A Liquid Chromatography-Mass SpectrometryMethod for the Quantification of Bioavailability andBioconversion of â-Carotene to Retinol in HumansYan Wang,† Xiaoying Xu,† Machteld van Lieshout,‡ Clive E. West,‡ Johan Lugtenburg,§Michiel A. Verhoeven,§ Alain F. L. Creemers,§ Muhilal,| and Richard B. van Breemen*,†

Department of Medicinal Chemistry and Pharmacognosy, University of Illinois at Chicago, Chicago, Illinois 60612, Division ofHuman Nutrition & Epidemiology, Wageningen University, The Netherlands, Leiden Institute of Chemistry, Leiden University,The Netherlands, and Nutrition Research and Development Centre, JI Dr Sumeru 63, Bogor 16112, Indonesia

A method based on high-performance liquid chroma-tography-atmospheric pressure chemical ionization massspectrometry (APCI LC-MS) was developed for thequantification of the bioavailability of retinyl palmitateand â-carotene and the bioconversion of â-carotene toretinol in humans. Following oral administration of[8,9,10,11,12,13,14,15,19,20-13C10]-retinyl palmitateand [12,13,14,15,20,12′,13′,14′,15′,20′-13C10]-â-caro-tene at physiological doses to children between 8 and 11years of age, blood samples were drawn and serum wasprepared. Retinol and â-carotene were extracted from 0.2-and 1.0-mL serum samples, respectively, and analyzedusing reversed-phase HPLC with a C30 column interfacedto an APCI mass spectrometer. Unlike other LC-MSassays for carotenoids, no additional purification stepswere necessary, nor was any derivatization of retinol orâ-carotene required. APCI LC-MS showed a linear detec-tor response for â-carotene over 4 orders of magnitude.Using selected ion monitoring to record the elution profileof protonated circulating â-carotene at m/z 537 and[13C10]-â-carotene at m/z 547, the limit of detection wasdetermined to be 0.5 pmol injected on-column. To assessthe ratio of labeled to unlabeled retinol, selected ionmonitoring was carried out at m/z 269, 274, and 279.These abundant fragment ions corresponded to the lossof water from the protonated molecule of circulatingretinol, [13C5]-retinol (metabolically formed from orallyadministered [13C10]-â-carotene), and [13C10]-retinol(formed by hydrolysis of [13C10]-retinyl palmitate). Theratios of labeled to unlabeled retinol and the ratio oflabeled to unlabeled â-carotene were calculated. Com-bined with standard HPLC measurement of â-caroteneand retinol concentration and a mathematical model,these results showed that this simple LC-MS method canbe used to quantify â-carotene bioavailability and itsbioconversion to retinol at physiologically relevant doses.

Carotenoids are the primary source of vitamin A in the humandiet in developing countries. Lack of vitamin A increases risk of

morbidity and mortality.1 There has been concern that thebioavailability of â-carotene and retinol and the bioconversion ofâ-carotene to retinol are less than previously thought.2 In nutrition,the term bioavailability means the fraction of an ingested nutrientthat is available for utilization in normal physiologic functions andfor storage.3 In nutrition, bioconversion means the fraction of abioavailable nutrient (here: absorbed provitamin A carotenoids)that is converted to the active form of a nutrient (here: retinol).Among the more than 600 carotenoids, ∼50 have provitamin Aactivity in humans, and among them, â-carotene is the mostnutritionally active.4 Besides its provitamin A activity, â-caroteneis also a singlet oxygen scavenger,5 and epidemiological studiesindicate that consumption of fruits and vegetables rich in â-car-otene is associated with reduced risks of heart disease6 andcancer.7 Despite its nutritional significance, â-carotene bioavail-ability and bioconversion to retinol remain poorly characterizedin humans because of the lack of suitable bioanalytical methods.Unlike bioavailability (as defined in pharmaceutical research) andpharmacokinetic measurements of typical drugs, studies withnutrients such as â-carotene and retinol are complicated by highcirculating concentrations (0.1-1.0 µM) of these nutrients in bloodand tissues. Isotopic labeling may be used to distinguish anadministered dose from circulating material. However, labelingwith stable isotopes is preferred to labeling with radioactiveisotopes, especially in studies involving children, because of safetyissues. Therefore, recent studies on the bioavailability and bio-conversion of â-carotene have used labeling with stable isotopes.

Current approaches to the measurement of the bioavailabilityof â-carotene and retinol involve either administration of highdoses of â-carotene (12-30 mg/d)8,9 or deuterated compounds

* Corresponding author: (telephone) (312) 996-9353 ; (fax) (312) 996-7107;(e-mail) breemen@uic.edu).

† University of Illinois at Chicago.‡ Wageningen University.§ Leiden University.| Nutrition Research and Development Centre.

(1) Beaton, G. H.; Martorell, R.; Aronson, K. J. Effectiveness of vitamin Asupplementation in the control of young child morbidity and mortality indeveloping countries. United Nations. 13. 1993 (ACC/SCN state of the artseries: nutrition policy discussion paper).

(2) de Pee, S.; West, C. E.; Muhilal; Karyadi, D.; Hautvast, J. G. A. J. Lancet1995, 346, 75-81.

(3) Jackson, M. J. Eur. J. Clin. Nutr. 1997, 51 (Suppl, 1), S1-S2.(4) Demming-Adams, B.; Gilmore, A. M.; Adams, W. W. FASEB J. 1996, 10,

403-412.(5) Burton, G. W.; Ingold, K. U. Science 1984, 224, 569-573.(6) Kritchevsky, S. B. J. Nutr. 1999, 129, 5-8.(7) Smith, T. A. Br. J. Biomed. Sci. 1998, 55, 268-275.(8) Brown, E. D.; Micozzi, M. S.; Craft, N. E.; Bieri, J. G.; Beecher, G.; Edwards,

B. K.; Rose, A.; Taylor, P. R.; Smith, J. C. Am. J. Clin. Nutr. 1989, 49,1258-1265.

Anal. Chem. 2000, 72, 4999-5003

10.1021/ac000454e CCC: $19.00 © 2000 American Chemical Society Analytical Chemistry, Vol. 72, No. 20, October 15, 2000 4999Published on Web 08/24/2000

such as [10,10′,19,19,19,19′,19′,19′-2H8]-â-carotene.10,11 In the firstapproach, the large dose facilitated the measurement of theadministered compound in the serum, but total plasma â-caroteneincreased 110-2600 times at the same time. This raises thequestion of whether absorption and bioconversion would be thesame at physiological doses that do not perturb total â-carotenelevels. In the second approach, the HPLC retention times ofâ-carotene and [2H8]-â-carotene were different. Since deuterationof â-carotene changes its HPLC retention time, there is concernthat other physicochemical properties might be altered. 13C-Labeling and GC/MS with an isotope ratio mass spectrometerhave also been used to study â-carotene kinetics in humans.12 Inthat study, sample preparation included semipreparative HPLCfollowed by saponification, hexane extraction, liquid-liquid parti-tion, reversed-phase HPLC, and finally hydrogenation catalyzedby platinum oxide. The analytical step required gas chromatog-raphy and then combustion of the purified sample followed bymeasurement using a specialized isotope ratio mass spectrometer.The elaborate sample preparation and specialized equipmentrequired for the approach limit its accessibility and applicabilityto large-scale human intervention studies.

Based on our liquid chromatography-mass spectrometry(LC-MS) methods for the analysis of retinol and carotenoids inhuman serum,13,14 a quantitative method was developed to quantifythe bioavailability of â-carotene and its bioconversion to retinolin humans. In our approach, 13C was used instead of deuteriumto minimize any physicochemical differences between labeled andunlabeled compounds. Furthermore, extrinsically specificallylabeled â-carotene and retinyl palmitate (See Figure 1) wereadministered to human volunteers at physiological doses, so thatcirculating levels of these compounds would not be perturbed.Here, we report our LC-MS method and preliminary data froma human intervention study investigating â-carotene bioavailabilityand its bioconversion to retinol in children in Indonesia.

EXPERIMENTAL SECTIONChemicals. Unlabeled all-trans-retinol and all-trans-â-carotene

were purchased from Sigma Chemical (St. Louis, MO), and [13C10]-â-carotene and [13C10]-retinyl palmitate were synthesized on onthe basis of published methods.15 The structures of thesecompounds and the sites of 13C-labeling are shown in Figure 1.HPLC grade or better solvents were purchased from FisherScientific (Fair Lawn, NY).

Sample Preparation. Although the study design and com-plete results will be reported separately, examples of our studyof â-carotene and retinol bioavailability and bioconversion ofâ-carotene to retinol in Indonesian school children will be used

to illustrate the practical application of the LC-MS method.Children (8-11 years old) were administered physiological dosesof [13C10]-retinyl palmitate and [13C10]-â-carotene (80 µg each) twicea day as supplements to a low-carotenoid and low-retinol diet.Blood samples were drawn at various times up to 10 weeks afterstarting the experiment. Serum was prepared by centrifugationof whole blood (750g for 10 min at room temperature) and storedon dry ice or in a -80 °C freezer until analysis. Retinol andâ-carotene were extracted from serum without saponification. Eachserum sample (1.0 mL for â-carotene and 0.2 mL for retinolanalysis) was mixed with 1 mL of 30% NaCl (aq) and 1 mL of 70%EtOH and then extracted three times with 3-mL portions ofhexane. The hexane extracts were combined and evaporated todryness under vacuum. The extraction procedure was carried outunder subdued light. The residue was redissolved in 200 µL ofmethanol/methyl tert-butyl ether (1:1; v/v) for LC-MS analysis.

LC-MS Analysis of â-Carotene. LC-MS was carried outusing a Hewlett-Packard (Palo Alto, CA) G1946A LCMSD quad-rupole mass spectrometer equipped with a series 1100 HPLCsystem consisting of a binary pump, automatic solvent degasser,autosampler, a YMC (Wilmington, NC) C30 column (3 µm; 250 ×4.6 mm), and a C30 guard cartridge. The solvent system consistedof a gradient from 15 to 30% methyl tert-butyl ether in 12 min,followed by 30% methyl tert-butyl ether for 13 min. The cosolventwas methanol containing 1 mM ammonium acetate, and the flowrate was 1 mL/min. The column was then flushed with 100%methyl tert-butyl ether for 5 min and then equilibrated with 15%methyl tert-butyl ether for 10 min before the next injection. For

(9) Micozzi, M. S.; Brown, E. D.; Edwards, B. K.; Bieri, J. G.; Taylor, P. R.;Khachik, F.; Beecher, G. R.; Smith, J. C. Am. J. Clin. Nutr. 1992, 55, 1120-1125.

(10) Dueker, S. R.; Jones, A. D.; Smith, G. M.; Clifford, A. J. Anal. Chem. 1994,66, 4177-4185.

(11) Novotny, J. A.; Dueker, S. R.; Zech, L. A.; Clifford, A. J. J. Lipid Res. 1995,36, 1825-1838.

(12) Parker, R. S.; Swanson, J. E.; Marmor, B.; Goodman, K. J.; Spielman, A. B.;Brenna, J. T.; Viereck, S. M.; Canfield, W. K. Ann. N. Y. Acad. Sci. 1993,691, 86-95.

(13) van Breemen, R. B.; Nikolic, D.; Xu, X.; Xiong, Y.; van Lieshout, M.; West,C. E.; Schilling, A. B. J. Chromatogr., A 1998, 794, 245-251.

(14) van Breemen, R. B.; Huang, C. R.; Tan, Y.; Sander, L. C.; Schilling, A. B. J.Mass Spectrom. 1996, 31, 975-981.

(15) Lugtenburg, J. Eur. J. Clin. Nutr. 1996, 50 (Suppl. 3), S17-S20.

Figure 1. Structures of all-trans-[12,13,14,15,20,12′,13′,14′,15′,-20′-13C10]-â-carotene and all-trans-[8,9,10,11,12,13,14,15,19,20-13C10]-retinyl palmitate and their conversion to all-trans-[13C5]- and[13C10]-retinol in the body. (* denotes position of 13C-label.)

5000 Analytical Chemistry, Vol. 72, No. 20, October 15, 2000

each analysis, 60 µL of the serum extract (200 µL total volume)was injected onto the HPLC column. Positive ion atmosphericpressure chemical ionization (APCI) and electrospray werecompared, and APCI was selected for all subsequent studiesbecause of the necessity for a wide dynamic range (see Figure 2and Results and Discussion). The optimum positive ion APCIconditions for â-carotene analysis included a nitrogen nebulizerpressure of 45 psi (3.1 bar), a vaporizer temperature of 325 °C, anitrogen drying gas temperature of 200 °C at 11 L/min, a capillaryvoltage of 2800 V, a corona current of 4.0 µA, and a fragmentorvoltage of 70 V. Selected ion monitoring (SIM) was used to recordthe abundances of the protonated molecules of unlabeled and[13C10]-â-carotene at m/z 537 and 547, respectively. The ratio oflabeled to unlabeled (circulating) â-carotene was calculated. Theserum concentration of â-carotene was measured using HPLC withUV/visible absorbance detection according to published proce-dures.16

LC-MS Analysis of Retinol. Retinol was analyzed using LC-MS as described above with the following modifications. Com-pared to the carotenoid analysis, a more polar solvent system wasused for the HPLC analysis of retinol. The solvent systemconsisted of gradient elution at 0.2 mL/min from 30 to 46% solventB in 8 min, 46 to 50% B in 4 min, and then isocratic elution at 50%B for 8 min. Solvent A consisted of methanol/water/acetic acid(50:50:0.5; v/v/v), and solvent B was methanol/methyl tert-butyl

ether/acetic acid (50:50:0.5; v/v/v). The C30 column (3 µm; 100× 2.1 mm) was flushed out with 100% solvent B for 5 min toremove strongly retained compounds and reequilibrated at 30%B for 10 min before the next injection. For each analysis, 20 µLof the serum extract (200 µL total volume) was injected onto theHPLC column. The positive ion APCI conditions for retinolanalysis included a nitrogen nebulizer pressure of 35 psi (2.4 bar),a vaporizer temperature of 150 °C, a nitrogen drying gas temper-ature of 275 °C at 5 L/min, a capillary voltage of 2200 V, a coronacurrent of 2.0 µA, and a fragmentor voltage of 50 V. Becauseretinol fragments during positive ion APCI form a base peak ofm/z 269,13 SIM was used to record the signals at m/z 269 forunlabeled retinol and the corresponding ions of m/z 274 and 279for [13C5]-retinol and [13C10]-retinol, respectively. [13C5]-Retinol wasexpected to be formed in vivo by metabolic conversion of [13C10]-â-carotene (Figure 1). The ratios of [13C5]-retinol and of [13C10]-retinol to unlabeled (circulating) retinol were calculated. Theconcentrations of retinol and â-carotene in serum were determinedat the same time using the HPLC method of Craft et al.16

RESULTS AND DISCUSSIONSince it is preferable to administer doses of labeled compounds

that do not perturb the steady state, an LC-MS method isrequired that can measure low concentrations of these labeledcompounds in serum. Furthermore, these measurements mustbe carried out in the presence of higher concentrations ofcirculating compounds coeluting from the HPLC. Therefore, anLC-MS method was required with a wide dynamic range for thesimultaneous determination of trace amounts of labeled retinoland â-carotene in the presence of large quantities of the corre-sponding unlabeled compounds. Over the range of â-caroteneconcentrations, 0.4968-99.36 pmol/µL (0.25-1987 pmol on-column), positive ion APCI produced a linear response with R2 )0.9984 (Figure 2A). In contrast, positive ion electrospray did notproduce a linear response for â-carotene over the same range ofconcentrations (Figure 2B). Similar results were observed forretinol.13 Therefore, APCI was used throughout this investigationinstead of electrospray, because of its wider dynamic range andgreater linearity of detector response.

The positive ion APCI mass spectrum of â-carotene (Figure3) showed a base peak at m/z 537 corresponding to the protonatedmolecule. The abundance of all the other ions was less than 25%.Therefore, SIM of the abundant protonated molecule of â-carotenewas used in all subsequent measurements. A low limit of detectionof 0.25 pmol on-column was obtained (defined as a signal-to-noiseratio of 3:1). The lower limit of quantification was determined tobe 560 fmol on-column with a coefficient of variation of 4.82% (n) 6). Comparable limits of detection were obtained for retinoland retinyl palmitate by monitoring the fragment ion at m/z 269(13). Because all human serum samples contained circulatingâ-carotene, no blank serum was available for limit of detectiondeterminations. Therefore, standard solutions were prepared inmethanol/methyl tert-butyl ether (1:1, v/v) instead of serum. Ourprevious report validated the use of mobile phase instead of serumfor the preparation of retinol standards for similar studies.13

Examples of LC-MS analyses of â-carotene in serum samplesobtained at day 0 and day 21 are shown in Figure 4. At day 0(Figure 4A), only unlabeled all-trans-â-carotene was observed inthe serum sample, and there was no compound with m/z 547 (that(16) Craft, N. E.; Wise, S. A.; Soares, J. H. J. Chromatogr. 1992, 589, 171-176.

Figure 2. Standard curves for the analysis of all-trans-â-caroteneusing (A) positive ion APCI mass spectrometry and (B) positive ionelectrospray mass spectrometry. The base peak of the mass spectrumat m/z 537 [M + H]+ was recorded using SIM. Note that theelectrospray curve shows a nonlinear response, while APCI gave alinear response over the entire range of concentrations investigated.

Analytical Chemistry, Vol. 72, No. 20, October 15, 2000 5001

is the m/z of labeled â-carotene) eluting at the same retentiontime. At day 21 (Figure 4B), both labeled and unlabeled â-carotenewere observed coeluting at 22.8 min. There was little differencein the retention time between labeled and unlabeled â-carotene,which confirmed that the physical properties of all-trans-[13C10]-â-carotene resemble those of unlabeled â-carotene. This is unlikethe results with [2H8]-â-carotene and unlabeled â-carotene reported

by Dueker et al.,10 where there was a distinct difference inretention time of the two compounds. Another advantage of 13C-labeling over deuterium labeling is the elimination of the pos-sibility of isotope scrambling during bioconversion of â-caroteneto retinol.

Structures of [13C10]-â-carotene and [13C10]-retinyl palmitate andtheir conversion to retinol in the body are shown in Figure 1.Whether cleavage of â-carotene to retinol is centric or eccentric,

Figure 3. Positive ion APCI mass spectrum of â-carotene. The protonated molecule was detected as the base peak at m/z 537.

Figure 4. LC-MS SIM chromatograms of human serum extractsshowing elution of all-trans-â-carotene and all-trans-[13C10]-â-caroteneat m/z 537 and 547. (A) Serum sample drawn immediately beforeadministration of all-trans-[13C10]-â-carotene; (B) serum sample ob-tained on day 21 of dietary administration of all-trans-[13C10]-â-carotene.

Figure 5. LC-MS analysis of retinol in a hexane extract of humanserum recorded using SIM of m/z 269 (circulating all-trans-retinol),274 (all-trans-[13C5]-retinol), and 279 (all-trans-[13C10]-retinol). (A)Serum sample obtained immediately before administration of labeledcompounds; (B) serum sample obtained on day 21 of dietaryadministration of labeled compounds. Note the appearance of 13C-labeled retinol in serum after administration.

5002 Analytical Chemistry, Vol. 72, No. 20, October 15, 2000

bioconversion of [13C10]-â-carotene would produce [13C5]-retinol.[13C10]-Retinyl palmitate was administered instead of [13C10]-retinolbecause nearly all retinol in food is present as esters, which aremore lipid soluble and stable. Nevertheless, circulating retinolbound to albumin is unesterified. Examples of LC-MS analysesof retinol from human serum are shown in Figure 5. During LC-MS, the elution profiles of circulating retinol and labeled retinolwere recorded using SIM of m/z 269, 274, and 279. The ion ofm/z 269 corresponded to the loss of water from the protonatedmolecule of retinol ([MH - H2O]+), the ion of m/z 274 cor-responded to the [MH - H2O]+ ion of [13C5]-retinol formed inthe body from bioconversion of [13C10]-â-carotene, and the ion ofm/z 279 corresponded to the [MH - H2O]+ ion of [13C10]-retinolformed in the body from hydrolysis of [13C10]-retinyl palmitate.At day 0 (Figure 5A), only unlabeled retinol was observed at m/z269 with a retention time of 17.7 min. The absence of coelutingsignals at m/z 274 and 279 indicates that there is no interferencefor the LC-MS analysis of labeled retinol. After 21 days of dietaryadministration of labeled â-carotene and retinyl palmitate, [13C5]-and [13C10]-retinol were detected coeluting with unlabeled retinolduring LC-MS analysis of a serum extract (Figure 5B), theseresults are typical of the data obtained on other days and for othersubjects.

CONCLUSIONAn APCI LC-MS method has been developed and applied to

the measurement of â-carotene and retinol in human serum.Utilizing administered 13C-labeled compounds, this method is

suitable for the investigation of â-carotene bioavailability and itsbioconversion to retinol. By combining simple sample preparationwith straightforward LC-MS analysis, the procedure is convenientand sufficiently sensitive and specific for measuring isotopicenrichment of â-carotene and retinol in human serum. The limitof detection and quantification for retinol and â-carotene (<1 pmolon-column) is suitable for human intervention studies utilizing lowdoses of labeled â-carotene and retinol that do not alter circulatinglevels of these compounds. A significant advantage of using 13C-labeling instead of deuterated compounds is the greater physi-cochemical similarity of the 13C-labeled compounds to theirunlabeled analogues, thus avoiding possible isotope effects.Finally, our LC-MS method uses standard equipment with nospecial need for high resolution or tandem mass spectrometry.Data on the bioavailability of â-carotene and its bioconversion toretinol obtained from a series of human intervention studies willbe published separately.

ACKNOWLEDGMENTWe thank the Hewlett-Packard Co. for providing the mass

spectrometer used in this investigation. Partial funding wasprovided by the National Institutes of Health Grant R24 CA83124(to R.B.v.B).

Received for review April 24, 2000. Accepted July 21,2000.

AC000454E

Analytical Chemistry, Vol. 72, No. 20, October 15, 2000 5003