Abstract
We examined three primary variables in the preparation of human liver microsomes. In three experiments, each using three livers, we manipulated 1) the force of the first centrifugation (9,000, 10,500, or 12,000g); 2) the presence of sucrose in the homogenization buffer; and 3) the number of homogenizing strokes (6, 8, or 10). Sedimentation plots for the marker enzymes succinate dehydrogenase, NADPH cytochrome P450 reductase (reductase), and glutathione S-transferase in the resulting premicrosomal, microsomal, and cytosolic fractions suggest that enhanced purity of microsomes can be obtained by reducing force of centrifugation, including sucrose, and increasing the number of homogenization strokes. Each microsomal fraction was also assayed for protein content, cytochrome P450, NADH cytochromeb5 reductase, cytochromeb5, absorbance at 420,p-nitrophenol hydroxylation, tolbutamide hydroxylation, dextromethorphan N- and O-demethylation, glucuronidation of morphine and 1-naphthol, and ester cleavage ofp-nitrophenolacetate. These microsomal indicators were ranked and tested for statistical differences. The use of 9000g statistically increased optimal recovery (per gram of liver) and specific activity (per milligram of protein). The inclusion of sucrose improved activity specific to reductase activity. Ten homogenization strokes improved activity specific to reductase activity. Substrate-dependent activities of dextromethorphanO-demethylation to dextrorphan and theN-demethylation of l-α-acetylmethadol (LAAM) to norLAAM and dinorLAAM were compared in microsomes prepared with or without sucrose and microsomes prepared using 9,000 or 12,000g force, respectively. No significant differences were found in the concentration-dependent activities. Variation of the methods used to prepare human liver microsomes can significantly affect the recovery and specific activity of microsomal components; however, they do not appear to affect enzyme kinetics.
Human liver microsomes (HLM2) are used widely to characterize the role of cytochrome P450s (P450) and other enzymes in drug metabolism. The generalized, differential centrifugation procedure used to prepare HLM is as follows. Typically, liver samples are homogenized and centrifuged at a lower force to form a crude pellet of cell debris, nuclei, peroxisomes, lysosomes, and mitochondria (premicrosomal pellet). The resulting supernatant is then centrifuged at a higher force to precipitate the microsomes. The microsomal pellet is resuspended in a final suspension buffer and is then ready for use in incubation studies. There are, however, many different procedural variables associated with this preparation (Table1) (Boobis et al., 1980; Raucy and Lasker, 1991; Kharasch and Thummel, 1993; Guengerich, 1994; Rodrigues et al., 1994). These variations include the number of strokes, or passes, used to homogenize the liver samples (e.g., 4–8); the forces with which the samples are centrifuged (e.g., the centrifuge force for the first spin ranges from 9,000g for 20 min to 18,000g for 10 min and between 100,000 and 143,000g for 60–90 min for the second spin); and content of the homogenization and final suspension buffers (EDTA, potassium chloride, glycerol). While not presented in Table 1, inclusion of sucrose in the homogenization buffer was considered critical in initial experiments on differential centrifugation (de Duve, 1971), and it is used in the preparation of microsomes in experimental animals (Papac and Franklin, 1988). In some procedures, steps are repeated for more thorough extraction. Relative volumes, concentrations, and dilutions of the samples also vary.
In this study, we examined three variables in the preparation of HLM that we felt might have the greatest impact on microsomal recovery and purity. We manipulated the following HLM preparatory variables: the force of the first centrifugation, the presence of sucrose in the homogenization buffer, and the number of homogenizing strokes. Cell fractions were evaluated for separation of marker enzymes. The microsomal data were analyzed for 12 positive indicators of microsomal purity (protein content, NADPH cytochrome P450 reductase, P450, NADH cytochrome b5 reductase, cytochromeb5, p-nitrophenol hydroxylation, tolbutamide hydroxylation, dextromethorphan N- andO-demethylation, glucuronidation of morphine and 1-naphthol, and ester cleavage of p-nitrophenolacetate) and three negative indicators of microsomal purity (absorbance at 420, succinate dehydrogenase, and glutathione S-transferase). The ideal fractionation protocol would maximize specific activity and recovery of endoplasmic reticulum proteins, minimize mitochondrial and cytosolic enzymes, and minimize denaturation. In two preparations exhibiting significant differences, the kinetics of two P450-mediated reactions, the P450 2D6-specific O-demethylation of dextromethorphan (Kupfer et al., 1984; Schmider et al., 1997) and the P450 3A4N-demethylation of l-α-acetylmethadol (LAAM) (Moody et al., 1997), were measured to determine whether purity influences kinetic variables in crude microsomal fractions. It is important to understand how current techniques used to prepare HLM may compare and how to improve microsomal purity.
Experimental Procedures
Materials.
p-Nitrophenol was obtained from Aldrich (Milwaukee, WI). Hydroxytolbutamide was provided by Hoechst AG (Frankfurt, Germany). Dextrorphan-d-tartrate was obtained from ICN (Irvine, CA). 1-Naphthol was obtained from J.T. Baker, Inc. (Phillipsburg, NJ). Morphine was obtained from Merck and Co., Inc. (West Point, PA). Chlorpropamide was obtained from Pfizer (Brooklyn, NY). Levallorphan tartrate and 3-methoxymorphinan were obtained from RBI (Natick, MA). Bovine serum albumin, 1-chloro-2,4-dinitrobenzene, cytochromeC, 2,6-dichloroindolephenol, dextromethorphan, Folin and Ciocalteu's phenol reagent, d-glucose 6-phosphate, glucose-6-phosphate dehydrogenase, glutathione, NADP, β-NADPH, β-NADH, p-nitrophenolacetate, potassium ferricyanide, sodium dithionite, succinic acid, sucrose, tolbutamide, trichloroacetic acid, and UDP-glucuronic acid were obtained from Sigma (St. Louis, MO). The liver samples were obtained from the International Institute for the Advancement of Medicine (Exton, PA) (Table2).
HLM Preparation.
To allow for sufficient microsomes to perform all assays, three livers were prepared for each of the three experiments (the force of the first centrifugation, homogenization buffer, and homogenizing strokes). Approximately 10 g of liver per experimental treatment was allowed to thaw in a room temperature homogenization buffer (0.1 M potassium phosphate buffer, pH 7.4, containing 0.125 M potassium chloride and 1.0 mM EDTA). After transfer to 25 ml of chilled homogenization buffer (plus or minus 0.25 M sucrose in the buffer experiment), livers were minced thoroughly with scissors and homogenized with 10 strokes (6, 8, or 10 strokes in the strokes experiment) using a Teflon-glass homogenizer (870 rpm). Strokes were even and steady, lasting approximately 15 s for passage, except for the first two strokes where greater pressure and time were spent on material on the bottom of the glass tube. The tube was submersed in a small bucket of ice and water during all homogenization. The homogenate was diluted to 4 volumes of sample weight (approximately 40 ml). The samples then were centrifuged at 12,000g (9,000, 10,500, or 12,000gin the force experiment) in a Sorvall RC-5B with a Sorvall SA-600 rotor for 20 min (Sorvall, Newton, CT). The supernatant from the first centrifugation was removed, the mitochondrial pellet was resuspended in 25 ml, and the centrifugation was repeated. The supernatants were combined and centrifuged at 138,000g in a Sorvall Ultra Pro 80 with a Sorvall T-1270 rotor for 60 min. The upper lipid layer was removed and the cytosolic supernatant collected. The microsomal pellet was resuspended in 0.125 M KCl, 0.1 M Tris (pH 7.4) with three homogenization strokes, and the 138,000gcentrifugation for 60 min was repeated. The microsomal pellet was resuspended in incubation buffer with six strokes and brought to a final volume of 26 ml. Samples were stored at −70°C.
Assays.
All resulting liver fractions (premicrosomal, microsomal, and cytosolic fractions) were subjected to the following four spectrophotometric assays performed in duplicate: protein content (Lowry et al., 1951), NADPH cytochrome P450 reductase (Masters et al., 1967), succinate dehydrogenase (Bachmann et al., 1966), and glutathioneS-transferase conjugation of 1-chloro-2,4-dinitrobenzene (Habig and Jakoby, 1981). Each microsomal fraction was tested with 11 additional assays performed in duplicate. P450, cytochromeb5, and absorbance at 420 contents were determined from difference spectra (Omura and Sato, 1964).p-Nitrophenol hydroxylation (Papac and Franklin, 1988), NADH cytochrome b5 reductase (Rogers and Strittmatter, 1973), and hydrolysis of p-nitrophenolacetate (Ashour et al., 1987) were determined colorimetrically. The glucuronidations of morphine and 1-naphthol were determined by HPLC (Liu and Franklin, 1984).
Tolbutamide hydroxylation was determined by using a modification of our previously published procedure (Ho and Moody, 1993) as follows. The incubation, in a final volume of 200 μl, consisted of 0.4 mg of microsomal protein, 500 μM substrate, a NADPH-generating system providing 1 mM NADP, 6 mM glucose 6-phosphate, 0.4 units of glucose-6-phosphate dehydrogenase, and 75 μl of 0.15 M Tris buffer with 5 mM magnesium chloride (pH 7.4). The assay was initiated by the addition of the NADPH-generating system in a 37°C shaking water bath and terminated after 0, 15, or 30 min by the addition of 25 μl of 3 N hydrochloric acid. The internal standard was added (25 μl of 0.05 mg/ml of chlorpropamide in buffer), and the mixture was vortexed and extracted with 2 ml of diethyl ether, vortexed, and centrifuged for 10 min at 600g (IEC model K, size 2; IEC, Needham Heights, MA). The organic layer was collected, air dried at room temperature, reconstituted with 150 μl of mobile phase, and vortexed for 5 min (IKA-Vibrax-VXR). The analysis was performed by HPLC using a Waters 996 photodiode array detector monitoring 233 nm, a Waters 600 Controller, a Waters 717 plus Autosampler (injecting 25-μl sample), and a Symmetry C18 column (250 × 4.6 mm, 5 μm; Waters, Milford, MA). The mobile phase consisted of methanol/0.01 M monobasic ammonium phosphate (pH 5.4) (60:40, v/v). The flow rate was 1 ml/min at room temperature.
Dextromethorphan N- and O-demethylation activity of HLM was modified from the method of Schmider et al. (1997) as follows. The incubation, in a final volume of 200 μl, consisted of 0.4 mg of microsomal protein, 250 μM substrate, a NADPH-generating system providing 1 mM NADP, 6 mM glucose 6-phosphate, 0.4 units of glucose-6-phosphate dehydrogenase, and 75 μl of 0.15 M Tris buffer with 5 mM magnesium chloride (pH 7.4). The assay was initiated by the addition of the NADPH-generating system in a 37°C shaking water bath and terminated after 0, 15, or 30 min by the addition of 50 μl of 30% trichloroacetic acid. The internal standard was added (50 μl of 6 μg/ml of levallorphan in buffer), and the mixture was vortexed and centrifuged for 10 min at 930g (IEC model K, size 2). Analysis of the supernatant was performed by HPLC using a Waters 474 scanning fluorescence detector monitoring 280-nm excitation and 310-nm emission wavelengths, a Waters 600 Controller, a Waters 717 plus Autosampler (injecting 15-μl sample), and a Luna C18(2) column (150 × 4.6 mm, 5 μm; Phenomenex, Torrance, CA). The mobile phase consisted of acetonitrile containing 0.06% triethylamine/0.05 M potassium phosphate (pH 2.8) containing 0.06% triethylamine (28:72, v/v). The flow rate was 1 ml/min at room temperature.
Determination of Kinetic Parameters.
Dextromethorphan O-demethylation was determined as described above, except that the substrate concentration was varied from 1 to 1500 μM. For each treatment group compared, each substrate concentration was performed in duplicate. LAAMN-demethylation to norLAAM and dinorLAAM was measured for substrate concentrations ranging from 0.3 to 1000 μM using our previously described gas chromatography/mass spectrometric method (Moody et al., 1995).
For both reactions monitored, the substrate concentrations were accurately determined at the lower concentrations. When there was evidence of substrate depletion over the incubation time, the substrate concentration used in calculations was the average of the added and measured concentration. LAAM N-demethylation was measured as the combined formation of norLAAM and dinorLAAM. Over the concentration ranges monitored, product formation from LAAM did not fit linear kinetics. A sigmoidal Vmax equation,v = (VmaxSn)/(S50n+ Sn), proposed by Ueng et al. (1997) was used to fit the data where n was determined from plots of log [v/Vmax − v] versus log S. For the latter, theVmax was initially estimated from a graph of S versus v, with reiterative determinations ofn as the Vmax was approached with the sigmoidal equation that was determined using the nonlinear regression function of KaleidaGraph (version 3.09, Synergy Software, Reading, PA). For concentrations of LAAM less than 100 μM, linear results were found in Eadie-Hofstee plots.Vmax and Kmvalues were determined from the y-axis intercept and the negative of the slope of linear regression analysis of these plots (Cricket Graph 1.3, Cricket Software, Malvern, PA). TheO-demethylation of dextromethorphan to dextrorphan was biphasic in Eadie-Hofstee plots. Vmax andKm values were determined using the nonlinear regression function of Systat (version 5.2.1, Statistical Solutions, Saugus, MA) to solve the following equation:v = [(Vmax1)(1 +Km1/S)−1] + [(Vmax2)(1 +Km2/S)−1].
Data Analysis and Statistics.
The microsomal data for 12 of the assays were considered positive indicators of microsomal purity: protein content, NADPH cytochrome P450 reductase, cytochrome P450, NADH cytochromeb5 reductase, cytochromeb5, p-nitrophenol hydroxylation, tolbutamide hydroxylation, dextromethorphan N- andO-demethylation, glucuronidation of morphine and 1-naphthol, and ester cleavage of p-nitrophenolacetate. The microsomal data for three of the assays were considered negative indicators of microsomal purity: absorbance at 420, succinate dehydrogenase, and glutathione S-transferase. For each experiment, the 15 indicators were expressed for recovery (per gram of liver), specific activity (per mg of protein), and for the ratio of activity to cytochrome P450 reductase activity. For each expression system, the activities (or contents) were ranked across an experimental group (e.g., 1, 2, 3 for the highest to lowest activities of positive indicators versus the 6, 8, and 10 strokes). To be ranked differently, the activities must differ by at least 5%. For the two-variable sucrose in homogenate experiments, ties were ranked as 1.5, 1.5. For the three-variable experiments, the tie rankings were established such that a difference of 1 would exist between groups and the total of all three rankings was always 6. Therefore when the two highest activities were tied, a ranking of 1.7, 1.7, 2.7 was assigned. When the two lowest activities were tied, a ranking of 1.3, 2.3, 2.3 was assigned. When the middle activity did not differ from the others by 5% but the extremes did, a ranking of 1.5, 2.0, 2.5 was assigned. Once rankings were assigned, experimental groups were then compared using one-way analysis of variance (p < 0.05) with the Tukey post hoc test (p < 0.05) (Zar, 1984).
The Student's paired t test (Zar, 1984) was used to compare the substrate concentration-dependent experiments with LAAM and dextromethorphan. Activities determined at any one concentration for the compared groups from a human liver were treated as matched pairs (e.g., the activity determined at 100 μM dextromethorphan for liver CO 06 prepared without sucrose was a matched pair to the activity determined at 100 μM dextromethorphan for liver CO 06 prepared with sucrose). To permit comparison between different concentrations, the percentage of difference between matched pairs was determined and used as the variable as previously described (Moody et al., 1999). Therefore, a comparison could be made between microsome preparations at different substrate concentrations using all three livers for each comparison. Combining the data in this manner increased the power of our test.
Results and Discussion
Sedimentation Profiles.
The separation of organelle marker enzymes was initially evaluated using the qualitative plots first used by de Duve and coworkers (1955)to demonstrate patterns of enzyme distribution following differential centrifugation. The specific activities (per mg of protein) of mitochondrial succinate dehydrogenase, endoplasmic reticular NADPH cytochrome P450 reductase, and cytosolic glutathioneS-transferase were plotted against the percentage of total protein recovered in the initial pellet, the microsomal pellet, and the cytosol (Fig. 1). In this manner, not only is the specific activity displayed, but the area for each fraction is proportional to the percentage of activity recovered. Increasing the number of homogenization strokes appears to release more cytosolic material from the pellets, but it decreases specific activity of the reductase in the microsomes. In the force experiment, the specific activity and recovery (per gram of liver) in the microsomal and cytosolic fractions appear to be optimal at 9,000g. In the homogenization buffer experiment, the addition of sucrose presents a tradeoff. The sucrose buffer appears to decrease both NADPH cytochrome P450 reductase specific activity and recovery in the microsomal fraction but also decreases succinate dehydrogenase contamination in the microsomal fraction (Fig. 1).
Recovery and Specific Activity in Microsomes.
When direct comparison (one-way analysis of variance) of preparation variables was made between any one of the 15 assays performed, no significant differences were found (data not shown). This was due greatly to the large variation between any three liver samples. To statistically evaluate the different treatments, a method was needed to group the different assays to increase the power of the test. We therefore used a ranking scheme that allowed us to group the results of all the assays and each liver per assay. An example of the ranking scheme is provided in Table 3. A summary of the statistical results is provided in Table4. In the force experiment, both optimal recovery and specific activity (or specific content) increased significantly with decreasing force. The addition of sucrose to the homogenization buffer significantly increased the relative microsomal specific activity (ratio of activity or content to NADPH cytochrome P450 reductase activity). In the strokes experiment, the relative specific microsomal activity was positively influenced by an increasing number of strokes.
Kinetics in Different Microsome Preparations.
As a further test of the impact of the microsomal preparation method on drug-metabolizing enzymes, detailed substrate concentration-dependent experiments were performed using a P450 2D6-specific pathway, dextromethorphan O-demethylation (Kupfer et al., 1984;Schmider et al., 1997), and a P450 3A4-specific pathway, LAAMN-demethylation to norLAAM and dinorLAAM (Moody et al., 1997). Dextromethorphan O-demethylation was compared in microsomes from livers prepared in homogenization buffer that did, or did not, contain sucrose. LAAM N-demethylation to norLAAM and dinorLAAM was compared in microsomes from liver homogenates that were subjected to initial centrifugation at 9,000 or 12,000g.
The kinetics of dextromethorphan O-demethylation in HLM have been well studied. As previously described (Schmider et al., 1997), dextromethorphan O-demethylation displayed biphasic properties in Eadie-Hofstee plots (Fig.2), and kinetic parameters could be estimated using nonlinear regression (Table5). For livers CO 07 and CO 08 the high-affinity (Km1) and low-affinity (Km2) components of the biphasic equation for dextromethorphan O-demethylation were similar (Table 5). While the Km1 for liver CO 06 was in the same range, its Km2 was much higher. The possibility that a third enzyme with intermediate affinity may have been involved in this liver preparation could not be adequately approached with the software available, but could represent an explanation for this extreme variation. No readily apparent differences in kinetic parameters were noted for livers prepared with sucrose versus without sucrose in the homogenization buffer (Table 5).
We have recently found that both the N-demethylation of LAAM and subsequent N-demethylation of norLAAM to dinorLAAM are carried out by P450 3A4 at substrate concentrations of 1 to 10 μM (Moody et al., 1997). The kinetics of LAAM N-demethylation, however, have not yet been studied in HLM. Because the previous data suggest that both LAAM and norLAAM are P450 3A4 substrates, we analyzed the combined formation of both products, norLAAM and dinorLAAM. Over a substrate concentration range of 0.3 to 1000 μM, nonhyperbolic curves were found for product formation (Fig.3A). These curves were amenable to fitting with a sigmoidal Vmax equation that permitted an initial estimate of kinetic parameters and a comparison in livers prepared using different centrifugation forces (Table6). While theVmax values for livers CO 04 and CO 05 were similar, a much higher Vmax was estimated for liver CO 03. The estimated S50 values were different for each of the three livers. No readily apparent differences between livers prepared using either 9,000 or 12,000g centrifugation forces were noted.
While these sigmoidal curve parameters are useful for this initial comparison, their value for future studies to examine in vitro clearance of LAAM is not so clear. Although other equations have been described to fit nonhyperbolic kinetics (Korzekwa et al., 1998), further studies are needed to ascertain the mechanism for these kinetics with LAAM metabolism before a reasonable choice can be made. The nonhyperbolic substrate versus activity curves forN-demethylation of LAAM may have arisen from a number of factors that could include substrate inhibition, product inhibition, involvement of other P450s, or the involvement of other metabolic pathways (e.g., ester hydrolysis, ring hydroxylation). Our initial studies on P450 3A4 involvement focused on lower substrate concentrations that were of therapeutic relevance. When we restrict the analysis to substrate concentrations less than 100 μM, the curves approach linearity in Eadie-Hofstee plots (Fig. 3B). Under these limitations, the Vmax values for livers CO 03, CO 04, and CO 05 prepared using 12,000g force were 420, 225, and 101 (pmol)(min)−1(mg of protein)−1, respectively, with correspondingKm values of 10.0, 10.7, and 45.6 μM. As it would be highly unlikely for plasma, or even liver, concentrations of LAAM to exceed 100 μM in even a toxic situation, these latter parameters may prove more useful for in vitro-in vivo modeling.
A statistical comparison of kinetic parameters, particularly when biphasic kinetics are present, is a complex matter, with no readily available solutions. We therefore took the approach that incubations with the same substrate at the same concentration represented matched pairs, and used Student's paired t test for evaluation (Zar, 1984). As an initial test of this statistical approach, we first compared all of the replicate results for the dextromethorphanO-demethylations with the first replicate being compared with the second replicate. As expected, there was no significant difference between the replicates (Table7). No significant difference was found in either dextromethorphan O-demethylation activities in microsomes prepared with homogenates with or without sucrose or in LAAMN-demethylation activities from microsomes prepared with the initial centrifugation at 9,000 versus 12,000g (Table 7).
Conclusions.
HLM are used widely to characterize the role of P450s and other enzymes in xenobiotic metabolism. There are, however, many procedural variables associated with the preparation of microsomes. We examined three of these variables to better understand how preparatory differences might affect microsomal purity and the comparison of microsomal data between research laboratories. The three variables chosen, centrifugation force, presence of sucrose in the homogenization buffer, and homogenization strokes, are among those that can have the largest impact on differential centrifugation (de Duve, 1971) and that have shown the widest variation between laboratories (Table 1). Although none of our experiments indicated a significant difference between procedures by all three measures of microsomal purity and recovery, three trends did develop.
Our data suggested that the use of 9,000g (rather than 10,500 or 12,000g) in the first centrifugation statistically increased optimal recovery (per gram of liver) and activity (or content) per milligram of protein. The inclusion of 0.25 M sucrose in the homogenization buffer improved activity (or content) per reductase activity. Ten homogenizing strokes (rather than six or eight) improved activity (or content) specific to reductase activity.
As a further test of the impact of microsomal preparation method on drug-metabolizing enzymes, two detailed substrate concentration-dependent experiments were performed. Using a P450 2D6 pathway, we compared the effect of sucrose in the homogenization buffer on the kinetic parameters of dextromethorphanO-demethylation. We also compared the effect of initial centrifugation forces of 9,000 and 12,000g on the kinetics of a P450 3A4-specific pathway, LAAM N-demethylation. Neither preparatory treatment significantly affected the substrate concentration-dependent activities of dextromethorphanO-demethylation or LAAM N-demethylation.
It is desirable to compare microsomal data from research programs using different HLM preparatory procedures. Our data suggest that certain aspects of this procedure could be standardized to improve recovery and purity. More importantly, these results suggest that interlaboratory comparisons are not greatly affected due to the preparation variables we studied.
Our results have also presented the first substrate concentration versus activity studies for LAAM N-demethylation in human liver micrososomes. They show that there are interliver differences in the kinetics and that when a wide range of substrate concentrations is considered, the kinetics are complex. When supratherapeutic concentrations of LAAM (>100 μM) are omitted, the kinetics then appear to approach linearity in Eadie-Hofstee plots. Kinetic parameters from the latter may be most valuable for in vivo-in vitro extrapolations; however, further studies are required to delineate the mechanisms of the complex kinetics.
Acknowledgments
We thank Michael Franklin, University of Utah, for performing the glucuronidation assays, and Kent Kunze, University of Washington, for assistance in evaluating the biphasic kinetic data.
Footnotes
-
Send reprint requests to: David E. Moody, Ph.D., University of Utah, Center for Human Toxicology, 20 S. 2030 E. RM 490, Salt Lake City, UT 84112-9457. E-mail: dmoody{at}alanine.pharm.utah.edu
-
↵1 Current Address: Secor International, Inc., Salt Lake City, UT 84107.
-
This research was supported by U.S. Public Health Service Grant R01 DA10100. A preliminary report of these results was presented at the Experimental Biology 1999 meeting in Washington, D.C.
- Abbreviations used are::
- HLM
- human liver microsomes
- P450
- cytochrome P450
- LAAM
- l-α-acetylmethadol
- HPLC
- high-performance liquid chromatography
- Received November 9, 2000.
- Accepted December 5, 2000.
- The American Society for Pharmacology and Experimental Therapeutics