Abstract
RO5263397 [(S)-4-(3-fluoro-2-methyl-phenyl)-4,5-dihydro-oxazol-2-ylamine], a new compound that showed promising results in animal models of schizophrenia, is mainly metabolized in humans by N-glucuronidation. Enzyme studies, using the (then) available commercial uridine 5′-diphosphate-glucuronosyltransferases (UGTs), suggested that UGT1A4 is responsible for its conjugation. In the first clinical trial, in which RO5263397 was administered orally to healthy human volunteers, a 136-fold above-average systemic exposure to the parent compound was found in one of the participants. Further administration in this trial identified two more such poor metabolizers, all three of African origin. Additional in vitro studies with recombinant UGTs showed that the contribution of UGT2B10 to RO5263397 glucuronidation is much higher than UGT1A4 at clinically relevant concentrations. DNA sequencing in all of these poor metabolizers identified a previously uncharacterized splice site mutation that prevents assembly of full-length UGT2B10 mRNA and thus functional UGT2B10 protein expression. Further DNA database analyses revealed the UGT2B10 splice site mutation to be highly frequent in individuals of African origin (45%), moderately frequent in Asians (8%) and almost unrepresented in Caucasians (<1%). A prospective study using hepatocytes from 20 individual African donors demonstrated a >100-fold lower intrinsic clearance of RO5263397 in cells homozygous for the splice site variant allele. Our results highlight the need to include UGT2B10 when screening the human UGTs for the enzymes involved in the glucuronidation of a new compound, particularly when there is a possibility of N-glucuronidation. Moreover, this study demonstrates the importance of considering different ethnicities during drug development.
Introduction
The uridine 5′-diphosphate-glucuronosyltransferases (UGTs) are a family of endoplasmic reticulum membrane enzymes that catalyze glucuronic acid transfer from uridine 5′-diphosphate-α-d-glucuronic acid (UDPGA) onto nucleophilic functional groups (typically containing hydroxyl, carboxylic acid, amine, or heteroaromatic nitrogen functionalities) (Guillemette et al., 2010; Rowland et al., 2013). UGT-catalyzed conjugation is an important metabolic pathway for the detoxification and elimination of bilirubin, as well as many therapeutic drugs and environmental carcinogens (Guéraud and Paris, 1998; Nowell et al., 1999; Ren et al., 2000; Miners et al., 2006). The 19 human UGTs are divided into subfamilies UGT1A, UGT2A, and UGT2B according to sequence similarity and gene structure (Mackenzie et al., 2005). In addition, there are few nucleotide-sugar transferases that mainly use cosubstrates other than UDPGA and they are not further discussed in this study (Rowland et al., 2013).
Despite the important role that UGT enzymes have in drug metabolism (Evans and Relling, 1999; Testa et al., 2012), they have rarely been reported as the key determinants in drug exposure variability or drug–drug interactions. This is most likely because 1) compared with cytochrome P450 enzymes (P450s), there are considerably fewer drugs in clinical practice for which glucuronidation is the primary clearance route; and 2) UGT substrates are often conjugated by more than one UGT enzyme. The likelihood of significant drug–drug interaction or polymorphic clearance due to UGT activity, which requires dose adjustment, is therefore often assumed to be low (Williams et al., 2004). Nevertheless, case reports exist in the literature related to UGT1A1 (Meza-Junco et al., 2009; Riedmaier et al., 2010), UGT1A3 (Riedmaier et al., 2010), and UGT2B17 (Wang et al., 2012). In a recent report (Wang et al., 2012), UGT2B17 poor metabolizers had a mean exposure to MK-7246 ([(7R)-7-{[(4-fluorophenyl)sulfonyl](methyl)amino}-6,7,8,9-tetrahydropyrido[1,2-a]indol-10-yl]acetic acid) that was 82-fold higher than that of extensive metabolizers.
Until recently, UGT2B10, an enzyme mainly expressed in the liver and not detected in the intestine (Ohno and Nakajin, 2009; Court et al., 2012; Fallon et al., 2013; Sato et al., 2014), was not commercially available and little was known about its activity. UGT2B10 has not typically been included in phenotyping work and has often been omitted from, or only mentioned in passing, in reviews of the area (Mackenzie et al., 2000; Tukey and Strassburg, 2000; Miners et al., 2002; Harbourt et al., 2012). Thus, until 2007, UGT2B10 was something of an “orphan” UGT. The identification of UGT2B10 involvement in cotinine, nicotine, and medetomidine N-glucuronidation reactions (Chen et al., 2007; Kaivosaari et al., 2007, 2008) stimulated new interest in the enzyme and showed that it could catalyze conjugation of nitrogen-containing heterocycles. Moreover, the high affinity of UGT2B10 for some tricyclic antidepressant drugs, such as imipramine and amitriptyline (Zhou et al., 2010; Kato et al., 2013), previously classified as UGT1A4 substrates (Breyer-Pfaff et al., 2000; Nakajima et al., 2002), highlighted the importance of UGT2B10 in the metabolism of compounds with more “mainstream” physicochemical characteristics.
Similar to other UGT2Bs, the UGT2B10 gene consists of six exons and is located in chromosome 4q13.2. HapMap data indicate that the UGT2B10 gene is located within one haplotype block in the Caucasian population (CEPH samples) and that four common haplotypes with allelic prevalence of 71, 14, 11, and 2.5% exist for this gene in this population (Chen et al., 2008a). Haplotype C (11% prevalence in Caucasians) encodes the Asp67Tyr mutation, which significantly reduced UGT2B10 activity (Chen et al., 2007, 2008a; Erickson-Ridout et al., 2012). Population studies have shown that this poor-metabolizer genotype (in combination with the UGT2B17 genotype) is associated with reduced circulating concentrations of nicotine and cotinine glucuronides in smokers and consequently reduced nicotine consumption (Berg et al., 2010a). In a small study considering both Caucasian and African-American populations, African-American ethnicity was associated with low nicotine glucuronidation in addition to the UGT2B10 Asp67Tyr allele, but the genetic factor behind this could not be explained at that time (Berg et al., 2010b).
In this article, we show that a UGT2B10 poor-metabolizer phenotype caused by a previously uncharacterized genetic polymorphism is responsible for the significantly higher exposure of RO5263397 [(S)-4-(3-fluoro-2-methyl-phenyl)-4,5-dihydro-oxazol-2-ylamine], which we observed in three subjects. The frequency of this polymorphism is highly population dependent: It is frequent in populations of African origin, moderately frequent in the Asian population, and rare in Caucasians. Our data illustrate the need to include UGT2B10 in phenotyping studies for drugs that are principally cleared by N-glucuronidation, and the importance of considering different ethnicities early enough in drug development.
Materials and Methods
Alamethicin, 17β-estradiol, saccharolactone, UDPGA, Williams’ Eagle’s medium, zidovudine, 7-hydroxy-4-trifluoromethylcoumarin, 4-methylumbelliferone, and Tris buffer were purchased from Sigma-Aldrich (including Fluka) (St. Louis, MO). RO5263397, [14C]RO5263397, [14C]amitriptyline, [14C]trifluoperazine, and dexmedetomidine were synthesized at Roche (Basel, Switzerland). Hecogenin was from ABCR (Karlsruhe, Germany), S-nicotine was from Alfa Aesar (Karlsruhe, Germany), and trifluoroperazine was from LKT Laboratories (St. Paul, MN). The BCA protein concentration measurement kit was from Pierce Biotechnology (Rockford, IL). Dexmedetomidine, levomedetomidine, and entacapone were a generous gift from Orion Pharma Ltd. (Espoo, Finland) to the University of Helsinki laboratory. The other used chemicals were of the highest purity available.
The enzyme activity studies, both UGT screening (phenotyping) and kinetics, were done in two different laboratories, in the Roche Laboratory of Drug Disposition and Safety, Pharmaceutical Research and Early Development at F. Hoffmann-La Roche Ltd. (Basel, Switzerland; referred to hereafter as Roche) and at the University of Helsinki Faculty of Pharmacy (Helsinki, Finland). The recombinant UGTs and some of the reagents used by each laboratory are listed below.
For the studies performed at Roche, the recombinant UGT1A1, UGT1A4, UGT1A6, UGT1A7, UGT1A8, UGT1A9, UGT1A10, UGT2B4, UGT2B7, UGT2B15, and UGT2B17, as well as UGT expression control membranes, pooled human liver microsomes (HLMs), and pooled human intestine microsomes (HIMs) were purchased from BD Biosciences (Woburn, MA; now acquired by Corning). The University of Helsinki laboratory has expressed all of these UGTs, as well as UGT2B10, locally, as previously described (Kaivosaari et al., 2007). The HLMs and HIMs used in Helsinki were from the same supplier as in Basel (Supplemental Table 6).
Enzymology Methods
Descriptions of the high-performance liquid chromatography (HPLC) system and methods are given in Supplemental Table 7. The “quench reagent” used was a 3:2 mixture of 10% trichloroacetic acid and acetonitrile. Quenched samples showed no degradation of glucuronide at room temperature over 24 hours. All quenched samples were worked up by the same procedure unless otherwise stated.
HLM Kinetics.
One-milliliter incubations were made up containing (finally) 10–250 µM [14C]RO5263397, 0.5 mg/ml pooled HLMs (pretreated for 30 minutes on ice with 67 µg alamethicin/mg HLM), 4 mM UDPGA, 5 mM saccharolactone, and 10 mM MgCl2 in 100 mM Tris buffer, pH 7.5. Incubations were warmed to 37°C before initiation by the addition of UDPGA. Incubations were stopped after 30 minutes by the addition of 333 µl quench reagent. Kinetic models (Michaelis–Menten, two-enzyme Michaelis–Menten, or substrate inhibition) were applied based upon inspection of the Eadie–Hofstee plots.
UGT Phenotyping at Roche.
Recombinantly expressed UGT enzymes were obtained from BD Biosciences. Three hundred–microliter incubations were made up containing 50 µM RO5263397 (or control substrate), 0.5 mg/ml membrane protein, 2 mM UDPGA, 5 mM saccharolactone, and 10 mM MgCl2 in 100 mM Tris buffer, pH 7.5. Reactions were warmed to 37°C and started by addition of UDPGA. After 60 minutes shaking at 37°C, reactions were stopped by addition of 100 µl quench reagent.
UGT Phenotyping and Kinetics at University of Helsinki.
One hundred–microliter incubations contained 0.2–2.0 mg/ml membrane protein [UGT2B10 cell homogenate was used as enzyme activity was almost completely lost if membrane preparation was performed (Kaivosaari et al., 2007)], 10 mM MgCl2, and UDPGA in 50 mM phosphate buffer, pH 7.4. Alamethicin (5% protein concentration) was added to samples containing HLMs or HIMs and the samples preincubated for 30 minutes on ice. Incubates were warmed to 37°C and reactions initiated by addition of UDPGA (5 mM). After 60 or 120 minutes, the reactions were terminated by addition of 60 µl of a 1:5 mixture of 4 M perchloric acid and methanol. Quenched incubates were placed on ice for 30 minutes and then centrifuged at 16,000g for 10 minutes. Supernatants were analyzed by HPLC-radioflow or ultra-performance liquid chromatography (LC) coupled with UV detection. For phenotyping experiments, concentrations of 2, 10, and 50 µM RO5263397 were used (10 µM data shown). In kinetics experiments, RO5263397 concentrations between 2 and 250 µM were used.
Chemical Inhibition Experiments.
Test substrates were 3 µM [14C]RO5263397, 5 µM amitriptyline, and 60 µM trifluoperazine, prepared as dimethylsulfoxide stock solutions. Test inhibitors were 10 µM hecogenin, 100 µM dexmedetomidine, and 500 µM S-nicotine (prepared as methanol stock solutions, with methanol as 100% activity control). HLMs were pretreated with 67 µg alamethicin per milligram protein on ice for 30 minutes. Five hundred–microliter incubations were made up containing substrates, inhibitors, 0.5 mg/ml HLM, 4 mM UDPGA, 5 mM saccharolactone, and 10 mM MgCl2 in 100 mM Tris buffer, pH 7.5. Final solvent concentrations were 0.3% (v/v) dimethylsulfoxide and 0.3% (v/v) methanol. Incubations were warmed to 37°C before initiation by UGPGA addition. Incubation times were 15, 30, and 40 minutes for RO5263397, trifluoperazine, and amitriptyline incubations, respectively. Incubations were quenched by addition of 167 µl quench reagent.
Hepatocyte Experiments.
Cryopreserved human hepatocytes from were obtained BD Biosciences (San Jose, CA), Celcis IVT (Brussels, Belgium), or XenoTech (Lenexa, KS). For batch information, see Supplemental Table 5. Cells were thawed and prepared following the supplier’s protocol (Celcis IVT). Two hundred–microliter incubations were made up containing 1 million viable cells per milliliter incubate in Williams’ Eagle’s medium and incubated with 3 µM [14C]RO5263397 (20 minutes), 60 µM trifluoperazine (60 minutes), 15 µM 17β-estradiol (10 minutes), 5 µM [14C]amitriptyline (60 minutes), and 100 µM 7-hydroxy-4-trifluoromethylcoumarin (20 minutes). Reactions were stopped by the addition of 67 µl quench reagent. Triplicate incubations were performed and the intrinsic clearance was calculated on the basis of percentage total radioactivity in direct glucuronide metabolite peak or calibration of the UV or fluorescence peak area against authentic external standard materials. In the case of trifluoperazine, a follow-up experiment using pooled cryopreserved hepatocytes and [14C]trifluoperazine was used to determine UV response factors.
Sample Workup and HPLC Analysis.
Quenched incubates were cooled on ice for 30–60 minutes and then centrifuged at 20,000g for 10 minutes. Supernatants were removed and analyzed by HPLC. Chromatogram fractions were collected in 96-well LUMA solid scintillation plates (PerkinElmer, Waltham, MA), dried down in a stream of nitrogen, and the radioactivity in each well was determined using a TopPlate NTX scintillation counter (PerkinElmer). HPLC methods and conditions can be found in Supplemental Table 7.
Clinical Study
This was a single-center (Institut de Pharmacologie Clinique Roche, Strasbourg, France), single-ascending dose, randomized, double-blind, placebo-controlled study in 49 healthy male subjects aged 18–45 years, to investigate safety, tolerability, pharmacokinetics (PK), and pharmacodynamics after oral administration of the novel compound RO5263397. Single doses of 1, 3, 10, 30, 60, and 125 mg RO5263397 were administered orally in capsules after an overnight fast, with six subjects receiving active drug per dose and two subjects receiving placebo, in a leap-frog, three-period crossover design. Two cohorts of eight subjects were dosed alternatingly with ascending higher doses, and all subjects were randomly assigned to receive placebo in one period and two active doses at different dose levels between 1 and 125 mg (in fasted state) in the two other periods. In the food effect arm, subjects were administered a single dose of 60 mg once in the fasted state and once in the fed state. The low dose of 5 mg was implemented via a protocol amendment, after PK data for subject 1008 became available, in order to probe the RO5263397 exposure phenotype, as a screening tool for all future subjects to be included into this study. The dose of 5 mg was selected to allow proper measurement and evaluation of PK well above the limit of quantification, while being safe and well tolerated. The exposure after 5 mg was expected to stay as well for poor metabolizers below the maximum highest exposure [dose 125 mg; mean area under the plasma concentration versus time curve extrapolated to infinity (AUC∞) 143 ng⋅h/ml, mean Cmax 50 ng/ml], which had been shown already in this study to be safe and well tolerated in all subjects. The study was conducted at the Institut de Pharmacologie Clinique Roche in concordance with the ethical principles in the Declaration of Helsinki and with applicable local regulations. This study was approved by an independent ethics committee (Nancy, France). All subjects gave written informed consent before participating.
Pharmacokinetic blood samples for the main study part were collected predose, and 0.5, 1, 2, 3, 4, 5, 6, 8, 12, 24, and 48 hours postdose. After PK data for subject 1008 became available, all subsequent newly recruited subjects had to undergo a phenotypic screening via administration of a probe dose of 5 mg as part of the screening assessments. After administration of this 5-mg probe dose, PK samples were obtained at 1, 3, and 6 hours postdose.
The concentration of RO5263397 in plasma was quantified using a validated LC–tandem mass spectrometry method, at Roche Analytical Laboratory (Basel, Switzerland). The lower limit of quantification was 0.1 ng/ml; the calibration range was 0.1–1000 ng/ml. The precision and accuracy of the assay were 4.1–11.7%, and 96.3–105.8%, respectively. The metabolite was also assessed. However, because no synthesized RO5263397-glucuronide was available as the reference compound, absolute quantitation was not possible. Pharmacokinetic parameters were estimated by noncompartmental analysis using WinNonlin 5.2.1. software (Pharsight Corporation, Mountain View, CA); actual blood sampling times were used. Values below the limit of quantification before Cmax were set to zero and after Cmax to missing. The area under the concentration-time curve from time zero to the last measurable concentration postdose (AUClast) was calculated using the linear trapezoidal rule. The t1/2 and AUC∞ were calculated by estimating λz. The terminal elimination rate constant λz was assessed by applying a linear regression over a minimum of the last three logarithmically transformed data points, allowing a fit with residual square R2 value of ≥0.7. The following parameters are reported: Cmax, tmax (time to maximum observed plasma concentration), AUClast, AUC∞, and t1/2.
mRNA Analyses
Relative concentration of UGT2B10 mRNA was measured using real-time reverse transcriptase quantitative PCR in cryopreserved human hepatocytes. Experimental conditions are described in the Supplemental Materials and Methods.
Genotyping Methods
All subjects who enrolled in this study and gave consent for DNA analysis were included in this genetic analysis. Subjects’ data were anonymized. Participating subjects signed an additional separate informed consent form before donating blood samples for DNA analysis. After the protocol amendment, clinical genotyping was included for all subjects within the main protocol and informed consent form. Genotyping methods are described in the Supplemental Materials and Methods.
UGT2B10 DNA Sequencing
UGT2B10 gene exons and intron-exon boundary regions were sequenced in clinical samples and in commercially available cryopreserved hepatocytes.
Results
Compound Structure and Properties
The chemical structure of RO5263397 is shown in Fig. 1 (Revel et al., 2013). Glucuronidation occurred on the dihydrooxazole ring nitrogen (see the Supplemental Materials and Methods for NMR structure assignment data). RO5263397 plasma protein binding was concentration independent with free fractions of 52% (rat), 53% (dog), 60% (cynomolgus), and 55% (human). The compound’s absolute bioavailability was 84% (rat), 32% (dog), and 27% (cynomolgus); no human absolute bioavailability study was conducted. RO5263397 is classified as a high-permeability and high-solubility compound (class I in the Biopharmaceutics Classification System), and is not a substrate of P-glycoprotein. On the basis of in vitro data, moderate systemic clearance was expected in humans.
Clinical Findings
After a single 10-mg oral administration of the novel compound RO5263397 (Revel et al., 2013) in a first-in-human study, one individual subject (subject number 1008) showed 136-fold higher exposure (AUClast; AUC∞ 171-fold) and 22-fold higher peak concentration versus all other subjects receiving the same 10-mg dose (Fig. 2A; Supplemental Table 1). This subject 1008 had significantly higher plasma concentrations of the parent compound, but no metabolism product, an N-glucuronide. LC–tandem mass spectrometry analysis confirmed the absence of the only major metabolite of RO5263397 in subject 1008, whereas the metabolite was present extensively in all other individuals. The finding in subject 1008 was confirmed via a second administration of a slightly higher dose of 15 mg to this particular individual. After PK data for subject 1008 became available, all subsequently recruited subjects had to receive a probe dose of 5 mg as part of the screening assessment. During this phenotypic screening with the probe dose of 5 mg, limited PK sampling was carried out in 30 new subjects to identify poor metabolizers to exclude them from the study for safety reasons. The PK profile in subject 1008 indicated a significant reduction in first-pass metabolism with increased bioavailability and increased Cmax, and therefore limited PK sampling up to 6 hours postdose was considered sufficient to identify similar subjects during this screening assessment. Upon inspection of the PK data, two additional individuals (subjects 1019 and 1036) with much higher plasma concentrations were identified (Fig. 2B; Table 1). All three poor metabolizers were of African origin.
In the general study population (i.e., after exclusion of identified poor metabolizers), exposure increased slightly more than dose proportionally from 1 to 60 mg, when comparing the two administered dose levels within an individual subject (Table 1; see Supplemental Table 2 for summary statistics ). Starting from 60 to 125 mg, an even larger overproportional exposure increase occurred. AUC increased from 32.3 ± 20.7 to 210 ± 196 ng⋅h/ml (i.e., 6.5-fold versus the 2-fold dose increase), and Cmax increased 9-fold from 10.7 ± 9.74 to 89.4 ± 103 ng/ml. Marked intersubject variability was observed at 60 mg, because individual subjects seemed to have different metabolic capacity for this specific pathway. The range of individual values at 60 mg was 57-fold for Cmax (0.644–36.9 ng/ml; coefficient of variation in percentage [CV%] 144) respectively 30-fold for AUC (4.34–130 ng⋅h/ml; CV% 117). A comparison of dose-normalized AUC and dose-normalized Cmax across all three genotype groups is shown in Supplemental Table 8. Administration of RO5263397 was well tolerated in all subjects at all doses, without any relevant safety findings.
Enzymology
RO5263397 was metabolized by human hepatocytes to give a single direct glucuronide as the major metabolite (>95%; data not shown). The kinetics of glucuronidation by HLMs was clearly biphasic (Fig. 3, A and B), indicating the involvement of two or more UGT enzymes. A Km value of 5.3 ± 1.0 µM was estimated for the high-affinity enzyme, whereas that of the low-affinity enzyme could not be calculated due to lack of saturation (Supplemental Table 3).
An extensive panel of recombinant UGT enzymes was tested for RO5263397 glucuronidation activity. Turnover was only initially observed with UGT1A4 (Fig. 4A, left; control data shown in Supplemental Table 4). Because RO5263397 was known to be N-glucuronidated and considering the similarity between RO5263397 and medetomidine, experiments were performed at the University of Helsinki to assess the potential involvement of UGT2B10, which had been expressed in-house (Kaivosaari et al., 2007). These experiments demonstrated that recombinant UGT2B10 could metabolize RO5263397 and confirmed the UGTs activities previously seen (Fig. 4A, right).
Kinetics experiments showed that recombinant UGT1A4 has a low affinity for RO5263397 (Km > 1000 µM; Fig. 3C), whereas recombinant UGT2B10 has a high affinity with a Km of 3.1 ± 0.8 µM (Fig. 3D), very similar to the estimated Km value in HLMs (5.3 ± 1.0 µM). The involvement of UGT1A4 and UGT2B10 was also consistent with the high relative activity of HLMs compared with HIMs in which these enzymes are not expressed (Ohno and Nakajin, 2009; Court et al., 2012; Fallon et al., 2013)
Chemical inhibition experiments, using HLMs, focused on UGT1A4 and UGT2B10 activities. These experiments utilized a low concentration of RO5263397 (3 µM) to ensure that the high-affinity enzyme relevant for the in vivo glucuronidation was probed. In parallel to RO5263397, we tested sensitivity of the glucuronidation of the marker substrates amitriptyline (UGT2B10) and trifluoperazine (UGT1A4) (Uchaipichat et al., 2006; Zhou et al., 2010) to three inhibitors. The inhibitors were the UGT1A4-specific hecogenin (Uchaipichat et al., 2006; Kato et al., 2013), and the almost UGT2B10-specific dexmedetomidine and nicotine (Kaivosaari et al., 2008). At a concentration of 100 µM, dexmedetomidine selectively inhibited UGT2B10 activity, reducing both RO5263397 and amitriptyline N-glucuronidation by 79%, while only inhibiting UGT1A4-mediated trifluoperazine glucuronidation by 5% (Fig. 4B). Similarly, 500 µM S-nicotine caused partial inhibition of RO5263397 (29%) and amitriptyline (25%) glucuronidation but no inhibition of UGT1A4. Conversely, 10 µM hecogenin inhibited trifluoperazine glucuronidation by 50% but only caused 5% inhibition of amitriptyline or RO5263397 glucuronidation. The profile of RO5263397 glucuronidation inhibition thus exhibited a close match to amitriptyline but a marked difference from trifluoperazine. The combination of individual recombinant UGT enzymes activity profiling, kinetics, and inhibition data all indicated that UGT2B10 is the principal enzyme involved in RO5263397 metabolism.
Genetic Variation in UGT2B10
UGT2B10 exon and exon-intron boundary sequencing revealed a total of 14 single nucleotide polymorphisms (SNPs) in the 48 clinical samples studied. No other known SNPs or different type of variations such as insertions or deletions were found in this sample set. Table 1 lists the SNPs found. Homozygous variants were found for only five SNPs (rs294777, rs9917968, rs861340, rs2942857, and rs61301802) in six subjects (subjects 1008, 1019, 1025, 1027, 1036, and 2011). Subjects 1008, 1019, and 1036, who showed poor metabolism of RO5263397, were found among those participants. They have two homozygous variants in common: rs861340 (GG) in intron 2 and rs2942857 (CC) at the juncture between intron 2 and exon 3. Only genotype CC (rs2942857) correlated with poor metabolism because genotype CC (rs861340) was also found in subjects 1025, 1027, and 2011, who had extensive metabolism. The rs2942857 has been described as splice acceptor site between intron 2 and exon 3 (Abecasis et al., 2010). A defect in this splice site would most probably result in a UGT2B10 mRNA without exon 3. That such an mRNA has not been described in publically available databases may indicate that this SNP disrupts the mRNA. Attempts to identify UGT2B10 mRNA subjects carrying the CC genotype by quantitative PCR in human liver hepatocytes failed, supporting this hypothesis (data not shown). Similar effects due to SNPs at splice sites resulting in undetectable mRNA species have been observed in other genes, such as TDP-43 (Avendaño-Vázque et al., 2012). In addition, while this article was in preparation, a new study was published, reporting that the same splicing mutation affects nicotine glucuronidation in African Americans (Murphy et al., 2014).
The three poor-metabolizer subjects were of African origin. The rs2942857 variant allele C was found in Africans with a frequency of 45%, whereas the frequency was only 7% in Native Americans, 8% in Asians, and 0% in Europeans (Abecasis et al., 2010). Among the African subpopulations, similar frequencies are described in Americans of African ancestry in the southwestern United States (ASW, 39%), Luhya in Webuye, Kenya (LWK, 44%), and Yoruba in Ibadan, Nigeria (YRI, 50%). These frequencies clearly indicate that this SNP is abundantly present in subjects of African origin. On the contrary, rs861340 allele variant G is found in all populations (73% in Africans, 18% in Native Americans, 28% in Asians, and 10% in Europeans). Interestingly, all subjects carrying the rs2942857 CC genotype also carried the rs861340 variant GG, but not vice versa. These two SNPs are located only 23 bases apart on chromosome 4 and only in moderate linkage disequilibrium, as shown for example in African ASW (r2 = 0.329), LWK (r2 = 0.264), and YRI (r2 = 0.333) populations and others, such as Chinese (r2 = 0.209), Japanese (r2 = 0.236), and Mexican (r2 = 0.387) populations as described in Abecasis et al. (2010).
No other SNPs found in the clinical samples correlated with poor metabolism. Interestingly, rs61750900 (Asp67Tyr), a variant with an allelic prevalence of approximately 10% in Caucasians and approximately 3% in Africans, has been described to cause significantly reduced UGT2B10 activity toward tobacco-specific nitrosamines, nicotine, cotinine, and olanzapine (Chen et al., 2007, 2008a,b; Erickson-Ridout et al., 2012). This variant was found in our study only in the heterozygous state (GT) in subjects 1002, 1013, 1016, 1018, 1022, 1024, 1026, 1030, 1032, and 2005 (Table 1), but none of these subjects showed reduced RO5263397 metabolism.
UGT2B10 Genotyping and Phenotyping Using Human Hepatocytes
The relationship between RO5263397 metabolism and rs2942857 genotype was tested prospectively in vitro using single-donor cryopreserved hepatocytes. To enhance the chance of obtaining rs2942857 homozygous cells, the 20 batches of hepatocytes were all sourced from African origin donors. Results of rs2942857 genotyping, UGT2B10 mRNA measurements, and RO5263397 metabolism are shown in Fig. 5 and Supplemental Table 5. RO5263397 glucuronidation activity was highly variable (83-fold) within this set of hepatocyte samples. By contrast, UGT1A4-mediated trifluoperazine glucuronidation varied only 17-fold. This was consistent with the high variability reported for UGT2B10 expression in HLMs (Fallon et al., 2013). The only sample carrying the rs2942857 CC genotype (subject 03C) displayed the lowest RO5263397 glucuronidation activity (0.16 µl/min per million cells) and undetectable mRNA. Subjects with a heterozygous genotype (AC) (n = 10) tended to have intermediate activities (average 3.13 µl/min per milligram) and only AA genotypes (n = 9) reached the highest activity (average 7.54 µl/min per milligram). The relationship with UGT2B10 mRNA expression is less clear but the samples showing higher expression have the reference AA allele. The mRNA and activity may be influenced by the fact that the UGT2B10 measured was relatively low. The rs61750900 Asp67Tyr was also genotyped but no correlation could be found with glucuronidation activity as shown in Supplemental Table 5. For the UGT2B10 substrates RO5263397 and amitriptyline, there was a clear association of metabolic activity with rs2942857 genotype (Fig. 6). Conversely, no association between marker reactions of UGT1A1 (17β-estradiol 3-glucuronidation), UGT1A4 (trifluoperazine glucuronidation), or overall glucuronidation activity (7-hydroxy-4-trifluoromethylcoumarin glucuronidation) was apparent.
Discussion
After first-in-human oral administration of the novel compound RO5263397, an unexpectedly dramatic difference in exposure of 136-fold was observed in one individual subject of African origin. Two further subjects of African origin were identified as poor metabolizers via phenotypic screening. A combination of enzyme phenotyping, inhibition, and kinetics investigations has generated a highly consistent picture of RO5263397 as an almost selective and high-affinity substrate of UGT2B10.
The fraction of RO5263397 metabolism mediated by glucuronidation is very high, in contrast with other UGT2B10 substrates. For instance, typically between 15 and 35% of amitriptyline metabolism by cryopreserved human hepatocytes was via direct glucuronidation, the majority of the metabolism coming from other enzymes, mainly P450s. Similarly, the primary route of nicotine metabolism is via oxidation rather than glucuronidation, making UGT2B10 phenotype a contributing, rather than a determining, factor in nicotine clearance. RO5263397 metabolism was therefore both selectively via glucuronidation and selectively via the UGT2B10 enzyme, making clearance highly sensitive to UGT2B10 activity. This, coupled with high first-pass extraction, led to the greatly elevated exposure in poor-metabolizer individuals.
We identified a UGT2B10 SNP (rs2942857) that correlated with poor RO5263397 clearance in our clinical study. rs2942857 is found predominantly in subjects of African origin (45%) and has been referred to as a splice acceptor variant. The relationship of the UGT2B10 rs2942857 alleles, mRNA expression, and RO5263397 and amitriptyline clearance in a panel of hepatocytes from individual African donors revealed that subjects carrying the homozygous genotype CC had poor metabolism and undetectable mRNA. Interestingly, a study on nicotine metabolism in different ethnic groups that was published shortly before our work was submitted for publication (Murphy et al., 2014) has identified the same polymorphic variant of UGT2B10 in African Americans. In other words, two independent studies, ours on a new drug candidate and theirs on nicotine metabolism, although initiated from very different starting points, reached the same finding and they can thus be viewed as supporting and validating each other.
It has often been assumed thus far that elimination by glucuronidation is typified by enzyme redundancy and low binding affinity, making glucuronidation a “safe” pathway for metabolism, lacking the polymorphism risks known for P450 enzymes. However, the trend toward non-P450 metabolism as a result of oxidative metabolic stability optimization, combined with the increased frequency of nitrogen-containing heterocyclic moieties being introduced into drug candidates, is likely to result in increasing numbers of UGT2B10 substrates being developed. Encountering a 136-fold higher exposure in a poor metabolizer in a first-in-human dose escalation study as presented is a significant safety risk, especially for narrow therapeutic range drugs. It is therefore important that less well understood metabolic enzyme classes, such as UGTs, become better characterized, because they may indeed be implicated as the principal routes of clearance of a new drug compound. A better understanding of ethnic variation in metabolic polymorphisms is also required. Phase 1 studies are typically conducted in a small subject number in a single clinical center, and the subject population is not representative of the overall patient population. To ensure patient safety when moving from the initial, small-scale, well controlled clinical trial to large clinical studies and the patient population as a whole, inclusion of different ethnicities in early clinical studies is highly recommended.
Acknowledgments
The authors thank Sandrine Simon for technical assistance in performing the human hepatocyte incubations; Christine Brodhag, Nadine Kumpesa, and Debora Souza da Costa for the DNA sequencing and mRNA expression experiments; Martin Binder for the NMR structural elucidation of RO5263397-glucuronide; Johanna Mosorin (University of Helsinki) for recombinant UGT expression; and the laboratory of Dr. Thomas Hartung, where radiolabelled RO5263397, amitriptyline, and trifluoperazine were prepared.
Authorship Contributions
Participated in research design: Fowler, Kletzl, Finel, Tuerck, Spleiss, Iglesias.
Conducted experiments: Manevski, Schmid.
Contributed new reagents or analytic tools: Norcross, Hoener.
Performed data analysis: Fowler, Kletzl, Finel, Tuerck, Spleiss, Iglesias.
Wrote or contributed to the writing of the manuscript: Fowler, Kletzl, Finel, Manevski, Schmid, Tuerck, Norcross, Hoener, Spleiss, Iglesias.
Footnotes
- Received September 26, 2014.
- Accepted December 10, 2014.
↵1 Current affiliation: Novartis Pharmaceuticals Ltd., Basel, Switzerland.
S.F. and H.K. contributed equally to this work.
This research was supported by F. Hoffmann-La Roche Ltd.
M.F. is a coauthor on a patent application (no. 20100087493) claiming rights related to the use of UGT2B10 modulators to improve the pharmacokinetics of UGT2B10 substrates, but has no conflict of interest with respect to this work.
↵This article has supplemental material available at jpet.aspetjournals.org.
Abbreviations
- AUC∞
- area under the plasma concentration versus time curve extrapolated to infinity
- AUClast
- area under the concentration-time curve from time zero to the last measurable concentration postdose
- HIM
- human intestinal microsome
- HLM
- human liver microsome
- HPLC
- high-performance liquid chromatography
- LC
- liquid chromatography
- MK-7246
- [(7R)-7-{[(4-fluorophenyl)sulfonyl](methyl)amino}-6,7,8,9-tetrahydropyrido[1,2-a]indol-10-yl]acetic acid
- P450
- cytochrome P450 isozyme
- PK
- pharmacokinetics
- RO5263397
- (S)-4-(3-fluoro-2-methyl-phenyl)-4,5-dihydro-oxazol-2-ylamine
- SNP
- single nucleotide polymorphism
- UDPGA
- uridine 5′-diphosphate-α-d-glucuronic acid
- UGT
- uridine 5′-diphosphate-glucuronosyltransferase
- Copyright © 2015 by The American Society for Pharmacology and Experimental Therapeutics