Chem. Res. Toxicol. 2009, 22, 299–310
299
Articles Nuclear Magnetic Resonance Spectroscopy as a Quantitative Tool To Determine the Concentrations of Biologically Produced Metabolites: Implications in Metabolites in Safety Testing Robert Espina, Linning Yu, Jianyao Wang, Zeen Tong, Sarvesh Vashishtha, Rasmy Talaat, JoAnn Scatina, and Abdul Mutlib* Drug Safety and Metabolism, Wyeth Research, 500 Arcola Road, CollegeVille, PennsylVania 19426 ReceiVed July 11, 2008
Nuclear magnetic resonance (NMR) spectroscopy has traditionally been considered as an indispensable tool in elucidating structures of metabolites. With the advent of Fourier transform (FT) spectrometers, along with improvements in software and hardware (such as high-field magnets, cryoprobes, versatile pulse sequences, and solvent suppression techniques), NMR is increasingly being considered as a critical quantitative tool, despite its lower sensitivity as compared to mass spectrometry. A specific quantitative application of NMR is in determining the concentrations of biologically isolated metabolites, which could potentially be used as reference standards for further quantitative work by liquid chromatography/mass spectrometry. With the recent demands from regulatory agencies on quantitative information on metabolites, it is proposed that NMR will play a significant role in strategies aimed at addressing metabolite coverage in toxicological species. Traditionally, biologically isolated metabolites have not been considered as a way of generating “reference standards” for further quantitative work. However, because of the recent FDA guidance on safety testing of metabolites, one has to consider means of authenticating and quantitating biologically or nonbiologically generated metabolites. 1H NMR is being proposed as the method of choice, as it is able to be used as both a qualitative and a quantitative tool, hence allowing structure determination, purity check, and quantitative measurement of the isolated metabolite. In this publication, the application of NMR as a powerful and robust analytical technique in determining the concentrations of in vitro or in vivo isolated metabolites is discussed. Furthermore, to demonstrate the reliability and accuracy of metabolite concentrations determined by NMR, validation and cross-validation with gravimetric and mass spectrometric methods were conducted. Introduction The recent guidance on “Safety Testing of Drug Metabolites” issued by the U.S. Food and Drug Administration Center for Drug Evaluation and Research (CDER) has highlighted the importance of identifying and characterizing drug metabolites as early as possible in drug discovery and development (1). Identification of the potentially significant drug metabolites (>10% of the parent exposure), especially those anticipated in humans after multiple dosing, has become a very critical activity during our search for therapeutic agents. In an effort to predict the possible metabolites that could be found in humans, in vitro studies are often conducted with human liver microsomes and * To whom correspondence should be addressed. Tel: 484-865-7525. Fax: 484-865-9408. E-mail:
[email protected]. 1 Abbreviations: ADME, absorption, distribution, metabolism and excretion; APAP, acetaminophen; APAP-Gluc, acetamidophenyl β-D-glucuronide sodium; BID, twice a day (bis in die); DMPK, drug metabolism and pharmacokinetics; ESI, electrospray ionization; 1H NMR, proton nuclear magnetic resonance; 12OH-lauric acid, 12-hydroxydodecanoic acid; LC/ MS, liquid chromatography/mass spectrometry; LOQ, limit of quantitation; MS/MS, mass spectrometry/mass spectrometry; NBC, 4-nitrobenzyl chloride; NB-GSH, S-(p-nitrobenzyl)glutathione; PCA, 4-pyridinecarboxaldehyde; PCA-NO, 4-pyridinecarboxaldehyde N-oxide; QC, quality control; RSD, relative standard deviation.
hepatocytes (fresh or cryopreserved). Frequently, these studies are conducted with nonlabeled “cold” compounds, and the identification of metabolites is often achieved through the application of liquid chromatography/mass spectrometry (LC/ MS)1techniques. With the recent advancements made in LC/ MS hardware and software, identification and characterization of metabolites in complex matrices have become a fairly routine practice in drug metabolism laboratories. While LC/MS is used routinely to obtain structures of metabolites, it has limitations in its ability to precisely identify sites of metabolic modifications in a number of cases. NMR, especially in combination with mass spectrometry and other hyphenated techniques, is customarily used to fully elucidate structures of previously uncharacterized metabolites (2). NMR offers the best hope of not only providing definitive structural data on the predicted major in vitro human metabolites but also quantitative information on the levels of the isolated metabolites. NMR has been used in the past as an analytical tool to determine concentrations of synthetic and biosynthetic products (3-14), metabolites, catabolites, and endogenous compounds in biological fluids (15-35), impurities present in products (7, 36-38), or endogenous components in metabolomic studies (39-41). The application of NMR as a tool to obtain quantitative
10.1021/tx800251p CCC: $40.75 2009 American Chemical Society Published on Web 11/04/2008
300
Chem. Res. Toxicol., Vol. 22, No. 2, 2009
information on the relative levels of endogenous compounds in metabolomic studies has been reviewed and will not be elaborated in this article (42-44). The theoretical aspects and validation of quantitative NMR have been discussed and reviewed in the literature (45-48). It is a routine practice that metabolites are initially isolated in sufficient quantities from a biological system such as liver microsomes, hepatocytes, urine, or bile for the purposes of structural elucidation. Upon confirmation of the structures by NMR, requests are then made to medicinal chemists to scale up the synthesis of these prominent in vitro human metabolites. However, depending on the resources and complexity of the synthesis, the receipt of these metabolite standards may be protracted, taking anywhere from several weeks to months. Most pharmaceutical organizations are not geared to meet these demands, especially now in response to the new FDA guideline on safety testing of metabolites. A practical solution is sought that will enable us to obtain a quantitative assessment of metabolite exposure in humans and coverage in toxicology species. While the use of radiolabeled materials is one option, there are often limitations in terms of availability of compound and study designs. To determine the exposure to a metabolite at steady state requires multiple dosing; hence, studies with radiolabeled compounds to address metabolite coverage need to take that point into consideration. Another viable option is to use biologically produced metabolites as analytical standards to obtain quantitative information on the exposure levels of these metabolites at steady state. However, gaining a reliable estimate of the amount of biologically generated metabolites is challenging due to the low (typically microgram) levels of compounds that are isolated. Depending on the source from which the metabolite is isolated, the amount of material that has traditionally been submitted for NMR characterization has ranged from approximately 10 µg to several hundred micrograms. Nowadays, with robust in vitro metabolite generating systems, it is often possible to isolate between 50-200 µg of the metabolites fairly easily. In recent years, analytical techniques have been improved to a point where pure samples of low-level metabolites can be obtained and structures determined. The NMR analyst can easily determine the degree of purity of the samples, a practice that is also used by analytical chemists to validate the purity of synthetic standards. From the experience that we have gained recently, it appears that reliable quantitative measurements of isolated metabolites can be made using NMR. We are proposing that the biologically produced and isolated metabolites be used as initial “reference standards” to obtain further quantitative information on their circulating levels at steady state in humans and toxicology species. Obviously, the reference standards obtained via this approach cannot undergo the rigorous testing required of an analytical standard intended for a GLP study. It is the intent of this publication to demonstrate that NMR can be used as a reliable tool to obtain concentrations of the biologically produced metabolites that could be used further for quantitative studies by other techniques such as LC/MS. This strategy provides us with an option to address metabolite coverage in toxicity species using nonradiolabeled compounds.
Materials and Methods Chemicals and Supplies. Acetaminophen (4-acetamidophenol) (APAP), acetamidophenyl β-D-glucuronide sodium (APAP-Gluc), phenacetin, dodecanoic acid (lauric acid), 12-hydroxydodecanoic acid (12OH-lauric acid), 4-nitrobenzyl chloride (NBC), S-(pnitrobenzyl)glutathione (NB-GSH), 4-pyridinecarboxaldehyde (PCA),
Espina et al. 4-pyridinecarboxaldehyde N-oxide (PCA-NO), (()-verapamil hydrochloride, (()-norverapamil hydrochloride, NADPH, and phosphate buffer (pH 7.4) were purchased from Sigma Chemical Co. (St. Louis, MO). Rat and human liver microsomes were obtained from Xenotech (Kansas City, KS). Bond-Elut C18 cartridges (10 g/60 cc and 500 mg/10 mL) were obtained from Varian Sample Preparation Products (Harbor City, CA). Analytical HPLC columns were obtained from various commercial suppliers. HPLC grade water, methanol, and acetonitrile were purchased from Mallinckrodt Chemicals (Phillipsburg, NJ). All deuterated solvents were purchased from Cambridge Isotope Laboratories, Inc. (Andover, MA). All solvents and chemical standards were of the highest grade commercially available with purities greater than 99%. The purity of the commercially obtained standards was confirmed by HPLC, LC/MS, and NMR analysis prior to their use in the studies. NMR Instrument. NMR data were recorded at 30 °C on a Bruker Avance III 600 spectrometer operating at the basic frequency of 600.13 MHz. All spectra were collected using a 5 mm CPTCI CryoProbe (Bruker BioSpin Corporation, Billerica, MA). The spectrometer was controlled using TopSpin (Bruker BioSpin, v 2.0 pl 5). Each sample contained the same volume (0.5 mL) of the solvent and was placed in the magnet at the same depth. All data collection was performed without spinning the samples. The probe was tuned and matched to the specific frequency using the automatic tuning method (ATMA) from Bruker BioSpin. The samples were field frequency locked using the 2H signal of the deuterated solvent. Shimming was performed on each sample prior to data acquisition using the TopShim automatic shimming method from Bruker BioSpin on the solvent peak. Each data set was collected under automation with the same parameter set using the ICON-NMR software (Bruker BioSpin, v 3.7) and manual sample changes. All spectra were referenced to the solvent signal. NMR Quantitative Analysis. Each proton spectrum was acquired with 65536 data points over a 7212 Hz spectral window using a 30° (2.7 µs) pulse and standard pulse sequence. Experiments were also performed with one set of samples using a 90° (8.0 µs) pulse to determine if there was any difference related to the pulse width. Eight scans with an acquisition time of 4.5 s were collected for each spectrum. Using the inversed recovery method, the relaxation time constant (T1) values were measured for all of the protons used in the quantitation analysis. The longer T1 value (∼12 s) was found for dimethyl sulfoxide (DMSO); therefore, a preacquisition delay of 60 s was used during the data acquisition. The receiver gain was set automatically using the more concentrated solution and was kept constant throughout each of the data sets. Data were Fourier transformed with a 1 Hz line broadening window function. A polynomial function was applied during automatic baseline correction prior to integration. Integration of the first spectrum (highest concentration) of the series was performed manually to define the regions to be integrated. Subsequently, the serial integration routine macro in TopSpin was used to integrate the spectra in the series. The signal of the protonated portion of the deuterated solvent was arbitrarily set to 100 and used as an internal reference to normalize the integral values within each spectrum. Because this protonated portion of the deuterated solvents changes for different solvents, it was important to use the same batch from the same vendor in all of the experiments within a series. The measured integral values were transferred to Microsoft Excel and normalized per proton for each different concentration. An average of the normalized integral value for each sample was plotted against the concentration value to create a calibration curve. Similarly, studies were conducted whereby only one proton signal (aromatic or aliphatic) was chosen from the spectrum of an analyte (APAP) and integral values obtained. The results from this approach were similar to what was achieved by averaging integral values from all protons. Hence, for consistency, unless otherwise stated, the results obtained only via the averaging method are reported for these analytes. Method Validation. Studies were conducted with compounds, which were previously shown to produce metabolites through various common metabolic reactions (hydroxylation, N-oxidation,
NMR for Quantitating Biologically Produced Metabolites
Chem. Res. Toxicol., Vol. 22, No. 2, 2009 301
Figure 1. Structures of compounds and their metabolites quantitated by NMR.
Table 1. Theoretical Concentrations of Metabolites Obtained if 50 µg of Each Metabolite Was Isolated from Biological Source and Dissolved in 500 µL of Solvent compound
metabolite
phenacetin APAP verapamil lauric acid PCA NBC theoretical compound
APAP APAP-Gluc norverapamil 12OH-lauric acid PCA N-oxide NBC-GSH
metabolite MW
concentration (mM)a
151 327 441 216 123 442 1000
0.66 0.31 0.23 0.46 0.81 0.23 0.10
a Fifty micrograms of metabolites was dissolved in 500 µL to give the theoretical concentrations. These metabolites were subsequently quantitated from NMR calibration curves (0.1-5 mM) generated from parent compounds.
N-demethylation, O-deethylation, O-glucuronidation, and glutathione conjugation) resulting in diverse metabolite structures ranging in molecular weights from 123 to 442 amu (see Figure 1 and Table 1). Quantitation of metabolites by proton nuclear magnetic resonance (1H NMR) (see above) was validated by a series of studies designed to investigate linearity, limit of quantitation (LOQ), and intra- and interday precision and accuracy. The acceptance criteria for accuracy and precision values were the same as those defined in the FDA guidance on bioanalytical method validation (49). Calibration curves were created using respective authentic standards of compounds whose metabolites were being
evaluated. Concentrations of metabolites of these compounds were subsequently calculated using these calibration curves. As part of method validation, studies were also conducted whereby the concentrations of metabolites were obtained using a calibration curve generated from a structural analogue. These concentrations, as determined by 1H NMR, were then compared against the expected nominal concentrations (based on the gravimetric method) for these metabolites. Preparation of Standards and Samples. Solutions of compounds (listed in Table 1) were prepared by weighing a few milligrams of each substance ((µg precision) and adding an appropriate volume of deuterated solvent to obtain 10 mM stock solution. If necessary, the samples were sonicated to obtain a clear solution. Serial dilutions were performed to obtain concentrations ranging from 0.01 to 10 mM. Replicate solutions of every sample were prepared for method validation and routine sample analysis. Each sample was dissolved in deuterated solvent, and an aliquot (0.5 mL) was transferred to a 5 mm NMR tube. Studies with APAP, NBC, phenacetin, PCA, verapamil, and their respective metabolites were performed using DMSO-d6 (D, 99.96%). Acetonitrile-d3 (D, 99.96%) was used in studies conducted with lauric acid and its metabolite. Compound X (proprietary) and its metabolites were dissolved in methanol-d4 (D, 99.8%). Linearity. Standard curves were prepared in the 0.1-5 mM (0.1-10 mM for APAP) range for all compounds whose metabolites were being quantitated. This range was chosen as it is expected to cover situations where one could have low amounts of isolated
302
Chem. Res. Toxicol., Vol. 22, No. 2, 2009
metabolites (approximately 50 µg that approaches NMR LOQ with acquisitions obtained in less than 30 min per sample) to high levels of metabolites (1-2 mg that approaches gravimetric LOQ whereby a sufficient amount of material can be weighed accurately on a sensitive analytical balance). Each sample was analyzed in duplicate for ∼12 min. The linear response was determined by obtaining the relative integral values of proton signals (average of all signals or a chosen single signal) plotted against the concentration of the compound. Intra- and Interday Precision and Accuracy. Studies were conducted to evaluate the intra- and interday precision and accuracy of the quantitative 1H NMR method. Metabolites were weighed accurately, and three sets of concentrations (low, medium, and high) were prepared as quality control (QC) samples. Subsequently, 1H NMR was used to obtain concentrations of these QC (metabolite) samples using the calibration curves prepared with the parent compounds. Hence, during these studies, the accuracy of the NMR generated data was evaluated by making comparisons against the gravimetrically (weighed) determined concentrations of the metabolites. Intraday precision and accuracy were carried out by analyzing a set of metabolite solutions (n ) 3) at each of the three concentrations. The interday precision and accuracy study was performed by analyzing these QC samples in triplicate at each concentration on three different days. LOQ for the Study. The LOQ was defined as the smallest quantity of the analyte that could be determined with acceptable accuracy and precision under a defined set of experimental conditions. At the LOQ, an acceptable precision was when a value of 30 min) may be necessary to obtain estimates of the metabolite concentrations. Intra- and Interday Accuracy and Precision. The intraday assay precision and accuracy results are summarized in Table 3. The % RSD values, which are a measure of the precision of the assay, were less than 5% for all metabolites at all three concentrations studied. The percent theoretical concentrations, which represent accuracy of the method, were within (15% for all metabolites measured at three different concentrations. These values are within the limits of acceptance for standard bioanalytical assays (49). As mentioned above, the integration of 1H NMR signals was generally obtained by averaging peak areas across the spectrum. However, in this study, we also showed that choosing one particular proton signal (either aromatic or aliphatic) and obtaining integral values of that proton over a range of concentration could also produce reliable quantitative values. Table 3 shows that comparable results were obtained for the quantitation of APAP glucuronide by using the two methods. The interday precision and accuracy data are listed in Table 4. Metabolites of all of the test compounds, quantitated on three different days, gave % RSD values ranging from 0.2 to 2.5% at all three concentrations. The accuracy of the method as determined by the percent difference from the theoretical gravimetrically determined concentrations ranged from 0.2-9.8% for all compounds at all three levels. LOQ for the Study. The LOQ was determined to be 0.1 mM (final concentration) for all compounds. This took into consideration the amount of time that could be practically spent on acquiring NMR data. Obviously, one could obtain a better signal-to-noise ratio by conducting lengthy experiments (creating large number of scans during one single analysis) using this technique and extend the LOQ well below the 0.1 mM concentration (see above). Furthermore, the signal-to-noise ratio
304
Chem. Res. Toxicol., Vol. 22, No. 2, 2009
Espina et al.
Figure 2. Calibration curve prepared for APAP with an extended LOQ of 0.01 mM.
Table 3. Intraday Assay Precision and Accuracy for the Analyses of Metabolites by Proton NMRa compound APAP-Gluc PCA-NO 12OH-lauric acid norverapamil NB-GSH
nominal concentration(mM) 0.19 1.92 7.47 0.20 2.00 4.00 0.20 2.00 4.00 0.20 2.00 4.00 0.20 2.00 4.00
mean observed concentration (mM) 0.18 1.79 7.00
0.19b 1.83b 7.10b 0.21 2.20 4.36 0.18 2.02 4.02 0.20 2.06 4.11 0.20 2.03 4.05
0.18c 1.80c 7.03c
% RSD 1.22 0.56 0.29
1.98b 0.61b 0.24b 1.10 0.02 0.05 0.18 0.23 0.15 0.82 0.53 0.07 0.42 0.02 0.21
% difference from nominal value 0.24c 0.60c 0.28c
7.07 6.98 6.26
3.34b 4.59b 4.94b 3.74 10.03 9.01 10.35 1.11 0.48 2.22 2.90 2.74 1.31 1.33 1.27
6.15c 6.33c 5.86c
a The concentrations of metabolites were obtained from the calibration curves constructed with parent compounds and compared with gravimetrically determined values. Calibration points and metabolite samples were analyzed in duplicates and triplicates, respectively. b Using a single aliphatic signal (1.97 ppm). c Using a single aromatic signal (7.33 ppm).
can also be improved by using 90° pulses instead of 30° pulses that were used in this study. In cases where only a small amount of material (
