HUANG ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 5, 2009 1327

DIETARY SUPPLEMENTS

Measurement of Vitamin D in Foods and Nutritional Supplements by Liquid Chromatography/Tandem Mass Spectrometry MIN HUANG, PAUL LALUZERNE, DOUG WINTERS, and DARRYL SULLIVAN1 Covance Laboratories Inc., 3301 Kinsman Blvd, Madison, WI 53704

Vitamin D is a fat-soluble vitamin with great nutritional interest. An HPLC/MS/MS method was developed to measure vitamin D with atmospheric pressure chemical ionization. Under the experimental parameters used, the LOQ was 0.018 IU/g or 0.45 ng/g, which greatly enhances the capability of measurement of vitamin D at low levels in foods and supplements. This method was validated with spike recovery of 100 ± 15% and the RSD of less than 10% for most sample matrixes, including infant formula, cheese, cereal and cerealbased foods, multivitamin supplements, and pet foods. The results for vitamin D were compared with those obtained by other methods.

V

itamin D is a key member of the fat-soluble vitamin group and is critical to good health. It has been shown to help efficient absorption of calcium (1). Usually, vitamin D has two forms, D3 (cholecalciferol) and D2 (ergocalciferol). D3 can be synthesized from provitamin D sterols when skin is exposed to sunlight for deriving provitamin D sterols by UV irradiation, while D2 can be produced in plants and fungi by the solar irradiation of ergosterol. Nevertheless, deficiency of vitamin D is quite common both in developed and developing countries, due to nutritional and geographic condition, climate, individual life style, and other reasons. Fortification of the vitamin is required in some foods and supplements to promote good health. For example, human breast milk may not supply enough vitamin D. Infant formula products are developed with suitable amounts of vitamin D and other nutrients. For adults, multivitamin supplements have become very popular. As a result, the measurement of vitamin D is often required for nutrition research, clinical study, and product development and quality control for foods and supplements. Currently, vitamin D in food is usually measured by HPLC (2–5). Since there are many sterols that have chemical formulas and structures similar to those of vitamin D, interferences with the measurement of trace levels of vitamin D are probable, especially given the complexity of food matrixes and the

Received December 12, 2008. Accepted by AP February 23, 2009. 1 Corresponding author’s e-mail: [email protected]

relatively low specificity of HPLC for isomeric compounds. Usually, a long sample preparation procedure for the elimination of interferences is required. Even given the thorough sample preparation procedures, there are still many failures in testing vitamin D in complicated food matrixes. HPLC/MS provides high sensitivity and selectivity and, therefore, has been increasingly used in applications for food and nutrition analysis (6–11). Heudi et al. (11) reported an HPLC/MS method to measure vitamin D3, along with other fat-soluble vitamins, in infant formula. Vitamin D2 was used as an internal standard to correct for the matrix effect. Sample saponification was done at 85°C for 30 min. The analytes were further separated by solid-phase extraction. Dimartino (10) used HPLC/MS to measure vitamin D3 in cheese and other foods. In sample preparation, overnight saponification was used to remove fat completely, followed by hexane extraction. The average spike recovery ranged from 98 to 105%. However, to develop a suitable method for a more complex sample matrix, tandem mass spectrometry (MS/MS) is preferred. Kamao et al. (7) used HPLC/MS/MS to measure fat-soluble vitamins in human breast milk. An isotope labeled D3 compound was synthesized and used as an internal standard, which was expected to serve better for correcting the matrix effect than vitamin D2. To measure vitamin D at an extremely low level in the milk, a derivatization was added to improve the sensitivity by increasing the ionization efficiency. With more HPLC/MS methods developed and more experienced HPLC/MS analysts in food laboratories, this technique has been used with greater frequency. In this study, an HPLC/MS/MS method for measurement of vitamin D in foods and supplements was developed. Overnight saponification was investigated to release and extract the analyte. Signal depression from matrixes was investigated, and isotope labeled vitamin D3 was used as an internal standard that was added before saponification to correct for signal depression. Experimental Chemicals and Reagents (a) Ethanol.—Sigma-Aldrich (St. Louis, MO). (b) Acetonitrile.—HPLC grade (Burdick & Jackson, Muskegon, MI).

1328 HUANG ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 5, 2009 Table 1.

Parameters for MS measurement Q1, amu

Q3, amu

DPa

EPb

CEc

CXPd

Vitamin D3

385.4

259.1

55.0

14.5

21.0

17.3

Vitamin D3

385.4

107.1

55.0

14.5

31.9

5.20

Vitamin D3

385.4

159.1

55.0

4.5

32.7

9.10

Vitamin D2

397.4

125.1

43.0

9.5

19.1

6.9

Vitamin D2

397.4

107.0

43.0

4.5

36.0

19.7

Vitamin D2

397.4

271.1

43.0

9.0

17.3

18.6

Isotope vitamin D3

388.4

259.1

55.0

14.5

21.0

19.1

Analyte

a b c d

DP = Declustering potential. EP = Entrance potential. CE = Collision energy. CXP = Collision cell exit potential.

(a)

(b)

Figure 1. MS/MS spectrum of the precursor ion at (a) 385 m/z of vitamin D3 and (b) 397 m/z of vitamin D2.

HUANG ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 5, 2009 1329

Figure 2. LC/MS spectra of vitamin D in cereal.

Figure 3. LC/MS/MS spectrum of vitamin D3 standard solution at 10 IU/mL obtained under the optimized conditions.

1330 HUANG ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 5, 2009

Figure 4. Chromatogram of vitamin D3 in liquid infant formula (0.44 IU/mL).

(c) Potassium hydroxide.—ACS grade (Fisher Chemical, Fairlawn, NJ). (d) Methanol.—HPLC grade (Fisher Chemical). (e) Hexane.—HPLC grade (Sigma-Aldrich). (f) Acetic acid.—Glacial, 99.9% (Fisher Chemical). (g) Pyrogallic acid.—ACS grade (J.T. Baker, Phillipsburg, NJ). (h) Butylated hydroxytoluene (BHT).—99.8% (ACROS Organics USA, Morris Plains, NJ). Reference Standards (a) Cholecalciferol (vitamin D3).—100% Pharmacopeial Convention, Rockville, MD).

(U.S.

(b) Ergocalciferol (vitamin Pharmacopeial Convention).

D2).—100%

(U.S.

Isotope Standard Isotope vitamin D3 – [2H3].—IsoScience (King of Prussia, PA). Apparatus (a) HPLC/MS/MS system.—MDS SCIEX API – 4000 QTRAP (Applied Biosystems, Concord, Ontario, Canada). (b) HPLC system.—Shimadzu LC20AD (Kyoto, Japan).

Figure 5. LC/MS/MS chromatogram of vitamin D3 in dog food.

HUANG ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 5, 2009 1331

Figure 6. Spectral interference of vitamin D3 (0.15 IU/g) in a complex pet food matrix.

(c) Analytical column.—YMC Carotenoid S-3 (150 ´ 2.0 mm, 3 mm particle size; Waters Corp., Milford, MA). Sample Preparation For infant formula, various other human foods, pet foods, and supplements, weigh 1 to 10 g in an Erlenmeyer flask, depending on the vitamin D concentration in these samples. Add 40 mL reagent grade alcohol with 2% pyrogallic acid, 0.3 mL 100 IU/mL isotope D3 internal standard, and 20 mL KOH (50%). Set for overnight saponification with magnetic stirring after removing air with nitrogen flow. Extract with 30 mL hexane containing 12.5 mg/L BHT and dry down about 10 mL of the extract for reconstitution in 1 mL 70% acetonitrile–H2O (70 + 30) with 5 min of sonication. Filter the sample solution through a 0.45 mm PTFE membrane before injection.

Table 2. Signal depression in different matrixes Signal depression, %

Sample

Signal depression, %

Chew

10

Cheese

45

Cereal product 1

29

Infant formula

69

Cereal product 2

31

Liquid infant formula

39

Pet food

44

Nutritional tablet

72

Sample

For liquid infant formula, weigh 30 g of the sample. Add 10 g of KOH pellets instead of a 50% solution. The rest of the sample preparation procedure is the same as that for powder infant formula. Instrument and Parameters In a general HPLC separation for most of the food matrixes, an isocratic mobile phase with a 10 min run time was used. The flow rate was 0.3 mL/min with a mobile phase consisting of 0.1% acetic acid in 75% acetonitrile/25% methanol (MeOH). For pet foods, the mobile phase used was 7% of 0.1% acetic acid in water and 93% of 0.1% acetic acid/25% acetonitrile/75% MeOH for better separation. The flow rate was the same, but the run time was 20 min. Because vitamin D is fat-soluble with low polarity, an atmospheric pressure chemical ionization (APCI) source was used to achieve the high sensitivity required. The analyte tends to be positively charged; therefore, it was measured in the positive ion mode. A 10 mL sample solution was injected for the measurement. Common parameters for APCI were optimized, including nebulizer current 2.0 mA, temperature 260°C, and ion source gas 27 psi. Other parameters of the mass analyzer for each analyte are listed in Table 1. Results and Discussion In various foods and supplements, vitamin D may be fortified in free form or encapsulated. Also, it may exist naturally in ingredients of animal meats (D3) and plants (D2). Different sample preparation procedures were investigated and compared. Our primary results showed that some

1332 HUANG ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 5, 2009

Figure 7. Chromatograms of multiple product ions of vitamin D3.

procedures worked for certain matrixes, but did not work well for the others. To efficiently release the analytes from all samples, overnight saponification was found to be necessary if a general method was to be developed that is applicable to most sample matrixes. Saponification was performed by adding strong KOH and ethanol with pyrogallic acid for protection of vitamin D from oxidation. In the following experiments, samples were prepared using the optimized

procedure described in the Experimental section, unless otherwise specified. Quantification and Confirmation of D3 and D2 Based on HPLC/MS/MS The MS/MS spectra of D3 and D2 are shown in Figure 1. Multiple reaction monitoring was used for confirmation and quantification under the parameters listed in Table 1. The

Table 3. Spike recovery (%) for some typical food matrixes Samples

50% level

100% level

150% level

Corn flake

103, 100

105, 101

101, 107

Multivitamin

115, 108

106, 108

93.9, 93.6

Cheese

114, 106

106, 109

94, 104

Cereal

97, 96

98, 108

115, 112

Wet pet food

100, 113

105, 111

106, 105

Dry cat food

96.5, 99.6

101, 104

109, 106

Liquid infant formula

95.4, 78.9

89.5, 92.4

95.9, 100

Fortified cereal

100.0, 91.6

98.5, 98.0

99.2, 93.9

HUANG ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 5, 2009 1333

Figure 8. Linear calibration curve for vitamin D measurement with the standard solutions at 50, 20, 10, 1, and 0.2 IU/mL; IS = internal standard.

identification of the analytes was based on monitoring three product ions for each analyte. They were amu 259.1, 107.1, and 367.2 from 385.4 for vitamin D3, and 125.1, 107.0, and 271.1 from 397.4 for vitamin D2. To reliably identify the analytes in complex food matrixes, it is necessary to positively identify multiple product ions and combine these data with other

evidence, such as relative abundance of the product ions in the mass spectra and peak shape and retention time in the chromatograms. The product ion of 367.3 produced from 385.4 by losing water was the most abundant one for D3. However, this response was also found from other cholesterols in some matrixes and caused spectral interferences. A similar situation

Table 4. Results comparison for food products (IU/g) Samples

HPLC/MS/MS

HPLC

Deviation, %

0.44

0.43

2.3

0.44

Infant formula 1

2.15 ± 0.12

2.33 ± 0.30

–8.4

2.25

Infant formula 2

3.22 ± 0.24

3.24 ± 0.57

Liquid infant formula

Infant formula 3 Nutritional tablet

3.48

–0.6

3.08

–8.0

NAa

24.3

21.1

NA

2.21

Interference

NA

NA

Cereal

1.52

NA

NA

1.38

Pet food 1

0.14

0.13

7.1

0.15

Pet food 2

0.14

0.16

–14

0.15

Pet food 3

0.43

0.35

20

0.46

Breakfast cereal

Pet food 4

30.8

3.76

Claim

0.45

0.48

Premix 1

12.2

14.2

Premix 2

22.7

Cheese

2.59

a

NA = Not available.

–6.7

0.46

–16.4

NA

22.5

0.8

15

NA

NA

2.11

1334 HUANG ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 5, 2009

was observed for the reaction of 397.4 to 379.3 for D2. Therefore, these reactions were not used for either quantification or identification, even though they had strongest signal intensities. In our experiments, the reactions of 385.4 to 259.1 and 397.4 to 125.1 were measured for quantification of D3 and D2, respectively. Because food matrixes are very complex, MS/MS should be the first choice if the instrumentation is available. Figure 2 gives LC/MS spectra of D3 and D2. It is obvious that both D2 and D3 have high baselines in LC/MS, due to its low selectivity. Under the optimized HPLC conditions, vitamin D and previtamin D were separated. The analytes can be measured simultaneously within 10 min. Figure 3 shows the chromatogram of vitamin D at 10 IU/mL in standard solution. Due to low level of vitamin D3 in liquid infant formula, a large sample size (30 g) was taken for the sample preparation and finally reconstituted in 1 mL of solution, gaining about 30 times concentration by this procedure. The chromatogram of a commercial liquid infant formula product with vitamin D3 at 0.44 IU/mL is shown in Figure 4. The signal was strong enough and exhibited good analyte peak shape and separation, which made accurate quantification possible. Vitamin D in pet food samples is usually present in low concentrations. However, the challenges associated with the measurement of vitamin D in pet foods is not only from the extremely low concentration, but also from its very complicated matrix. Even using MS/MS, multiple peaks were observed, as shown in Figure 5. By comparing the spectra of sample, vitamin D standard, and isotope labeled vitamin D, it was confirmed that the peak that eluted at 4.7 min was vitamin D. The other peaks could have been produced from sterols with the same molecular weight and similar chemical structure. There are dozens of compounds that have the same molecular weight as vitamin D3, and is difficult to differentiate these compounds from vitamin D3 by MS. In practice, we found that pet foods from the same manufacturer may have different matrix interferences, depending on their ingredients and the source of raw materials. As a result, much more effort was needed for method development and validation. As an example, by using the normal HPLC procedure described in the Experimental section, trace levels of vitamin D in some pet food products were found to be higher than the values expected, although the analyte peak shape was good and no interfering peak was observed. However, when the HPLC mobile phase was modified to 93% organic solvent and 7% water to increase the retention time from 4.6 to 13 min, interfering peaks were confirmed for some products and separated. As shown in the bottom chromatogram of Figure 6, the interfering peaks, which we believe may have been produced from other sterol compounds, showed the same response under the MS/MS monitoring. It was also found that the interferences were matrix-dependent. Our results showed that the interferences were different from product to product, indicating variances with raw materials or fortified ingredients. For some products with a low level of vitamin D, the interferences could cause

results that are several times higher. After removing the interferences with the longer 13 min HPLC run time, the results matched well with the values expected. Therefore, the longer HPLC run time was used to separate vitamin D from complex pet food matrixes, unless specified. Signal Depression in Different Matrixes Using the optimized procedure, signal depression in different matrixes was investigated. The depression was evaluated by comparing the signal response of the isotope labeled internal standard between the standard solution and the samples. The depression was found to be most severe for chew samples (90%) and much more modest for nutritional tablets (28%). The results for other matrixes are listed in Table 2. From the results, it is clear that an internal standard is required to correct the signal depression from sample matrixes. LOD and LOQ Under the optimized conditions, this method is very sensitive. For the lowest standard solution at 0.2 IU/mL or 5 ng/mL, vitamin D was positively identified by all three product ions at 259.1, 107.1, and 367.2, which provided strong evidence for the identification of vitamin D3 (Figure 7). The peak of the isotope labeled D3, with the same retention time at 4.6 min, further supported the identification. The LOD, defined as three times the S/N, was determined by 10 replicate analyses of a diluted pet food sample solution with a concentration close to the blank for the analytes. LOQ was calculated based on 10 times the S/N. It was determined that the LOD and LOQ for vitamin D3 were 0.0096 and 0.032 IU/mL and 0.006 and 0.02 IU/g, respectively. Spike Recovery and RSD The method was validated for accuracy by spiking vitamin D at three different concentration levels, 50, 100, and 150%, of the analyte concentration in samples. For fortified cereal food, the average spike recovery of duplicate spikes at each level was found to be 96, 99, and 97%, respectively. The RSD in that matrix was 3.3% for 18 replicates. The results for other matrixes are listed in Table 3. Linearity The linearity for the analytes was good, with an r value of 0.9990 or higher for the calibration curve made from standard solutions with concentrations of 50, 20, 10, 1, and 0.2 IU/mL (Figure 8). Sample Analysis Samples of typical food matrixes were analyzed by this HPLC/MS/MS method. The results are listed in Table 4. Infant formula 1 and 2 listed in the table were used as controls so multiple measurements were carried out. The others were a single measurement. The results by HPLC and the claim values are also given in the table, if they are available. The HPLC method used was an AOAC method that has an overnight saponification, followed by hexane extraction,

HUANG ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 5, 2009 1335

drying down, and gel permeation chromatography separation (12). In many cases, HPLC suffered from spectral interference and the comparisons are not listed in the table, although more purification steps were used. Conclusions The method was demonstrated to be accurate for measurement of vitamin D in many matrixes, from a relatively high concentration level in nutritional supplements to an extremely low level in liquid infant formula and pet foods. The high selectivity of HPLC/MS/MS helps identify and eliminate spectral interference in complex food matrixes. References (1) Ball, G.F.M. (2006) Vitamins in Foods, Analysis, Bioavailability, and Stability, CRC Press, Taylor & Francis Group, Boca Raton, FL (2) Blake, C.J. (2007) J. AOAC Int. 90, 897–910

(3) Hart, G.R., Furniss, J.L., Laurie, D., & Durham, S.K. (2006) Clin. Lab. 52, 335–343 (4) Lensmeyer, G.L., Wiebe, D.A., Binkley, N., & Drezner, M.K. (2006) Clin. Chem. 52, 1120–1126 (5) Olkowski, A.A., Aranda-Osorio, G., & McKinnon, J. (2003) Int. J. Vitam. Nutr. Res. 73, 15–18 (6) Pramank, B.N., Ganguly, A.K., & Gross M.I. (2002) Applied Electrospray Mass Spectrometry, Marcel Dekker, New York, NY (7) Kamao, M., Tsugawa, N., Suhara, Y., Wada, A., Mori, T., Murata, K., Nishino, R., Ukita, T., Uenishi, K., Tanaka, K., & Okano T. (2007) J. Chromatogr. B 859, 192–200 (8) Priego Capote, F., Jiménez, J.R., Granados, J.M., & de Castro M.D. (2007) Rapid Commun. Mass Spectrom. 21, 1745–1754 (9) Magalhães, P.J., Carvalho, D.O., Guido, L.F., & Barros, A.A. (2007) J. Agric. Food Chem. 55, 7995–8002 (10) Dimartino, G. (2007) J. AOAC Int. 90, 1340–1345 (11) Heudi, O., Trisconi, M., & Blake, C. (2004) J. Chromatogr. A 1022, 115–123 (12) Official Methods of Analysis (2005) 18th Ed., AOAC INTERNATIONAL, Gaithersburg, MD