Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

The major metabolites of P¹ observed in vivo and in vitro result from initial cytochrome P450-catalyzed N-dealkylation to DIP, catalyzed primarily by CYP1A2, and aromatic hydroxylation to regioisomeric ring hydroxylation products, principally 4'-OHP, catalyzed primarily by CYP2D6 (Fig. 1). In humans, 3-(1-naphthoxy)lactic acid is the major urinary metabolite (Walle et al., 1985) from P. It arises from the P450-catalyzed formation of DIP via oxidative loss of acetone (Bakke et al., 1973). Further metabolism of DIP catalyzed by monoamine oxidase affords 3-(1-naphthoxy)-2-hydroxypropionaldehyde (Nelson and Burke, 1978; Goldszer et al., 1981; Chen and Nelson, 1982) as an intermediate, with subsequent oxidation of the aldehyde to 3-(1-naphthoxy)lactic acid.

View larger version (15K): [in this window] [in a new window]   Fig. 1.   Major oxidative pathways of propranolol metabolism.

In human liver microsomes, the rate of N-dealkylation of P is highly correlated with the rate of phenacetin O-deethylation, a known CYP1A2-catalyzed process (Masubuchi et al., 1994; Yoshimoto et al., 1995), and the content of immunochemically determined CYP1A2 (Masubuchi et al., 1994). -Naphthoflavone, a potent inhibitor of CYP1A2, inhibited all of the P oxidation activities, and the IC₅₀ value for N-desisopropylase activity was much smaller than values for ring-hydroxylase activities (Masubuchi et al., 1994).

Studies in human lymphoblastoid cell- (Yoshimoto et al., 1995) and yeast-expressed (Ching et al., 1996) recombinant human CYP1A2 have confirmed that CYP1A2 catalyzes formation of DIP and of the 4'- and 5'-hydroxylated metabolites, as well. In the presence of CYP1A2-expressed in human lymphoblastoid cells, N-dealkylation of P occurred about 2 to 3 times as fast as aromatic hydroxylation at a high-substrate concentration (Yoshimoto et al., 1995). In a yeast expression system, CYP1A2 catalyzed both aromatic hydroxylation and N-dealkylation, with the V_max being about 2 times higher for N-dealkylation than for 4'-hydroxylation and 8 to 40 times higher than for 5'-hydroxylation when 2R-P and 2S-P were studied separately (Ching et al., 1996).

The active site for CYP1A2 accommodates a variety of substrates with partition coefficients that vary more than 6 orders of magnitude (e.g., from benzo[a]pyrene, log P = 6.25, to caffeine, log P = 0.01). Most are neutral or weakly basic, although exceptions occur. Many of the substrates have significant planar molecular portions (e.g., aromatic hydrocarbons, nitrogenous heterocyclic systems related to MeIQ, acetanilide, caffeine, and steroids). Collating of physical chemical factors, such as log P, log D, pK_a, molecular size and shape, dipole moment, and frontier orbital energies, has been done in attempts to differentiate between substrate selectivities of CYP isoforms (Lewis, 2000). According to this work, CYP1A2 substrates commonly include polar planar molecules with low log P values at physiological pH that are neutral or weakly basic and have rectangular surfaces, fairly high dipole moments, a hydrogen-bonding donor or acceptor site near the site of metabolism, and solvent-accessible volume of 200 Å³ or less (De Rienzo et al., 2000). Many known substrates of CYP1A2 do not meet all of these criteria.

Although CYP1A2 has been less extensively studied than CYP2D6, protein homology models based on the crystal structures of CYP102 and CYP101 have been reported. Models based on potential substrate-protein interactions, using CYP 102 as a template, have been reported (Lewis et al., 1999b; Lozano et al., 2000; De Rienzo et al., 2000). Theses models have relatively large binding pockets composed of mostly hydrophobic and aromatic residues, with a number of polar, potentially hydrogen-bonding residues also present near the heme. In these models, substrates are placed approximately orthogonal to the heme plane. More than one mode of binding is proposed to accommodate formation of multiple regioisomeric oxidation products from substrates like caffeine. Roles have been assigned to Phe226 and Tyr495 in key - stacking interactions with the CYP1A2 substrates caffeine, phenacetin, tacrine, 7-methoxyresorufin, 2-amino-3-methylimidazo[4,5-f]quinoline (IQ), 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP) and furafylline (Lewis and Lake, 1996; Lewis et al., 1999b). Residues Thr115, Thr124, and Thr498 participate in hydrogen-bonding interactions that act to orient the substrates in the active site in this model. Sequence alignment with CYP102 places these residues within various previously defined SRS regions (Gotoh, 1992).

A model for CYP1A2 based on the study of N-hydroxylation of several congeners of MeIQ, mutagens that are weakly basic planar heterocyclic nitrogen-containing ring systems, has been proposed and refined (Lozano et al., 1997, 2000). Regions that are assigned interactions include in SRS-1, Thr115, Asp119, Thr124, Phe125; in SRS-2, Thr223 and Phe226; in SRS-4, Thr321; in SRS-5, Leu382 and Pro383; and in SRS-6, Gly223. Based on catalytic efficiencies for metabolism of MeIQ, 7-ethoxyresorufin and phenacetin, a series of CYP1A2 mutants, have been studied (Parikh et al., 1999; Josephy et al., 2000). Mutants Phe226Ile, Phe226Thr, and Phe226Tyr had reduced catalytic efficiencies for metabolism of these substrates. Mutations at other sites produced varying activities depending on the substrate.

Earlier point mutation studies in SRS-3 of rat CYP1A2 showed that changes that increase the hydrophobic character of this region resulted in greater catalytic activity for methoxy- and ethoxyresorufin O-dealkylation (Krainev et al., 1992). Mutations of Thr319 (Thr321 in the human form) and to a lesser extent Glu318 (Asp320 in the human form) in SRS-4 were found to affect the K_d values of enantiomers of chiral amines (Hiroya et al., 1992; Krainev et al., 1992, 1993). Studies of the interactions of a series of aryl diazenes with rat liver 1A2 have suggested a relatively large active site for this isoform (Tuck et al., 1993).

Some work relating structural features of CYP1A2 inhibitors to their inhibitory potency has been reported. In the inhibition of phenacetin O-deethylation by a series of quinolone antibiotics, specific binding interactions similar to the suggested interactions of caffeine in the CYP1A2 active site were proposed, but no correlation with their lipophilicity was obtained (Fuhr et al., 1993). However, the range of inhibitory potency of these compounds was very limited. In a study of inhibitory potency of three structurally similar antiarrhythmics on 7-methoxyresorufin demethylation, a combination of factors, including a charge interaction of the amine nitrogen on each of these inhibitors with Asp313 of CYP1A2 and various possible hydrogen bonding and hydrophobic interactions, were suggested (Wei et al., 1999).

As a known substrate of CYP1A2, P meets most of the criteria proposed by Lewis (1999a) and can serve as a model on which to make systematic structural changes to test some of the requirements for metabolism by CYP1A2. The series of fluorinated and nonfluorinated analogs, which varied in physicochemical properties including basicity, lipophilicity, and substituent size (Tables 1 and 2) and showed significant changes in metabolism by baculovirus-insect cell expressed human recombinant CYP2D6 (Upthagrove and Nelson, 2001), were examined in CYP1A2. Changes in observed Michaelis-Menten kinetic parameters were correlated to changes in physicochemical properties, and the regioselectivity of oxidation was examined.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Propranolol-Related Substrates and Standards. Propranolol enantiomers (2S-P and 2R-P) were obtained from Sigma (St. Louis, MO). The fluorinated and steric analogs of P (Table 1) and the metabolic standards 4'-OHP, 5'-OHP, 7'-OHP, 4'-OHTFP, 4'-hydroxydeshydroxypropranolol, DIP, deshydroxydesisopropylpropranolol, and the internal standard propranolol O-methyl ether were obtained as described in the accompanying paper (Upthagrove and Nelson, 2001).

Basicity and Lipophilicity. The pK_a, log P, and log D values (Table 1) for this series of amines (Upthagrove and Nelson, 2001) were determined at pION, Inc. (Cambridge, MA), using their validated potentiometric method (Avdeef et al., 1993). The log P and log D values for these compounds were also determined at pION, Inc., according to their validated potentiometric method (Avdeef, 1992; Slater et al., 1994).

Incubations in Insect Cell Expressed Human CYP1A2. Baculovirus-insect cell microsomes coexpressing human CYP2D6 and human P450 reductase were purchased from GENTEST (Woburn, MA). Duplicate incubation mixtures, having a final volume of 200 µl, contained 10 pmol of CYP1A2/ml, 100 mM phosphate buffer (pH 7.4), and 4 µl of a solution of substrate in a 1:1 mixture of CH₃OH and H₂O (final CH₃OH concentration, 1%). After preincubation at 37°C for 3 min, NADPH in phosphate buffer (final concentration, 1 mM) was added to initiate the reaction. After 2 min, the reaction was stopped by addition of 20 µl of 7% perchloric acid, and 5 mg of ascorbic acid was added to stabilize ring-hydroxylated metabolites. Substrate concentrations of 0.24, 0.49, 0.98, 1.95, 3.91, and 7.81 µM were used. Preliminary incubations of P and TFP were performed to establish the linearity of metabolite formation with respect to time and the amount of enzyme.

HPLC Separation. Following addition of the internal standard, incubation samples were vortexed and centrifuged to precipitate protein. One hundred microliters of the supernatant was analyzed by HPLC using a Hewlett-Packard 1100 LC system (Palo Alto, CA) with Chemstation instrument control and data analysis software. For all substrates, separation was achieved on a Microsorb-MV phenyl column (250 × 4.6 mm; Varian Corp., Walnut Creek, CA). Mobile phase gradients consisted of an aqueous phase containing 1% by volume triethylamine, 0.8% by volume phosphoric acid (to bring the pH to 2.2), and acetonitrile. The HPLC gradient used for metabolite separation from all substrates except FP and TFE was 25% (v/v) acetonitrile (75% aqueous buffer described above) for 10 min, then a linear increase to 70% acetonitrile over the next 10 min, an additional 2 min at 70% acetonitrile, then returned to 25% acetonitrile over the next 5 min. For metabolites from FP, the gradient was 20% acetonitrile (80% aqueous buffer) for 10 min, then a linear increase to 60% acetonitrile over the next 10 min, held at 60% acetonitrile for 3 min, and then returned to 20% acetonitrile over the next 4 min. For metabolites from TFE, the gradient was 20% acetonitrile (80% aqueous buffer) for 10 min, then a linear increase to 70% acetonitrile over the next 10 min, held at 70% acetonitrile for 2 min, and then returned to 20% acetonitrile over the next 5 min. The internal standards used were propranolol O-methyl ether for metabolites of P, FP, DFP, TFP, and tBuMe, and P for metabolites of iPrMe, TFE, desOHP, and desOHTFP.

Fluorescence Detection. For fluorescence detection, tables were programmed to excitation wavelength _ex = 295 nm and emission wavelength _em = 370 nm over times when N-desalkyl metabolite, internal standard, and substrate eluted and continued to the end of the HPLC run. At other times during the run, fluorescence excitation and emission wavelengths were set to _ex = 320 nm and _em = 420 nm to increase sensitivity for the 4'-hydroxylated metabolite.

Metabolite Quantitation. DIP, 4'-OHP, 5'-OHP, 4'-OHTFP, deshydroxydesisopropylpropranolol, and 4'-hydroxydeshydroxypropranolol were quantitated by comparison of peak area ratios of these compounds formed metabolically with standard curves of peak area ratios of known amounts of synthetic standards. Amounts of other metabolites were estimated by comparison of the area ratios with standard curve ratios of 4'-OHP, 5'-OHP, or 4'-OHTFP.

Metabolite Identification. Fluorescence Excitation and Emission Spectra. Unknown metabolite peaks in the HPLC eluent were collected in silane-treated tubes. The fluorescence excitation and emission spectra of each fraction were obtained after transfer to a limited volume quartz cuvette, as describe in the accompanying work (Upthagrove and Nelson, 2001).

Mass Spectra. Fraction collection was repeated, and combined fractions (total approximately 2 ml) were treated with 150 µl of aqueous saturated Na₂CO₃, which was adequate to bring the eluent solution to approximately pH 8. The solutions were then extracted with 3 ml of EtOAc, vortexed for at least 15 s to mix, then centrifuged to separate the layers. The EtOAc layer was transferred to a silane-treated concentration tube, and the solvent evaporated using a stream of dry nitrogen and heat. The metabolite residues were dissolved in 50 µl of CH₃OH and 50 µl of H₂O. The metabolite residues dissolved in CH₃OH/H₂O were infused into a Finnigan LCQ mass spectrometer (Thermo Finnigan, San Jose, CA) equipped with a modified electrospray source at a rate of 300 nl/min. Tandem mass spectroscopy spectra were obtained by collision-induced dissociation in the ion trap.

Calculation of Michaelis-Menten Kinetic Parameters. K_m and V_max for each of the enzyme-catalyzed reactions were determined by nonlinear fitting of initial velocity versus substrate concentration curves using the k.cat program (Biometallics, Inc., Princeton, NJ). The robust weighting option was used. All velocity values deviating from the calculated curve by less than 10 times the average residual were included in the fit. The rate constants (k_cat) for metabolite formation were calculated by dividing the V_max values by the enzyme concentration. Substrate concentrations were corrected for substrate depletion by using the average of the initial substrate concentration and the final substrate concentration (Segel, 1975). Final substrate concentrations were calculated by subtracting the concentrations of all metabolites from the initial substrate concentrations.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Fluorinated and Steric Analogs of Propranolol. pK_a Values, log D, and Steric Parameters. The propranolol-related compounds for the metabolic studies are listed in Table 1. These included the 1"-mono, di-, and trifluorinated derivatives (FP, DFP, TFP) and steric congeners with two and three additional 1"-methyl groups (iPrMe, tBuMe). Additional analogs include the trifluoroethyl analog TFE and deshydroxy analogs desOHP and desOHTFP. Their pK_a values and log D (pH 7.4) values (Upthagrove and Nelson, 2001) are included in Table 1. A measure of the relative sizes of the substituents is given by the Taft E_s and Charton van der Waals radius r_v parameters also in Table 1.

Identification and Quantitation of Metabolites. Incubations of all of the P analogs in CYP1A2 resulted in formation of the product of N-dealkylation and, in most cases, the 4'- and 5'-hydroxylated metabolites also. No other metabolites were observed. Representative chromatograms showing the retention times of the ring-hydroxylated metabolites from two of the substrates appear in the accompanying report (Upthagrove and Nelson, 2001). Ring hydroxylation regioisomers were identified as described therein.

Kinetic Parameters. The K_m and k_cat values for formation of each of the metabolites in CYP1A2 appear in Table 1. Total k_cat values, which appear in Table 3, are the sums of the individual metabolite k_cat values. For each substrate, there is some variability in the K_m values determined for each metabolite; the K_m values used in correlations with physicochemical parameters are the averages of the K_m values for the individual metabolites. A small difference in the K_m values and catalytic efficiencies of 2R-P and 2S-P were observed. 2S-P has a 2-fold higher K_m and a 2-fold lower k_cat/K_m than 2R-P. Among the TFP stereoisomers studied, 3-fold differences in K_m and k_cat/K_m were observed.

Metabolite Ratios. For several of the P analogs, the regioselectivity of metabolism is different than in P. The amount of 4'- and 5'-hydroxylation compared with N-dealkylation increases as the number of fluorines increases (Table 2). Under the conditions of the experiments, we were unable to quantitate the aromatic hydroxylation products from 2R-P and 2S-P, but quantifiable amounts of aromatic hydroxylation products were observed in the metabolism of other analogs. Increasing the size of the alkyl group also increases the ratio of ring hydroxylation to N-dealkylation products. Removing the side chain hydroxyl group from P also increases the ratio of ring hydroxylation to N-dealkylation metabolites, and in desOHTFP, ring hydroxylation is highly preferred over N-dealkylation (Table 2).

Correlations with Physicochemical Parameters. Plots of log K_m and log k_cat/K_m versus pK_a and log K_m and log k_cat/K_m versus log D with equations for the best fit lines of the data appear in Figs. 2 and 3, respectively. For this series of compounds, a significant positive correlation of log K_m with pK_a (r² = 0.47) and a significant negative correlation of log k_cat/K_m with pK_a (r² = 0.63) are observed. The more basic compounds have higher K_m values (lower affinities) and lower catalytic efficiencies. The opposite trends appear in the correlations with log D. A significant negative correlation of log K_m with log D (r² = 0.50) and a significant positive correlation of log k_cat/K_m with log D (r² = 0.69) are observed. The more lipid soluble compounds have lower K_m values (higher affinity for CYP1A2) and higher catalytic efficiencies.

View larger version (15K): [in this window] [in a new window]   Fig. 2.   Plot of log Km () and log kcat/Km () for CYP1A2-catalyzed metabolism of propranolol analogs versus pKa. log Km = 0.085 pKa  0.75; r2 = 0.47 (p < 0.05); log kcat/Km = 0.13 pKa + 1.70; r2 = 0.63 (p < 0.05).

View larger version (16K): [in this window] [in a new window]   Fig. 3.   Plot of log Km () and log kcat/Km () for CYP1A2-catalyzed metabolism of propranolol analogs versus log D.  log Km = 0.20 log D + 0.42; r2 = 0.50 (p < 0.05); log kcat/Km = 0.32 log D  0.14; r2 = 0.69 (p < 0.05).

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

All of the compounds in this series are oxidized at sites analogous to the sites of metabolism in P. As would be expected for an enzyme with a relatively large active site that accommodates a variety of substrates, none of the small structural changes in the analogs prevented CYP1A2-catalyzed N-dealkylation and aromatic 4'- and 5'-hydroxylation.

The small difference in the K_m values and catalytic efficiencies of 2R-P and 2S-P (2S-P has a 2-fold higher K_m and a 2-fold lower k_cat/K_m than 2R-P) is in agreement with previously reported results in recombinant human 1A2 systems (Yoshimoto et al., 1995; Ching et al., 1996). The small (3-fold) difference in K_m and k_cat/K_m among the TFP stereoisomers indicates that regioselectivity of metabolism is only slightly dependent on the stereochemistry at C-2 in the side chain of TFP, like in P.

As shown in Tables 3 and 4, the sequential addition of fluorines leads to lower K_m values, by up to slightly more than 10-fold, indicating increased enzyme affinity. Rate constant values (k_cat) decrease only slightly with addition of fluorines, resulting in higher catalytic efficiencies. The catalytic efficiency (k_cat/K_m) varies across the series to a greater extent than the K_m value (about 25-fold versus 9-fold), providing a useful comparison of the relative rates of metabolism of these agents with each other under nonsaturating conditions.

Higher binding affinity (decreasing log K_m) and higher log catalytic efficiency values are associated with increasing log D values (Figure 3), suggesting that hydrophobic interactions play an important role in binding to and turnover by CYP1A2. Lower affinity and lower catalytic efficiency associated with increasing pK_a values (Fig. 2) mirror this trend almost exactly, reflecting the effect of protonation state of the amine nitrogen on partitioning of the substrates between lipid and aqueous phases. Compounds with lower pK_a values (increased proportion of nonionized species at pH 7.4) have lower K_m values indicating that increased lipophilicity increases enzyme affinity. The correlations of log D and pK_a show slightly less scatter in the plots against log k_cat/K_m (r² values, 0.63 and 0.69, respectively) than against log K_m (r² values, 0.43 and 0.50, respectively), accounting for a slightly greater proportion of the metabolic data.

The effects of the structural changes in P on CYP1A2-catalyzed metabolism are complex. Not only are the affinity (as reflected in the K_m value) and the total catalytic efficiency affected, but the regioselectivity of metabolism is changed as well. The aromatic hydroxylation/N-dealkylation ratio is greater than 1 for the tBuMe and desOHP analogs and for the trifluoro analogs, suggesting that small differences in preferred binding orientations may contribute to the differences in regioisomeric products. Changes in substrates that result in a decrease in the polarity (an increase in the lipophilicity) (Table 1) may alter the attraction of the aliphatic portion of the molecule to a hydrophobic region of the enzyme. Alternatively, there could be a specific interaction of the side chain hydroxyl group of P with a site in the enzyme that is absent in the deshydroxy analogs (desOHP and desOHTFP) or that is changed due to steric hindrance by substitution of a larger N-alkyl group (e.g., in tBuMe or TFP). Both of these possibilities are consistent with the current models for the CYP1A2 active site and with the view that individual residues of 1A2 control catalytic selectivity by altering kinetics of specific steps in the multistep P450 reaction cycle (Yun et al., 2000).

The possibility that the rate of electron/proton or hydrogen atom transfer is limiting the overall rate of N-dealkylation in this series of compounds seems unlikely. Across the series, the increased ratio of ring hydroxylation to N-dealkylation is due to an increase in k_cat/K_m for 4'- and 5'-hydroxylated metabolites rather than a decrease in catalytic efficiency for N-dealkylation (Table 2). Our observations do not distinguish between hydrogen atom transfer or proton/electron transfer mechanisms, but the presence of fluorines on the carbon adjacent to the site of abstraction would disfavor either process (i.e., an increase in the energy of activation for either electron or hydrogen atom removal would be expected with fluorine addition). The presence of fluorines on the carbon- to the amine nitrogen increases the one electron ionization potential, as reported for the analogous -mono-, di-, and trifluoroethyl amines (Staley et al., 1977), and is supported by electron impact-mass spectrometry data on P, FP, DFP, and TFP (Upthagrove et al., 1999). -Fluorination can also significantly affect radical abstraction reactions, and the effect depends on the nature of the abstracting radical. Fluorine inductively deactivates hydrogen atom abstraction by electrophilic radicals due both to increased carbon-hydrogen bond strength and to unfavorable polarization in the transition state. Fluorine can be activating for abstraction by nucleophilic radicals, however, due to a favorable polarization of the transition state (Smart, 1995).

At higher substrate concentrations, biphasic kinetics were observed for all the substrates with CYP1A2. Further work on this observation is beyond the scope of this study. For the kinetic parameters reported here, substrate concentrations were kept low enough to estimate only the higher affinity K_m and k_cat values. Biphasic binding curves have been reported for the binding of R- and S-1-cyclohexylethylamine in rat CYP1A2 (Krainev et al., 1993), and non-linear Michaelis-Menten kinetics have been reported for yeast recombinant rat CYP1A1-catalyzed metabolism of 7-ethoxyresorufin and aminopyrine (Inouye, et al., 2000). The biphasic behavior may be due to the simultaneous presence of two substrate molecules in the active site of CYP1A2, as has been suggested for other cytochrome P450 isoforms and other substrates. However, ratios of metabolites from P and TFP do not change significantly at higher substrate concentrations, as has been observed for other systems in which biphasic kinetics has been reported (Korzekwa et al., 1998). Differential source- and substrate-dependent autoactivation of CYP1A2 has also been reported (Ekins et al., 1998). Further studies would be required to explore this observation.

Comparison of Results in CYP1A2 with CYP2D6. Correlations of log D and pK_a values of these related substrates with low K_m values (and higher catalytic efficiency) are consistent with lipophilicity being a major contributor to substrate affinity for CYP1A2. These results are in contrast with results of the studies with CYP2D6 in which higher enzyme affinity (low K_m values) and higher catalytic efficiency are positively correlated with the increased degree of substrate protonation at pH 7.4 (higher pK_a) and smaller log D values (Upthagrove and Nelson, 2001).