Identification and Quantification of Flavonoids in Human Urine

Institute of Food Safety and Toxicology, Danish Veterinary and Food ... Helsinki, Finland, and Institute of Pharmaceutical and Analytical Chemistry, R...
0 downloads 0 Views 90KB Size
Anal. Chem. 2000, 72, 1503-1509

Identification and Quantification of Flavonoids in Human Urine Samples by Column-Switching Liquid Chromatography Coupled to Atmospheric Pressure Chemical Ionization Mass Spectrometry Salka E. Nielsen,*,† Riitta Freese,‡ Claus Cornett,§ and Lars O. Dragsted†

Institute of Food Safety and Toxicology, Danish Veterinary and Food Administration, Mørkhøj Bygade 19, DK-2860 Søborg, Denmark, Division of Nutrition, P.O. Box 27 (Viikki, Latokartanonkaari 9), FIN-00014, University of Helsinki, Finland, and Institute of Pharmaceutical and Analytical Chemistry, Royal Danish School of Pharmacy, Universitetsparken 2, DK-2100 Copenhagen, Denmark

A rapid and sensitive high-performance liquid chromatographic mass spectrometric (HPLC-MS) method is described for the determination and quantification of 12 dietary flavonoid glycosides and aglycons in human urine samples. Chromatographic separation of the analytes of interest was achieved by column-switching, using the first column (a Zorbax 300SB C-3 column) for sample cleanup and eluting the heart-cut flavonoid fraction onto the second column (a Zorbax SB C-18 column) for separation and detection by ultraviolet and atmospheric pressure chemical ionization MS using single ion monitoring in negative mode. The fragmentor voltage was optimized with regard to maximum abundance of the molecular ion and qualifier ions of the analytes. Calibration graphs were prepared for urine, and good linearity was achieved over a dynamic range of 2.5-1000 ng/mL. The inter- and intraassay coefficients of variation for the analysis of the 12 different flavonoids in quality control urine samples were 12.3% on average (range 11.0-13.7%, n ) 24, reproducibility) and the repeatability of the assay were 5.0% (mean, range 0.1-14.8%, n ) 12). A subset of 10 urine samples from a human dietary intervention study with high and low flavonoid content was analyzed, and the results are reported. Flavonoids are polyphenolic compounds with antioxidant properties that occur ubiquitously in foods of plant origin.1 A high intake of flavonoids has been inversely associated with subsequent heart disease in some prospective studies,2-4 and several other biological effects have been ascribed to flavonoids in many in vivo * Corresponding author: (e-mail) [email protected]; (fax) +45 33 95 66 96. † Danish Veterinary and Food Administration. ‡ University of Helsinki. § Royal Danish School of Pharmacy. (1) Formica, J. V.; Regelson, W. Food Chem. Toxicol. 1995, 33, 1061-1080. (2) Hertog, M. G.; Kromhout, D.; Aravanis, C.; Blackburn, H.; Buzina, R.; Fidanza, F.; Giampaoli, S.; Jansen, A.; Menotti, A.; Nedeljkovic, S.; Pekkarinen, M.; Simic, B. S.; Toshima, H.; Feskens, E. J. M.; Hollman, P. C. H.; Katan, M. B. Arch. Intern. Med. 1995, 155, 381-386. (3) Hertog, M. G. L.; Feskens, E. J. M.; Hollman, P. C. H.; Katan, M. B.; Kromhout, D. Lancet 1993, 342, 1007-1011. (4) Knekt, P.; Ja¨rvinen, R.; Reunanen, A.; Maatela, J. Br. Med. J. 1996, 312, 478-481. 10.1021/ac991296y CCC: $19.00 Published on Web 03/03/2000

© 2000 American Chemical Society

and in vitro studies.5 In plants, the flavonoids are mainly present as glycosides, and due to obsolete analytical methods, it has long been a controversy as to whether flavonoid glycosides are absorbable as intact glycosides in humans or only after release of the aglycon by endogenous enzymes or by the colonic microflora.6-10 Furthermore, research on flavonoids in relation to the human diet has mainly concentrated on the flavonol, quercetin, partly due to early methodologies enabling sensitive detection of this compound in biological fluids and in foods.11,12 However, it has recently been shown, that other subgroups of flavonoids may be more important to the human health, such as the citrus flavones and the isoflavones found in soy food, since they are presumably absorbed to a higher extent than the flavonols8,13,14 and in some populations the intake of these compounds is very high.15,16 Additionally, research on absorption and excretion of flavonoids in humans has mainly concentrated on single compounds given in high doses14,17 due to the lack of methodologies that meet the requirements of sensitivity and specificity for determination of flavonoid aglycons and glycosides in biological fluids from normal, unsupplemented human subjects. The few analytical methodologies that have been proven able to detect low flavonoid concentrations in biological samples suffer by limitations in the number of (5) Middleton, E. J. Int. J. Pharm. 1996, 34, 344-348. (6) Ku ¨ hnau, J. World Rev. Nutr. Diet. 1976, 24, 117-191. (7) Hollman, P. C. H.; Vries, J. H. M.; van Leeuwen, S. D.; Mengelers, M. J. B.; Katan, M. B. Am. J. Clin. Nutr. 1995, 62, 1276-1282. (8) Fuhr, U.; Kummert, A. L. Clin. Pharmacol. Ther. 1995, 58, 365-373. (9) Cova, D.; De Angelis, L.; Giavarini, F.; Palladini, G.; Perego, R. Int. J. Clin. Pharmacol. Ther. Toxicol. 1992, 30, 29-33. (10) Nielsen, S. E. Metabolism and Biomarker Studies of Dietary Flavonoids. . Ph.D. Thesis. Copenhagen, 1999; pp 1-51 (ISBN 87-90599-01-2, Quickly Tryk A/S). (11) Hertog, M. G. L.; Hollman, P. C. H.; Venema, D. P. J. Agric. Food Chem. 1992, 40, 1591-1598. (12) Hollman, P. H.; van-Trijp, J. P.; Buysman, M. P. Anal. Chem. 1996, 68, 3511-3515. (13) Xu, X.; Wang, H. J.; Murphy, P. A.; Cook, L.; Hendrich, S. J. Nutr. 1994, 124, 825-832. (14) Ameer, B.; Weintraub, R. A.; Johnson, J. V.; Yost, R. A.; Rouseff, R. L. Clin. Pharmacol. Ther. 1996, 60, 34-40. (15) Justesen, U.; Knuthsen, P.; Leth, T. Cancer Lett. 1997, 114, 165-167. (16) Cassidy, A. Proc. Nutr. Soc. 1996, 55, 399-417. (17) Hollman, P. C. H.; van Trijp, J. M. P.; Buysman, M. N. C. P.; van der Gaag, M. S.; Mengelers, M. J. B.; Vries, J. H. M.; Katan, M. B. FEBS Lett. 1997, 418, 152-156.

Analytical Chemistry, Vol. 72, No. 7, April 1, 2000 1503

Figure 1. Chemical structures of the 12 dietary flavonoids employed in the HPLC-APCI-MS assay and the internal standards, 5,7,8-trihydroxyflavone and morin.

structurally different flavonoids that they can detect. This is either due to selective detection of, for example, flavonoids that possess fluorescence when derivatized with AlNO3, such as the flavonols possessing a 3-OH group,12 or due to high method specificity.18-20 To evaluate the influence of flavonoids in our diet, it is crucial to monitor the concentration of all the major dietary flavonoid glycosides and aglycons simultaneously in human samples. We have therefore designed a simple assay, requiring limited sample preparation, that can simultaneously quantify low levels of important dietary flavonoids21 as both flavonoid glycosides and aglycons in human urine samples. The chosen flavonoids belong to the groups of flavanones (naringin, naringenin, hesperetin) found in citrus fruits, flavonols (quercetin-3-O-glucoside, quercetin3-O-galactoside, rutin, quercetin, kaempferol, isorhamnetin, tamarixetin) found in, for example, onions, apples, tea, crucifers, and wine, and the dihydrochalcones (phloridzin, phloretin) found in apples (see Figure 1). The assay was optimized to monitor these 12 structurally different flavonoids by column switching and highperformance liquid chromatography-mass spectrometry (HPLCMS) using atmospheric pressure chemical ionization (APCI) in selected ion monitoring (SIM) mode. EXPERIMENTAL SECTION Apparatus. The HPLC-MS system was from Hewlett-Packard (Waldbronn, Germany) and consisted of a Hewlett-Packard 1090 (18) Nielsen, S. E.; Dragsted, L. O. J. Chromatogr., B 1998, 707, 81-89. (19) Nielsen, S. E.; Dragsted, L. O. J. Chromatogr., B 1998, 713, 379-386. (20) Erlund, I.; Alfthan, G.; Siren, H.; Ariniemi, K.; Aro, A. J. Chromatogr., B 1999, 727, 179-189. (21) Dragsted, L. O.; Strube, M.; Leth, T. Eur. J. Cancer Prev. 1997, 6, 522528.

1504

Analytical Chemistry, Vol. 72, No. 7, April 1, 2000

HPLC system equipped with three pumps, a thermostatically controlled column compartment, an autosampler with a 250-µL loop, an automatic six-port column switching valve, and a diode array detector (see Figure 2). Column temperature was maintained constant at 40 °C. UV detection was carried out simultaneously at 260 and 350 nm, with peak scanning between 210 and 600 nm (2-nm step). Solvents were continuously degassed using helium. The outlet of the diode array detector was directly connected to a Hewlett-Packard MSD 1100 mass spectrometer allowing the mobile phase to enter the APCI ion source without splitting. 1H NMR spectra were obtained using a Bruker (Rheinstetten, Germany) AMX 400WB spectrometer equipped with a Bruker (Rheinstetten, Germany) 120-µL flow cell (1H detection only), connected to a Kontron (Tegimenta, Switzerland) HPLC system with 420 HPLC pumps controlled by a Kontron 450 software system and a Rheodyne 7725I injector with a 100-µL sample loop. Reagents and Standards. Acetonitrile and methanol were of HPLC grade and obtained from Rathburne Ltd. (Walkerburn, U.K.). The flavonoid standards (see Figure 1), quercetin-3-Ogalactoside, quercetin-3-O-glucoside, 5,7,8-trihydroxyflavone, kaempferol, isorhamnetin, and tamarixetin, were all obtained from Apin Chemicals Ltd. (Oxon, U.K.). Phloridzin, phloretin, naringin, hesperetin, and morin were purchased from Sigma Chemicals Co. (St. Louis, MO) and rutin, quercetin, and naringenin were from Aldrich (Steinheim, Germany). All standards were HPLC grade. A 500 µg/mL stock solution of a mixture of all the flavonoid aglycons and flavonoid glycoside standards (except the internal standards, morin and 5,7,8-trihydroxyflavone) was prepared in DMSO. Dilution of the stock solution with DMSO yielded the

Figure 2. Schematic diagram of the column-switching HPLC-MS system.

working solution at concentrations of 0.5, 5, 10, and 50 µg/mL. Stock solutions of the internal standards 5,7,8-trihydroxyflavone and morin were prepared in methanol and DMSO, respectively, at concentrations of 10 µg/mL. All stock and working solutions were stored at -20 °C and were stable for at least 3 months. The enzymes used for hydrolysis of the urine samples were β-glucuronidase (Escherichia coli, >200 standard units/mL) obtained from Boehringer Mannheim (Mannheim, Germany) and arylsulfatase (Aerobacter aerogenes, 16.8 standard units/mL) from Sigma Chemicals Co. All other chemicals used were of HPLC grade or reagent grade. Samples. Urine samples were obtained from 10 female subjects participating in a parallel intervention study with highly controlled isocaloric diets either high or low in flavonoids (to be published elsewhere). The subjects volunteered for the study and gave their written consent after receiving carefully information about the study. The study was approved by the Ethical Committee of the Faculty of Agiculture and Forestry, University of Helsinki. The intakes of flavonoids were controlled by varying the intake and selection of vegetables, berries, and fruits in the diets and otherwise excluding known rich sources of flavonoids. At the 10MJ daily energy intake level, the total amount of vegetables, berries, and fruits was 810 g in the high-flavonoid group and 220 g in the low-flavonoid group. The subjects collected 24-h urine samples over 3 consecutive days at the end of the 6-week dietary period. The urine samples were collected with special urinecollecting aliquot cups (Daisho Co. Ltd., Osaka, Japan). With this equipment, 1/21 of the total urine volume was sampled in small containers to which 70 mg of ascorbic acid had been added. The samples were immediately refrigerated after collection, and after measurement of the sample volume, 140 µL of 1 M HCl was added

to 7-mL aliquots of each sample, and the resultant mixture was stored at -40 °C until analysis. Prior to flavonoid analysis, aliquots of the three 24-h urine samples collected from each subject were pooled according to the daily volumes, giving 3 mL of a urine sample representing the average flavonoid excretion in the last 3 × 24 h of the 6-week dietary period. The pooled aliquot was filtered through a 0.2-µm filter, and to a 2-mL sample of filtered urine was added 25 µL (250 ng) of the stock solution of 5,7,8trihydroxyflavone as internal standard prior to enzymatic hydrolysis. Calibration Samples. Samples for the calibration curves were prepared by spiking blank urine samples (2 mL) from a previous intervention study.22 The addition of either 0.5, 5, 10, 20, or 40 µL of the working solutions resulted in flavonoid concentrations of 0.25, 2.5, 12.5, 25, 125, 250, 500, and 1000 ng/mL. The calibration samples were enzymatically hydrolyzed and otherwise treated similarly to the samples from the intervention study prior to HPLC-MS. Enzymatic Hydrolysis. The pH of the urine samples was adjusted to pH 5 (4.9-5.1) by addition of ∼100 µL of 2 M sodium acetate buffer containing 10 mg/mL ascorbic acid. Argon was blown over the sample, 2 µL of β-glucuronidase and 10 µL of arylsulfatase preparation (>2 and 0.7 units, respectively) were added, and the sample was incubated in a sealed vial for 1 h at 37 °C under continuous shaking. After hydrolysis, 2 mL of ice-cold methanol was added to the sample and it was evaporated to dryness under vacuum. Enzymatic hydrolysis of the urine samples using an HP-2 β-glucuronidase preparation from Sigma Chemicals Co. as reported by Erlund et al.20 was also assessed. An aliquot of 40 µL of the crude enzymatic preparation, containing 4000 units of β-glucuronidase and 200 units of sulfatase, was added to two different urine samples in duplicate and to blank urine samples containing 250 ng of all flavonoid standards, under the same conditions as described above, but with an incubation time of 17 h at 37 °C under continuous shaking. Chromatography. Prior to HPLC-MS analysis, the hydrolyzed samples were redissolved in 10% aqueous methanol and 25 µL of morin dissolved in DMSO (10 ng/µL) was added as an additional internal standard, giving a final volume of 250 µL, to assess the performance of the mass spectrometer. The sample was then centrifuged at 10000g in 5 min at 4 °C and all the supernatant was injected onto the HPLC-APCI-MS system. A Zorbax 300SB-C3 (4.6 × 50 mm, 3.5 µm) column with guard cartridge (4 × 4 mm, 5 µm) was used as column 1 and a Zorbax SB-C18 (4.6 × 150 mm, 5 µm) column was used as column 2 (all from Hewlett-Packard). The mobile phases used were as follows: A, 0.5% aqueous formic acid; B, acetonitrile; and C, methanol. Figure 2 and Table 1 show in detail the chromatographic conditions for the column-switch HPLC methodology developed. Briefly, a steep linear gradient of 1-45% methanol in (A) (v/v) applied to column 1 eluted polar impurities to waste prior to the elution of the flavonoids using isocratic conditions of 45% (C) in (A). Before the elution of the most polar target compounds, the flavonoid glycosides, from column 1, the automatic six-port valve was programmed to switch from position 1 to 2, selectively eluting a heart-cut fraction containing all the flavonoids onto column 2. (22) Nielsen, S. E.; Young, J. F.; Haraldsdottir, J.; Daneshvar, B.; Lauridsen, S. T.; Knuthsen, P.; Sandstro ¨m, B.; Dragsted, L. O. Br. J. Nutr. 1999, 81, 447-455.

Analytical Chemistry, Vol. 72, No. 7, April 1, 2000

1505

Table 1. Chromatographic Conditions for the HPLC-APCI-MS System time (min) 0.0 5.0 12.0 12.1 13.0 13.1 17.0 17.1 37.0 39.0 43.0 45.0 50.0

% mobile phase (v/v) B C 0 0 0 100 100 0 0 25 25 95 95 0 0

1 45 45 0 0 0 0 3 3 3 3 1 1

flow (mL/min)

column switch (min/position of valve)

1.0 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8

0.0/1 5.7/2 9.7/1

17.0/2

Table 2. Molecular Ions and Fragmentor Ions of the Flavonoids Monitored by SIM Using HPLC-APCI-MS compound naringin rutin quercetin-3-O-gal quercetin-3-O-glc phloridzin morin (IS)b quercetin naringenin phloretin hesperetin kaempferol 5,7,8-trihydroxyflavone (IS)b isorhamnetin tamarixetin b

retention M - 1-, m/z time (min) (intensity)a

fragment ions, m/z (intensity)a

17.99 18.04 18.08 18.18 18.62 23.12 25.80 29.87 31.07 34.72 34.81 36.37

579 (100) 609 (100) 463 (100) 463 (100) 435 (10) 301 (100) 301 (80) 271 (50) 273 (60) 301 (100) 285 (100) 269 (100)

271 (10) 301 (40) 301 (35) 301 (35) 273 (100) 151 (90) 179 (50) 151 (100) 229 (20) 151 (40)

38.15 38.85

315 (100) 315 (40)

300 (99) 300 (100)

300 (70) 300 (60) 300 (60) 167 (20) 151 (100) 167 (100)

a The intensity is determined relative to the maximum mass peak. IS, internal standard.

The more unpolar impurities remaining on column 1 after elution of the target compounds were eluted by a quick column wash with 100% acetonitrile, followed by reequilibration with 100% aqueous formic acid, with the six-port switching valve in position 1. Finally a ternary gradient of methanol, acetonitrile, and aqueous formic acid was applied after switching the flow back to column 2, eluting the target compounds through the UV and APCI-MS detector. The HPLC pumps were programmed to return to the initial conditions after 45 min, still with the solvent flow eluting through both columns, thus preparing the system for the next injection. APCI-MS. APCI was performed in negative mode using selected ion monitoring. The ions selected for SIM are seen in Table 2; dwell time was 22 ms. The capillary voltage, vaporizer temperature, drying gas temperature, corona current, and fragmentor voltage were all optimized with regard to maximum signal intensity of molecular ions and fragment ions by consecutive injections of 250-µL samples (1 ng/µL) containing all the flavonoid glycosides, aglycons, and internal standards. The following optimal conditions were used: APCI capillary voltage, 2500 V; corona current, 20 µA; vaporizer temperature, 500 °C; nebulizer pressure, 60 psig; drying gas temperature, 300 °C; fragmentor voltage, 140 V. 1506

Analytical Chemistry, Vol. 72, No. 7, April 1, 2000

Linearity, Limit of Detection, and Limit of Quantification. Standard curves for each of the 12 flavonoid standards were prepared over a concentration range of 0.25-1000 ng/mL with 7 different concentration levels and more than triplicate injections at each level. Peak areas were plotted against the corresponding standard concentration using weighed linear regression to generate the standard curves. Conclusions about the limit of detection (LOD) and the limit of quantification (LOQ) were drawn by inspecting the chromatograms obtained of the lowest calibration levels. Performance Verification, Precision, and Specificity. An aliquot of 25 µL of 5,7,8-trihydroxyflavone dissolved in MeOH (10 ng/µL) was added to the urine sample prior to enzymatic hydrolysis as an internal standard correcting for analytical loss. Furthermore, morin was added to the hydrolyzed samples prior to LC-MS analysis of the samples as an additional internal standard to assess the performance of the mass spectrometer. Prior to and after each series of analysis, it was controlled that retention time of the heart-cut fraction on column 1 was within the time frames of column switching, by injection of a 250-ng sample of all flavonoid standards, including the internal standards, in 250 µL of 10% aqueous methanol containing 1% formic acid, on column 1 alone. Additionally, another 250 µL of this mixture was injected to assess the performance of the entire HPLC-APCIMS method. The precision of the assay was evaluated by analyzing blank urine samples spiked with flavonoids at concentrations of 25, 125, 250, and 1000 ng/mL in triplicates on the same day (repeatability). Reproducibility was evaluated by analyzing three blank urine samples spiked with flavonoids at a concentration of 125 ng/mL within each series of samples over a period of 1 month (n ) 24). The specificity of the method with regard to the interference of endogenous compounds was demonstrated by analyzing blank urine samples (n ) 3) collected after intake of a diet devoid of flavonoids.22 Verification. The identity of the flavonoids determined in real samples was verified by consecutive collection of individual HPLC fractions obtained from 10 mL of urine from three different urine samples from both the high- and low-flavonoid groups. An aliquot of each fraction was analyzed by HPLC-APCI-MS using different HPLC parameters,23 with the APCI-MS detector in scan mode and otherwise using identical parameters as described above. Furthermore, the identity of naringenin and hesperetin was additionally verified by stop-flow proton NMR by introducing the purified fractions of naringenin and hesperitin (∼10 µg each) in 100 µL of acetonitril-d3 to a Bruker AMX-400 WB 400-MHz spectrometer equipped with a 120-µL flow probe. The sample plug was pumped into the flow probe using CH3CN at a flow of 1 mL/min. When the 100-µL flavanone sample entered the flow probe (transfer time 0.9 min), the solvent flow was stopped using a VICI (Valco Instruments Co. Inc.) model EHMA electric valve, and 1024 scans were performed. A sample of pure CD3CN was run after washing of the flow probe for 30 min with CH3CN in order to check for contamination between samples. The CH3CN signal was suppressed using a 1D NOESY sequence with presauration (4 s) and decoupling during the mixing time (300 ms). (23) Nielsen, S. E.; Breinholt, V.; Justesen, U.; Cornett, C.; Dragsted, L. O. Xenobiotica 1998, 28, 389-401.

Statistics. Since the data obtained from the intervention study were not normally distributed, a nonparametric Wilcoxon rank scores test (SAS statistical package GLM procedure) was performed to compare flavonoid excretion in the group of high- versus low-flavonoid intake. Safety Considerations. General guidelines for work with organic solvents were respected, and urine samples were handled as potentially infectious. RESULTS AND DISCUSSION By the use of a column-switching technique and APCI-MS we have developed a rapid HPLC-MS assay requiring limited sample amount and sample preparation, allowing the simultaneously detection of both flavonoid glycosides and aglycons of 12 important dietary flavonoids21 belonging to the groups of flavanones, flavonols, and dihydrochalcones in human urine samples. Previous methodologies developed for flavonoid detection in biological fluids have been limited to the flavonols possessing fluorescence after postcolumn derivatization12 or to a single flavonoid aglycon due to the high sensitivity and selectivity required of the method.18-20 The limit of quantification in the present assay is lower, but in the same order of magnitude as of those previously reported methodologies for flavonoid detection,12,18-20 due to a relatively high background arising from the limited sample cleanup prior to APCI-MS. However, this LOQ proved sufficient for analyzing real samples from subjects receiving a diet very low in flavonoid content, and the present assay would thus be applicable to analysis of urine from subjects on their habitual diets. Enzymatic Hydrolysis. Enzymatic hydrolysis of the glucuronic and sulfate conjugates of flavonoids present in urine was employed instead of acidic hydrolyses as previously used,12 avoiding the hydrolysis of any urinary flavonoid glycosides. The optimal incubation time for the enzymatic hydrolysis has been investigated previously,19 and a different type of β-glucuronidase has also been evaluated for hydrolysis.18 Furthermore, Erlund et al. recently published the advantages of using an HP-2 β-glucuronidase preparation for hydrolysis of plasma samples using 17 h of incubation.20 This enzymatic preparation was therefore also assessed for the hydrolysis of two different urine samples from the high fruit and vegetable groups in duplicate, using the same hydrolysis conditions as described by Erlund et al.20 However, this approach gave similar or lower amounts of the detected flavonoids (data not shown), and the HP-2 enzyme preparation hydrolyzed all flavonoid glycosides present in the flavonoid standard mixture when this was included in a blank urine sample. Arylsulfatase from A. aerogenes and β-glucuronidase from E. coli were therefore used for the enzymatic hydrolysis as described in the Experimental Section using previously optimized conditions.19 Chromatography. Simple HPLC, using a single column with different gradients and various organic solvents in combination with preceding solid-phase extraction, was investigated for the determination of flavonoids in human urine samples, but sufficient separation of the target compounds from interfering compounds in the urine was unattainable. Thus, the methodology of column switching previously used for flavonoid detection in human urine samples18,19 was further developed for HPLC-APCI-MS giving the required selectivity and allowing the exclusion of the step of solidphase extraction prior to HPLC. By the application of different

Figure 3. Chromatograms of individual ion tracks of the flavonoid glycosides generated by negative APCI in SIM mode after columnswitching chromatography of the samples. Upper chromatogram at each m/z, blank urine sample; middle chromatogram, standard mixture containing the four flavonoid glycosides; lower chromatogram, urine sample from high-flavonoid group. The numbers designated to each flavonoid glycoside corresponds to the numbers in Figure 1.

conditions to the two columns, adequate separation and sensitivity was achieved. A Zorbax 300SB-C3 column was selected as the first column for sample cleanup instead of C8 as previously used,18,19 and a Zorbax SB-C18 column as the second column, to increase the chromatographic differences between the first and the second columns. Furthermore, the gradient applied to column 1 was changed in order to elute the flavonoids in the smallest possible heart-cut fraction still having sufficient removal of impurities, preventing contamination of the APCI chamber. By the application of methanol as the organic solvent on column 1 and of acetonitrile on column 2, sufficient separation of the target compounds from interfering compounds in urine was achieved prior to APCI detection. It was necessary with acetonitrile as the major organic solvent on column 2 to achieve separation of isorhamnetin and tamarixetin, which have identical molecular ions and fragment ions (m/z 315 and 300, respectively), since these compounds were not separated in methanol on the selected stationary phases. However, the acetonitrile gradient resulted in only partial separation of quercetin-3-O-galactoside and quercetin3-O-glucoside, due to the close structural resemblance of these glycosides, but satisfactory integration of these compounds was still attainable. Typical SIM chromatograms of a blank urine, a standard sample, and a urine sample obtained after fruit and vegetable intervention are shown in Figures 3 and 4. Mass Spectrometry. The operation in SIM mode allies excellent sensitivity with high specificity, as only ions that correspond to a specific fragmentation route are monitored. This enabled the simultaneous detection of both molecular ions and selected qualifier ions of 12 different dietary relevant flavonoid glycosides and aglycons in human urine as seen in Figures 3 and 4. The MS conditions using SIM APCI in negative mode were optimized with regard to drying gas flow, fragmentor voltage, gas temperature, and capillary voltage. Useful fragmentation was achievable for the flavonoid glycosides, due to loss of the glycoside chain, of quercetin and naringenin and for the methylated flavonoids hesperetin, isorhamnetin, and tamarixetin (see Table 2). Especially the fragment ion of morin, quercetin, and the (24) Lin, Y. Y.J. Chromatogr., A 1993, 629, 389-393.

Analytical Chemistry, Vol. 72, No. 7, April 1, 2000

1507

Figure 4. Chromatograms of individual ion tracks generated by negative APCI in SIM mode after column-switching chromatography of the flavonoid aglycons. Upper chromatogram at each m/z, blank urine sample; middle chromatogram, standard mixture of all 12 flavonoids including internal standards; lower chromatogram, urine sample from high-flavonoid group with added internal standards. The numbers designated to each flavonoid correspond to the numbers in Figure 1.

flavanone aglycons at m/z 151, probably resulting from cleavage of the C-ring between carbons 3 and 4 and positions 1 and 2,24 and at m/z 300, from loss of either glycoside moieties or methyl groups in the flavonols, proved very useful as qualifier ions for flavonoid detection. In contrast, no fragmentation was seen for the flavonol kaempferol and the internal standard 5,7,8-trihydroxyflavone under the chosen conditions. Linearity, Limit of Detection, and Precision. The calibration graphs were linear in the range of 2.5-1000 ng/mL urine for all 12 flavonoids. The LOQs were thus determined as 2.5 ng/mL urine, since detection at the lowest concentration level, 0.25 ng/ mL, was impossible. Inspection of the SIM chromatograms of all the flavonoids at the LOQ revealed S/N ratios of >20. Hence we can conclude that the LOD is between 0.25 and 2.5 ng/mL urine 1508 Analytical Chemistry, Vol. 72, No. 7, April 1, 2000

for both flavonoid aglycons and flavonoid glycosides. Correlation coefficients of 0.995 or higher were obtained for the relationship between peak area ratios (flavonoid standard/internal standard) and the corresponding calibration concentration. The method proved specific for the flavonoids and the internal standards, since no interfering endogenous compounds could be seen at the elution positions of the compounds in the SIM chromatograms when blank urine samples were analyzed (see Figures 3 and 4). However, in some urine samples from the intervention study, compounds interfering with the flavonoid aglycon phloretin at m/z 273 were observed. In this case, the identity of phloretin was verified by the presence of fragment ions (m/z 229 and 167). The isoflavone, daidzein, was also assessed as a potential internal standard, but interfering peaks were seen in some of the urine samples, thus excluding the use of this compound in the present assay. The precision and accuracy of the assay was thoroughly investigated for the analysis of urine samples. Table 3 shows the repeatability, accuracy, and reproducibility data obtained for all the flavonoids determined in the assay at the different concentrations tested. Most of the coefficients of variation (CVs) was below 10% with the flavonoid glycosides having the highest CVs. Although a slightly higher CV was seen for the reproducibility of the assay when this was evaluated over a period of 1 month (n ) 24), it can still be considered acceptable and meets with our objective of a routinely applicable analysis for flavonoids in urine samples. Analysis of Samples from Dietary Intervention. To evaluate the applicability of the method, a subset of urine samples from an intervention study with a diet high or low in flavonoid content (unpublished study) was analyzed. The results of the samples analyzed are presented in Table 4. As can be seen, there are large differences in the levels of urinary excretion of flavonoid aglycons between the two groups. However, only the increase in quercetin and phloretin was significant using nonparametric test (p < 0.05), because one of the subjects in the high-flavonoid group had very low excretion of urinary flavonoids compared to the other four subjects. After exclusion of this subject, there was significant difference between the flavonoid excretion in the two groups (p < 0.05) for all flavonoid aglycons except for kaempferol. No flavonoid glycosides were detected in any of the urine samples, which is in accordance with previous studies.6,8-10 The identity of the eight flavonoid aglycons was positively verified by running HPLC-APCI-MS of consecutively collected fractions using different HPLC conditions23 with the APCI-MS detector in scan mode (data not shown). This allowed the comparison of UV spectrum, retention time, and the full-scan MS spectrum (m/z 100-700) of each urinary flavonoid with authentic standards. Furthermore, the high amounts of urinary naringenin and hesperetin were verified by conclusive identification of these two flavanones by proton NMR using stop-flow methodology. The 1H NMR spectra achieved were identical to those of authentic standards (data not shown). Literature data on flavonoid excretion in humans are very limited,10 and there are no previous reports of the urinary excretion of several different flavonoids in the same samples allowing a comparison of the amounts excreted. However, we previously (25) Young, J. F.; Nielsen, S. E.; Haraldsdottir, J.; Daneshvar, B.; Lauridsen, S. T.; Knuthsen, P.; Crozier, A.; Sandstro ¨m, B.; Dragsted, L. O. Am. J. Clin. Nutr. 1998, 69, 87-94.

Table 3. Presision and Accuracy Evaluation for the Analysis of Flavonoids in Human Urine Using HPLC-APCI-MSa concn (ng/mL) found

added 25 125 125 250 1000 a

repeatability (CV %, n ) 3)

26.2 (24.6-28.7) 122.0 (110.3-128.8) 117.5 (106.0-126.3) 247.0 (232.0-269.5) 981.5 (842.8-1121.4)

reproducibility (CV %, n ) 24)

accuracy (%)

6.8 (1.3-11.8) 4.7 (0.1-9.7)

6.5 (1.4-14.8) 5.8 (0.1-12.5) 5.1 (0.8-15.1) 3.8 (0.4-7.8) 7.4 (0.2-16.4)

12.3 (11.0-13.7) 3.3 (0.9-7.2) 5.2 (4.1-8.8)

Mean of determinations for all 12 flavonoid standards; range given in parentheses.

Table 4. Excretion of Flavonoids in Urine Samples Determined by HPLC-APCI-MS flavonoid intake low high

X h (SD) X h (SD)

energy intake (MJ/day)

quercetin

naringenin

9.88 (0.15) 9.98 (0.38)

44.0 (37.8) 349*b (198)

430 (333) 2810 (1990)

urinary excretion of flavonoids (nmol in 24 h)a phloretin hesperetin kaempferol isorhamnetin ndc 747* (423)

330 (309) 3290 (2035)

174 (99.1) 420 (231)

tamarixetin

nd 113 (90.7)

nd 127 (105)

a Determined in pooled 24-h urine samples from healthy female subjects collected on 3 consecutive days (n ) 5) at the end of a 6-week dietary intervention with diets either high or low in fruits, berries, and vegetables. b *, p < 0.05 using nonparametric test (Wilcoxon rank scores), n ) 5. cnd, not determined.

determined the urinary excretion of quercetin after intakes of very low amounts of quercetin (4.8-9.6 mg/day)25 as compared to the average daily intake in Denmark of 12 mg/day. In that study, we found an excretion of 72.4-116.9 nmol of quercetin in 24-h urine, which is in good accordance with the present findings. Others have reported urinary excretions of up to 4.5 µmol of quercetin17 and 80.9 µmol of naringenin14 after extremely high flavonoid doses. In the present study, small amounts of quercetin, naringenin, hesperetin, and kaempferol were determined even in the group having the low-flavonoid diet, which shows that the sensitivity of the present assay is sufficient to analyze samples from normal subjects with a low level of flavonoids in their habitual diet. CONCLUSIONS We have successfully developed a very rapid and sensitive HPLC-MS assay for the analysis of dietary flavonoids in human urine. The use of column switching and APCI-MS in SIM mode allowed limited sample preparation and gave specific, quantitative

determination of 12 different dietary flavonoid glycosides and aglycons in human urine samples. Application of small changes to the assay should allow the developed methodology to be used for the determination of different flavonoids or other dietary compounds in human urine samples. The method was validated with respect to linearity, precision, and accuracy, and analysis of real samples was demonstrated. ACKNOWLEDGMENT The authors thank Anni Schou for excellent technical assistance. This research was in part supported by a FØTEK 3 grant from the Danish Research Councils to the project, Antioxidative Defence.

Received for review November 10, 1999. Accepted January 5, 2000. AC991296Y

Analytical Chemistry, Vol. 72, No. 7, April 1, 2000

1509