang et al., 1995). There is one parameter to be fitted, namely, the elimination coefficient, km3. Because of the low extraction of M3G, the peak portion of the outflow profile is almost exclusively throughput, and only the tail portion is sensitive to km3. As a consequence, km3 is poorly identifiable. The fractional recovery activity-time integrals (AUC{mi}) and integrals of product of fractional recovery and time (AUMC{mi}) for the preformed M3G were computed to estimate the mean transit time. The distribution space for each label was determined as the product of the transit time and the appropriate flow rates.
From the fractional outflow recovery curve of the vascular reference (the labeled RBC curve), the transport function of the injection and collection system of the outflow profile for the sham experiments (conducted with injection of an MID dose into the inflow and outflow catheters without the liver) was deconvoluted. A linear flow-limited transformation of the deconvoluted RBC curve was then carried out to generate a calculated first pattern for each diffusible reference, by selection of trial values for the ratio of extravascular to vascular distribution spaces and of t0, the common large-vessel transit time (Geng et al., 1995). The resulting curve was convoluted with the system transport function. The generated diffusible reference curve (for labeled albumin, sucrose, or water) was compared with that obtained experimentally, and the parameter values were repetitively refined until a best fit was obtained using a least-squares procedure (International Mathematics Statistical library; Visual Numerics, Houston, TX). Parameter estimates and their uncertainties were obtained using a weighting strategy and Jacobian matrix procedure. A similar process was used to gain the best-fit values for influx (km1), efflux (km2), and sequestration (km3) coefficients for M3G.
Physiologically based liver model for morphine. The physiological model that was used to describe the disposition of M and its metabolite M3G in the rat liver is depicted in Fig. 1. The model is composed of three compartments: the summed sinusoid and interstitial spaces or the extracellular compartment (of volume VE), the liver or tissue compartment (of volume Vt), and the bile compartment (of volume Vbile and flow Qbile). The drug is delivered under constant blood flow (Q) into the liver sinusoid. There is binding of M to RBC and albumin, and the unbound M in plasma (Mplasma,u) exchanges with that in tissue (Mt). The influx and efflux clearances of M are characterized as first-order parameters CLin, and CLef, respectively. Once M traverses across the sinusoidal membrane, M may undergo biotransformation to M3G (denoted by saturable constants for metabolism, Vmax,met and Km,met), endure excretion across the canalicular membrane (denoted by saturable constants for biliary transport, Vmax,ex and Km,ex), or efflux back to the sinusoid (CLef). The M3G metabolite, once formed within the hepatocyte, may efflux out to the perfusate blood (described by rate coefficient km2) or exit into the bile (described by rate coefficient km3); M3G may also re-enter the hepatocyte (described by rate coefficient km1).
Mass balance equations were developed to relate the events occurring during the single-pass perfusion of M (see Appendix A). Under first order conditions, the six equations (A1, A2, A3, A4, A5, A6) may be used to solve for the steady-state outflow concentrations of M and M3G in plasma and bile. The procedure involved arrangement of the elements in the differential equations as coefficients in a 3 × 3 or a 6 × 6 square matrix. Inversion of the matrices and manipulation of the resulting equations were performed with the program Theorist (Prescience Corporation, San Francisco, CA) using a Macintosh computer (PowerMac 9500; Apple Computer, Cupertino, CA). Additional manipulation of the solved concentrations further provided the influx clearance [PinS{mi}] of M3G from the apparent extraction ratio of hepatically-generated M3G and the extraction ratio of preformed M3G (Appendix B, eq. B4).
Fitting and Simulations. The data for M and M3G were fit to the differential equations in Appendix A (Scientist; MicroMath Inc., Salt Lake City, UT). The liver physiologically based model (Fig. 1) was used to fit the experimental data. The Michaelis-Menten parameters for metabolism (Vmax,met and Km,met) and excretion (Vmax,ex and Km,ex) of M were first approximated by regression of the formation rate of M3G (summation of the total rates of appearance of M3G in perfusate and bile) and the excretion rate of M versus the unbound logarithmic average concentration of M, Ĉu, respectively. Both the unbound input concentration, fu,bCIn and the unbound output concentration of M in blood, fu,bCOut, were used to calculate Ĉu, which equals [fu,b(CIn - COut)/[ln(fu,bCIn)/(fu,bCOut)]; these served as initial estimates for fitting. The number of unknown parameters in the model was considerably reduced because the first-order rate constants (km1, km2, and km3) that describe the influx, efflux, and excretion processes of M3G were provided by the MID studies. The uptake influx clearance of M (CLin) was obtained from hepatocyte uptake studies (average of homogeneous, PP, and PV cells). The efflux clearance (CLef) may be set equal to CLin or may be fitted. These, together with known volumes, binding, flows, and the kinetic constants, were used for fitting (Table 4). Various weighting schemes (unity, 1/prediction, and 1/prediction2) were used to optimize estimation of the efflux clearance and the Michaelis-Menten parameters for excretion and metabolism of M. Moreover, perfusion data were simulated based on the regressed Vmax and Km for metabolism and excretion of M, the transport clearances inferred from hepatocyte uptake studies (CLin = CLef) and first-order rate constants pertaining to M3G (km1, km2, and km3).
Data Comparison. Statistical comparisons of data sets were carried out with the Student's t test, and a P value of 0.05 was viewed as significant. For paired data sets, the paired t test was used.
Results
Hepatocyte Uptake. The enrichment of the select populations of PP and PV cells for uptake studies was verified by the PP:PV ratio of the enzymes alanine aminotransferase (2.86) and glutamine synthase (32.7), respectively. These values were similar to those observed by others (Tirona and Pang, 1999; Liu et al., 2005). Uptake of M by PP, PV, and homogeneous isolated rat hepatocytes showed that the uptake data were linear during the 1-min of study (data not shown). When the slopes corresponding to the uptake velocity, v, were presented against the concentration, concentration-independent uptake clearances of 0.0117, 0.00119, and 0.0123 ml/min per 106 cells were observed for the PP, PV, and homogeneous cells, respectively (Fig. 2). Noting that 1 g of liver consists of 1.25 × 108 cells (Lin et al., 1980), the clearances correspond to 1.47, 1.49, and 1.54 ml min-1 g-1, respectively; the average value was 1.50 ml min-1 g-1 liver.
Plasma Binding and RBC Distribution in Perfusate. The unbound fractions in plasma of both M (fu = 0.89 ± 0.07) and M3G (fu{mi} = 0.98 ± 0.09) were close to unity and were concentration-independent, indicating negligible or low binding of either compound to perfusate albumin (Table 1). The distribution equilibrium of M between plasma and RBC was attained at the earliest time, 0.25 min for all M concentrations examined (up to 589 μM). The blood/plasma concentration ratio was 1.36; eq. 1); and according to eq. 2, the CRBC/Cp concentration ratio was 3.4 at the hematocrit of 0.15. Based on these constants, the unbound fraction of M in blood, fu,b, was estimated to be 0.654 according to eq. 3, and this value was assumed to be the same for the unbound fraction of M in tissues, fu,t. For M3G, there was no evidence of RBC binding or distribution.
M3G Uptake and Excretion in MID Studies. Representative outflow profiles for the labeled substances injected into the portal vein of the liver are shown in Fig. 3A. Because of the rapid transit through the liver, the outflow curve of the labeled RBC (tracing the M bound to RBC) had the steepest up-slope and the highest and earliest peak. The 125I-labeled albumin (tracing the M3G bound to albumin) and [14C]sucrose (tracing unbound M3G) curves rose less quickly and decayed with slower slopes, resulting in lower and later peaks, and were more dispersed than that for RBCs. The D2O profile displayed the greatest dispersion, with a markedly delayed up-slope and down-slope due to its ability to permeate cellular, vascular, and interstitial spaces. Despite differences in dispersion due to their space of distribution, the recoveries or areas of the noneliminated references were complete. The outflow profile of [3H]M3G behaved very similar to the extracellular reference, [14C]sucrose, among all of the experiments; the recovery was also similar to that of [14C]sucrose. The mean transit time for each indicator was calculated from their respective outflow curves (Table 2) according to moment analyses, and the distribution spaces for the noneliminated indicators were generally similar to those observed for other perfused rat livers (Geng et al., 1995). The complete dose-recovery in outflow perfusate for M3G and the similarity in mean transit time of M3G (23.5 ± 4.5 s) to that for sucrose (22.2 ± 4.4 s) supported the view that M3G failed to gain hepatocellular entry.
Fits to the [3H]M3G and reference [14C]sucrose profiles shown in Fig. 3A are presented as semilogarithmic plots (Fig. 3B). The optimized parameters are summarized in Table 3. There was no concentration dependence displayed for any of the fitted rate constants for influx (km1) and efflux (km2) and for excretion (km3). The tracer [3H]M3G outflow profile was further resolved into the throughput and returning components of which the major component was the throughput. The reference curve ([14C]sucrose fit) was very similar to the [3H]M3G throughput, indicating that the majority of the administered [3H]M3G dose that was unbound passed through the liver without entering the hepatocytes. The mean value for the influx permeability surface area product (PinS{mi}) for M3G was 0.0014 ± 0.0004 ml s-1 g-1 was approximately 1/10 of the flow rate (Q) of 0.018 ± 0.003 ml s-1 g-1, suggesting the presence of a diffusional barrier for M3G entry into the hepatocyte. When mean data for all of the experiments was considered, no significant difference with concentration was evident between km1 (0.0051 ± 0.0021 s-1 and km2 (0.0020 ± 0.0022 s-1)); a low km3 (0.0041 ± 0.0044 s-1) was also observed (Table 3). The ratio of km3/km2 was approximately 2 (0.0041/0.0020), suggesting that the excretion of M3G into bile (with km3), possibly mediated through Mrp2, exceeds that for basolateral efflux (with km2), possibly mediated through Mrp3.
Steady-State Morphine Studies. M and M3G were detected in both perfusate blood and bile. The steady-state CL displayed concentration dependence (Fig. 4), as did the biliary excretion of M (Fig. 5A) and formation rate of M3G, estimated as the summed outflow rates of M3G in perfusate and bile (Fig. 5B). Regression of these rates, based on the optimized weighting of unity, against the logarithmic average input and output unbound concentration of M, Ĉu furnished the Vmax,ex (6.76 ± 1.2 nmol min-1 g-1) and Km,ex (23.4 ± 16.4 μM) for excretion and Vmax,met (50.4 ± 12.6 nmol min-1 g-1) and Km,met (74.5 ± 13.6 μM) for glucuronidation (Fig. 5; Table 4). The parameters suggest that excretion is a higher affinity, lower capacity system, whereas glucuronidation is a lower affinity and higher capacity system. The ratio of the metabolic to excretory intrinsic clearances (Vmax/Km)is 0.677/0.289 or 2.34.
Diffusion-Limited Entry of M3G. The presence of a transport barrier for sinusoidal uptake of M3G was further confirmed by the excretion data of M3G, arising from M3G administration and from formed M3G (Fig. 6A). The slope of the excretion rate versus formation rate of M3G yielded the apparent extraction ratio of formed M3G (0.63). This value was 12.3 times the extraction ratio of preformed M3G (regressed slope of excretion rate versus presentation rate of preformed M3G or 0.051) (Fig. 6A). When the bile flow rates for the M experiments, associated with higher M3G formed within the cell, were compared with those for the preformed M3G studies, a higher bile flow rate was apparent for the M experiments versus the M3G MID studies (1.48 ± 0.36 versus 1.17 ± 0.25 μl min-1 g-1; P < 0.05) (Fig. 6B), suggesting that the formed M3G promoted the choleresis. A similar scenario was found for 4-methylumbelliferyl glucuronide, for which a transport barrier was also implicated (Ratna et al., 1993).
Solutions to the extraction ratios of preformed or hepatically generated metabolite (Appendix B) further pointed to the estimation of the influx clearance of M3G, PinS{mi} (eq. B4). Knowing that the ratio of the extraction ratios of both hepatically generated and preformed M3G (eq. B4) was approximately 12.3, substitution of the flow rate (0.964 ml min-1 g-1, average of M and M3G studies) provided the influx, sinusoidal transport clearance for M3G as 0.081 ml min-1 g-1 or 0.0014 ml s-1 g-1, a value virtually identical to that found from MID studies (0.0014 ml s-1 g-1; Table 3). From the excretion data, the value of the apparent extraction ratio of formed M3G was approximately 0.63 (Fig. 6A), a value consistent with that obtained from the MID studies [0.67 or km2/(km2 + km3), constants from Table 3].
Physiological Modeling. Fitting with the physiologically based liver model (Fig. 1) successfully provided estimates of constants for metabolism, excretion, and efflux. The number of unknown parameters was considerably reduced when the uptake influx clearance (CLin) of M was provided from the hepatocyte uptake studies (average of homogeneous, PP, and PV cells to give 1.50 ml min-1 g-1 liver), and the first-order rate constants km1, km2, and km3 for M3G were estimated from the MID studies. The regressed Michaelis-Menten parameters for excretion (Fig. 5A) and metabolism (Fig. 5B) were used as initial estimates, and known volumes, binding constants, and flow were used for fitting (Table 4).
In some fits, the efflux clearance (CLef) was set equal to CLin, whereas in other instances, CLef or both CLin and CLef were fitted. The fit with CLef as an estimated parameter showed a high efflux clearance (2.5 ml min-1 g-1), but the corresponding estimates for Vmax and Km were generally associated with higher coefficient of variations (data not shown). Other fits with both CLin and CLef as fitted parameter performed poorly as well, showing again high coefficients of variation and poor model selection criterion (data not shown). The best fit resulted when CLef was assigned equal to CLin, with a weighting scheme of 1/prediction. The fitted parameters were in general agreement with those obtained from regression (Table 5). However, the fitted Km,ex was associated with a high coefficient of variation due to the large scatter in the data. The Vmax,met was 7-fold that of Vmax,ex, and the Km,met was slightly greater than Km,ex.
Simulation was also performed based on the regression parameters (Table 5), in conjunction with the assigned CLin and CLef (1.5 ml min-1 g-1). The simulated data based on the regression parameters generally matched the observed data, although the output concentration of M, the excretion, and outflow concentration of M3G exhibited systematic trends in comparison with those obtained from fitting (compare Figs. 7 and 8). The COut was consistently overpredicted (Fig. 7A), and the extraction ratio was underpredicted (Fig. 7B). The same comment could be made for the M3G data (Fig. 8B); the appearance of M3G in outflow perfusate and bile was consistently underestimated as was the total formation rate of M3G. The trend for the biliary excretion of M was not apparent due to the scatter in the data (Fig. 8A). In comparison, fitting provided a better match to the observed data, namely for the outflow concentration of M, E, and for the excretion and metabolism of M (compare Figs. 7 and 8).
Discussion
Drug transport has been implicated to exert a rate-limiting role in the hepatic drug processing of some drugs. In such an instance, the influx clearance should be lower or comparable with those for elimination. This was found for ethacrynic acid (Tirona et al., 1999) but not for enalapril (Abu-Zahra and Pang, 2000) or digoxin (Liu et al., 2005), although transporters were involved for sinusoidal influx of all of these drugs. Strategies to discern the rate-determining step in liver often involve the appraisal of uptake from hepatocytes, in vitro metabolic experimentation, and finally, the integration of data in liver perfusion studies. Data in vitro would aid in modeling of the liver consisting of intact sinusoidal membranes for influx and efflux and metabolic enzymes and excretion transporters behind the sinusoidal barrier for drug elimination (Tirona et al., 1999; Abu-Zahra and Pang, 2000; Liu et al., 2005). Additional transport and removal data on the behavior of the preformed metabolite would further enhance prediction of kinetic data of the formed metabolite.
To mimic the in vivo situation, it was further necessary to ascertain the extent of binding of both species in both perfusate plasma and blood to rule out the limitation due to binding (Liu et al., 2005). Our results on morphine binding to perfusate BSA (4%) and distribution into RBCs (Table 1) were similar to those from other binding studies with human or animal plasma (Olsen, 1975; Leow et al., 1993) or RBCs (Mistry and Houston, 1987). The events were in rapid equilibrium and unlikely to be rate-limiting for the poor-bound morphine. The unbound fractions of morphine in rat plasma have been reported at 0.85 (Mistry and Houston, 1987) and 0.86 (Baggot and Davis, 1973). The absence of the endogenous proteins γ-globulin and α1-acid globulin in the perfusate, both of which bind morphine in humans (Olsen, 1975; Leow et al., 1993), may account for the slight difference observed.
We also ascertained the uptake of M3G since different methods, ranging from whole animal (Smith et al., 1973; Ouellet and Pollack, 1995) to recirculating liver perfusions (Ouellet and Pollack, 1995), have reported substantial biliary excretion of the administered M3G. However, it must be recognized that there are fundamental differences between single-pass and recirculating livers regarding interpretation of the formed metabolite data. The recirculating system entails the total excretion of M3G as a nascently formed as well as a preformed metabolite into bile, since the formed metabolite effluxes back into the circulation (de Lannoy et al., 1993). By contrast, the single-pass MID outflow curve for [3H]M3G in this study (Fig. 3, A and B; Tables 2 and 3) clearly showed that the major proportion of M3G passed through the liver (throughput component) without ever entering the hepatocytes. Consequently, only a small percentage of the administered M3G dose (3.69 ± 1.65%) was excreted into the bile. This barrier for morphine 3-glucuronide was further shown as different extraction ratios between the formed M3G (excretion rate versus formation rate of M3G) and preformed M3G (Fig. 6A) and the slightly higher choleresis associated with formed M3G (Fig. 6B). Such events are similar to those of 4-methylumbelliferyl glucuronide for which a sinusoidal barrier is implicated (Ratna et al., 1993). The existence of a membrane barrier at the sinusoidal membrane for M3G has been postulated by O'Brien et al. (1996) who claimed that the lack of change in the biliary extraction ratio of hepatically generated M3G with perfusate flow rate was due to a barrier effect at the sinusoidal membrane; the assumption made was that the influx and efflux membrane permeabilities were equal. Moreover, the MID studies provided a direct low estimate of the in influx clearance of 0.0014 ml s-1 g-1 for M3G, a value that is less than 10% the value of flow (Table 3). The octanol/water-partitioning studies conducted in our laboratory (unpublished data) further suggest that the ratio of M3G that partitioned into the octanol phase versus water was 0.0079, whereas that for M was 0.74. All of the above-mentioned information is consistent with the barrier-limited transport of M3G.
Uptake studies showed the concentration-independent uptake of morphine among even, periportal, and perivenous hepatocytes (Fig. 2). We failed to observe the saturable uptake with a low Km of approximately 48 μM observed by Iwamoto et al. (1978). The difference may possibly be attributed to the use of (-)-morphine in our case. By contrast, concentration-dependent glucuronidation and biliary excretion of M were observed, resulting in decreases in the steady-state clearance (Figs. 4 and 5). The precise roles of transport versus enzymes were further discerned with modeling. The simple model (Fig. 1) was able to provide fitted Km and Vmax values that were generally similar to those obtained upon regression. Lower first-order intrinsic clearances for metabolism or excretion (1.1 and 0.20 ml min-1 g-1) were obtained in comparison with that for transport (1.5 ml min-1 g-1). The closeness in values of the regression and fitted parameters suggest that Ĉu approximates well the tissue concentration and infers that, overall, transport is not the rate-limiting step in the removal of morphine.
It must be recognized that rat Ugt activities are often described to be concentrated in the perivenous region of the liver (Knapp et al., 1988). We attempted to study morphine glucuronidation with microsomes derived from the PP and PV hepatocytes. However, the rates were low, the disintegrations per minute were approximately 2 to 3 times background, and the results were inconclusive. Hence, the influence of Ugt heterogeneity was further ascertained by simulation only since the fit of the data to the more complex zonal liver model (Abu-Zahra and Pang, 2000; Liu et al., 2005) was compromised due to nonlinearity in the system. For simulation, the zonal model (zones 1, 2, and 3 for the PP, midzonal, and PV regions) proposed earlier (Abu-Zahra and Pang, 2000; Liu et al., 2005) was used. The Vmax for glucuronidation in zones 1, 2, and 3 were assigned 20, 35, and 45% of the total Vmax,met (estimated with fitting; Table 5),whereas the Vmax for the excretion of M in zones 1, 2, and 3 was set as one-third the total Vmax,ex (estimated with fitting; Table 5). Values of the Km for metabolism and excretion were held constant for all zones and set the same as the regressed values. The volumes were assigned one-third the total volume. Simulations so performed revealed that the outflow concentrations of M and M3G as well as the excretion rates were virtually unperturbed by the perivenous enrichment in UGT compared with an even distribution of the UGT (data not shown). The outcome is expected since the extraction ratio of morphine is low at increasing or saturating input concentrations, which would decrease E even further. Under this condition, all enzymes are recruited by the substrate and enzyme heterogeneity exerts little outcome on the kinetics of morphine removal.
The concentration-dependent excretion of M into bile (Vmax,ex and Km,ex) may be indicative of active transport by P-gp across the canalicular membrane. Although M has been implicated as a P-gp substrate in studies involving predominately brain penetration (Kharasch et al., 2003), altered pharmacokinetic or pharmacodynamic effects have not been consistent (Drewe et al., 2000; Wandel et al., 2002), thereby designating M a weak P-gp substrate. Such inconsistencies may indicate that simultaneous passive diffusion and P-gp transport are likely mechanisms of cellular transport of M in other organs, such as the brain and small intestinal mucosa. Although our data clearly show nonlinearity in the excretion rate of M (Fig. 8A), it is probable that P-gp plays a part on M excretion in the canalicular membrane and is readily saturable, implying differences in activity and expression compared with different organs/tissues.
In summary, a comprehensive investigation into the disposition of morphine in the in situ perfused rat liver preparation was presented. Perfusion of the glucuronidated metabolite M3G together with MID noneliminated reference indicators in the injection indicated the presence of a substantial diffusional barrier at the sinusoidal membrane, resulting in less than 4% of the M3G dose entering the bile. This was further supported by consideration of solved explicit equations based on the pharmacokinetic model and observations of increased bile flow due to the presence of hepatically generated M3G. Utilization of a simple, physiologically based pharmacokinetic model allowed us to probe M uptake, M3G formation, and sinusoidal and canalicular transport of M3G via fitting. Hepatocytic uptake of morphine is deemed to be rapid enough and does not rate-limit the metabolism and excretion M.
Appendix A: Transfer Equations to Describe M and M3G in the Physiological Model
Single-Pass Perfusion. The liver is subdivided into three compartments (sinusoid, tissue, and bile; Fig. 1). A constant input of drug, CIn, is delivered under constant hepatic blood flow rate (Q) into the rat liver. The assumptions were that M and M3G are unbound in tissue; Mbile and M3Gbile are the amounts in bile. The following mass balance equations describe drug and metabolite {mi} concentrations in a single-pass perfusion system.
For drug and metabolite in liver sinusoid,
For drug and metabolite in liver tissue,
For the excretion rates of drug and metabolite into bile,
Appendix B: Extraction Ratios of M and Preformed and Hepatically Generated M3G
The following equations were solved for the model presented in Fig. 1 under first-order conditions where the parameters for influx, efflux, excretion, and metabolism were presented as the intrinsic clearances CL1, CL2, CL3 or (Vmax,ex/Km,ext), and CL4 or (Vmax,met/Km,met), respectively. The steady-state extraction ratio for parent drug morphine (E) in the liver is The steady-state extraction ratios for preformed morphine 3β-glucuronide [E{pmi}] in the liver is The steady-state extraction ratio for hepatically generated morphine-3β-glucuronide [E{mi}] in the liver is
All other parameters are as previously defined, and the equations were the same as those from de Lannoy et al. (1993). Since PinS{mi} = VEkm1, division of eq. B3 by eq. B2 and rearrangement yields the following:
Acknowledgments
We thank Dr. Andreas J. Schwab (McGill University, Montréal, QC, Canada) for contribution to the fit to the M3G data and for invaluable discussion of the data and Ford Barker for excellent technical assistance.
Footnotes
-
This study was supported by the Medical Research Council of Canada Grant MOP64350.
-
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
-
doi:10.1124/jpet.105.100446.
-
ABBREVIATIONS: M, morphine; M6G, morphine 6β-glucuronide; M3G, morphine 3β-glucuronide; (-)M, (-)-morphine; MDR, multidrug resistant; P-gp, P-glycoprotein; Mrp, multidrug resistance-associated protein; PSC833, valspodar; MID, multiple indicator dilution; HPLC, high-performance liquid chromatography; BSA, bovine serum albumin; PP, periportal; PV, perivenous; LSC, liquid scintillation counting; RBC, red blood cell; Hct, hematocrit; CL, clearance; {mi}, metabolite.
-
↵1 Current affiliation: Victoria College of Pharmacy, Monash University, Melbourne, Australia.
- Received December 27, 2005.
- Accepted February 2, 2006.
- The American Society for Pharmacology and Experimental Therapeutics