Identification and Quantification of Cardiac Glycosides in Blood and

Journal of Chemical Education · Journal of Chemical Information and Modeling .... An LC–MS/MS method for the determination of digitoxigenin in skin ...
0 downloads 0 Views 120KB Size
Anal. Chem. 1999, 71, 4034-4043

Identification and Quantification of Cardiac Glycosides in Blood and Urine Samples by HPLC/ MS/MS Fuyu Guan,†,‡ Akira Ishii,† Hiroshi Seno,† Kanako Watanabe-Suzuki,† Takeshi Kumazawa,§ and Osamu Suzuki*,†

Department of Legal Medicine, Hamamatsu University School of Medicine, 3600 Handa-cho, Hamamatsu 431-3192, Japan, and Department of Legal Medicine, Showa University School of Medicine, 1-5-8 Hatanodai, Shinagawa-ku, Tokyo 142-8555, Japan

Cardiac glycosides (CG) are of forensic importance because of their toxicity and the fact that very limited methods are available for identification of CG in biological samples. In this study, we have developed an identification and quantification method for digoxin, digitoxin, deslanoside, digoxigenin, and digitoxigenin by highperformance liquid chromatography tandem mass spectrometry (HPLC/MS/MS). CG formed abundant [M + NH4]+ ions and much less abundant [M + H]+ ions as observed with electrospray ionization (ESI) source and ammonium formate buffer. Under mild conditions for collision-induced dissociation (CID), each [M + NH4]+ ion fragmented to produce a dominant daughter ion, which was essential to the sensitive method of selected reaction monitoring (SRM) quantification of CG achieved in this study. SRM was compared with selected ion monitoring (SIM) regarding the effects of sample matrixes on the methodogoly. SRM produced lower detection limits with biological samples than SIM, while both methods produced equal detection limits with CG standards. On the basis of the HPLC/MS/MS results for CG, we have proposed some generalized points for conducting sensitive SRM measurements, in view of the property of analytes as well as instrumental conditions such as the type of HPLC/MS interface and CID parameters. Analytes of which the molecular ion can produce one abundant daughter ion with high yield under CID conditions may be sensitively measured by SRM. ESI is the most soft ionization source developed so far and can afford formation of the fragile molecular ions that are necessary for sensitive SRM detection. Mild CID conditions such as low collision energy and low pressure of collision gas favor production of an abundant daughter ion that is essential to sensitive SRM detection. This knowledge may provide some guidelines for conducting sensitive SRM measurements of very low concentrations of drugs or toxicants in biological samples. Cardiac glycosides (CG) have been used for many years in the management of congestive heart failure and other cardiac †

Hamamatsu University School of Medicine.

4034 Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

diseases.1 Digoxin and digitoxin are the most commonly prescribed CG but the less clinically useful CG include R-acetyldigitoxin, deslanoside, β-methyldigoxin, ouabain, proscillaridin, and strophatin-K.2,3 These compounds are among the cardiac drugs with very narrow toxic/therapeutic margins.3,4 Overdose, suicidal, or accidental ingestion of CG may cause severe toxicity.5-8 Identification of these drugs in biological samples is of significance in forensic toxicology. To date, only limited techniques are available for the detection of CG in biological samples. Radioimmunoassay (RIA) is so far the most sensitive method for the detection of CG. However, RIA has been reported to be nonspecific and subject to cross-reactivity with active and inactive metabolites of CG,9-11 as well as with endogenous, digoxin-like immunoreactive substances in the biofluids of pregnant women, newborn infants, or patients with renal, diabetic, or hepatic disease conditions.12-17 High-performance ‡ On leave from Beijing Institute of Pharmacology and Toxicology, Beijing 100850, China. Present address: PA Equine Toxicology and Research Laboratory, Department of Chemistry, West Chester University, 220 East Rosedale Ave., West Chester, PA 19382 § Showa University School of Medicine. (1) Reynolds, J. E. F., Ed. Martindale-the Extra Pharmacopoeia, 29th ed.; The Pharmaceutical Press: London, 1989; p 822. (2) Ellenhorn, M. J., Barceloux, D. G., Ed. Medical Toxicology-Diagnosis and Treatment of Human Poisoning; Elsevier: Amsterdam, 1988; p 200. (3) Scherrmann, J. M.; Bourdon, R. Lett. Pharmacol. Clin. 1988, 5, 1. (4) Godfraind, T. In Pharmacologie-Des Concepts Foundamentaux aux Applications Therapeutiques; Schorderet, M., Ed.; Frisons-Roche/Slatkine: Paris, 1989; p 167. (5) Lacer, S.; Scholz, H.; Buschmann, I.; Thoenes, M.; Meinertz, T. Eur. J. Clin. Pharmacol. 1998, 54 (1), 95. (6) Caspi, O.; Zylber-Katz, E.; Gotsman, O.; Wolf, D. G.; Caraco, Y. Ther. Drug Monit. 1997, 19 (5), 510. (7) Grellner, W.; Kaferstein, H.; Sticht, G. Forensic. Sci. Int. 1997, 89 (3), 211. (8) Laberge, P.; Martineau, P. Ann. Pharmacother. 1997, 31 (9), 999. (9) Baselt, R. C.; Cravey, R. H. Disposition of Toxic Drugs and Chemicals in Man; 4th ed.; Chemical Toxicology Institute, Foster City, CA, 1995; p 802. (10) Stone, J. A.; Soldin, S. J. Clin. Chem. 1989, 35, 1326. (11) Stoll, R. G.; Christensen, M. S.; Sakmar, E.; Wagner, J. G. Commun. Chem. Pathol. Pharmacol. 1972, 4, 503. (12) Valdes, R.; Graves, S. W.; Brown, B. A. Landt, M. J. Pediatrics, 1983, 102, 947. (13) Graves, S. W.; Brown, B. A.; Valdes, R. Ann. Intern. Med. 1983, 99, 604. (14) Spiehler, V. R.; fischer, W. R.; Richards, R. G. J. Forensic Sci. 1985, 30, 86. (15) Nanji, A. A.; Greenway, D. L. Br. Med. J. 1985, 290, 432. (16) Stone, J.; Bentur, Y.; Zalstein, E.; Soldin, S.; Giesbrecht, E.; Koren, G. J. Pediatr. 1990, 117, 321. (17) Tzou. M. C.; Reuning R. H.; Sams, R. A. Clin. Pharmacol., Ther. 1997, 61 (4), 429.

10.1021/ac990268c CCC: $18.00

© 1999 American Chemical Society Published on Web 08/17/1999

measurements of very low concentration of analytes in biological samples without pitfalls.

Figure 1. Chemical structures of digoxin and other cardiac glycosides.

liquid chromatography (HPLC) methods are useful in the resolution of CG followed by RIA or fluorescence detection. HPLC methods avoid interference from the metabolites and endogenous substances.18-21 These methods are useful for monitoring plasma concentrations of CG in clinical laboratories, although the methods could not be used to provide unequivocal identification of CG. Although gas chromatography coupled with mass spectrometry (GC/MS) is the accepted method for confirmation of the presence of many drugs and toxicants, CG are not amenable to this technique, due to their nonvolatile and thermolabile characteristics. An alternative technique to GC/MS is HPLC/MS, which is capable of analyzing nonvolatile and thermolabile compounds and providing unequivocal identification.22-24 LC/LC/MS/MS detection of oleandrin, a cardiac glycoside not included in the present study, has been reported.25 A method for the quantification of acetyldigoxin, digoxin, digitoxin, lanatoside C, oleandrin, and ouabain in plasma by HPLC/MS has been published.26 In the present study, we have developed a sensitive method for the identification and quantification of digoxin, digitoxin, deslanoside, digoxigenin, and digitoxigenin (Figure 1) in whole blood and urine by HPLC/MS/MS. In addition, some suggestions are proposed for conducting sensitive selected reaction monitoring (SRM) (18) Plum, J.; Daldrup, T. J. Chromatogr. 1986, 377, 221. (19) Kwong, E.; McErlane K. M. J. Chromatogr. 1986, 381, 357. (20) Oosterkamp, AJ.; Rrth, H.; Beth, M.; Unger, K. K.; Tjaden, U. R.; van de Greef, J. J. Chromatogr., B: Biomed. Appl. 1994, 653, 55. (21) Tzou, M. C.; Sams R. A.; Reuning R. H. J. Pharm.,. Anal. 1995, 13 (12), 1531. (22) Gelpi, E. J. Chromatogr., A 1995, 703, 59. (23) Henion, J.; Wachs, T.; Mordehai, A. J. Pharm., Biomed. Anal. 1993, 11 (11-12), 1049. (24) Huang, E. C.; Wachs, T.; Conboy, J. J.; Henion, J. D. Anal. Chem. 1990, 62, 713A. (25) Rule, G.; Mclaughlin L. G.; Henion, J. Anal. Chem. 1993, 65, 857A. (26) Tracqui, A.; Kintz, P.; Ludes, B.; Mangin, P. J. Chromatogr., B: Biomed. Appl. 1997, 692, 101.

EXPERIMENTAL SECTION Chemicals. Digoxin was purchased from Aldrich (Wilwaukee, WI), digoxigenin hydrate and digitoxin were from Wako Pure Chemical Industries (Osaka, Japan), digitoxigenin was from Tokyo Chemical Industries (Tokyo, Japan), and proscillaridin was from Sigma (St. Louis, MO). Methyldigoxin was a kind donation from Yamanouchi Pharmaceuticals (Tokyo, Japan), and deslanoside was a kind grant from Fujisawa Pharmaceuticals (Osaka, Japan). Each CG was dissolved in HPLC grade methanol (Wako Pure Chemical Industries, Osaka, Japan) to prepare a 1.0 mg/mL stock solution and stored at 4 °C for no longer than one month. Dilute working solutions were prepared daily. Acetonitrile (HPLC grade) was obtained from Wako Pure Chemical Industries. Distilled water was re-deionized by a Mili-Q water system (Millipore, Bedford, MA). Ammonium formate aqueous solution (0.20 M) was prepared from ammonium hydroxide and formic acid and adjusted to pH 3.4 with formic acid. Dilution of a 0.20 M stock solution gave a 2 mM HCOONH4 solution to be used as a component of the HPLC mobile phase. Other chemicals used were of analytical grade. Human blood and urine samples were obtained from healthy donors. The blood samples were stored at 4 °C in the presence of Na2EDTA, and urine samples were stored at -20 °C until analysis. Instruments. The HPLC/MS consisted of a SpectraSystem P 4000 HPLC pump (Thermo Separation Products, Fremont, CA) and a Finnigan Mat TSQ 7000 triple-stage quadrupole tandem mass spectrometer equipped with a Finnigan Mat “high flow” elctrospray ionization (ESI) or atmospheric pressure chemical ionization (APCI) interface (San Jose, CA). Data were processed by Finnigan Mat ICIS software on a Digital DECstation 3000 computer (Maynard, MA). A TSQ 7000 mass spectrometer was automaticallly tuned and calibrated with infusion of a mixture of apomyoglobin (5 nM/ mL) and 1-methionyl-arginyl-phenylalanyl-alanine acetate (MRFA) (20 nM/mL) in methanol/water (50:50, v/v) with 1% acetic acid added. The voltage of the tube lens, which focuses the spray leaving the heated capillary for the skimmer, and the voltage of lens L11 were optimized with loop injection of digoxin (10 ng/5 µL in acetonitrile/2 mM HCOONH4, 50:50, v/v) using ICIS software, after ESI interface parameters were optimized and HPLC flow rate was chosen. The dynode voltage of TSQ mass spectrometer was 15 kV, and electron multiplier voltage was set to 1400 V for total ion current (TIC) or selected ion monitoring (SIM) measurements or 1700 V for SRM measurements. ESI interface parameters were as follows: spray voltage, 4.5 kV (applied to the needle); sheath gas (N2), 70 psi (1 psi ) 6894.74 Pa); auxiliary gas (N2), 10 mL/min; and capillary heater temperature, 200 °C. TIC, SIM, and SRM Measurements. All mass spectrometric experiments were conducted in positive ion mode. Flow injection of 5-15 ng of each compound in 5 µL of acetonitrile/2 mM HCOONH4 (50:50, v/v) was used to introduce the sample into the mass spectrometer, except for on-line HPLC/MS experiments. Collision-induced dissociation (CID) of selected precursor ions in the rf-only quadrupole collision cell was achieved using argon as the collision gas and offset voltage as collision energy. SRM acquisition was conducted by selecting the desired precursor ions Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

4035

using the first quadrupole while monitoring specific daughter ions with the third quadrupole. Time-scheduled SRM conditions for HPLC/MS/MS experiments were as follows: HPLC time 0-8.5 min, m/z 408.3 f 391.2; 8.5-9.5 min, m/z 960.7 f 651.0; 9.510.5 min, m/z 798.6 f 651.0; 10.5-11.8 min, m/z 812.6 f 651.0; 11.8-12.6 min, m/z 392.2 f 375.2; 12.6-15.0 min, m/z 782.6 f 635.1; pressure of collision gas (argon), 1.5 mTorr (1 Torr ) 133.32 Pa); CID offset voltage, -15 V for all the reactions except m/z 960.7 f 651.0 (-20 V). Daughter ion m/z window was (0.3 amu; dwell time was 0.6 s for each selected reaction. SIM experiments were carried out by monitoring, with the first quadrupole, the precursor ions of the analytes in the scheduled time intervals as described above. HPLC Conditions. A Mightysil RP-18 column (Kanto Chemical, Tokyo, Japan), 150 × 2.0 mm i.d., packed with 5-µm reversedphase C18, protected by a 35 × 2.0 mm i.d. 5-µm C18 precolumn (Shiseido, Tokyo, Japan), was used for separation of CG. HPLC separations were carried out at ambient temperature, using a binary gradient composed of mobile phase A (20% acetonitrile/ 80% 2 mM HCOONH4 in water, v/v) and mobile phase B (80% acetonitrile/20% 2 mM HCOONH4 in water, v/v). The gradient expressed as changes in mobile phase B was as follows: 0-2 min, hold at 0% B; 2-12 min, a linear increase to 100% B; 12-15 min, hold at 100% B; and 15.1-20 min, switch to and hold at initial condition (0% B). The mobile phase flow rate was 0.20 mL/min. A Rheodyne injector (Rheodyne, Cotati, CA) with an injection loop of 5 µL was employed, and a 5 µL of sample or standard was injected. Sample Preparation and Extraction. To 1.0 mL of whole blood in a 15-mL disposable plastic centrifuge tube, was added an aliquot of CG mixture (in 2-30 µL of methanol) containing digoxin, digitoxin, deslanoside, digoxigenin, and digitoxigenin. And methyldigoxin (IS, in 10 µL of methanol) was added. The mixture was briefly vortexed and then 3.5 mL of distilled water and 0.50 mL of ammonium acetate buffer (2 M, pH 9.5) were added, followed by centrifuging at 2000 rpm for 5 min. The supernatant was collected by aspiration and subjected to solidphase extraction (SPE). Blood samples spiked with CG were extracted by SPE using Oasis HLB cartridges, 3 cm3/60 mg of sorbent (Waters, Milford, MA). SPE was manually performed with a VacElute SPS 24 batch extraction device (Varian, Harbor City, CA). The cartridges were sequentially conditioned with 1.0 mL each of methanol and distilled water and 3.0 mL of ammonium acetate buffer (0.1 M, pH 9.5). The sample supernatant was loaded onto the SPE cartridge and eluted by gravity. The cartridge was rinsed with 2 mL of ammonium acetate (0.1 M, pH 9.5) and dried under vacuum of 5 in. of Hg (1 in. of Hg ) 3386.39 Pa) for 2 min. The cartridge was eluted with 3 mL of chloroform/2-propanol (95:5, v/v) into a glass tube. The eluent was transferred to a clean 4-mL vial (Supelco, Bellefonte, PA) and evaporated to dryness on a heating block at 50 °C, under nitrogen. Dried extract was reconstituted in 30 µL of acetonitrile plus 70 µL of 2 mM HCOONH4 and briefly ultrasonicated. An aliquot of 5 µL was injected onto the HPLC/ MS. Urine samples (2 mL) spiked with CG were mixed with 0.2 mL of ammonium acetate buffer (2 M, pH 9.5) and extracted as described above. 4036 Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

Animal Experiments. Six male Sprague-Dawley rats weighing about 200 g were put under temporary anaesthesia by inhalation of diethyl ether and orally administered 20 µg of digoxin suspended in 0.5 mL of 2% methyl cellulose (25 cP, Sigma) aqueous solution. Three hours postdrug administration, blood (about 5 mL) samples were drawn from the abdominal aorta of the rats that were reanaesthesized with another inhalation of diethyl ether. The blood samples were stored in the presence of Na2EDTA at 4 °C until analysis. The control group (3 rats) was similarly treated except that no digoxin was administered. The control blood samples were used to prepare calibration curves for quantification of digoxin in administration samples. Rat blood samples were extracted as described above for human blood samples. RESULTS Mass Spectra of CG. The mass spectra of CG measured under different instrumental and solution chemistry conditions were compared in order to choose abundant quasi-molecular ions for further study. The ESI interface gave rise to mass spectra of CG quite different from those obtained with the APCI interface. Figure 2 shows the mass spectra of digoxigenin and digitoxigenin obtained by ESI and APCI. In the spectra acquired with ESI and shown in Figure 2a and c, there were a base peak (m/z 392.2 or 408.3) and also a small peak (m/z 374.8 or 391.3) related to [M + NH4]+ and [M + H]+ ions, respectively. In the spectra obtained with APCI and shown in Figure 2b and d, there were several peaks, m/z 339.1, 357.2, and 375.2 for digitoxigenin in (b) and m/z 337.2, 355.2 and 391.3 for digoxigenin in (d), which were attributed to [M + H - nH2O]+ (n ) 0, 1, 2, 3) ions. And it should be noted that the base peak (m/z 339.1 or 355.2) was the fragment [M + H - 2H2O]+ ion rather than the quasi-molecular [M + NH4]+ or [M + H]+ ion. For digitoxin, digoxin, methyldigoxin, and deslanoside, the mass spectra obtained with ESI were similar to those of digoxigenin and digitoxigenin, namely [M + NH4]+ ion produced a base peak (m/z 782.6, 798.6, 812.6, or 960.7) and [M + H]+ ion produced a very weak peak (m/z 764.6, 780.6, 794.6, or 943.0). However, with APCI, no spectrum with a recognizable peak was obtained for digitoxigenin, digoxin, methyldigoxin, and deslanoside. Unlike other CG, proscillaridin formed an abundant [M + H]+ ion with a weak [M + NH4]+ ion in the ESI source. Thus, we chose the ESI interface for HPLC/MS measurements of CG. The pH of the buffer and its concentration were examined to observe their effects on the intensity of each [M + NH4]+ ion of CG. Ammonium formate buffers (2 mM) of pH 3.4 and pH 5.7 were compared; pH 3.4 buffer resulted in a higher [M + NH4]+ signal intensity to all CG in the study except proscillaridin than did pH 5.7 buffer. Ammonium formate (pH 3.4) concentrations of 0.5, 1, 2, 5, and 10 mM were also evaluated. The intensity of the [M + NH4]+ ion of CG decreased slightly with increase in buffer concentration from 0.5 to 2 mM, whereas the intensity significantly dropped with increase in the concentration from 5 to 10 mM. In considering the buffering capacity of the ammonium formate buffer, 2 mM ammonium formate buffer (pH 3.4) was chosen for this study. CID Fragmentation of CG. Fragmentation of CG was examined under CID conditions from mild (-15 eV of CID offset energy and 1.5 mTorr of collision gas) through vigorous (-40

Figure 2. Comparison of mass spectra of digitoxigenin (5 ng) and digoxigenin (5 ng) obtained by ESI with those by APCI. Graphs a and c are mass spectra of digitoxigenin and digoxigenin obtained with ESI, and graphs b and d are those obtained with APCI, respectively. The left vertical axes represent relative percentage intensity of peaks, the right vertical axes show “absolute” intensity of peaks recorded by the mass spectrometer, and the horizontal axes indicate m/z value (amu). CG were loop injected into a 0.20 mL/min flow of 50% acetonitrile/50% HCOONH4 (2 mM, pH 3.4) heading to the ESI or APCI interface. APCI conditions: temperature of vaporizer and heating capillary was 400 and 175 °C, respectively, sheath gas (N2) pressure was 70 psi, and auxiliary gas (N2) flow was 2 mL/min.

eV of CID offset energy and the same pressure of collision gas). In the daughter ion spectra of [M + NH4]+ ions of CG acquired under the mild CID conditions, there was one main peak (m/z 651.0, 651.2, or 635.1) dominating the spectra, as shown in Figure 3a-c. The dominating daughter ion for digoxin, methyldigoxin, and deslanoside had the same m/z value (m/z 651) as indicated in Figure 3a and b and was considered to have the same structure as indicated for “CID fragment I” in Figure 1. The [M + NH4]+ ion for digitoxin fragmented in a similar manner, and the structure of its dominating daughter ion is indicated by “CID fragment II” in Figure 1. The [M + NH4]+ ions formed by digoxigenin and digitoxigenin fragmented to produce abundant [M + H]+ ions with m/z of 391.2 and 375.2, respectively. On the basis of the above results, we chose the precursor [M + NH4]+ ion, the dominating daughter ions, and the mild CID conditions for SRM quantification of CG so that the desired sensitivity could be achieved. When collision energy increased, more daughter ions were produced from one precursor ion. The typical daughter ions and fragmentation patterns of CG are listed in Table 1. These daughter ions are characteristic of parent CG and thus provide the identity of the CG. It should be noted that proscillaridin, which has a structure slightly different from those of other CG mentioned above,

underwent fragmentation in a different manner. Its [M + H]+ ion, the abundant molecular ion formed in the ESI source, could only be fragmented under vigorous CID conditions, producing many daughter ions (m/z 513.8, 366.6, ...) of which none was dominating or abundant, as shown in Figure 3d. Thus, proscillaridin was excluded from HPLC-SRM quantification in this study. Optimization of ESI Interface Parameters. There are three main parameters affecting the ESI process. Sheath gas assists aerosol formation of HPLC eluent entering the ESI interface. The capillary is a channel through which ions pass from the atmospheric pressure region to the vacuum manifold of the mass analyzer, and it is heated to a certain temperature to desolvate the analyte particle ions. Auxiliary gas helps to focus the aerosols toward the capillary and to keep the spray chamber dry. The effect of changes in each parameter setting on the response of a given CG was examined, keeping the other two parameters constant. At each setting, each compound was loopinjected six times; the peak heights were recorded in the SIM mode and averaged. Capillary temperatures at 175, 200, 225, and 250 °C were checked; at 200 °C, the SIM signal of each CG was the highest. Sheath gas pressures of 50, 60, 65, 70, and 75 psi were examined; the highest SIM signals of CG were observed at 70 psi. Auxiliary gas flow rates of 2, 5, 8, and 10 mL/min were Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

4037

Figure 3. CID daughter ion spectra of methyldigoxin (a), deslanoside (b), digitoxin (c), and proscillaridin (d) acquired with ESI, showing one main ion (m/z 651.0, 651.2, and 635.1, respectively) dominating the daughter ion spectra except in (d). CID conditions are as follows: (a and c) 1.5 mTorr of collision gas (Ar) and -15 V of offset voltage; (b) 1.5 mTorr of Ar and -20 V of offset voltage; (d) 2.1 mTorr of Ar and -30 V of offset voltage. The left and right vertical axes and horizontal axes mean the same as those in Figure 2. Ten nanograms of methyldigoxin and digitoxin each and 15 ng of deslanoside and proscillaridin each were loop injected.

also examined, and the flow rate of 8-10 mL/min produced the highest SIM signal. Extraction of CG. CG were extracted by liquid-liquid extraction (LLE) with chloroform/2-propanol (95:5, v/v).19,22 Published results have shown that digoxin and its metabolites were extracted from serum by C18 or cyclodextrin SPE columns27-29 and that oleandrin was extracted by immunoaffinity columns.25 In this study, we compared LLE by chloroform/2-propanol (95:5, v/v) or ethyl acetate with SPE by C18 columns or Oasis HLB columns (Waters) for the extraction of CG from human whole blood and urine samples. The extraction efficiency decreased in the following order: chloroform/2-propanol (95:5) > or = ethyl acetate = Waters Oasis HLB column > C18 column. The HLB column was chosen for this study since SPE uses less organic solvents and can be rapidly accomplished with batch extraction devices, although the extraction efficiency with HLB columns was a little lower than that with chloroform/2-propanol. The percentage recovery of CG from whole blood and urine by HLB columns is

shown in Table 2. The recovery from urine was generally higher than that from whole blood. The recovery of deslanoside was lower than that of any other CG. Comparison of SRM with SIM for Detection of CG. Theoretically, SRM produces lower noise levels, higher signalto-noise ratios and lower detection limits than SIM, because SRM selectively monitors an ion reaction with two mass analyzers set to specific mass windows while SIM detects just the selected precursor ions with only one mass analyzer. Some investigators have reported that lower detection limits were achieved with SRM,30 while another report stated that SRM produced higher detection limits in some anionic herbicides than SIM.31 In the present study, SRM and SIM were compared for the detection of CG in a mixture of CG standards and in whole blood samples. The detection limits of CG in an authentic standard mixture achieved by SRM were slightly lower than those by SIM. For whole blood samples, as shown in Figure 4, SRM reached detection limits that were 10 times lower than SIM. This observa-

(27) Longerich, L.; Vasdev, S.; Johnson, E.; Gault, M. H. Clin. Chem. 1988, 34 (11), 2211-2216. (28) Stone, J. A.; Soldin, S. J. Clin. Chem. 1988, 34 (12), 2547-2551. (29) Tzou, M. C.; Sams, R. A.; Reuning, R. H. J. Pharm., Biomed. Anal. 1995, 13 (12), 1531-1540.

(30) Gilbert, J. D.; Olah, T. V.; Morris, M. J.; Schwartz, M. S.; Mcloughlin, D. A. In Methodological Surveys in Bioanalysis of Drugs; Reid, E., Hill, H. M., Wilson, I. D., Eds.; Royal Society of Chemistry: Cambridge, UK, 1994; Vol. 23, pp 157-167. (31) Koeppen, B.; Spliid, N. H. J. Chromatogr., A 1998, 803, 157-168.

4038 Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

4039

392.2 (2.6) 407.9 (5.3)

digitoxigenin

[M + NH4 - 1]+ 375.6 (47) 391.5 (100) 531.6 (66)

[M + H]+ 374.7 (72) 390.5 (66) 530.9 (100)

[M]+

782.3 (59)

765.8 (19) 781.9 (67) 780.9 (39)

635.5 (100) 652.1 (30)

634.7 (28)

337.1 (54)

[M + H - 3H2O]+

336.4 (23)

[M - 3H2O]+

366.6a (74)

other

1.5 1.5 2.1

-20 -20 -30

CID conditions

Ar, mTorr

offset energy, eV

504.8 (38) 521.9 (31) 521.1 (13)

375.5 (10)

374.7 (15) 390.9 (16)

243.1 (38) 243.3 (9.0)

373.0 (29) 373.4 (8.9)

1.5

-20

CID conditions

1.5

-15

812.9) (9.4)

811.9 (24)

795.9 (48)

795.0 (14)

651.8 (100)

651.0 (94)

522.4 (6.0)

391.6 (24)

256.6 (7.7)

1.5

CID conditions

-15

offset Ar, [M + NH4]+ [M + NH4 - 1]+ [M + H]+ [M]+ [M + H - mD]+d [M - mD]+d [M + H - mD -D]+ [M - mD - D]+ [M + H - mD - 2D]+ [M - mD - 2D]+ [2D - OH - H2O]+c energy, eV mTorr

783.1 (41) 799.4 (9.6)

338.5 (99) 354.9 (82)

[M - 2H2O]+

960.9 (8.2)

943.6 (6.3)

780.9 (14)

651.1 (100)

522.7 (18)

520.9 (17)

390.7 (21)

243.0 (32)

373.4 (8.4)

-25

1.5

a [M + H - H O - C H O ]+. b [M + H - D]+, [M + H - 2D)+, and [M - H - 3D]+ represent the ions formed by sequential losses of one digitoxose shown in the bracket in the molecular structures 2 6 11 4 in Figure 1. c 2D means two digitoxoses combined together in the way shown in the bracket in Figure 1. d [M + H - mD]+ and [M - mD]+ represent the loss of a 4-methyldigitoxose. e [M - G]+ stands for the loss of a β-D-glucose.

deslanoside

offset Ar, compound [M + NH4]+ [M + NH4 - 1]+ [M + H]+ [M - G]+e [M + H - G - D]+ [M - G - D]+ [M + H - G - 2D]+ [M - G - 2D]+ [M + H - G - 3D]+ [M - G - 3D]+ [2D - OH - H2O]+c other energy, eV mTorr

methyldigoxin

compound

digoxin

digitoxin

339.4 (100) 355.9 (47)

[M + H - 2H2O]+

CID conditions

offset Ar, [M + H - 2D]+b [M - 2D]+b [M + H - 3D]+b [M - 3D]+b [2D - OH - H2O]+ c others energy, eV mTorr

512.9 (27)

373.0 (73)

[M - H2O]+

[M - D]+b

651.4 (100)

374.1 (27) 513.5 (30)

357.4

[M + H H2O]+

compound [M + NH4]+ [M + NH4 - 1]+ [M + H]+ [M]+ [M + H - D]+b

proscillaridin

digoxigenin

[M + NH4]+

compounds

Table 1. CID Fragmentation Ions of CG, Mass-to-Charge Ratio (m/z), and Relative Abundance (Shown in Parentheses)

Table 2. Recovery, Accuracy, Precision, and Detection Limits of CGa whole blood (n ) 7)

digoxigenin deslanoside digoxin digitoxigenin digitoxin methyldigoxin (IS)

added (ng/mL)

recovery (%)

8.0 8.0 1.0 8.0 8.0 1.0 8.0 5.0

64 ( 7.4 11 ( 1.2 32 ( 2.0 43 ( 2.2 36 ( 4.2 34 ( 2.0 39 ( 5.6 36 ( 2.6

urine (n ) 7)

accuracyc found CV (ng/mL) (%) 4.3 8.6 0.99 9.0 6.5 1.1 9.2

17 11 7.4 7.1 5.5 12 7.0

detection limitc (ng/mL) 0.4 0.8 0.05 2 0.1

added (ng/mL)

recoveryb (%)

10 10 1.0 10 10 1.0 10 5.0

79 ( 14 20 ( 2.1 72 ( 8.1 71 ( 5.4 70 ( 8.0 37 ( 6.6 46 ( 4.9 66 ( 6.9

accuracyc found CV (ng/mL) (%) 7.6 9.3 0.89 9.8 11 0.88 9.0

14 21 5.0 5.5 8.0 9.5 7.8

detection limitd (ng/mL) 1 0.6 0.05 1 0.1

a CG was spiked to human whole blood and urine. b Recovery was estimated by external standard calibration so that the recovery of IS could be measured. c Spiked CG were quantified by internal standard calibration. d A concentration series of CG spiked to whole blood or urine were prepared and measured, and the lowest concentration that produced an analyte signal greater than 3 times the noise level was defined as detection limits in this work.

Figure 4. Comparison of SIM (left) and SRM (right) chromatograms obtained from 4.0 and 0.4 ng of CG spiked into 1-mL whole blood samples, respectively. The left vertical axes represent relative percentage intensity of peaks, the right vertical axes show “absolute” intensity of peaks recorded by the mass spectrometer, and the horizontal axes indicate HPLC running time in minutes. The concentration of methyldigoxin (IS) was 5.0 ng/mL. It should be noted that the position order of each compound in the left graphs was different from that in the right ones. In the SIM chromatogram (left), from top to bottom, are digitoxigenin ([M + NH4]+ ) 392.3), digoxigenin ([M + NH4]+ ) 408.3), digitoxin ([M + NH4]+ ) 782.5), digoxin ([M + NH4]+ ) 798.5), methyldigoxin (IS, [M + NH4]+ ) 812.5), and deslanoside (M + NH4+ ) 960.7), respectively, while in the SRM chromatogram (right), from top to bottom are digitoxin, digitoxigenin, methyldigoxin, digoxin, deslanoside, and digoxigenin, respectively.

tion shows the detection power and the usefulness of SRM over SIM for the analysis of CG in biological samples. Linearity, Accuracy, Precision, and Detection Limits of the Method. Extracts of CG from human whole blood and urine samples were separated by HPLC and quantified by SRM, using methldigoxin as IS. The chromatograms of CG in whole blood and urine are shown in Figure 5. SRM chromatograms of blank 4040 Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

blood and urine samples were clean except in channels for digitoxigenin (m/z 392.2) and digoxigenin (m/z 408.5), where background noises were noted. In blood samples, the digitoxigenin peak was not well resolved from those of impurities, which obstructed the quantification of this CG below 2 ng/mL. SRM has been considered to be a highly selective detection technique and capable of detection of low concentration of analytes in

Figure 5. SRM chromatograms of 1.0 mL of blank whole blood (top left), 2.0 mL of blank urine (bottom left), and CG (1.0 ng) spiked to 1.0 mL of whole blood (top right) and 2.0 mL of urine (bottom right). The left and right vertical axes and horizontal axes mean the same as those in Figure 4. Methyldigoxin (IS) concentration was 5.0 ng in 1.0 mL of blood or 2.0 mL of urine. The peak identities are the same as those in the SRM chromatograms in Figure 4. Table 3. Calibration Equation and Quantitation Range

m

a

y ) mx + ba b

r2

SE

n

quantitation range, ng/mL

digoxiginen deslanoside digoxin digitoxin

(4.01 ( 0.06) × 10-1 (8.23 ( 0.38) × 10-3 (8.21 ( 0.15) × 10-2 (3.50 ( 0.10) × 10-2

Whole Blood (7.03 ( 2.36) × 10-2 (1.06 ( 1.71) × 10-3 (2.67 ( 6.01) × 10-3 (-0.24 ( 4.21) × 10-1

0.999 0.986 0.997 0.994

0.053 0.0039 0.016 0.010

8 7 11 10

0.4-10 0.8-10 0.05-10 0.1-10

digoxiginen deslanoside digoxin digitoxiginen digitoxin

(6.84 ( 0.25) × 10-1 (1.29 ( 0.11) × 10-2 (1.68 ( 0.03) × 10-1 (3.40 ( 0.09) × 10-1 (5.88 ( 0.19) × 10-2

Urine (-0.30 ( 1.41) × 10-1 (5.19 ( 6.72) × 10-3 (1.40 ( 1.18) × 10-2 (1.97 ( 0.58) × 10-1 (1.59 ( 8.32) × 10-3

0.996 0.97 0.997 0.997 0.991

0.18 0.0086 0.032 0.074 0.021

5 6 12 6 11

1-10 1-10 0.1-10 1-10 0.1-10

The slope (m) and intercept (b) are expressed as “average ( SD”.

complicated matrixes without interference. In this study, interference was observed with SRM detection. This observation suggests the necessity for efforts in sample cleanup and HPLC resolution for biological samples prior to SRM detection. The peak area ratio of CG to IS was linear; the calibration equations and quantification range are listed in Table 3. For most CG, the coefficients of determination (a statistical term, the square of the correlation coefficient, r2) of the calibration equations were greater than 0.99. Deslanoside, with low extraction recovery (11

( 1.2 and 20 ( 2.1%) from blood and urine, respectively, showed lower coefficients (less than 0.99) of determination. The accuracy and precision (Table 2) of the present method for whole blood and urine samples were less than 20 and 21%, respectively, for CG except for deslanoside. The detection limit varied with sample matrix and with compounds studied. With CG standard mixtures, as low as a few picograms of CG could be detected with a signalto-noise ratio greater than 3. With blood or urine samples, a detection limit for digoxin and digitoxin as low as 0.05-0.1 ng/ Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

4041

Figure 6. SRM chromatograms of control blood sample obtained from rat (left) and the blood sample postadministration of 20 µg of digoxin (right). The left and right vertical axes and horizontal axes mean the same as those in Figure 4. DA of m/z 798.5 and DA of m/z 812.5 represent signal channels monitoring digoxin and methyldigoxin (IS), respectively. Here DA means daughter ion.

mL was achieved, which was equivalent to 2-4 pg injected onto the column, as summarized in Table 2. For digoxigenin and digitoxigenin, however, the detection limits were high (Table 2), 1 ng/mL in urine for both CG and 2 ng/mL in blood for digitoxigenin. The high detection limits were due to the high background noise level with urine sample or poor resolution of digitoxigenin from impurities in blood sample. It is reported that the toxic and lethal plasma levels of digoxin are 3 and 5 ng/mL and those of digitoxin are 30 and 30-100 ng/ mL, respectively, and that the therapeutic level of digoxin in urine is 25-125 ng/mL.32 Deslanoside, digitoxigenin, and digoxigenin are less toxic than digoxin, and thus their toxic and lethal plasma levels are expected to be higher than those of digoxin. On the basis of these data, the present method is considered sensitive for the detection, identification, and quantification of CG in biological samples. Verification of the Method. Digoxin (20 µg) was orally administered to each of six rats, and blood samples were collected 3 h postdrug administration for determination of digoxin. Typical SRM chromatograms are shown in Figure 6. There was no peak due to impurities, and thus, the chromatograms were clean. The blood concentration of digoxin 3.0 h postdrug administration in the rat, as determined by IS calibration, was 4.3 ( 0.78 ng/mL (n ) 6). The results showed that the method we have developed is reliable, reproducible, and easily applied to biological samples. DISCUSSION SRM is a highly selective and sensitive technique for the detection of very low concentrations of drugs in biological samples. In the application of this technique, some considerations should be taken into account. First, m/z values of precursor and daughter ions should be as high as possible so that less chemical noises interfere with them. Second, efforts should be made to increase ionization efficiency at the API source. Besides these two basic considerations, the effects of chemical properties of the analytes, CID parameters, and API interfaces on sensitivity of SRM detection were examined in this study. The stability of precursor and daughter ions of an analyte under CID conditions is the most important factor affecting SRM (32) Repetto, M. R.; Repetto, M. Clin. Toxicol. 1997, 35 (4), 345.

4042 Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

sensitivity. SRM measurements with a triple-stage quadrupole tandem mass spectrometer are conducted in such a way that precursor ions are selected to pass through the first quadrupole and fragment at the second quadrupole that is instrumentally designed as a CID zone, and specific daughter ions are monitored with the third quadrupole. Obviously, the yield of the daughter ions determines the intensity of the SRM signals and the sensitivity of an SRM measurement. The percent yield can be estimated from the intensity of the daughter ion and its precursor ion obtained under certain CID conditions, by the following equation:

yield (%) ) intensity of daughter ion × 100/intensity of precursor ion The intensity of precursor ions can be obtained when no collision gas and energy are supplied to the CID cell. Apparently, the yield is a value that can be used to assess whether an SRM measurement with a pair of precursor and daughter ions is sensitive. If the yield is high, the SRM detection is sensitive and vice versa. We have demonstrated this concept by our experimental results. The [M + NH4]+ ions of most CG were easily fragmented in CID, and the yield of the abundant daughter ions was as high as 2040%; SRM detection of CG was sensitive, and the detection limit was very low. In contrast, proscillaridin with a structure slightly different from those of other CG (see Figure 1) produced abundant molecular [M + H]+ ions and weak [M + NH4]+ ions; the yield from the [M + H]+ ion to its most intensive daughter ion (m/z 349) was as low as 1%, resulting in a weak SRM signal and low sensitivity. To obtain a high yield of specific daughter ions, a great amount of precursor ions capable of fragmentation should be available for CID. The precursor ions should be a balance between stable and fragile. The precursor ions should be stable until CID fragmentation so that they can pass through the atmospheric region, the capillary heater, and the first quadrupole to the CID region without being decomposed, yet be fragile under CID conditions so that they can be fragmented to produce abundant daughter ions with high yield. It is our assumption that adduct ions such as [M + NH4]+ formed in the ESI source fall into this category of the precursor ions that are fragile and possibly capable

of producing a high yield of abundant daughter ions. Precursor ions that easily lose neutral groups such as H2O, HCl, NH3, CO, and CO2 to form stable daughter ions may also give a high yield of abundant daughter ions. Another possible way to estimate the yield of daughter ions from parent compounds is to evaluate electron impact (EI) mass spectra of the parent compounds; weak or absent molecular ion and/or abundant daughter ion in the mass spectrum means a high yield of the daughter ion. CID parameters such as collision energy and pressure of collision gas also affect the yield of daughter ions and intensity of SRM signals. Mild CID conditions such as low collision energy and low pressure of collision gas favor a high yield of one abundant daughter ion from a fragile precursor ion and thus are suitable for sensitive SRM measurements of compounds capable of forming fragile precursor ions, which is the generalization of our results that under mild CID conditions a high yield of the dominating daughter ions from [M + NH4]+ ions was observed and sensitive SRM determination of CG was achieved. On the other hand, vigorous CID conditions such as high collision energy and high pressure of collision gas have to be used for fragmentation of stable precursor ions that cannot be fragmented by mild CID conditions. In such a case, the precursor ion may be broken down to many daughter ions of which none is abundant. CID fragmentation of the [M + H]+ ion of proscillaridin at vigorous CID conditions (-30 eV of collision energy and 2.1 mTorr of Ar) is an example where many daughter ions formed but none of them was abundant (see Figure 3d). SRM measurements of compounds that form stable precursor ions, conducted under vigorous CID conditions, may not be sensitive.

ESI and APCI were compared regarding SRM measurements, because these are the two major interfaces widely used in coupling HPLC to MS on-line. ESI is the most soft ionization technique available so far. It is only in an ESI source that “fragile” [M + NH4]+ ions of CG formed that were essential to the highly sensitive SRM measurements of CG achieved in this study. APCI, however, may break down thermally labile compounds due to the high temperature of the heating vaporizer in this source, resulting in product ions of the compounds. In this study, abundant ions of the thermally decomposed products of digitoxigenin and digoxigenin were produced from the APCI source, and no molecular or product ion was observed with other CG. These product ions may be less easy to undergo CID fragmentation and are not suitable for SRM measurements. Thus, ESI is the choice for SRM measurements of compounds thermally labile or capable of forming fragile molecular ions. ACKNOWLEDGMENT F.G. is grateful to the Ministry of Education, Science and Culture of Japan for its financial support with the fellowship without which this study would not have been undertaken to completion.

Received for review March 9, 1999. Accepted June 29, 1999. AC990268C

Analytical Chemistry, Vol. 71, No. 18, September 15, 1999

4043