A Sensitive and Specific Stable Isotope Assay for Warfarin and its Metabolites
E.D.Bush,L.K.Low and W.F.Trager
Department of Medicinal Chemistry, BG-20, University of Washington, Seattle, Washington 98195, USA
A capillary gas chromatographic mass spectrometric method for the quantification of warfarin and its known metabolites from microsomal incubations is described. Deuterium labelled 4′, 6-, 7- and 8-hydroxy warfarins are used as internal standards and the method has detection lmits of 1 ng ml’ with 20 ng ml-being the lower limit for accurate quantification.
INTRODUCTION
The oral anticoagulant warfarin (4-hydroxy-3-(1-phenyl-3-oxobutyl)-2H-1-benzopyran-2-one),Fig.1,

Warfarin 8
x R R HH R5
H
4′-OH H H H
6-OH H OH H H H
7-OH
8-OH H
H H
H H H
OH H
H

alcohol 1 RS; SR
alcohol 2 RR; SS
Figure 1. The structure of warfarin and its known metabolites.
has found extensive clinical use in the treatment of such pathological conditions as thrombophlebitis,pulmonary emboli, and myocardial infarction.’It has also been widely used as a rodenticide to help control rat popula-tions’ and more recently has found use as a probe to investigate the multiplicity and catalytic activity of microsomal and purified cytochrome P-450 prepar-ations.Because of its clinical and pharmacological importance, considerable effort has been expended toward the development of analytical methods to quan-tify both warfarin and its metabolites in various biologi-cal matrices. A partial list of these published methods includes spectrophotometric, fluorometric,°thin-layer chromatographic,10 gas-liquid chromatographic,” gas chromatographic mass spectrometric12 and high perfor-

mance liquid chromatographic13 approaches.In our hands each of these methods suffers from one or more serious disadvantages. Some of these disadvantages are: (1) low or variable recoveries of warfarin;(2) inability to quantify all warfarin metabolites;(3) poor sensitivity; (4)poor reproducibility;(5) quantification problems when the dynamic range between warfarin and its meta-bolites exceeds 100; and (6) inability to differentiate between various stable isotopically labeled analogs. Therefore,a new methodology was sought which would circumvent these problems. In this paper we report the development of a capillary gas chromatographic mass spectrometric method which permits quantification of warfarin and its known metabolites from microsomal incubations at the low nanogram level.
EXPERIMENTAL
Materials and methods
Racemic warfarin was obtained from Sigma Bio-chemicals and was resolved as previously described.’ Unlabeled phenolic warfarin metabolites (4′-,6-,7-and 8-hydroxy warfarins) were synthesized according to the method of Hermodson et al.15 The synthesis of 2′,3′,4′,5′,6′-pentadeuterio-6-,7- and 8-hydroxywar-farins and 5,6,7,8-tetradeuterio-4-hydroxywarfarin will be presented elsewhere.Diazomethane was generated from Diazald(Aldrich Chemical Co.)according to the directions supplied by the manufacturer. N,O-bis-(Trimethylsilyl)-trifluoroacetamide (BSTFA) was pur-chased from Pierce Chemical Co.Concentration tubes and caps were purchased from Laboratory Research Co.,Los Angeles,California. pH Measurements were performed on an Orion Research model 701A digital pH meter. Microsomes and microsomal stock solutions were prepared as detailed previously. A standard sol-ution containing all the warfarin phenolic metabolites was prepared by dissolving 150 μg of each metabolite in 2ml of 40% 0.1 N KOH (80 ml 0.1 N KOH with120 ml 0.05 M phosphate buffer,pH 7.4) followed by addition of 6 ml of the 0.05 M phosphate buffer and then 2 ml of 40% 0.1 N HCI (prepared in a manner analogous to the KOH). A standard solution of the deuterated internal standards was similarly prepared
© Wiley Heyden Ltd,1983

CCC-0306-042X/83/0010-0395$02.00
BIOMEDICAL MASS SPECTROMETRY, VOL. 10, NO. 7, 1983 395 
E.D. BUSH, L.K. LOW AND W. F. TRAGER
using 150 ug of each of the deuterated phenolic meta-bolites.These solutions were kept frozen at-20Cuntil analysis. Low speed centrifugations were conducted on an IEC HN-SII benchtop centrifuge from Davon/IEC Division. The evaporator used wasa Multivap (model No. 113) from Organomation. Capillary gas chromatography was performed on either wide or nar-row bore DB-5 bonded phase (normal film thickness) fused silica columns (30 m) from J & W Scientific Co. Capillary gas chromatographic mass spectrometric measurements were conducted on either a Hewlett-Packard 5985 mass spectrometer or on a VG micromass 7070H mass spectrometer interfaced to a VG 2000 data system and fitted to a Hewlett-Packard 5700 GC.
Extraction and derivatization scheme for warfarin and metabolites
The extraction scheme is based on 1 ml of a microsomal incubation mixture.Upon completion of the incubations the reactions were terminated by addition of acetone (0.60 ml)followed by the addition of the internal stan-dard solution (0.02 ml of the deuterated standard sol-ution described above). The mixture was then vortexed for 30s, capped and stored at -20℃ until analysis. After thawing, the solutions were centrifuged,the supernatant removed and the protein precipitate discar-ded.To these supernatants was added 0.5 M NaH2PO4 (0.20 ml).The resulting slightly acidic solutions were extracted with cyclohexane (3x2ml) followed by Et2O/EtOAc, 1:1(2x2ml).The cyclohexane layer was removed by an aspirator with a foreline trap and discarded. The Et2O/EtOAc extracts were combined in a concentration tube and solvent removed in the Multivap evaporator under a stream of N2. To the dry extracts was added ethereal diazomethane (approxi-mately 1 ml) and the tubes capped.After 15 min the caps of the tubes were removed and methanol (0.1 ml) added to increase the rate of methylation of the phenolic groups.Typically the tubes,after being recapped,were allowed to stand at room temperature for 24 h, then the ether and excess diazomethane removed under a stream of nitrogen.The samples were analyzed by gas chromatography mass spectrometry(GC/MS) generally within a few days.During the course of these experi-ments it was found that the samples need not be analyzed immediately,since they would remain stable for several weeks if kept at-20℃.
Gas chromatographic mass spectrometric assay for warfarin and metabolites
The gas chromatographic conditions for the assay on either GC/MS instrument were:(1) helium carrier gas head pressure, 14 psi wide bore, 20 psi narrow bore; (2) splitless injection, injector temperature 250℃;(3) oven temperature program 1-2min isothermal at 160℃,then the temperature increased at a rate of 30°℃min-’ to 250°C and held there for 8-12 min; (4) columns run directly into mass spectrometric source; (5) direct inlet transfer line held at 240-250°C. The mass spectrometer conditions were:(1)source tem-
396 BIOMEDICAL MASS SPECTROMETRY, VOL.10,NO.7,1983

perature 200°℃; (2) 70 eV EI, 200 μA VG and 300μA HP emission current; (3) SIM or SIR mode, multiplier at 2400-2800 V.
Just prior to gas chromatographic mass spectrometric analysis,BSTFA (50-100 μl) was added to the dry extracts in the concentration tubes and allowed to stand for 1 h.Between 1 and 3 μl of this solution was injected onto the gas chromatograph mass spectrometer for analysis. Standard curves were constructed by adding 300ng of each of the deuterated internal standards (20 μl of the standard solution)and 3,15,30,150,300, 1500 and 3000 ng of each of the unlabeled metabolites (0.2-200 μl of the unlabeled standard solution) to 1 ml of the microsomal preparation (minus cofactors). The results when using the HP 5895 gas chromatograph mass spectrometer were based on selected ion mass chromatogram peak areas. Generally the base peaks ([M-CH3CO]+) at 309 for unlabeled metabolites and 313 (4′-OH) or 314 (6,7 or 8-OH) for the labeled internal standards were monitored. Results from the VG 7070H gas chromatograph mass spectrometer were based on an average of peak height and peak area for the base peak. The combination of peak height and area for data from the VG 7070 gave more consistent and precise results. This did not appear to be necessary for the data obtained from the HP 5895,since more reproducible gas chromatogram peak profiles were gen-erally obtained from this instrument. To calculate unknown concentrations of sample compound the ratio of areas (and/or height) of sample compound to internal standard was compared to previously run calibration curves. Equation (1) was used for these calculations.
(1)
where:M=ion intensity of the base peak of unlabeled sample.M+5=ion intensity of the base peak of the 2Hs standard. A and B=known concentrations of unlabeled and labeled materials.The linear regression coefficients (m and b) were calculated from the standard curve measurements by nonweighted linear regression analysis.
RESULTS AND DISCUSSIONS
Of the assays outlined in the introduction, only two have the capability of separating and quantifying each of the warfarin metabolites shown in Fig. 1. The two assays are the 4CTLC assay developed by Pohl et al.1 and the reverse phase HPLC assay developed by Fasco et al.13 However,both assays as published suffer from two main weaknesses. The first, at least in our hands,is a problem with reproducibility; and the second is difficulty in adapting either method to include mass spectrometry. The latter problem originates from the fact that the extraction of nanogram levels of compound from TLC plates or from collected HPLC fractions gives a low yield and is a tedious process.
Recent reports in the literature suggested that fused silica capillary columns, in addition to providing increased resolution, could also be advantageous in analyzing compounds previously found to be intractable by traditional gas chromatographic analysis. Thus, capil-lary GC/MS appeared to be an approach that warranted 
ISOTOPE ASSAY FOR WARFARIN AND METABOLITES

Figure 2. Ion chromatogram obtained on a Hewlett-Packard gas chromatograph mass spectrometer of warfarin and warfarin metabolites, utilizing BSTFA as an injection solvent and derivatizing agent for the alcohols. The chromatograms are (from top to bottom), derivatized alcohols (396), phenolic meta-bolites (309),warfarin (279) and total ion current.
serious investigation provided that an apparent dynamic range problem could be solved.This problem results from the fact that warfarin is commonly present in a large excess (100-1000 fold in comparison with its metabolites in rat liver microsomal incubations) and fused silica capillary columns are easily overloaded. These columns,therefore,cannot tolerate such wide differences in relative concentrations for quantitative measurements. Thus, an extraction scheme was developed which preferentially extracted warfarin(up to 85% removal) from the microsomal mixture but left the metabolites to be recovered during a subsequent extraction (greater than 95% recovery of the meta-bolites in the second extraction). This approach dramatically reduced the concentration ratio between warfarin and metabolites. The extraction scheme has the added advantage of removing lipophilic con-taminants from the microsomal mixture which have a tendency to interfere with gas chromatographic mass spectrometric quantification.
Development of the extraction scheme was based on the observation that warfarin was preferentially extrac-ted from microsomal incubations by cyclohexane. Several experiments were conducted to optimize this result.The most successful treatment, and thus the one adopted for our work, was to acidify the solutions with 0.5 M NaH2PO4 to pH 5.8 and then extract with cyclo-hexane to remove warfarin and lipophilic impurities. The residual aqueous phase was further extracted with

Weight ratio
Figure 3. Standard curves for the four warfarin phenolic meta-bolites. Weight ratio=ratio of gram weights of the metabolite standards. Unlabeled standard weights varied between 1.5 and 1500 ng ml,the labeled standard was fixed at 150 ng ml-.
Et2O:EtOAc(1:1)to recover metabolites. The extrac-tion scheme is reproducible and efficient and is of utility for almost any warfarin assay. For instance, it has been used to greatly improve the reproducibility of the 14C warfarin metabolite TLC assay of Pohl et al.10 since it effectively removes the high background counts due to parent drug.”An added advantage is that the extraction scheme permits the ready recovery of unused (radio-labeled)substrates.
Gas chromatographic conditions were developed which yield baseline resolution of warfarin,the four phenolic metabolites and the two diastereomeric alcohols (Fig. 2). Briefly,the assay consists of extracting the samples from biological media (as described above), adding ethereal diazomethane and allowing the sol-utions to stand overnight. After removal of the excess diazomethane,the samples are dissolved in BSTFA. Since diazomethane does not methylate the alcoholic hydroxyl groups of the warfarin alcohols, BSTFA serves to silylate these groups and thereby stabilize them for gas chromatographic analysis. In addition, the BSTFA appears to stabilize the samples in general so that they do not decompose over a period of weeks if stored in the freezer.
Table 1.Slope,intercept and correlation coefficients calculated by linear regression analysis of standard curve data
Table 1.Slope,intercept and correlation coefficients calculated by linear regression analysis of standard curve data
4′-OH 6-OH 7-OH 8-OH
(a)Intraday nyn=8
Slope±SD
lope±SD 0.7806±.005 0.8513±.006 0.891·.002 0.8382±.006
0.1547±.16 0.1057±.14 0.2195±.19 0.1323±0.19
0.988 0.991 0.990 0.989
(b) Interdayrday n=3
Slope±SD 0.7903±.031 0.8422±.034 0.8310±.012 0.8479±.040
y-intercept±SD 0.1851±.210 0.1153±.225 0.2205±.36 0.1394±.35
(b) Interdayrday n=3
BIOMEDICAL MASS SPECTROMETRY, VOL. 10, NO.7,1983 397 
E. D. BUSH, L. K. LOW AND W. F. TRAGER
The discussion to follow will focus on the methodology developed to quantify the phenolic meta-bolites from microsomal incubations. Since the alcohols are not significant microsomal metabolites and since an internal standard for the benzylic hydroxywarfarin was not available,quantification of these materials was not pursued further.To quantify the phenolic metabolites the pentadeuterio or tetradeuterio labeled internal stan-dards were utilized. With these materials, standard cur-ves were determined which covered a range of ratios (metabolite/internal standard) of 0.1-10. The range of sample compound measured in the standard curves was 3-3000 ng ml-’ of microsomal incubate.Figure 3 shows the results of a standard curve run and Table 1 gives the calculated linear regression parameters for each metabolite.Table 1 also lists both the intra-and interday variation of the calibration curve parameters. In gen-eral,variation of the slope was less than 2% intraday and less than 5% interday. As can be seen, the points at the low end of the curve (Fig. 3) begin to deviate from linearity. This deviation, while large in relative amount (in error by as much as 100%), is actually small in absolute amount (in the order of 2-3ng).This phenomenon generally indicates that the limits of quantification for the assay have been reached. There-fore,microsomal incubations were kept sufficiently large to insure that metabolite levels were on the linear portion of the curve.
The assay has been in use in our laboratory for approximately one year. It has proved to be reliable, sensitive,and relatively easy to run.The assay involving GC/MS has a cycle time of less than 20 min and the complete sample work-up time is between 25 and 30 min. Levels as low as 1 ng ml ‘can be detected and levels at 20 ng ml ‘accurately quantified.In order to compare the reliability of the gas chromatographic mass spectrometric assay with that of the previously used 1C TLC assay a series of microsomal incubations were conducted and analyzed by both procedures. Figure 4 illustrates the results of this investigation as the measured rate of metabolite production from incuba-

Figure 4.Comparison of results for the quantification of the phenolic warfarin metabolites, utilizing the 14CTLC assay and the GC/MS assay; C=microsomes from non-induced rats, PB=microsomes from phenobarbital-induced rats and BNF= microsomes from β-naphthaflavone-induced rats.
tions with microsomes from non-induced, phenobar-bital-induced and β-naphthoflavone-induced rat livers. Clearly,the relative ratios of one metabolite to another are approximately the same between the two assays. However,with the use of deuterated internal standards, the gas chromatographic mass spectrometric assay, which corrects for recovery losses,consistently yielded greater absolute amounts of metabolites.
Acknowledgments
We appreciate helpful discussions and the technical assistance of Dr James Callis (Department of Chemistry,University of Washington) and Mr William Howald (this department).
This research was supported in part by Research Grants GM22860 and GM25136 and Biomedical Research Development Grant 1508 RRO9082.
REFERENCES
1.R. A. O’Reilly and P.M. Aggeler, Pharmacol.Rev. 22, 35 (1970).
2.W.B. Jackson, P.J. Spear and C. G. Wright, Pest Control 39,13(1971).
3.W.T.Porter, C.Wheeler and W. G. Trager, Biochem. Phar-macol.30,3099(1981).
4. M. J. Fasco, K. P. Vatsis, L. S. Kaminsky and M.J.Coon. J.Biol.Chem.253,7813(1978).
5.L.S. Kaminsky, M. J. Fasco and F. P. Guengerich,J.Biol. Chem.254,9657 (1979).
6.M. J.Fasco,L.J.Piper and L. S. Kaminsky, Biochem. Phar-macol.28,97(1978).
7.L.R.Pohl,S.D.Nelson,W.R.Porter,W. F. Trager, M. J. Fasco,F.D.Baker and J.W.Fenton Il,Biochem.Pharmacol. 25,2153(1976).
8. R. A. O’Reilly, P. M. Aggeler, M. S. Hoag and L. L. Leong, Thromb. Diath. Heamorrh. 8, 82(1962).
9.M.Corn and R. Berberich, Clin.Med.13, 126 (1967).
398 BIOMEDICAL MASS SPECTROMETRY, VOL 10, NO.7,1983

10. L. R. Pohl, R. Bales and W.F.Trager,Res.Commun.Pathol. Pharmacol.15,233(1976).
11. K.K.Midha,I.J.McGilveray and J.K. Cooper, J. Pharmacol. Sci. 63, 1725 (1974).
12. A.M. Duffield, P. H. Duffield, D. J. Birkett,M.Kennedy and D.N.Wade,Biomed. Mass Spectrom. 6, 209 (1979).
13.M. J.Fasco, M. J. Cashin and L. S. Kaminsky, J. Liq. Chromatogr. 2,565 (1979).
14.B.D.West,S.Preis, C. H. Schroeder and K. P. Link, J. Am. Chem.Soc.83,2676(1961).
15. M.A. Hermodson, W. M. Barker and K.P.Link,J.Med.Chem. 14,167(1971).
16. E.D.Bush and W.F. Trager, Biomed.Mass Spectrom.8,211 (1981).
17.R. Remmel and G. Elmer, personal communication.BMS-794833