Abstract
Conjugation of a structurally diverse set of 53 catechol compounds was studied in vitro using six recombinant human sulfotransferases (SULTs), five UDP-glucuronosyltransferases (UGT) and the soluble form of catechol O-methyltransferase (S-COMT) as catalyst. The catechol set comprised endogenous compounds, such as catecholamines and catecholestrogens, drugs, natural plant constituents, and other catechols with diverse substituent properties and substitution patterns. Most of the catechols studied were substrates of S-COMT and four SULT isoforms (1A1, 1A2, 1A3, and 1B1), but the rates of conjugation varied considerably, depending on the substrate structure and the enzyme form. SULT1E1 sulfated fewer catechols. Only low activities were observed for SULT1C2. UGT1A9 glucuronidated catechols representing various structural classes, and almost half of the studied compounds were glucuronidated at a high rate. The other UGT enzymes (1A1, 1A6, 2B7, and 2B15) showed narrower substrate specificity for catechols, but each glucuronidated some catechols at a high rate. Dependence of specificity and rate of conjugation on the molecular structure of the substrate was characterized by structure-activity relationship analysis and quantitative structure-activity relationship modeling. Twelve structural descriptors were used to characterize lipophilicity/polar interaction properties, steric properties, and electronic effects of the substituents modifying the catechol structure. PLS models explaining more than 80% and predicting more than 70% of the variance in conjugation activity were derived for the representative enzyme forms SULT1A3, UGT1A9, and S-COMT. Several structural factors governing the conjugation of catechol hormones, metabolites, and drugs were identified. The results have significant implications for predicting the metabolic fate of catechols.
The catechol group, two vicinal hydroxyls in an aromatic ring, is present in the structure of catecholamine neurotransmitters and hormones, and many of their metabolites, and is also found in metabolites of phenolic hormones such as estrogens. The group is an important pharmacophore, present in drugs with varying pharmacological action, e.g., the dopamine agonist apomorphine, the β1-receptor agonist dobutamine, the antihypertensive α-methyldopa, the dopadecarboxylase inhibitor carbidopa, and the catechol O-methyltransferase inhibitor entacapone. Catechol groups are also present in numerous dietary compounds (coumarins, flavonoids, anthocyanins, etc.) and may thereby interfere with the pharmacokinetics and/or pharmacodynamics of endogenous catechols and catechol drugs. Conjugation regulates the activity of catechol hormones and neurotransmitters, and these reactions usually dominate the metabolism of catechol drugs and other exogenous catechols.
The most important conjugation reactions of catechols are sulfation, glucuronidation, and methylation catalyzed by sulfotransferases (SULTs2), UDP-glucuronosyltransferases (UGTs), and catechol O-methyltransferases (COMTs), respectively. Human cytosolic sulfotransferases form a family of 11 known isoenzymes expressed in liver, intestine, placenta, platelets, and many other tissues (Weinshilboum et al., 1997; Coughtrie and Johnston, 2001). The human UGT family presently comprises 16 individual distinct, expressed isoenzymes, not including variant allozymes (Mackenzie et al., 1997). UGTs are lumenally facing in the endoplasmic reticulum and are localized predominantly in hepatic, renal, and intestinal tissues. Two forms of human COMT have been identified, a cytoplasmic soluble form (S-COMT) and a membrane-bound form (MB-COMT) located in the cytosolic side of the rough endoplasmic reticulum (Ulmanen et al., 1997). Primary structures of the two COMT forms are otherwise identical, but MB-COMT has an N-terminal extension of 50 amino acids, presumably for membrane anchoring. S-COMT is the predominant form in most tissues. Highest COMT activities have been found in liver, kidney, intestine, and brain.
Although the catechol group, in principle, can readily form sulfates, glucuronides, and O-methylated products, remarkable differences in the conjugation profile of different catechol-type drugs have been observed. The case of catechol drugs frequently coadministered in the triple therapy of Parkinson's disease, levodopa, a dopadecarboxylase inhibitor, and a COMT inhibitor, is a representative example. The major metabolic pathway of levodopa in humans is decarboxylation to dopamine (about 70%). About 10% of levodopa is O-methylated by COMT (Männistö et al., 1992), and 20 to 70% circulates as sulfoconjugates (Johnson et al., 1980). More than 95% of circulating dopamine is found as sulfates in humans (Johnson et al., 1980; Eisenhofer et al., 1999), and a specific SULT isoform (SULT1A3) exists for this purpose (Dajani et al., 1998; Eisenhofer et al., 1999). The decarboxylase inhibitor benserazide is rapidly metabolized to the active metabolite trihydroxybenzylhydrazine, which is found mainly as glucuronides and O-methylated metabolites in human urine (Schwartz and Brandt, 1978). Benserazide itself is an excellent substrate of COMT (Lautala et al., 2001). COMT inhibitors entacapone and tolcapone show subnanomolar affinity for the catechol-binding site of COMT but are extremely poor substrates (Lotta et al., 1995). Their predominant metabolic reaction in humans is direct glucuronidation (Wikberg et al., 1993; Jorga and Fotteler, 1996). It is obvious that coadministration of catechol-type drugs results in complex metabolic interactions involving several enzyme systems and endogenous and dietary catechols.
The availability of expression systems for individual SULT, UGT, and COMT isoforms allows detailed characterization of reactions competing for catechol conjugation and facilitates molecular-level insight into substrate selectivities. Understanding the molecular determinants for conjugation would be invaluable in the modern drug discovery process seeking a balance between potency and absorption, distribution, metabolism, and excretion properties. Such knowledge can be used in virtual screening, for guiding synthesis during lead optimization, in evaluating drug candidates for metabolic interactions, and in planning metabolism studies. Some structural factors affecting catechol conjugation have been identified in recent studies on SULT1A3 (Dajani et al., 1999) and S-COMT (Lautala et al., 2001), but systematic studies on catechol glucuronidation by human UGT isoforms are not available. A large number of compounds have been tested as substrates of various recombinant UGT isoenzymes (e.g., Tukey and Strassburg, 2000), but it is difficult to find a consistent set for structure-activity analysis, and published data on catechols as UGT substrates are scarce.
Here we have measured the conjugation rate for a structurally diverse set of catechol compounds. SULT, UGT, and COMT forms expected to be most important in the conjugation of catechols were used as catalyst. SAR and QSAR analyses were performed to identify structural factors controlling substrate specificity and conjugation rate, and to derive models and rules for predicting the behavior of catechol-type compounds as substrates of these enzymes. To the best of our knowledge, this is the first time that competitive metabolic reactions of a specific functional group have been systematically studied at the level of individual isoenzymes representing different enzyme families.
Materials and Methods
The catechols screened and the other reagents used were purchased from Aldrich Chemical Co. (Steinheim, Germany), Roche Diagnostics (Mannheim, Germany), Fluka (Buchs, Switzerland), ICN Pharmaceuticals (Costa Mesa, CA), Merck (Darmstadt, Germany), Rathburn Chemicals (Walkerburn, Scotland, UK), Riedel-de Haën (Seelze, Germany), and Sigma-Aldrich (St. Louis, MO), and were of the highest grade available. Entacapone and 3,5-dinitrocatechol were kindly supplied by Orion Pharma (Espoo, Finland). S-Adenosyl-l-[methyl-14C]methionine, PAP35S, and [14C]UDPGA were obtained from PerkinElmer Life Sciences (Boston, MA). PAPS (>99% purity) was purchased from H. R. Glatt (German Institute for Human Nutrition, Potsdam, Germany).
COMT Expression. Cloning of the human catechol O-methyltransferase cDNA and production of the human S-COMT in Escherichia coli have been described in detail earlier (Lundström et al., 1991, 1992). In brief, the E. coli cells carrying the expression plasmids were induced with isopropyl-β-d-thiogalactopyranoside, and after 3 h, the cells were harvested by centrifugation and disrupted by sonication in ice-cold phosphate buffer at pH 7.4. After centrifugation (10,000g, 10 min) the supernatants were divided into aliquots, frozen, and stored at -70°C before being used for the enzyme assays. Total protein concentration was determined according to the method of Bradford (1976) using bovine serum albumin as a standard. Samples of the bacterial culture were characterized with SDS-polyacrylamide gel electrophoresis (10% acrylamide) and Coomassie Brilliant Blue staining. The COMT-specific proteins were also visualized by immunoblotting using a guinea pig polyclonal antiserum (Ulmanen et al., 1997) and the ECL detection system (Amersham Biosciences UK, Ltd., Little Chalfont, Buckinghamshire, UK). The EMBL Accession Number for COMT is Z26491.
COMT Assays. Catechol O-methylation activities were determined using a recently published method (Lautala et al., 1999). Part of the activity data for structure-activity analysis was taken from that paper. Additional data were measured to complete the substrate set corresponding to the other enzymes. Briefly, typical incubation conditions for the screening assays contained 5 mM MgCl2, 20 mM l-cysteine, 0.15 mM AdoMet, human S-COMT bacterial lysate (0.2–20 μg protein), 0.5 mM catechol substrate, and [14C]AdoMet (0.1 μCi) in 100 μl of 100 mM Na2HPO4/NaH2PO4 buffer (pH 7.4). Blanks contained no acceptor substrate; the same volume of solvent was substituted. The samples were preincubated at 37°C for 5 min before initiation by addition of the AdoMet/[14C]AdoMet mixture. In case of unexpectedly low activities, lower catechol concentrations were used to reveal possible substrate inhibition. The reaction was terminated after 15 to 30 min by addition of 10 μl of ice-cold 4 M perchloric acid. Proteins were precipitated on ice for 10 min and then removed by centrifugation (5 min, 22,000g); 100 μl of the supernatant were injected into HPLC. Methylated products and AdoMet were separated by HPLC (1090; Hewlett Packard, Boeblingen, Germany), and the radiolabeled products were detected and quantified using a flow scintillation analyzer (150TR; PerkinElmer Life Sciences) fitted with a 300-μl flow-cell packed with silanized cerium-activated lithium glass as scintillant (PerkinElmer Life Sciences) or with a 500-μl cell into which scintillation liquid (3 ml/min; National Diagnostics, Atlanta, GA) was pumped. The isocratic HPLC system was composed of phosphate/citrate buffer (50 mM Na2HPO4, 20 mM citric acid, 0.15 mM Na2EDTA, pH adjusted to 3.2 with o-phosphoric acid) and methanol. In some cases 1-octanesulfonic acid (1.25 mM) was added to the buffer. The flow rate was 1 ml/min. The amount of methanol in the mobile phase varied between 3 and 60% depending on the substrate. The column used was a Hypersil BDS-C18, 125 × 4 mm, 5 μm (Hewlett Packard) and was heated to 40°C.
UGT Expression. Recombinant V79 cell lines were grown in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, 100 units/ml penicillin, and 0.1 mg/ml streptomycin. Cell cultures were grown in 75-cm2 flasks (Costar, Cambridge MA) fitted with vented caps in humidified incubators at 37°C with the atmosphere maintained at 5% CO2. V79 cells expressing human UGTs were maintained under optimized constant selection concentrations of geneticin (G418; Invitrogen, Paisley, Scotland, UK): V79/UGT1A1, 1 mg/ml; V79/UGT1A6, 100 μg/ml; and V79/UGT1A9, 200 μg/ml. HEK293 cells expressing UGT2B15, used by kind permission of Dr. T. Tephly, University of Iowa, were grown in the media described above supplemented with 10 mM HEPES, pH 7.4, and 700 μg/ml geneticin. The cloning and expression of these UGT isoforms are reported elsewhere (Fournel-Gigleux et al., 1991; Sutherland et al., 1992; Green et al., 1994). EMBL Accession Numbers for the UGT isoforms are: UGT1A1, M57899; UGT1A6, J04093; UGT1A9, AF056188; UGT2B7, J05428; and UGT2B15, U08854.
UGT Assays. UGT assays were performed by an adaptation of a previously published method (Ethell et al., 1998). Assays comprised 100 mM Tris/maleate buffer, pH 7.4, containing 5 mM MgCl2, typically 500 μM substrate, 2 mM UDPGA (containing 0.1 μCi of [14C]UDPGA), and 350 to 500 μg of cellular sonicate (standard cellular sonication described in Ethell et al., 1998) in a total volume of 100 μl. Blanks contained no acceptor substrate, with the same volume of solvent substituted. Screening assays were incubated for 60 min at 37°C. Each batch of screening assays was accompanied by positive controls for the cell lines as follows: UGT1A9, 500 μM propofol (Aldrich, Gillingham, Dorset, UK); UGT1A6, 500 μM 1-naphthol (Fluka); UGT1A1, 250 μM 17α-ethinylestradiol (Sigma, Gillingham, Dorset, UK); and UGT2B15, 500 μM 8-hydroxyquinoline (BDH, Poole, Dorset, UK). Incubations were terminated by the addition of 100 μl of methanol prechilled to -20°C. The precipitated proteins were removed by centrifugation at 1000g. The resulting supernatant was then transferred to an HPLC vial and 150 μl of this was directly injected onto HPLC (Shimadzu; Dyson Instruments, UK). The gradient conditions were 0 to 100% acetonitrile in 0.05 M ammonium acetate developed over 13 min on a Techsphere 5ODS2 column (HPLC Technology, Macclesfield, UK). Radioactive UDPGA and glucuronides were detected using model 9701 radioactivity monitors (Reeve Analytical, Glasgow, UK) fitted with a 200-μl flow cell packed with silanized cerium-activated lithium glass as scintillant.
Cloning of Human Sulfotransferases. Cloning and E. coli expression of human sulfotransferases 1A3 (vector pET11a/BL21 pLysS cells), 1A1 wildtype (1A1*1) (pET17b/BL21 cells), and 1E1 (pCW/JM109 cells) have been described previously (Jones et al., 1995; Dajani et al., 1998; Rubin et al., 1999). Human SULT1C2 was obtained as an expressed sequence tag (EMBL Accession Number U66036) from the UK HGMP Resource Centre, Cambridge UK, cloned into the vector pET11a (Novagen, Nottingham, UK) for expression in E. coli, and resequenced to confirm identity. Human sulfotransferases 1A2 and 1B1 were cloned using polymerase chain reaction (PCR) from Human Liver Quick Clone cDNA (BD Clontech Laboratories, Basingstoke, UK). Forward and reverse primers (BioLine, London, UK) used in the PCR cloning reactions for SULT1A2 were AAGAGCTCAGGAACATGGAG and CCCCTCTCACAGCTCAGAGC, respectively. Forward and reverse primers used in PCR cloning reactions for SULT1B1 were CAATCTGGTATTAAATGCTTTCCC and TTTAAATCTCTGTGCGGAATTG, respectively. PCR reactions contained 0.2 to 0.4 μM primer, 0.02 ng/μl DNA, 0.2 μM deoxynucleoside-5′-triphosphates, 1 to 2 mM MgCl2, 1× reaction buffer, and 4 units of Bio-X-Act DNA Polymerase (BioLine). PCR reactions (Robocycler; Stratagene, La Jolla, CA) were performed at 94°C for 5 min, followed by 30 cycles of 94°C for 30 s, 56°C (SULT1A2) for 30 s, or 50°C (SULT1B1) for 60 s, 72°C for 2 min, and finally 72°C for 10 min, which yielded the expected 0.9-kilobase fragment. PCR products were ligated into the vector pCR2.1 (Invitrogen, Paisley, UK) and transformed into Top10 cells. Transformed Top10 cells were grown, plasmid DNA-purified (Wizard Plus SV DNA Purification System; Promega, Southampton, UK), and sequenced to confirm the identity of SULT1A2 and SULT1B1 clones. cDNAs were cloned into the expression vector pET17b, sequenced, and transformed into BL21(DE3) cells as previously described (Dajani et al., 1998). EMBL Database Accession Numbers representing SULT cDNAa used are as follows: SULT1A1*1, AJ007418; SULT1A2*1, U28169; SULT1A3, X84653; SULT1B1, U95726; SULT1C2, U66036; SULT1E1, Y11195.
Preparation of E. coli/Sulfotransferase Cell-Free Extracts. Sulfotransferase activity was determined using cell-free extracts prepared from E. coli over-expressing human SULTs 1A1, 1A2, 1A3, 1B1, 1C2, or 1E1. One hundred-milliliter cultures of BL21(DE3) cells expressing SULT1A1, 1A2, 1B1, or 1C2; BL21 pLysS cells expressing SULT1A3; or JM109 cells expressing SULT1E1 were grown in LB (0.1 mg/ml ampicillin) and induced with isopropyl β-d-thiogalactoside (1 mM final concentration) when cultures reached an O.D. reading of 0.5. Cultures were grown for an additional 15 h, then centrifuged at 5,000g, and the pellet was frozen at -70°C. Chloramphenicol (34 μg/ml final concentration) was also included in the culture medium for BL21 pLysS cells expressing SULT1A3. Cell-free extracts were prepared by lysing the cell pellets with 40 mM Tris-HCl (pH 8.0) and lysozyme (0.5 mg/ml final concentration). BL21 pLysS cells expressing SULT1A3 did not require the addition of lysozyme. Lysed cells were then centrifuged at 14,000g for 30 min at 4°C, and the cell-free supernatant was removed and stored at -70°C. Protein concentrations were determined by the method described by Bradford (1976).
SULT Assays. Sulfotransferase activity was determined using a modification of the PAP35S method described by Foldes and Meek (1973). Assays for each isoform were optimized with respect to incubation time and protein content using a reference substrate (4-nitrophenol for 1A1, 1A2, 1B1, and 1C2; dopamine for 1A3; and ethinylestradiol for 1E1). Assays were performed in duplicate in a final volume of 160 μl containing 10 mM potassium phosphate (pH 6.8), 20 μl of substrate (10 or 100 μM, final concentration), 20 μl PAPS containing 0.04 μCi PAP35S and PAPS (to give a final concentration of 20 μM), and 20 μl of enzyme protein (10 μg of 1A1, 1A2, 1A3, and 1B1; 25 μg of 1C2 and 1E1). Blank reactions contained 20 μl of water in place of substrate. Reactions were incubated at 35°C and terminated by the addition of 200 μl barium acetate (0.1 M), 200 μl of barium hydroxide (0.1 M), and 200 μl of zinc sulfate (0.1 M). The reaction mixtures were centrifuged at 14,000g for 4 min, and 500 μl of supernatant were mixed with 4 ml of scintillation fluid. Radioactivity was quantified by liquid scintillation spectrometry. Sulfation rates were measured at two substrate concentrations. The values determined at the higher concentration were used for SAR analysis, when substrate inhibition was not observed. The value determined at the lower concentration was used when it was clearly higher, taking into account the experimental uncertainty.
QSAR Analysis. The following descriptors were used in QSAR modeling: octanol/water distribution coefficient (logD), pKa of the most acidic catechol hydroxyl (pKacat), molar volume (Vm), the number of rotatable bonds (Rot), count of rings (R), total count of nonhydrogen atoms in the substituents located ortho to one of the catechol hydroxyls (oT), indicator for nonconjugated amino group (N+), indicator for carboxyl group (COO-), indicator for third hydroxyl in the catechol ring (OH), count of hydrogen bond donors, count of hydrogen bond acceptors (HBA). ClogP values were calculated with ChemDraw Ultra 6.0 (CambridgeSoft Corp., Cambridge, MA) and were corrected for ionization at pH = 7.4 using calculated pKa values to obtain the distribution coefficients log D = log P- log(1 + 10pH-pKa). The pKa values and molar volumes (Vm) were calculated using the ACD/LogD 4.0 program (Advanced Chemistry Development Inc., Richmond, Toronto, ON, Canada). Catechol group pKa values for dihydroxybenzylidene derivatives were calculated using ΔpKa = 2.7 and 1.0 for cyanoacrylate substituent and acrylate substituent, respectively. Substructure descriptors were counted from the two-dimensional molecular structure. Terminal CX3, OH, or NH2 were not counted as rotors. NH and OH groups were counted as hydrogen bond donors (amine and carboxyl groups ionized). Hydroxyl, ether carbonyl, and carboxyl oxygens were counted as hydrogen bond acceptors. In case of fused ring systems, the definition of ortho-position to the catechol group is ambiguous. Atoms corresponding to a six-membered ring substituent in ortho-position were counted in these cases. PLS analysis was made using the Simca-P 8.1 program (Umetrics AB, Umeå, Sweden). The PLS method was used in QSAR analysis because, unlike multiple linear regression, it allows all variables of possible relevance to be used in the model and is independent of the number of cases and colinearities in the data. In PLS modeling, the structural variables are combined to a few new variables (PLS components), which reflect only the information in the original structural variables that is relevant for modeling and predicting activity (Sjöström and Eriksson, 1995). The number of components in the model is optimized for best predictive ability using cross-validation. Models with good explanatory and predictive power were obtained for SULT1A3, UGT1A9, and S-COMT. The regularities found form the basis for discussing the other enzyme reactions.
Results
Conjugation Rate Studies. Conjugation rates of compounds containing the catechol function were measured in vitro with recombinant human enzymes including six sulfotransferase isoforms (SULTs 1A1, 1A2, 1A3, 1B1, 1C2, and 1E1), five UDP-glucuronosyltransferases (UGTs 1A1, 1A6, 1A9, 2B7, and 2B15), and the soluble form of catechol O-methyltransferase (S-COMT). Several classes of catecholtype compounds were studied: 15 endogenous compounds (catecholamines, their metabolites, and catecholestrogens), 11 drugs (4 dopamine receptor ligands and 7 enzyme inhibitors), 9 plant catechols, and 18 other compounds with varying substituents and substitution patterns to complement the data for structure-activity analysis (Figs. 1 and 2).
Four sulfotransferases (SULTs 1A1, 1A2, 1A3, and 1B1) conjugated catechols from all substrate classes (Table 1). SULT1A3, which has been called the dopamine or catecholamine sulfotransferase, sulfated most of the catechols studied, many of them at a higher rate than the prototypical substrate dopamine (8.5 nmol/mg/min). The highest rate (22.4 nmol/mg/min) was obtained with 4-isopropylcatechol. Control activities measured with 4-nitrophenol for SULTs 1A1, 1A2, and 1B1 were 19, 17, and 28 nmol/mg/min, respectively. Many small neutral catechols were sulfated at comparable rates by these enzymes, but only SULT1A1 sulfated compounds from all substrates classes at high rates. The SULT1E1 isoform, also called estrogen sulfotransferase, sulfated fewer compounds, but several catechols reacted at higher rates than did estrogens. Highest activity was measured for 4-tert-butyl-5-methoxycatechol (7 nmol/mg/min). With SULT1C2, only five catechols were sulfated, and the rates were just above the detection limit.
UDP-glucuronosyltransferases, with a few exceptions, accepted only drugs and small neutral chemicals as substrates (Table 2). UGT1A9 glucuronidated some catechols from all classes, and many compounds reacted at a higher rate than did the phenolic prototype compound propofol (approximately 1 nmol/mg/min) (Lautala et al., 2000). UGT1A1 glucuronidated only 13 compounds, but the rates measured for tetrachlorocatechol and the 12-lipoxygenase inhibitors ethyl-3,4-dihydroxybenzylidinecyanoacetate and 2-(1-thienyl)ethyl-3,4-dihydroxybenzylidinecyanoacetate were comparable to the reference activity. UGT1A6, 2B7, and 2B15 glucuronidated only catechols with small neutral substituents at rates comparable to those of their respective reference substrates.
Most catechols studied were substrates of S-COMT. Compounds representing all substrate classes reacted at a high rate comparable to that of catechol itself (32.2 nmol/mg/min) (Table 2). Of the drugs studied, only dobutamine was O-methylated at a high rate. The other drugs reacted at a clearly lowered rate, or were not substrates.
QSAR Analysis. Analysis of structure-conjugation relationships was carried out for selected enzyme forms (SULT1A3, UGT1A9, and S-COMT) to characterize the substrate specificity differences exhibited by the three enzyme families and to identify structural factors that determine selectivity or control the conjugation rate by a specific route. SULT1A3 and UGT1A9 were selected because these enzymes conjugated a larger number of catechols at high activity and showed less zero activities than the other enzymes in their respective families.
Eleven descriptors to characterize the substituents modifying the catechol structure were calculated. Properties characterized and the corresponding descriptors used were: lipophilicity (octanol/water distribution coefficient at physiological pH, log D), polar interactions properties [counts of hydrogen bond donors (HBD) and acceptors (HBA); indicator for basic nitrogen, N+, and carboxyl group, COO-; and indicator for third hydroxyl in the catechol ring, OH], steric properties and conformational freedom (molar volume, Vm; count of heavy atoms in ortho-substituent, oT; count of rotable bonds, Rot; count of rings, R), and electronic effects (acidity of the catechol group, pKacat). The PLS method was used to model the relationship between conjugation activity and the structural variables. In the case of SULT1A3 and S-COMT, logarithm of the activity was used as the response variable. In the case of UGT1A9, the distribution of the glucuronidation activity values was markedly uneven, displaying 19 zero values, whereas the majority (31 cases) of the activity values were clustered at 2 orders of magnitude above the detection limit. Only three cases of low activity were observed between these two groups. Due to the nature of the data, the cases were assigned to two activity classes: good substrates and nonsubstrates. PLS discriminant analysis (PLS-DA) was carried out to identify structural factors that are effective in discriminating the two classes.
A two-component PLS model containing the 11 structural descriptors and one or more expanded terms were derived for each enzyme reaction. The R2 and Q2 values of the model derived for SULT1A3 were 0.806 and 0.690 including all compounds (n = 50). Leaving out 4-hydroxyestrone, for which the fitted value deviated about 1 log unit from the experimental log activity, gave R2 and Q2 values of 0.831 and 0.729, respectively. The PLS weight plot (Fig. 3) shows how the structure variables are combined to form the quantitative relation between sulfation activity and the substrate structure. Variables located close to each other are correlated. Structure variables in the upper right quadrant (pKacat, N+, log D, and the cross-term N+ × COO-) and the lower left quadrant (OH, HBA, COO-, and oT) are, respectively, positively and negatively correlated with sulfation activity. Variables located in the upper left quadrant are less important for the model, especially R, which is located close to the origin. The relative effect of each structure variable on the predicted activity can be seen from the regression coefficient plot (Fig. 4). The model for S-COMT explained 84% and predicted 72% of the variation in O-methylation rate (R2 = 0.839, Q2 = 0.722, n = 49). Two compounds, tyrphostin A and 2,3-dihydroxybenzoic acid, were removed from the model as outliers. The model contained two second-order terms: square of the count of ortho-substituent heavy atoms (oT2) and the cross-term for the presence of charged groups (N+ × COO-). Catechol group pKa and the steric variable oT displayed the highest positive and negative weights, respectively, in the first PLS dimension, which explains 75% of the response variation (Fig. 5).
PLS discriminant analysis for classification of UGT1A9 substrates gave a two-component model with R2 = 0.830 and Q2 = 0.765 (n = 49). One compound, epicatechin gallate, was removed as an outlier. The model contained one cross-term, log D × Vm. Separation of good substrates from catechols not glucuronidated by UGT1A9 is displayed as a score plot (Fig. 6). Comparison with the weight plot shows that separation in the horizontal dimension, which explains 75%, is most strongly influenced by log D and oT coupled to active compounds, and the presence of carboxylate anion and count of hydrogen bonding groups was reflected in poor reactivity.
Discussion
Catechol itself was a good substrate of S-COMT and most SULT enzymes, but a poor substrate or nonsubstrate of all UGT isoforms. In contrast, all UGT isoforms and all SULT isoforms except SULT1C2 conjugated some catechols at high rates. The presence or absence of the second phenolic hydroxyl adjacent to the reacting hydroxyl is evidently not of crucial importance in the case of SULTs (with the exception of SULT1C2) or UGTs, although it is a strict requirement for substrate specificity in the case of COMT. Therefore, it is probable that many factors found to be important for catechol conjugation apply also for sulfation and glucuronidation of phenolic compounds in general. QSAR analysis was carried out to identify significant factors using an approach similar to that reported recently for modeling glucuronidation of catechols by rat liver microsomes (Antonio et al., 2002). A somewhat simpler parameterization scheme was found adequate in this work.
The effect of substrate lipophilicity and the presence of ionic or hydrogen bonding groups was markedly different for the three enzymes. These factors had a rather small influence in the COMT model. On the contrary, lipophilicity was the most important factor for UGT1A9 activity, whereas the presence of charged or polar groups was strongly correlated with lack of activity. However, good substrates of UGT1A9 included compounds with an ionized group and a low log D value. A positively charged amino group and a small number of hydrogen-bonding groups favored glucuronidation of such compounds. In the case of SULT1A3, lipophilicity was positively correlated with sulfation rate, but the specific effects of polar functional groups were more important than general lipophilicity. The presence of a positively charged amino group favored sulfation, whereas the presence of a carboxylate anion and, especially, a third hydroxyl in the catechol ring, strongly decreased reactivity. The third hydroxyl was better tolerated by UGT1A9 and S-COMT. The crossterm indicating zwitterionic species was significant both in the SULT1A3 model and the COMT model, but the effects were opposite.
The PLS model derived for S-COMT suggests that acidity of the catechol group and the size of the adjacent substituents strongly correlate with decreased methylation rate. A similar (weaker) effect was found for SULT1A3, but the same two factors were modeled to have an opposite effect favoring glucuronidation in the case of UGT1A9. The acid dissociation constant of catechol is similar to that of phenol (pKa = 9.5 and 9.9, respectively). Consequently, the catechol group is un-ionized at physiological pH. One strong electronwithdrawing group, however, can lower pKa by more than 2 log units, resulting in a half-ionized compound, and two such substituents can lower pKa 4 to 5 log units, resulting in complete ionization of the more acidic hydroxyl (Perrin et al., 1981). Electron density on the other hydroxyl is also affected. This may have a strong effect on stability of the Michaelis complex or the transition state leading either to increased or decreased rate, depending on the catalytic mechanism. The effect on catechol O-methylation has been observed previously and explained by excessively increased stability of the Michaelis complex and catecholate anion (Lautala et al., 2001). The reason for the opposite pKa effect in the case of UGT1A9 cannot be deduced based on the available data. It is possible that there is an optimal pKa for the reacting hydroxyl. The highest glucuronidation rates were determined for partially ionized catechols, although even the completely ionized entacapone was glucuronidated at twice the rate of propofol, the neutral phenolic standard substrate of UGT1A9. However, it has been shown that only the less acidic hydroxyl group of entacapone is glucuronidated by this isoform (Lautala et al., 2000).
The size of the substituent adjacent to the catechol group was clearly the most significant of the steric factors modeled. COMT data were modeled best with the square term for this descriptor included. The relationship predicts a steep decrease of methylation activity with more than two heavy atoms in the ortho-substituent.
One compound was excluded as an outlier from each model: 4-hydroxyestrone for SULT, epicatechin gallate for UGT, and tyrphostin A for S-COMT. The reasons for the divergent behavior may be three-dimensional steric effects, which could not be accounted for by the descriptors used, or in case of epicatechin gallate, the presence of impurity. In addition, 2,3-dihydroxybenzoic acid was excluded from the COMT model, because the effects of carboxyl groups in ortho- and meta-positions were different. A separate parameter for ortho-carboxyl could not be used, because there was only one case.
The other UGT isoforms showed narrower substrate specificity for catechols compared with UGT1A9. Nevertheless, several compounds were glucuronidated at high activities by UGT1A1 and 1A6. Three compounds were excellent substrates of UGT1A1: a compact, hydrophobic molecule, tetrachlorocatechol, and two 12-lipoxygenase inhibitors (benzylidenecyanoacetates) with a larger flexible side chain containing polar functional groups. Octyl gallate was one of the control substrates of UGT1A1, but the shorter homolog, methyl gallate, was not glucuronidated. A common feature of good substrates is the lowered pKa of the catechol hydroxyls. The more acidic catechol, entacapone, was a weak substrate, and the two other COMT inhibitors, tolcapone and 2,5-dinitrocatechol, were not glucuronidated. UGT1A6 is considered specific for planar phenols, and indeed, all catechols glucuronidated with high activity possessed a small planar substituent in the meta/para-position. Catechol itself was not a substrate, but a methoxy substituent ortho to a catechol hydroxyl turned it to an excellent substrate. The optimal geometry of the methoxy substituent is not in the aromatic plane. The best substrate, 3-methoxy-5-bromocatechol, exhibits clearly lowered pKa of the catechol group.
We suggested previously (Dajani et al., 1999) that hydrogen bonding of the nonreacting catechol hydroxyl to Tyr240 would explain the relatively low Km observed for certain catechols in SULT1A3, although Tyr240 is not visible in the crystal structures of SULT1A3. The results of this work support Tyr240 as the catechol-bonding residue. Tyr240 is highly conserved among SULTs, but SULT1C2 has Arg in this position. Accordingly, all the other SULTs sulfated at least some catechols with high activity, whereas only marginal sulfation was observed in the case of SULT1C2.
A recent publication reported kinetic analysis of catecholestrogen sulfation by recombinant human SULTs expressed in COS cells (Adjei and Weinshilboum, 2002). These authors found sulfation of catecholestrogens by most human SULTs with the exception of SULT1B1 and SULT1C2. We also could detect no sulfation of the three catecholestrogens studied by SULT1C2; however, unlike Adjei and Weinshilboum (2002), we found that SULT1B1 was able to sulfate these compounds. This apparent difference may reflect the much lower expression level obtained by transient transfection of mammalian cells compared with expression in E. coli.
Some of the catechols studied contain other functional groups, the conjugation of which could interfere with the interpretation of the results. The natural compounds contained several hydroxyl groups, which may be conjugated by SULT and UGT enzymes. However, most of these compounds were poor substrates. The same applies to those compounds that could have been glucuronidated in the carboxyl group.
We are aware of limitations of this approach: use of enzyme activities measured at a single enzyme concentration and the term rate, instead of proper kinetic parameters. However, the number of substrate concentrations we could examine was limited by practical necessity, and we decided to take this approach to obtain the large amount of information needed to characterize catechol conjugation as a whole. More detailed kinetic analysis of certain key individual enzymes will be reported elsewhere.
In conclusion, many structural factors controlling specificity of sulfation, glucuronidation, and O-methylation of catechol-type substrates were identified. The structure-activity relationships and QSAR models derived allow the discrimination of good substrates of a specific enzyme from poor substrates. This level of predictive ability may be useful in the early phases of drug discovery and in planning absorption, distribution, metabolism, and excretion studies.
Acknowledgments
We are grateful to Rana Dajani, Sheila Sharp, and Nicola Rose for valuable assistance and advice with the cloning, expression, and purification of sulfotransferases.
Footnotes
-
↵2 Abbreviations used are: SULT, sulfotransferase; UGT, UDP-glucuronosyltransferase; COMT, catechol O-methyltransferase; S-COMT soluble catechol O-methyltransferase; MB-COMT, membrane-bound catechol O-methyltransferase; PAPS, 3′-phosphoadenosine 5′-phosphosulfate; EMBL, European Molecular Biology Laboratory; AdoMet, S-adenosyl-l-methionine; UDPGA, UDP-glucuronic acid; HPLC, high-performance liquid chromatography; PCR, polymerase chain reaction; oT, SAR, structure-activity relationship; QSAR, quantitative structure-activity relationship; HBA, hydrogen bond acceptors; PLS, partial least squares; DA, discriminant analysis.
-
This work was funded by the Commission of the European Communities (Contract number BMH4-CT97–2621).
-
↵1 Present address: Orion Pharma, Department of Pharmacokinetics, Espoo, Finland.
- Received April 30, 2002.
- Accepted May 15, 2003.
- The American Society for Pharmacology and Experimental Therapeutics