738 Biomed Environ Sci, 2015; 28(10): 738-750 Original Article

Application of UPLC-MS/MS Method for Analyzing B-vitamins in Human Milk*

REN Xiang Nan1, YIN Shi An1, YANG Zhen Yu1, YANG Xiao Guang1,#, SHAO Bing2,#, REN Yi Ping3, and ZHANG Jing2

1. National Institute for Nutrition and Health, Chinese Center for Disease Control and Prevention, Beijing 100050, China; 2. Beijing Key Laboratory of Diagnostic and Traceability Technologies for Food Poisoning, Beijing Research Center for Preventive Medicine, Beijing 100013, China; 3. Zhejiang Provincial Centre for Disease Prevention and Control, Hangzhou 310051, Zhejiang, China

Abstract Objective To determine ten B-vitamins in human milk by ultra-high performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS). Methods The pretreated human milk samples were adequately separated and quantified within 11 min by UPLC-MS/MS with an Acquity UPLC HSS T3 column (2.1×100 mm, 1.8 µm). The mobile phase was a gradient of 2.5 mmol/L ammonium formate aqueous solution and acetonitrile at a flow rate of 0.35 mL/min. Stable isotope internal standards were used in the analysis, to correct for the method variability, including matrix and ionization effects. The homogenized human milk samples were deproteinzed using methanol, unknown contaminants were extracted with diethyl ether and hydrophobic phase was discarded. The analytes were monitored via ESI+ionization and detected in multiple reaction monitoring (MRM) with three acquisition functions. Results Calibration curves ranged from 0.5-160 ng/mL (thiamin, riboflavin, biotin, nicotinic acid, pyridoxine, pyridoxamine, pyridoxal), and 2.5-800 ng/mL (pantothenic acid, FAD and nicotinamide) (R2=0.990-0.999). The relative recovery ranged from 80.1% to 120.2%; accuracy was determined to be 98.3% to 108.0%. Intra-day and inter-day variation were 3.4%-19.9% and 5.9%-18.1%, respectively. The limit of quantification (LOQ) for all vitamins was between 0.25 and 3 µg/L. Conclusion This method was successfully applied for simultaneous analysis of ten B-vitamins in human milk. Key words: B-vitamins; Human milk; UPLC-MS/MS Biomed Environ Sci, 2015; 28(10): 738-750 doi: 10.3967/bes2015.104 ISSN: 0895-3988 www.besjournal.com (full text) CN: 11-2816/Q Copyright ©2015 by China CDC

INTRODUCTION meet all nutritional requirements for exclusively breastfed infants aged 0 to 6 months of life[1-3]. It is -vitamins are essential nutrients for important to analyze B-vitamins in human milk growth and development of infants, most because inadequate vitamin intake could lead to of which participate in the metabolism in infantile vitamin deficiencies[4-8]. B [1] the form of coenzyme . Human milk is thought to Adequate intakes (AI) of nutrients, including

*This study was supported by the National High Technology Research and Development Program of China (863 Program) (No. 2010AA023004). #Correspondence should be addressed to YANG Xiao Guang, Professor, Tel: 86-10-83132798, Fax: 86-10-83132808, E-mail: [email protected]; SHAO Bing, Professor, Tel: 13381081679; E-mail: [email protected] Biographical note of the first author: REN Xiang Nan, female, born in 1984, PhD, majoring in food nutrition.

B-vitamins in human milk by UPLC-MS/MS 739

B-vitamins, are generally estimated by the content in phenomenon of cross-talk may occur if the former human milk from well-nourished mothers and an method is used for analyzing ten B-vitamins, average daily breast milk intake[3,9-10]. The previous which leads to greater deviation of results[29-31]. methods of analyzing B-vitamins, e.g. microbiological, Therefore, it is essential to monitor two or more chemical, radioactive methods, enzyme channels for MRM analysis and optimize the mobile immunoassays, and more recently high performance phase to better separate components to avoid liquid chromatography (HPLC) were usually used for cross-talk[29-34]. the analysis of B-vitamins in human milk[9,11-16]. In To our knowledge, no analytical method has general, each method could be used for only one been published to quantify simultaneously ten individual vitamin with labor-intensive and B-vitamins in human milk. Therefore, we have time-consuming pretreatments[10]. Moreover, a developed a rapid, sensitive, efficient UPLC-MS/MS larger volume of human milk was required for these method for simultaneous analysis of ten B-vitamins determinations[9,11,14-18]. Because of the complexity in human milk. Moreover, this method has been of B-vitamins, constraints of traditional methods and successfully applied to the quantification of thiamin, difficulty of operability of sampling, research studies riboflavin, pyridoxal, pyridoxine, pyridoxamine, of B-vitamins in human milk have been studied in nicotinamide, nicotinic acid, flavin adenine various countries with a limited sample size ranging dinucleotide (FAD), biotin, and pantothenic acid in from 5 to 152 samples[9-16]. human milk. More recently, only Hampel et al.[10] described the analysis of five B-vitamins (thiamin, riboflavin, MATERIALS AND METHODS flavin adenine dinucleotide (FAD), nicotinamide, and [19] pyridoxal) and Tao et al. in China analyzed five Chemicals and Reagents B-vitamins (thiamin, riboflavin, nicotinamide, pantothenic acid, and pyridoxal) in human milk by Thiamin hydrochloride (99% purity), biotin (99% ultra-high performance liquid chromatography- purity), pantothenic acid (99.9% purity), nicotinic tandem mass spectrometry (UPLC-MS/MS). However, acid (98% purity), riboflavin (98% purity), flavin they were not suitable for simultaneously analyzing adenine dinucleotide (FAD) (95% purity), ten B-vitamins, which are necessary in order to nicotinamide (99.5% purity), thiamin- 13 13 establish the data base of human milk and direct the (4-methyl- C-thiazol-5yl- C3) hydrochloride (99 13 13 diet for lactating women. Even though pyridoxal is atom % C, 98% CP), riboflavin- dioxopyrimidine- C4, 15N (99 atom %13C, 98 atom% 15N, 97% CP), the principal form of vitamin B6, pyridoxine and 2 pyridoxamine exist in human milk and the pyridoxal-(methyl-d3) (98 atom % D, 98% CP) were percentages are 2%-16%, 6%-13%, obtained from Sigma-Aldrich Trading Co., Ltd respectively[11,15,20]. Pyridoxine has been used to (Shanghai, China). Pyridoxine hydrochloride (98% treat neonates with early onset seizures. It can purity) and pyridoxamine dihydrochloride (98% improve infants’ intelligence and prevent the purity) were purchased from TCI (Tokyo, Japan). neurologic function damaged[21-23]. The different Pyridoxal hydrochloride (99% purity) was obtained existed forms of each vitamin can be transformed. from Acros Organics (Morris Plains, NJ, USA). Except nicotinamide, nicotinic acid is found in human Acetonitrile and methanol were bought from Dikma milk[9]. Pantothenic acid mostly exists in its free form (HPLC buffer, Beijing, China) and ammonium formate (85%-90%) and bound form (10%-15%)[24]. Pantothenic was obtained from Fluka (Shanghai, China). acid is the universal precursor for coenzyme A (CoA) Ultrapure water was prepared with a Milli-Q and acyl carrier protein and it is closely associated Ultrapure water system (Millipore, MA, USA). with infants’ key metabolic and energy-yielding The human study was approved by the ethics pathways[25-26]. Biotin is found in human milk and can committee of the National Institute of Nutrition and enhance the appetite and prevent the neonatal Food Safety, China CDC. Written informed consent anemia[2,27-28]. These forms are the constituents of was obtained from all participants. Human milk B-vitamins and also contribute to the vitamin samples, included colostrums, transitional milk and functions significantly, but were less studied[10]. mature milk, were collected from apparently healthy Some different components have the same women in Huangpu city in the Guangdong Province quantification ion, e.g. nicotinamide and nicotinic of China and stored at -80 °C freezer in darkness acid, pyridoxamine and pyridoxine. So the until analysis.

B-vitamins in human milk by UPLC-MS/MS 741 confirmed by directly injecting the standards (1.0 remove the hydrophobic molecules. The hydrophilic μg/mL) into the mass spectrometer. The optimum phase was transferred to centrifuge tubes and was conditions were as follows: capillary voltage, 3.5 kV; centrifuged at 14,480 xg for 10 min at 4 °C. Finally, source temperature, 120 °C; desolvation 90 µL of the clear supernatant was transferred into temperature, 350 °C; desolvation gas (nitrogen, 99% 150 µL glass insert in a 1.5 mL amber glass vial and purity) flow, 682 L/h; collision gas (argon, 99.99% analyzed. -3 purity) pressure, 4.04×10 mbar. Detection was Both working solutions and samples were acquired in multiple reaction monitoring (MRM) at prepared at the same day and all samples were three functions (time segments). Cone voltage, analyzed within 24 h to ensure the stability of the collision energies and dwell times were optimized analytes. for each compound to get the highest sensitivity and In order to monitor stability of the system accuracy (Table 1). during the process of run, the standard curve, quality control sample in triplicate and pooled milk Quality Control sample were repeatedly injected every 25 injections 0.1 g of NIST SRM 1849a infant formula was to evaluate the possible influence and ensure the dissolved in 10 mL water of LC-MS grade and accuracy of analysis. aliquots were stored at -80 °C freezer in amber tubes Method Validation until analysis. Pooled breast milk which mixed 2 L human milk samples was also used as the quality The method was validated by linearity, limit of control material for each analytical process. 5 detection (LOD), limit of quantification (LOQ), replicates of the NIST samples and the pooled breast accuracy, precision, recovery, matrix effects and milk were analyzed with each run. stability in accordance with ‘Guideline on bioanalytical method validation’ of the European Sample Pretreatment Medicines Agency[36] and Commission Decision of [35] Breast milk samples stored at -80 °C and were the European Communities . thawed over night at 4 °C. On the following day, the Quantification of the analytes was carried out samples were prepared under the dim light and on using the isotope internal standards dilution method ice bath to protect the vitamins from degradation. and by the peak-area ratios of the analytes to the The samples were homogenized using sonicator respective internal standard (100 ng/mL). The (MIRIS, Sweden) which was specially designed for parameters of the linear regression equations, homogenization of milk and it has a metallic probe. including slope, intercept and the coefficient of 2 The time of homogenization was tested and 2 s for determination (R ) were determined. LOD and LOQ 1 milliliter was shown to be optimal. Due to of each compound were evaluated by a signal-to generating heat, samples were kept on ice bath noise ratio of 3:1 and 10:1 respectively. when homogenized the milk in order to prevent the The accuracy of the method was evaluated by loss of nutrients. It was homogenized for 20 s for 10 the recovery assay using a three level standard milliliter milk samples, which could make milk addition on four days and 9 replicates for each samples more homogeneous and it did not affect the standard addition level on each day (Table 2), and remanent samples used for analyzing the other analyzing the NIST SRM 1849a infant formula. The components. 10 µL of internal standard and 250 µL milk control samples were unavailable and the most of methanol were added to 50 µL of homogenized similar substance was infant formula. The breast milk or 20 µL of the NIST SRM 1849a infant concentrations of vitamins in breast milk sample formula[10]. The samples were mixed for 1 min were determined. The spiked concentrations were and centrifuged at 14,480 xg for 10 min at 4 °C in carried out using aliquots from the same sample. order to remove solid particles from the supernatant. The low, middle and high concentration were spiked in the pooled milk sample and were determined in The supernatant was transferred into amber [35,37-38] tubes and evaporated to dryness (nitrogen-stream, nonuplicate (Table 2). Precision of the method was evaluated by intra-day and inter-day relative room temperature). The residue was reconstituted [38] and mixed for 1 min in 100 µL of pure water. standard deviation (RSD) . The results of NIST SRM 1849a were used to assess the accuracy of our The samples were then back-extracted with 100 µL method, which was expressed as the ratio of of diethyl ether by mixing for 30 s and placing determined concentration to theoretical them at 4 °C refrigerator for 10 min in order to 742 Biomed Environ Sci, 2015; 28(10): 738-750

[32] concentration . The results of pooled milk samples A meant the peak area of the pooled milk were compared on different days in order to sample, B and C represented the peak area of the evaluate the stability and accuracy of the method, pre-preparation and post-preparation after adding expressed as the ratio of the obtained results to the the standards, and D expressed the peak area of the mean value to reflected the dispersion degree neat standard solution with the same concentration among results of different days[10]. The extraction recovery and matrix effects were as C. evaluated by the test design described[10,32,37-39]. Application Extraction recovery reflects the efficiency of whole pretreatment. Matrix effects indicate the influence According to the lactation stage, 183 of human of existing system. The equations for the calculations are as follows: milk samples from Guangdong province were divided Extraction recovery (RE)=[(B-A)/(C-A)] ×100% (1) into seven groups: 0-3 d, 4-7 d, 8-10 d, 11-15 d, 16- Matrix effects (ME)=[(C-A)/D] ×100% (2) 30 d, 31-100 d, and 101-300 d, respectively. Statistical

Table 2. Recoveries, Precision and Matrix Effects for Target Compounds at Three Spiked Levels

Spiked Intra-day Inter-day Matrix Effects Isotope Internal Relative Analyte Concentration RSD (%) RSD (%) (%) Standards Recovery (%) (ng/mL) (n=9) (n=36) (n=9)

100 95.6 4.8 98.9 (4.1) 13 Thiamin Thiamin- C3 200 80.1 14.1 7.9 91.7 (9.5) 500 93.7 11.0 85.7 (5.2) 100 86.9 13.7 101.9 (2.4) 13 15 Riboflavin Riboflavin- C4, N2 200 87.9 18.6 13.3 109.6 (2.5) 500 97.8 15.5 104.3 (5.7) 100 80.2 8.3 87.2 (4.5) 13 15 Nicotinamide Riboflavin- C4, N2 200 80.7 10.1 12.0 85.1 (0.6) 500 80.2 8.7 93.4 (9.8) 100 81.4 8.1 16.6 (12.6) 13 15 Nicotinic acid Riboflavin- C4, N2 200 98.1 9.2 14.9 22.9 (6.9) 500 111.4 17.7 17.1 (2.7) 10 91.7 10.5 104.3 (2.1)

Pyridoxine Pyridoxal-d3 20 120.2 7.7 7.3 105.9 (10.2) 50 82.5 4.0 106.1 (8.2) 10 104.7 20.0 99.5 (7.9)

Pyridoxamine Pyridoxal-d3 20 87.3 10.1 18.0 89.0 (0.9) 50 98.7 15.2 88.9 (6.6) 50 95.1 7.0 89.2 (4.3)

Pyridoxal Pyridoxal-d3 100 80.5 7.9 5.9 93.9 (15.8) 200 87.1 7.6 99.7 (8.6) 10 109.1 13.8 88.7 (8.0) 13 Biotin Thiamin- C3 20 116.7 10.3 18.1 86.3 (4.8) 50 105.8 4.8 98.9 (1.6) 500 92.3 9.2 89.8 (9.3) 13 Pantothenic acid Thiamin- C3 1000 85.1 17.3 14.9 85.5 (9.5) 2000 85.2 14.0 87.7 (8.5) 500 61.8 19.7 54.1 (7.5) 13 15 FAD Riboflavin- C4, N2 1000 67.4 14.9 15.2 63.2 (9.7) 2000 76.8 13.2 62.0 (5.5) B-vitamins in human milk by UPLC-MS/MS 743 analyses among different groups were performed The column A is UPLC HSS T3 (2.1×50 mm, 1.8 µm, by using the SPSS software (Ver. 19.0) (IBM, Waters), column B is UPLC BEH C18 column (2.1×100 USA). Differences were considered significant at mm, 1.7 µm, Waters), and column C is UPLC HSS T3 P<0.05. (2.1×100 mm, 1.8 µm, Waters). It was found that column A and B, could neither separate these RESULTS AND DISCUSSION vitamins, nor generate reproducible peak areas in parallel runs. Only column C could adequately Optimization of UPLC-MS/MS Conditions separate these vitamins and support constant retention times and peak areas in parallel In preliminary tests, three columns were experiments (Figure 1). So 100 mm HSS T3 was compared regarding the separation of ten vitamins. selected in this experiment.

Figure 1. The quantitative daughter ion chromatograms of 10 target components and 3 internal standards: (1) pyridoxamine; (2) nicotinic acid; (3) pantothenic acid; (4) pyridoxal; (5) pyridoxal-d3; (6) 13 13 15 pyridoxine; (7) nicotinamide; (8) thiamin- C3; (9) thiamin; (10) biotin; (11) FAD; (12) riboflavin- C4, N2; (13) riboflavin. 744 Biomed Environ Sci, 2015; 28(10): 738-750

Formic acid and ammonium formate were difference for the contents of other vitamins considered as the constituents of mobile phase, between sonicated samples and vortexed samples which led to different effects of separation. (P>0.05). The RSD of the sonicated samples ranged Increasing concentration of ammonium formate between 6.5% and 18.6% (n=18), the RSD of the resulted in complete separation. Thiamin, vortexed samples ranged from 10.0% to 23.2% pantothenic acid and biotin were not separated (n=18). The RSD of thiamin, riboflavin, nicotinamide, using 0.1 mmol/L ammonium formate. Nicotinic acid, pyridoxal, FAD and pantothenic acid between pyridoxamine, biotin and FAD were not separated sonicator and vortex were significant differences fully using 0.5 mmol/L ammonium formate. Nicotinic (12.2% vs. 14.7%, 10.7% vs. 20.5%, 9.8% vs. 13.3%, acid, pyridoxamine and thiamin were not separated 6.5% vs. 10.0%, 14.8% vs. 21.8%, 16.8% vs. 23.2%) fully using 1.0 mmol/L ammonium formate. Except (P<0.05). The samples homogenized using the nicotinic acid and pyridoxamine, other components sonicator had better RSD using than vortexed were separated ideally using the 2.5 mmol/L samples, indicating that sonication contributed to ammonium formate (Figure 1). Either 0.1% formic the homogenization of the human milk samples. The acid or 0.1% formic acid in 2.5 mmol/L ammonium sonicator which was designed for homogenizing milk formate did not improve separation effects further. samples had the metallic probe, so it can fully touch The peak of FAD disappered when it was eluted by milk samples and make it more homogeneous. the 0.1% formic acid. Therefore, 2.5 mmol/L Hence, sonicator was used to homogenize human ammonium formate was considered as ideal milk samples in present study. constituents of mobile phase. Methanol (MeOH) and acetonitrile (ACN) were A total of 10 target components and 3 internal applied to remove the protein (Figure 2). The standards required 26 ion monitoring channels, so extraction recovery rates were 80.31%-118.63% for the whole run time (11 min) was divided into 3 MRM using either MeOH or ACN for thiamin, riboflavin, acquisition functions to gain the collection points of nicotinamide, pyridoxine and biotin. However, the each chromatographic peak (10-25 points) to ensure recovery rates were better by using MeOH the accuracy of quantification (Figure 1). Adequate (80.7%-93.9%) than ACN (22.0%-66.7%) for nicotinic separation and two channels for MRM analysis can acid, pyridoxamine, pyridoxal and pantothenic acid. contributed to avoid cross-talk, gain the collection Hampel et al[10]. had used MeOH to remove proteins points of each chromatographic peak (10-25 points) in analysis of five B-vitamins, the results of analysis [29-34] and ensure the accuracy of quantification . of 10 B-vitamins were in accordance with former [10] Optimization of Sample Pretreatment conclusions . More particularly, the extraction recoveries of FAD by using ACN were rather low A sonicator or vortex was used to homogenize (4.51%). So MeOH was selected to remove proteins. human milk samples before sampling. They were The other step of pretreatment was extracting the analyzed in sextuplicate in three distinct days. The non-polar constituents. Diethyl ether and ethyl acetate contents of nicotinamide (529.6 µg/L vs. 437.2 µg/L) were tested in our experiments. The response of using were higher of sonicated samples than vortexed diethyl ether was nearly twice higher than that of samples (P=0.03). But there was no significant using ethyl acetate for some target components, such

Figure 2. Comparison of extraction recoveries of different solvents. B-vitamins in human milk by UPLC-MS/MS 745 as pyridoxal, pyridoxine and riboflavin. Considering were evaluated by recovery experiments. The losses the effects of extraction, sensitivity of target of process and system errors were corrected using components and the reference[10], diethyl ether was the isotope internal standards. Because not all applied to remove the non-polar constituents to labeled vitamins are available, the replacements assure the efficency of analysis. were chosen for the labeled thiamin, pyridoxal and riboflavin based on retention time, matrix effects Method Validation and reference[10]. The labeled thiamin was used as Linearity The R2 of FAD and pantothenic acid the internal standard of thiamin, biotin, pantothenic ranged from 0.990 to 0.999 over a concentration acid, the labeled pyridoxal for three B6 vitamers and range from 0.2 to 2000 ng/mL and the R2 of other the labeled riboflavin was used to analyze the compound ranged from 0.990 to 0.999 over a riboflavin, FAD, nicotinamide, nicotinic acid. The concentration range from 0.2 to 900 ng/mL (Table 3). detailed spiked concentrations, recovery, RSD are Considering the actual concentrations of B-vitamins shown in Table 2. Except for FAD, the recoveries of in human milk, the working standard solution of all other vitamins were in the range of 80.1%-120.2% thiamin, riboflavin, biotin, nicotinic acid, pyridoxine, (n=9), RSD of intra-day (n=9) and RSD of inter-day pyridoxamine, pyridoxal ranging from 0.5 to 160 (n=36) ranged between 4.8%-19.9%, and 5.9%-18.1%, ng/mL (0.5, 1, 5, 10, 25, 50, 80, 100, 125, and 160 respectively. The recoveries and RSD were ng/mL) and the working standards of pantothenic acceptable for these vitamins at three levels, acid, FAD, nicotinamide ranging from 2.5 to 800 reflecting good accuracy and precision of the ng/mL (2.5, 5, 25, 50, 125, 250, 400, 500, 625, and method based on theory of the paper of 800 ng/mL) were used in the analysis considering the methodology. content of actual milk. Moreover, the NIST SRM 1849a and pooled milk LOD and LOQ Because it is difficult to simulate sample as the quality control sample were analyzed the breast milk matrix, the LOD and LOQ of thiamin, to assess the accuracy of the method. They were riboflavin and pyridoxal were evaluated adding the analyzed in sextuplicate in six distinct days (Table 4). isotope internal standards to the human milk. The The determined results of seven vitamins in NIST LOD and LOQ of others were estimated using the SRM 1849a were in good agreement with the pooled milk sample. For Quattro Premier XE, LOQ for theoretical concentrations, appearing in a range of all vitamins was between 0.25 and 3 µg/L, the LOD accuracy between 98.3% and 108.0% with the for all vitamins was between 0.05 and 1.50 µg/L inter-day RSD 6.4%-10.3%. We have prepared (Table 3). However, for Xevo TQ-S, LOQ for all pooled human milk samples as reference material vitamins was between 0.01 and 0.50 µg/L. for internal quality control. Even though the true Accuracy and Precision Accuracy and precision contents of vitamins in pooled human milk sample were

Table 3. Linear Relationships and Sensitivity of Compounds Analyte Coefficient LOD 1 (µg/L) LOQ 1 (µg/L) LOQ 2 (µg/L) Reference Data (µg/L)

Thiamin 0.997 0.05 0.25 0.01 2-221[10] Riboflavin 0.996 0.20 1.00 0.03 0-845[10] Nicotinamide 0.995 0.10 0.50 0.10 2-3179[10] Nicotinic acid 0.990 0.50 2.00 0.20 Pyridoxine 0.999 0.10 0.50 0.06 Pyridoxamine 0.991 1.50 5.00 0.04 6-692[10] Pyridoxal 0.999 0.20 0.70 0.03 FAD 0.993 0.20 1.00 0.05 29-818[10] Biotin 0.994 1.00 3.00 0.40 2.8-5.9[9] Pantothenic acid 0.997 0.50 1.00 0.50 360-2900[9,16]

Note. LOD 1: data obtained from Quattro Premier XE; LOQ 1: data obtained from Quattro Premier XE; LOQ 2: data obtained from Xevo TQ-S. 746 Biomed Environ Sci, 2015; 28(10): 738-750 unknown, the concentrations obtained by each the human milk. Lactose and protein was the major analysis with the samples were quite stable. The components of human milk. Through comparison analyzed results of pooled human milk sample in with response of standards in 7% lactose added 1.5% different days should be relatively fixed at one point. human serum albumin and water, except nicotinic The accuracy of pooled human milk was expressed acid, standards of other compounds in two matrixes as the ratio of determined concentration to mean. had the same effects. Because of using standards in The range of accuracy was between 80.3% and water matrix, the determined data of nicotinic acid 120.6% with the inter-day RSD 8.7%-14.5% for multiplied by the ratio of standards response in pooled human milk (Table 4). The good accuracy and lactose to response in water (0.21) was actual data precision of methods were demonstrated again by of nicotinic acid. the results of analysis of the quality control sample. Stability The pooled human mature milk Matrix Effects Matrix effects are critical for samples were stored at -80 °C in amber tubes and analysis of UPLC-MS/MS. The possible effects were analyzed over three months (Figure 3). The thiamin, considered in our study and were shown in Table 2. riboflavin, pantothenic acid and pyridoxal were We consider that, if the value was in the range of ± relatively stable (RSD=9.6%-14.4%), nicotinamide, 15%, the matrix effect could be ignored; if the value nicotinic acid, pyridoxine, FAD revealed a greater was lower than -15%, it could show matrix dispersion degree (RSD=25.5%, 20.4%, 26.8%, and suppression effect[32,36-38]. No significant matrix 21.0% respectively), this meant that these vitamins effects (85.1%-109.6%) were observed for thiamin, were less lusty than others and need more accurate pyridoxine, pyridoxal, pantothenic acid, riboflavin, operation. The trends were similar to literature[10]. nicotinamide, pyridoxamine and biotin. Obvious Matrix Effects Matrix effects are critical for matrix effects were found for FAD (54.1%-63.2%) analysis of UPLC-MS/MS. The possible effects were and nicotinic acid (16.6%-22.9%) at three considered in our study and were shown in Table 2. concentrations. Because of simultaneous analysis of We consider that, if the value was in the range of ± more B-vitamins, determined data divided by the 15%, the matrix effect could be ignored; if the value process efficiency of FAD (44.69%) was actual data of was lower than -15%, it could show matrix FAD. suppression effect[32,36-38]. No significant matrix Matrix effects may affect the ionization of the effects (85.1%-109.6%) were observed for thiamin, target compounds, appearing the inhibition or pyridoxine, pyridoxal, pantothenic acid, riboflavin, enhancement of the response compared with the nicotinamide, pyridoxamine and biotin. Obvious standards in solvents by the complex components in matrix effects were found for FAD (54.1%-63.2%) the samples, such as proteins or sugars[32,37-38]. The and nicotinic acid (16.6%-22.9%) at three pretreatment of milk samples included protein concentrations. Because of simultaneous analysis of precipitation and hydrophobic molecules removing, more B-vitamins, determined data divided by the so the matrix effects were considered. The human process efficiency of FAD (44.69%) was actual data of milk is very complex, so it is impossible to simulate FAD.

Table 4. The Accuracy of NIST SRM 1849a and Pooled Human Milk

Theoretical Determined Inter-day Mean of Accuracy of Inter-day Accuracy Analyte Concentrations of Data of NIST RSD (%) Pooled Pooled Milk RSD (%) of NIST (%) NIST (µg/L) (µg/L） (n=36) milk (µg/L) (%) (n=36) 125.7 Thiamin 135.8±12.1 108.0 8.9 33.2 84.9-118.9 9.8 (106.1-145.3) 203.7 Riboflavin 208.4±13.2 102.3 6.4 64.3 85.2-114.2 10.3 (193.3-214.1) 1090 Nicotinamide 1137.6±117.3 104.4 10.3 220.8 85.7-113.2 9.2 (890-1290) 134.6 Pyridoxine 139.3±9.1 103.5 6.5 3.3 80.3-119.7 12.2 (116-153.2) 19.9 Biotin 19.6±1.3 98.4 6.6 (17.3-22.5) 682 Pantothenic acid 700.9±58.4 102.8 8.3 910.7 82.8-120.6 14.5 (644-720) Pyridoxal 62.6 81.0-111.5 8.7 Nicotinic acid 183.3 85.4-115.9 11.4 FAD 205.1 84.2-118.0 9.9 B-vitamins in human milk by UPLC-MS/MS 747

Application to Actual Field Human Milk Samples FAD, and nicotinamide can be fully obtained using the optimized UPLC-MS/MS method. This meant This method was applied to analyze 183 human that this high-throughput quantitative method was milk samples collected from Guangzhou of China suitable to analyze the free forms of ten B-vitamins (Table 5). Analysis of variance for multiple compari- in human milk. sons was performed using SPSS statistical software. [10] As reported in reference , both pyridoxine and CONCLUSION pyridoxamine was not detected in some human milk samples. However, in analysis of 183 human milk A high-throughput method for fast samples, pyridoxine and pyridoxamine can be quantification of thiamin, riboflavin, pyridoxal, detected in nearly 60% samples and can be analyzed pyridoxine, pyridoxamine, nicotinamide, FAD, in 40% samples. Nicotinic acid can be detected in nicotinic acid, biotin and pantothenic acid has been nearly 75% samples and can be analyzed in 65% firstly developed and validated, and found to be samples. Pantothenic acid can be analyzed in all accurate and precise for the analysis of B-vitamins in human milk samples. The analysis of these new human milk. The method has been applied components contributed to establish the data base successfully, for the first time, to simultaneously of human milk and direct the diet for lactating analyze ten B-vitamins in human milk. Acquiring the women. content of ten B-vitamins of human milk could In general, the concentrations of thiamin, helpfully establish the data base of human milk and pyridoxine, pyridoxamine and pyridoxal were direct the diet for lactating women. relatively stable, even though there were significant differences among individual groups. There was no ACKNOWLEDGMENTS significant difference in the contents of biotin of different lactational stage (P>0.05). There were The authors thank Daniela Hampel, Lindsay H. significant differences in the contents of riboflavin, Allen, and ZHANG Shuang Qing for their support in nicotinamide, nicotinic acid, FAD and pantothenic this work. The authors would like to acknowledge acid from each other of most groups (P<0.05). The volunteers who supplied with human milk samples in concentrations of analyzed B-vitamins were in Guangdong province and Abbott laboratory accordance with literature[9-11,14-18]. (Singapore) who donated the National institute of Through analyzing the actual human milk standards and technology (NIST) Standard Reference samples, thiamin, riboflavin, nicotinic acid, Material (SRM) 1849a Infant/Adult Formula for their pyridoxine, pyridoxal, pyridoxamine, pantothenic acid, support and assistance in this work.

Figure 3. Dispersion degree (%) of breast milk for three months (n=50). 748 Biomed Environ Sci, 2015; 28(10): 738-750

B-vitamins in human milk by UPLC-MS/MS 749

Received: April 28, 2015; 18. Bohm V, Peiker G, Starker A, et al. Vitamin B1, B2, A and E and Accepted: August 7, 2015 beta-carotene content in transitional breast milk and comparative studies in maternal and umbilical cord blood. Z REFERENCES Ernahrungswiss, 1997; 36, 214-9. 19. Tao BH, Huang T, Lai SY, et al. Determination of thiamin, riboflavin, nicotinamide, pantothenic acid and pyridoxal in 1. Elmadfa I, Meyer AL. Vitamins for the first 1000 days: human milk by UPLC-MS/MS. Journal of Food Safety and preparing for life. Int J Vitam Nutr Res, 2012; 82, 342-7. Quality, 2014; 5, 2087-94. (In Chinese) 2. Bowman BA, Russell RM. Present knowledge in nutrition. China: 20. Morrison LA, Driskell JA. Quantities of B6 vitamers in human People's Medical Publishing House, 1998; 236. milk by high-performance liquid chromatography. Influence of 3. Chinese Nutrition Society. Chinese DRIs. Beijing, Science Press. maternal vitamin B6 status. J Chromatogr, 1985; 337, 249-58. 2000; 356. (In Chinese) 21. Yang Z, Yang X, Wu Y, et al. Clinical diagnosis, treatment, and 4. Stamm RA, Houghton LA. Nutrient intake values for folate ALDH7A1 mutations in pyridoxine-dependent epilepsy in three during pregnancy and lactation vary widely around the world. Chinese infants. PLoS One, 2014; 9, e92803. Nutrients, 2013; 5, 3920-47. 22. Jain-Ghai S, Mishra N, Hahn C, et al. Fetal onset 5. Powers HJ. Riboflavin (vitamin B-2) and health. Am J Clin Nutr, ventriculomegaly and subependymal cysts in a pyridoxine 2003; 77, 1352-60. dependent epilepsy patient. Pediatrics, 2014; 133, e1092-6. 6. Ornoy A, Tekuzener E, Braun T, et al. Lack of severe long-term 23. Badoe EV. Pyridoxine-dependent epilepsy: an often outcomes of acute, subclinical B1 deficiency in 216 children in unrecognized but treatable cause of intractable seizures: case Israel exposed in early infancy. Pediatr Res, 2013; 73, 111-9. report from Korle Bu Teaching Hospital, Accra, Ghana. West 7. Ooylan LM, Hart S, Porter KB, et al. Vitamin B-6 content of Afr J Med, 2013; 32, 231-4. breast milk and neonatal behavioral functioning. J Am Diet 24. Song WO, Chan GM, Wyse BW, et al. Effect of pantothenic acid Assoc, 2002; 102, 1433-8. status on the content of the vitamin in human milk. Am J Clin 8. Oguz SS, Ergenekon E, Tumer L, et al. A rare case of severe Nutr, 1984; 40, 317-24. lactic acidosis in a preterm infant: lack of thiamine during total 25. Boaz SM, Champagne CD, Fowler MA, et al. Water-soluble parenteral nutrition. J Pediatr Endocrinol Metab, 2011; 24, vitamin homeostasis in fasting northern elephant seals 843-5. (Mirounga angustirostris) measured by metabolomics analysis 9. Sakurai T, Furukawa M, Asoh M, et al. Fat-soluble and and standard methods. Comp Biochem Physiol A Mol Integr water-soluble vitamin contents of breast milk from Japanese Physiol, 2012; 161, 114-21. women. J Nutr Sci Vitaminol (Tokyo), 2005; 51, 239-47. 26. Andrieux P, Fontannaz P, Kilinc T, et al. Pantothenic acid 10. Hampel D, York ER, Allen LH. Ultra-performance liquid (vitamin B5) in fortified foods: comparison of a novel chromatography tandem mass-spectrometry (UPLC-MS/MS) ultra-performance liquid chromatography-tandem mass for the rapid, simultaneous analysis of thiamin, riboflavin, spectrometry method and a microbiological assay (AOAC flavin adenine dinucleotide, nicotinamide and pyridoxal in Official Method 992.07). J AOAC Int, 2012; 95, 143-8. human milk. J Chromatogr B, 2012; 903, 7-13. 27. Widness JA, Nalbant D, Matthews NI, et al. Tracking donor RBC 11. Yagi T, Iwamoto S, Mizuseki R, et al. Contents of all forms of survival in premature infants: agreement of multiple vitamin B6, pyridoxine-beta-glucoside and 4-pyridoxic acid in populations of biotin-labeled RBCs with Kidd mature milk of Japanese women according to antigen-mismatched RBCs. Pediatr Res, 2013; 74, 689-97. 4-pyridoxolactone-conversion high performance liquid 28. Mock DM, Mock NI, Stratton SL. Concentrations of biotin chromatography. J Nutr Sci Vitaminol (Tokyo), 2013; 59, 9-15. metabolites in human milk. J Pediatr, 1997; 131, 456-8. 12. Tamura T, Picciano MF. Folate determination in human milk. J 29. Alves MN, Zanchetti G, Piccinotti A, et al. Determination of Nutr Sci Vitaminol (Tokyo), 2006; 52, 161. cocaine and metabolites in hair by column-switching 13. Lu B, Ren Y, Huang B, et al. Simultaneous determination of four LC-MS-MS analysis. Anal Bioanal Chem, 2013; 405, 6299-306. water-soluble vitamins in fortified infant foods by 30. Zanchetti G, Floris I, Piccinotti A, et al. Rapid and robust ultra-performance liquid chromatography coupled with triple confirmation and quantification of quadrupole mass spectrometry. J Chromatogr Sci, 2008; 46, 11-nor-Delta9-tetrahydrocannabinol-9-carboxylic acid 225-32. (THC-COOH) in urine by column switching LC-MS-MS analysis. J 14. Hampel D, Shahab-Ferdows S, Domek JM, et al. Competitive Mass Spectrom, 2012; 47, 124-30. chemiluminescent enzyme immunoassay for vitamin B12 31. Zhao L, Sun X, Xiong Z, et al. Simultaneous determination of analysis in human milk. Food Chem, 2014; 153, 60-5. mosapride and its active des-p-fluorobenzyl and 4'-N-oxide 15. Hamaker B, Kirksey A, Ekanayake A, et al. Analysis of B-6 metabolites in rat plasma using UPLC-MS/MS: An application vitamers in human milk by reverse-phase liquid for a pharmacokinetic study. Talanta, 2015; 137, 130-5. chromatography. Am J Clin Nutr, 1985; 42, 650-5. 32. Ansermot N, Brawand-Amey M, Kottelat A, et al. Fast 16. Ford JE, Zechalko A, Murphy J, et al. Comparison of the B quantification of ten psychotropic drugs and metabolites in vitamin composition of milk from mothers of preterm and human plasma by ultra-high performance liquid term babies. Arch Dis Child, 1983; 58, 367-72. chromatography tandem mass spectrometry for therapeutic 17. Sandberg DP, Begley JA, Hall CA. The content, binding, and drug monitoring. J Chromatogr A, 2013; 1292, 160-72. forms of vitamin B12 in milk. Am J Clin Nutr, 1981; 34, 33. Taylor PJ. Matrix effects: the Achilles heel of quantitative 1717-24. 750 Biomed Environ Sci, 2015; 28(10): 738-750

high-performance liquid chromatography-electrospray-tandem 37. Zhu Y, Liu X, Xu J, et al. Simultaneous determination of mass spectrometry. Clin Biochem, 2005; 38, 328-34. spirotetramat and its four metabolites in fruits and vegetables 34. Avery MJ. Quantitative characterization of differential ion using a modified quick, easy, cheap, effective, rugged, and safe suppression on liquid chromatography/atmospheric pressure method and liquid chromatography/tandem mass ionization mass spectrometric bioanalytical methods. Rapid spectrometry. J Chromatogr A, 2013; 1299, 71-7. Commun Mass Spectrom, 2003; 17, 197-201. 38. Wu X, Xu J, Dong F, et al. Simultaneous determination of 35. Commission E. Commission Decision of 12 August 2002 oxathiapiprolin and two metabolites in fruits, vegetables and implementing Council Directive 96/23/EC concerning the cereal using a modified quick, easy, cheap, effective, rugged, performance of analytical methods and the interpretation of and safe method and liquid chromatography coupled to results. Official Journal of the European Communities, 2002; tandem mass spectrometry. J Chromatogr A, 2014; 1329, 30-7. 17.8, 8-36. 39. Matuszewski BK, Constanzer ML, Chavez-Eng CM. Strategies 36. European Medicine Agency, http://www.ema.europa.eu/docs/ for the assessment of matrix effect in quantitative en_GB/document_library/Scientific_guideline/2011/08/WC50 bioanalytical methods based on HPLC-MS/MS. Anal Chem, 0109686.pdf (acessed at 14.01.13) 2003; 75, 3019-30.