Professional Documents
Culture Documents
PII: S0308-8146(15)30079-0
DOI: http://dx.doi.org/10.1016/j.foodchem.2015.10.077
Reference: FOCH 18272
Please cite this article as: Woollard, D.C., Bensch, A., Indyk, H., McMahon, A., Determination of Vitamin A and
Vitamin E Esters in Infant Formulae and Fortified Milk Powders by HPLC: Use of Internal Standardisation, Food
Chemistry (2015), doi: http://dx.doi.org/10.1016/j.foodchem.2015.10.077
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
1 Determination of Vitamin A and Vitamin E Esters in Infant Formulae and
5
a
6 R.J. Hill Laboratories Ltd, 1, Clyde Street, Hamilton 3216, New Zealand
b
7 New Zealand Laboratory Services, Auckland, New Zealand
c
8 Fonterra Cooperative Group Ltd, Waitoa, New Zealand
d
9 Wyeth Nutrition, Askeaton, County Limerick, Republic of Ireland
10
13
14 Abstract
15
16 An HPLC method is described using normal phase conditions with an unbonded silica
19 fluoresence for vitamins A and E and visible absorbance for β-carotene. An attribute of
20 the method is the use of retinol propionate, α-tocopheryl propionate and all-E-β-apo-8’-
21 carotenoic acid ethyl ester internal standards to compensate for analytical variations
23 with the aid of protease to remove vitamin encaspsulation and facilitate vitamin partition
24 into hydrocarbon solvent. Figures of merit indicate the method is suitable for its
25 intended purpose in the highly regulated infant formula environment, including liquid
26 fomulations. The method is extendable to whole milk powders where total vitamin A
27 content data can be calculated by summing the innate long-chain vitamin A esters with
30
31 Methods for the determination of fat-soluble vitamins are widely reported and
32 summarized in many authoritative reviews (De Leenheer, Lambert & Van Bocxlaer,
33 2000; Rucker, Suttie, McCormick & Machlin L, 2001; Eitenmiller, Landen & Ye, 2008).
34 In general, and except for vitamin K, sample preparation for foods including milk
35 products commence with alkaline digestion, where the matrix is disrupted to liberate
37 Silvera, 2002 Eitenmiller et al., 2008; Gentili et al., 2013). However, where the nature of
38 the food allows simplified extraction, the lengthy saponification procedure is avoided to
39 minimise oxidative and manipulative losses of the target analyte. Such alternative
40 methods have been reported for vitamin A in liquid milk (Hite, 2003) and milk powders
41 (Woollard & Woollard, 1984) using total fat extraction and subsequent normal phase
42 liquid chromatography.
44 of long-chain fatty-acid retinyl esters (Woollard & Indyk, 1989), whereas the
45 supplementary forms are invariably retinyl palmitate or retinyl acetate, both available
48 milk almost exclusively as d-α-tocopherol, albeit in low concentration, while the principal
51 succinate and α-tocopheryl nicotinate but neither is commonly used for infant nutrition.
52 Vegetable oils contain significant amounts of non-esterified d-α- and other tocopherols
56 interested in both total vitamin content and concentration of the supplementary forms
57 must select methods that avoid saponification, such as those described by Woollard &
58 Blott (1986), Woollard, Blott & Indyk (1987), Hewavitharana, van Brakel & Harnett M.
59 (1996), Chase, Ye, Stoakes, Eitenmiller & Long (2004) or Chávez-Servín, Castellote &
61 (Thompson et al., 2013) and 2012.10 (McMahon, Christiansen, Shine, Loi & Dowell,
62 2013), have focused on normal-phase separation of total lipid extracted from infant and
63 adult formulations. The former uses on-line column switching between dual 3 µm silica
64 columns to create a clean extract for concurrent UV detection (UVD) of retinyl palmitate,
65 retinyl acetate, vitamin E acetate and vitamin E alcohol. The latter procedure employed
66 selective use of UVD at 325 nm and fluorescence detection (FLD) to create efficient
68 which allowed gradient elution of co-extracted fats. This method incorporated a papain
70 Following the first reports of fluorescence for vitamin E acetate (Woollard, Blott &
71 Indyk, 1987), the present study improves previous methodology for the quantitation of
72 vitamin A and E esters using modern fluorescence detectors and a 3 µm silica column
73 for efficient separations, coupled with analyte quantitation by the internal standard
74 technique. The use of internal standards for vitamins A, E and β-carotene has been
76 serum (Woollard & Woollard, 1984; Nierenberg & Nann, 1992; Epler, Ziegler & Kraft,
77 1993; Aust, Sies, Stahl & Polidori, 2001: Liu Z., Lee H-J., Garofalo F., Jenkins D.J.A. &
78 El-Sohemy A. 2011; Kand’ár et al., 2013) and for cosmetics (Baruffini et al., 1992) and is
79 similarly applied here where surrogate materials for each measurand improved relative
82 protease was incorporated in the method to efficiently liberate the embedded vitamins
83 from their encapsulating materials. The described method also facilitated an estimation
84 of total vitamin A content through summation of the added and natural esters.
85
86
87 2. Material and Methods
88
89 2.1 Chromatography
91 LC20AT quaternary pump and SIL-20AHT autosampler was operated in normal phase
92 mode with an Alltech Econosphere 3 µm silica column (150 x 4.6 mm, part number
95 Channel 1 of the fluorescence detector was set at 325 nm (excitation) and 450 nm
96 (emission) with medium sensitivity (x4 gain) to monitor the retinyl esters, while channel
97 2 was set at 280 nm (excitation) and 330 nm (emission) with medium sensitivity (x4
98 gain) to detect α-tocopheryl acetate. The diode-array detector was used at 450 nm to
99 monitor β-carotene signals but also collected data at other wavelengths, notably 325
100 nm for an alternative vitamin A option. Typical flow-rate of 1.0 mL/ min was used with
103 Other equipment was proven suitable by repeating analyses on an Agilent 1100
104 (Santa Clara, CA, USA) with a single channel programmable fluorescence detector
105 (FLD) and diode-array detector. The configuration was either i) FLD at 280 nm / 330 nm
107 325 nm (retinyl esters) and 450 nm (β-carotene) or, ii) programmed FLD wavelength
108 initially at 325 nm / 450 nm for retinyl esters, changed prior to α-tocopheryl acetate
110
112 Varian Cary 50 UV-vis. spectrophotometer (Palo Alto, CA, USA) with 1cm quartz
113 cuvettes was used to calibrate retinyl propionate (325 nm), tocopherol propionate (284
114 nm) and apo-carotenoic acid ethyl ester (450 nm). Protease incubation was performed
115 at 50 ±3 oC in a Grant water bath with GD100 controller (Chelmsford, UK). Sample
116 clarification was performed in a Gerber centrifuge (Funke-Gerber, Berlin, Germany) at
117 1000 revs/min (~ 350 rcf) and buckets to contain 50 mL polypropylene Centristar®
118 centrifuge tubes (Corning, NY, USA). pH control was done with a Metrohm 867 (Herisau.
119 Switzerland).
120
122 Thermostable protease (EC 3.4.21.14), subtilisin A from Bacillus licheniformis was
123 sourced from Megazyme (Bray, County Wicklow, Eire), product code E-BSPRT in 50% glycerol,
125 Tris.HCl buffer (1 M) was prepared by dissolving 121 g tris base (Sigma Aldrich, St Louis,
126 MO, USA) in about 800 mL water, then bringing to pH 8.0 with drop-wise addition of
127 concentrated hydrochloric acid. This was made to 1L and diluted to 0.1 M for use.
128 Solvents used were HPLC-grade hexane, ethanol and isopropanol (IPA) from Fisher
129 Scientific (Loughborough, UK). An extraction solution of 50% ethanol in hexane was prepared.
130 To prepare the mobile phase, various volumes of IPA were tested to achieve optimum separation
131 of vitamin A and E esters, the preferred being 600 µL in 1 L hexane. The mobile phase was
132 filtered through anhydrous sodium sulphate (Scharlau, Barcelona, Spain) on a 0.45 µm
133 membrane filter with standard vacuum apparatus thus minimising its humidity and stabilising
134 retention times. The mobile phase was continuously recycled to reduce reagent costs and
136 Saturated sodium chloride was prepared by adding 320 g sodium chloride (Merck,
138 undissolved.
139 Vitamin A propionate (2,500,000 IU/g) was sourced from BASF (part 50096515,
140 Basel, Switzerland) and stored in a freezer under nitrogen gas. A stock solution of about
141 625 IU/mL was prepared by dissolving 25 mg in 100 mL isopropanol (IPA) containing 100
142 – 200 mg butyrated hydroxytoluene crystals (BHT, Sigma Aldrich, St Louis, MO, USA) to
143 minimise vitamin degradation. This was stored in the dark in a freezer (-20 °C) for up to
144 six months and used as internal standard for vitamin A palmitate and vitamin A acetate.
145 It was calibrated, using equation 1, by dilution of 100 µL to 10 mL with subsequent
146 absorbance measurement against IPA at 325 nm. Recalibration was performed on a
147 weekly basis. Calculations were converted to IU/mL to permit unitary (1.00) response
148 factors no matter which retinyl esters were being measured, including the natural long-
150
154 where
160
161 Retinyl palmitate and retinyl acetate were obtained from Sigma-Aldrich (St Louis,
162 MO, USA), containing 1,750,000 IU/g and 2,800,000 IU/g, respectively, and prepared in
164 α-Tocopheryl propionate, internal standard for α-tocopheryl acetate, was not
166 Bonrath and Cirillo (2002): 2.5 g (equivalent to 5.8 mmol) of α-tocopherol (Sigma-
167 Aldrich) was weighed into a 50 mL round-bottomed flask containing 1.5 g (11.5 mmol, bp
168 167 oC) of propionic anhydride (Merck & Co., NJ, USA). 200 µL (2.5 mmol, bp 115 oC)
169 of pyridine (Merck & Co) was added as catalyst and a water-cooled condenser fitted.
170 The flask and condenser were held in a boiling water for 3 hr covered with aluminium foil
171 protection from light. The product was cooled and 20 mL water added. The stoppered
172 flask was shaken and left overnight to hydrolyse excess propionic anhydride to propionic
173 acid. The lower aqueous phase was withdrawn and the oily propionate ester washed
174 twice with 20 mL water. The flask was then left under rotary evaporation (Buchi R3 with
175 V700 pump and F105 chiller, Flawil, Switzerland) at 70 °C under 80 mbar vacuum until
176 dry. The vitamin E propionate was stored in the flask under nitrogen gas at -20 oC until
177 required and was stable for > 1 year under such conditions.
178 A stock solution of α-tocopheryl propionate was prepared by dissolving about 250
179 mg in 100 mL IPA containing 100 – 200 mg BHT, to yield nominally 2.5 mg/mL, and
180 stored for up to six months at -20 oC. It was calibrated, using equation 2, after dilution
181 of 100 µL to 10 mL with IPA and subsequent absorbance measurement against IPA at
182 284 nm. Recalibration was performed on a weekly basis to guard against degradation.
185 where
190
191 Vitamin E acetate (Sigma-Aldrich) solution was prepared at about the same
193 Pure all-E-β-apo-8’-carotenoic acid ethyl ester (CAS 1109-11-1; CI 40825; E160f),
194 abbreviated here as β-apo-ester, was gifted in vials from DSM (Heerlen, Netherlands),
195 and stored in a freezer until opened. A stock solution of about 1000 µg/mL was
197 200 mg BHT to minimise oxidation, and stored for up to one month at -20 oC. It was
199 measurement against IPA at 449 nm using equation 3. Recalibration was performed
201
209
214 standards, with no further dilution. These ratios suited most applications but were
215 varied as required, taking care to monitor the exact amount of each internal standard
221
223 20.0 g of milk powder or infant formula was accurately weighed into a 200 mL
224 beaker and 80.0 g warm water (~ 40 oC) added. The mixture was stirred to dissolve the
225 powder and 5.0 g of reconstituted sample (equivalent to 1.0 g powder) weighed into a 50
227 formulations, 5.0 g was used and treated the same way.
228 5 mL of 0.1 M Tris.HCl buffer, pH 8 was added to the centrifuge tube followed by
230 3 oC for 30 min with occasional mixing. After cooling, 200 µL of mixed internal standard
231 solution was accurately added, followed by 20 mL hexane : ethanol mixture (50:50) and
232 mixed for one minute by vortex or shaking. The extraction was completed by adding 5
233 mL saturated sodium chloride and vortex mixing or shaking again for one minute.
234 Phase separation was continued by standing in the dark for 30 – 60 min until
235 complete, otherwise centrifuged for 10 min at ~350 rcf. The upper hexane layer was
236 filtered through a 0.45 µm nylon filter and 2 mL collected in an amber crimp-topped
238
240 Sample extracts (10 µL) were injected into the HPLC with the eluent monitored
241 simultaneously by FLD and UVD. FLD gain settings were adjusted as required to keep
243 The usual flow-rate was 1.0 mL/min with the 3 µm, 150 x 4.6 mm ID, column,
244 kept at 25 oC, the low viscosity of hexane producing minimal back-pressure. The flow
245 was raised to 2 mL/ min after the elution of the last peak of interest (α-tocopheryl
246 acetate) and continued for the duration of the chromatogram to elute lipids in readiness
247 for the next injection. Mobile phase recycling was implemented in order to conserve
248 solvent.
249
251 The concentrations of each internal standard in the mixed standard were calculated from their
252 concentrations, as determined by equations 1 - 3 (in IU/mL, mg/mL or µg/mL, as stated), multiplied by
253 their proportions in the mixed standard (0.40 for vitamins A and E or 0.20 for β-apo-Ester). The
254 amount of each internal standard added to the samples was calculated by multiplication with 0.2 mL,
256 This data was used to calculate the vitamin concentrations in reconstituted and liquid samples
258
259 Vitamin A esters, IU/100 g = ASMP x A propionate added (IU) x 100 x recon (4)
260 AIS x W (g) x RF
261
262
263 Vitamin E acetate, mg/100 g = ASMP x E propionate added (mg) x 100 x recon (5)
264 AIS x W (g) x RF
265
266
267 β-Carotene, µg/100 g = ASMP x apo-ester added (µg) x 100 x recon (6)
268 AIS x W (g) x RF
269
270 where
274 recon = reconstitution factor for powdered samples = 20 (1 for liquid samples)
277
278 Response factors used were 1.000 for vitamin A esters, the units (IU) being independent of
279 molecular weights. Response factors, as divisors, of 0.971 were applied to vitamin E acetate and
280 0.985 for β-carotene respectively, based upon literature E1% values in hexane (FAO, 1991; JECFA,
281 2000). Experimentation proved these response factors were valid during the current study.
282 To express vitamin A in weight units, IU/100 g was multiplied by 0.3 to give ug/100 g retinol
283 equivalents. In the case of vitamin E acetate, IU/100 g and mg/100 g are equivalent for synthetic
284 forms.
285
287 The method was subjected to normal intra-laboratory challenges of precision and
288 accuracy. Repetitive testing of a range of commercial infant formulations yielded relative
289 standard deviations (RSD) that were compared for acceptability using the Horwitz
290 function. Certified Reference Materials (CRM) were used where applicable but were
291 unsuitable in the case of vitamin E because certificates reported total tocopherol, rather
292 than the α-tocopheryl acetate entity. Spiked recovery data were generated for each of
294
296
298 Manufacturers of infant formulae increasingly require data about the level of
299 supplementary vitamins added during the fortification processes. Where these differ
300 structurally from the endogenous forms, and where the analytical process does not
301 modify analyte structure, this aim is possible. Thus, an unambiguous quantitation is
302 possible for the acetate esters of vitamin A and E. In the case of retinyl palmitate and β-
303 carotene, both are innate to milkfat and potentially exist in infant formulae, but
304 estimations of the additive concentrations remain feasible because most infant formulae
305 are milkfat-free and have minimal endogenous levels of either compound.
306 Although not the main purpose of this study, the near absence of retinol itself in
307 milk meant the described method allowed a determination of total vitamin A by
308 summation of all the esters present. Vitamin A exists as a collection of medium- and
309 long-chain esters, in a similar pattern to triglyceride fatty acids (Woollard and Indyk,
310 1989). The identity and variable molecular weight of each ester is not important when
312 units, typically µg/100 g retinol equivalents, were then performed at the conclusion of the
313 calculations.
314 Total vitamin E determinations were not possible because of the presence of
315 considerable free α-tocopherol (and other tocopherol congeners) from those vegetable
316 oils incorporated to replace milkfat, with an estimation of added α-tocopheryl acetate
317 significantly lower than total vitamin E content. Chavez-Servın et al, (2006) used an
318 amine-column to detect α-tocopherol and its congeners using solvent gradients, a
319 strategy not possible in this study where a silica column was used.
320
321 3.2 Selection of internal standards
322 The use of internal standards for vitamins A, E and β-carotene is not new in
323 clinical chemistry where the acetate esters are used for retinol and α-tocopherol in
324 serum (Kand’ár et al., 2013). In the absence of saponification, other esters of vitamins A
325 and E make ideal surrogates for the measurands and when added early in the extraction
326 process, compensate for oxidative and manipulative losses of the labile analytes, while
328 The propionate esters of retinol and α-tocopherol were well resolved with respect
329 to retinyl acetate, retinyl palmitate and α-tocopheryl acetate, thereby supporting their use
330 as internal standards. Retinyl propionate was commercially available, but α-tocopheryl
332 Borath & Cirillo (2002) involving the pyridine-catalysed reaction of α-tocopherol with
333 propionic anhydride. 3 h heating at 100 oC was used rather than using a microwave
334 reactor for 30 min but esterification was complete as demonstrated by the absence of
335 free α-tocopherol during chromatography. An extra water washing step was used to
336 ensure all propionic anhydride was removed. A similar esterification has been reliably
337 demonstrated in the compendial analysis of tocopherol food additives by GC-FID where
339 Commercially available vitamin E esters (succinate and nicotinate) had poor
340 chromatographic properties on silica due to their polar nature so could not be considered
342 Non-polar β-carotene elutes near the void volume on normal-phase columns,
344 Many candidate xanthophylls were expensive or highly retained on silica columns which
345 reduced the number of practical choices. β-apo-8’-carotenoic acid ethyl ester, all-E
346 (trans) isomer, is manufactured in large quantity as a feed additive for egg colouration so
347 was chosen for further study, although all-E-β-apo-8’-carotenal (CAS 1107-26-2),
350 moiety does not influence the chromophoric conjugated structure, although their weight-
351 based coefficients (E1%) are a function of molecular weight. For simplicity, calculations
352 of vitamin A concentrations were kept in international units (IU), where, like ε, all forms
353 are equal. A response factor of 1.00 was therefore applicable independent of which
354 supplementary ester (acetate or palmitate) or natural ester was being calibrated by the
356 Similarly, α-tocopheryl acetate, the sole measurand for vitamin E, and α-
357 tocopheryl propionate have same molar extinction coefficient. Calculations were in
358 weight units so a response factor corresponding to their molecular weights ratio (0.971)
359 was applied. This response factor was confirmed experimentally from standard
360 solutions, proving the extra carbon in the side chain has no measurable impact on
361 spectral absorptivity. Additionally, the stereoisomeric variation of the α-tocopherol moiety
362 was ignored, since a single chromatographic α-tocopheryl acetate peak was observed
365 Despite the disrupted β-ionone ring of β-apo-ester and lower molecular weight of
366 this C30 carotenoid (460.7 mol/g) compared to C40 β-carotene (536.9 mol /g), it has a
367 similar reported E1% (2550) as β-carotene (2590) in hexane yielding a theoretical
368 response factor of 0.985. There is a small hypsochromic shift of the principle peak in
369 hexane of β-apo-ester (449 nm) compared with β-carotene (453 nm), the net result of
370 one fewer conjugated bond and an extra carbonyl (Schwieter et al., 1969). This had no
372 standards thus confirming the theoretical response factor could be reliably used.
373
375 Studies performed in the 1980’s reported a useful method for testing vitamins A
376 and E in dairy powders using a fast extraction of the fat in a single tube, followed by
377 HPLC with selective fluorescence detection (Woollard & Indyk, 1986; Woollard & Indyk,
378 1989). The current method followed a similar strategy but used advances in HPLC
379 instrumentation and internal standardisation to improve data quality. The original
380 method, exploiting extraction with dimethysulphoxide (DMSO), was modified in the
381 present study with protease digestion to facilitate release of micro-encapsulated vitamins
382 that are increasingly present in modern dry-blended products. The hydrophobic fat-
384 gelatin or gum arabic polysaccharide and therefore required effective release into
385 solution prior to extraction and detection. Gum arabic has a peptide backbone that is
389 weights as required. Sample sizes up to 50 g were trialled with concomitant increases in
390 volumes, internal standards and protease but no improvement in precision was noted
393 supplemental vitamins added pre-evaporation and therefore have no remaining intact
395 were treated identically to dry-blended products. Besides, protease digestion has a
396 secondary purpose in digesting milk proteins that potentially interfere with phase
397 separation of subsequent hydrocarbon extractions. Similarly, UHT liquid milks and
398 ready-to-drink infant formulations were treated with protease, although vitamin
400 The choice of proteolytic enzyme was made based upon the well established
401 properties of alkaline proteases, also called proteinases, from bacillus licheniformis
402 (Ferrero, Castro, Abate, Baigorı & Sińeriz, 1996) and previous knowledge of the dietary
403 fibre test where complete removal of protein is mandatory. Serine protease, Subtilisin A,
404 from Bacillus licheniformis sourced from Megazyme proved to be a convenient and
405 stable protease, although papain was used in AOAC 2012.10 (McMahon et al., 2013).
406 During extraction, a 5 g portion was utilised from the total 100 g digest,
407 equivalent to 1 g of powder. This was extracted into 10 mL hexane, as a 50:50 mixture
408 with ethanol, and salt solution to improve phase separation. Higher boiling solvents,
409 heptane and isooctane, were equally effective extractants, although hexane was easier
410 to reduce in volume for improved sensitivity if required, albeit not routinely needed in this
411 study.
412
414 Unbonded silica columns of various geometry were successfully tested, including
415 5 µm columns, but a 3 µm (150 x 4.6 mm) Alltech column with 80 Å pore size was
416 chosen for subsequent studies. Hewavitharana et al. (1996) successfully used a similar
417 column but with different polar modifiers. Injection volumes were typically 10 µL but
419 The diameter of the selected column was kept to a maximum to allow increased
420 flow-rates for faster lipid clearance. The mobile phase had low viscosity, allowing flow
422 This promoted rapid clearance of triglycerides from the column, creating a more stable
423 chromatographic environment. The work of Thompson, Schimpf & Baugh (2013)
424 overcame the same problem through column switching, thereby diverting lipid to waste
425 from a short pre-column. Chromatography was continued here at elevated flow rate until
426 the relatively polar unesterified α-tocopherol had eluted to prevent it being carried over
427 to the next chromatogram. Late eluting β- and γ-tocopherols did not cause interference
429 Vitamins A and E both have natural fluorescence, and β-carotene a selective 450
430 nm viewing wavelength, so are amenable to simplified detection in the presence of other
431 lipid material. Co-extracted lipids passed through the detectors without response, and
432 exited the column quickly enough to prevent accumulation and associated
434 retinyl esters could be satisfactory detected by UVD at 325 nm with a multi-wavelength
435 detector. However, vitamin E was necessarily detected by fluorescence because UVD
437 It was important to avoid incursion of water onto the silica column with
438 consequently diminished retention times. As part of this process, the mobile phase was
439 passed through anhydrous sodium sulphate desiccant before use, placed above the
440 0.45 um solvent clarification filter. Refiltration through anhydrous sodium sulphate was
441 also an option to maintain low humidity. To conserve solvents, the mobile phase was
442 recycled for a week, otherwise replaced when retention or baseline shifts were noted.
443 Retention times were highly sensitive to the concentration of the polar modifier
444 isopropanol (IPA), which was necessarily pipetted into hexane in an accurate manner. A
445 number of IPA concentrations were tested over the range 0.04 - 0.08%, with 0.06%
446 shown to give good isomer separations on the selected Alltech column. This was
448 separation.
449 Chromatographic selectivity for the analytes was also dependent on IPA
451 increasing IPA until it eluted near the retinyl or α-tocopheryl esters. However, the unique
452 FLD detection modes for vitamins A and E esters overcame any potential
454 absorbed in the visible region at 450 nm. Similarly, at higher IPA concentrations α-
455 tocopheryl acetate also eluted near retinyl propionate, but without spectral interferences.
456 Fig 1 shows the chromatographic output for a typical infant formula extract
457 containing supplementary retinyl palmitate, α-tocopheryl acetate and β-carotene using
458 concurrent dual-channel fluorescence and visible detection. In the latter detection
459 mode, only β-carotene (combined Z (cis) - and E (trans) isomers) and β-apo-ester
460 internal standard are visible. It should be noted that α-carotene, if present, is not
461 separated from β-carotene and calculations will essentially represent total carotenoid. In
462 addition, the xanthophylls lutein, astaxanthin and canthaxanthin did not
463 chromatographically interfere with the non-polar carotenes, although lycopene co-eluted
466 supplementary retinyl acetate and α-tocopheryl acetate, with naturally occurring retinyl
467 esters and β-carotene. When geometric Z-isomer (cis) peaks were present they eluted
468 prior to the corresponding all-E (all-trans) isomer peaks and were included to give a total
469 peak area, but without corrections for their lower biopotency.
470
472 Four commercially available formulations were selected for repeatability studies,
473 each with supplementary α-tocopheryl acetate and β-carotene as judged by label
474 declarations. Two of the formulations contained retinyl palmitate while the other two
475 contained retinyl acetate, the latter also known to be dry-blended. In addition, a
476 vitaminised whole milk powder was repetitively tested for its total vitamin A content. The
477 results are given in Table 1 where repeatability relative standard deviations (RSDr) were
478 observed in the range 3 - 10% across the three vitamins. HorRat values (observed
479 RSD/ calculated RSD from Horwitz equation) are also given that confirm acceptability of
480 the observed RSDr at the observed concentrations. Observed repeatability was equal or
481 better than generally accepted guidance limits for repeatability based on blind duplicates
482 (HorRat 0.3 – 1.3; Horwitz & Albert, 2006). Vitamin A calculations were converted from
484 Two infant formulae were obtained from a New Zealand manufacturer and tested
485 in duplicate (n=2), twice per week for five weeks (d = 10 pairs), both manufactured by
487 acetate and β-carotene. Table 2 summaries the observed intermediate precision
488 (RSDiR) obtained using standard statistics (Chesher, 2008) and associated HorRat,
489 historically expected to lie 0.5 – 1.5 (Horwitz & Albert, 2006). Re-estimation of RSDr was
490 also performed based on the differences (squared) between duplicates which showed
494 To estimate recoveries, each of the three vitamins, prepared in calibrated IPA
495 solution, were spiked at three levels into suitable matrices prior to protease digestion.
496 For vitamin A, suitable blank matrices were available for the palmitate and acetate
497 esters by selection of infant formulae without the targeted analyte. For vitamin E, a
498 whole milk powder without supplementary α-tocopheryl acetate was used, while a fully
499 oil-filled infant formula without fortified carotenoids was used for β-carotene recovery
500 estimates. Table 3 shows 89 -104% recovery with an average 98.5 % across all the
501 vitamins. Each vitamin has a parallel internal standard added to the digest that
502 compensates for potential losses from degradation, volume changes and extraction
503 shortcomings, thereby leaving spectral calibration and dispensing of the internal
505 Two infant formula certified reference materials with vitamin data were available
506 from National Institute of Standards and Technology (NIST). Triplicate (n =3) testing of
507 NIST 1849 and NIST 1849a yielded vitamin A data shown in Table 4. When compared
508 with certified concentrations agreement was good, confirming method accuracy. NIST
509 1849 and NIST 1849a both contained retinyl palmitate ester, with no retinyl acetate, the
511 Direct comparison of vitamin E acetate data with data assigned in NIST 1849
512 was not possible because the latter included total α-tocopherol, most of which was oil-
513 derived and therefore unesterified. However, the NIST 1849 certificate noted that about
514 5 mg /100 g (50 mg/kg) α-tocopherol acetate is expected, which concurs with the
515 current observations as shown in Table 4. NIST 1849a had a certified range for vitamin
516 E acetate.
517 Similarly, a β-carotene value was not assigned in NIST 1849 or NIST 1849a.
518 Although highly variable in NIST 1849 documents, (200 – 1200 µg/100 g, Sharpless et
519 al., 2010), concentrations observed here were 239 µg/100 g (SDr = 35.4 µg/100 g), at
520 the lower end of this range although it is known that β-carotene is particularly unstable.
521 Results obtained for vitamin A and β-carotene in whole milk powder agreed well
522 with published data, although both vitamins are known to be seasonally variable in New
524
525 Conclusion
526
527 In response to the needs of infant formula manufacturers or regulators to determine the
528 quantity and nature of supplementary vitamins A and E esters, a method has been
529 developed using modern HPLC equipment with dual-channel fluorescence detection.
530 Benefits of the method are avoidance of saponification, which allows other esters to be
531 used as internal standards, namely retinyl propionate, and α-tocopheryl propionate. The
532 method was extended to include β-carotene with all-E-β-apo-8’-carotenoic acid ethyl
533 ester as internal standard, although the limitations of its early elution from normal phase
534 columns is acknowledged. The extraction procedure, using simple protease digestion is
535 considerably faster than traditional methods involving alkaline digestion with multiple
536 washing and extractions, and provides adequate precision and recovery. A corollary of
537 the method is the rapid determination of total vitamin A, including natural esters and
538 provitamin A (β-carotene) in milk and milk powders, although the small amount or
541
542 Aust O., Sies H., Stahl, W. & Polidori M. C. (2001). Review. Analysis of lipophilic
543 antioxidants in human serum and tissues: tocopherols and carotenoids. Journal of
545
546 Baruffini A., De Lorenza E., Gandini C., Kitsos M. and Massolini G.(1992). High-
549
550 Bonrath W. & Cirillo F (2002). Process for Preparation of Tocol Acylates and Tocopherol
552
555 performance liquid chromatography- diode array detection using a short narrow-bore
557
558 Chesher D. (2008). Evaluating assay precision. The Clinical Biochemist Reviews, S23 –
559 S26.
560
561 Chase W. G., Ye L., Stoakes V. C., Eitenmiller R. R. & Long A. R. (2004). An
562 interlaboratory-verified method for the determination of vitamins A and E in milk- and
563 soy-based infant formula by liquid chromatography with matrix solid-phase dispersion
565
566 DeVries J.W. & Silvera K.R. (2002). Determination of vitamins A (retinol) and E (alpha-
567 tocopherol) in foods by liquid chromatography: collaborative study. Journal AOAC International,
570 De Leenheer A. P., Lambert W.. E. & Van Bocxlaer J. F. editors (2000). Modern
572 volume 84. Marcel Dekker, New York, Basel. ISBN 0-8247-0316-2
573
574 Eitenmiller R. R., Landen W. O. & Ye L. editors (2008). Vitamin Analysis for the health
575 and food sciences, Second edition. CRC Press, Taylor and Francis Group, Boca Raton,
577
578 Epler K. S., Ziegler R. G. and Kraft N. E. (1993). Liquid chromatographic method for the
579 determination of carotenoids, retinoids and tocopherols in human serum and in food.
581
582 FAO (1991). FAO Food and Nutrition Paper 19. Spefications for Identity and Purity.
583 Food and Agricultural Organisation of the Unitech Nations, Rome, Italy, ISBN 92-5-
584 101126-5
585
586 Ferrero M. A., Castro G. R. Abate C. M., Baigorı M. D. and Sińeriz F. (1996).
587 “Thermostable alkaline proteases of Bacillus licheniformis MIR 29: isolation, production
588 and characterization”. Applied Microbiology and Biotechnology, 45, 327- 332
589
590 Gentili, A., Caretti, F., Bellante, S., Ventura, S., Canepari, S., & Curini, R. (2013).
591 Comprehensive profiling of carotenoids and fat-soluble vitamins in milk from different
592 animal species by LC-DAD-MS/MS hyphenation. J. Agric. Food Chem., 61, 1628-1639.
593
594 Hewavitharana A. K., van Brakel A. S. & Harnett M. (1996). Simultaneous liquid
597
598 Hite D. (2003). Determination of retinyl palmitate (vitamin A) in fortified fluid milk by
599 liquid chromatography: collaborative study. Journal AOAC International, 86, 375 – 385
600
601 Horwitz, W., & Albert, R. (2006). The Horwitz ratio (HorRat): A useful index of method performance
602 with respect to precision. Journal of AOAC International, 89, 1095 - 1109.
603
604 Indyk H., Lawrence R. & Broda D. (1993). The micronutrient content of bovine whole
605 milk powder: influence of pasture feeding and season. Food Chemistry, 46, 389-396
606
608 FAO/WHO Expert Committee on Food Additives. 55Th session, p 121-127, Geneva,
610
611 Kand’ár R., Novotná P. & Drábková (2013). Determination of retinol, α-tocopherol,
612 lycopene, and β-carotene in human plasma using HPLC with UV-Vis detection:
614
615 Liu Z., Lee H-J., Garofalo F., Jenkins D.J.A. & El-Sohemy A. (2011). Simultaneous
619
620 McMahon, A., Christiansen, S., Shine, L., Loi, C., & Dowell, D. (2013). Simultaneous
621 determination of 13-cis and all-trans vitamin A palmitate (retinyl palmitate), vitamin A
622 acetate (retinyl acetate), and total vitamin E (α-tocopherol and DL-α-tocopherol acetate)
623 in infant formula and adult nutritionals by normal phase HPLC: First action 2012.10.
625
626 Nierenberg D. W. & Nann S. (1992). A method for determining concentrations of retinol,
627 tocopherol, and five carotenoids in human plasma and tissue samples. American
629
630 Sharpless K. E., Lindstrom R. M., Nelson B. C., Phinney K. W., Rimmer C. A., Sander L.
631 C., Schantz M. M., Spatz R. O., Thomas D. B., Turk G. C., Wise S. A., Wood L. & Yen
633 Infant/Adult Nutritional Formula. Journal of AOAC International, 93, 1262 – 1274
634
635 Schwieter U., Englert N., Rigassi N. and Vetter W. (1969). Physical Organic Methods in
637
638 Rucker R.B., Suttie J. W., McCormick D. B. & Machlin L. (2001). Handbook of vitamins,
640
641 Thompson, L. B., Schimpf, K & Baugh, S. (2013). Determination of vitamins A and E in
642 infant formula and adult/pediatric nutritional formula by HPLC with UV and fluorescence
643 detection: First action 2012.09. Journal of AOAC International,., 96, 1407-1413.
644
645 Woollard G. A. & Woollard D. C. (1984). The Determination of Serum Retinol by High
648
650 powders using high performance liquid chromatography. New Zealand Journal of Dairy
652
653 Woollard D. C. & H. Indyk H. (1986). The HPLC analysis of vitamin A isomers in dairy
657 Woollard D. C. & Blott A.D. (1986). The routine determination of vitamin E acetate in
658 milk powder formulations using high performance liquid chromatography. Journal of
660
661 Woollard D. C., Blott A.D. & H. Indyk H. (1987). Fluorometric detection of tocopheryl
662 acetate and it’s use in analysis of infant formulae. Journal of Micronutrient Analysis, 3, 1
663 – 14
664
665 Woollard D. C. & H. Indyk H. (1989). The distribution of retinyl esters in milks and milk
667
668
669
670 Fig 1: Chromatograms with dual-channel florescence (plots A and B) and visible
671 detection (plot C) of a typical infant formula extract containing supplementary retinyl
672 palmitate, α-tocopheryl acetate and β-carotene.
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699 Fig 2: Chromatograms with dual-channel fluorescence (plots A and B) and visible
701 retinyl acetate and α-tocopheryl acetate, with naturally occurring retinyl esters and β-
702 carotene.
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728 Table 1: Repeatability studies for vitamin A esters, vitamin E acetate and β-carotene (provitamin A)
729 in manufactured formulations by same-day repetition (n=5)
730
731
Sample Vitamin A Vitamin E β-Carotene
Mean RSDr Mean RSDr Mean RSDr
HorRat HorRat HorRat
ug/100 g % mg/100 g % ug/100 g %
a,c
Formula 1 502 3.09 0.15 5.28 3.95 0.28 118 8.93 0.40
a,d
Formula 2 379 4.56 0.22 7.40 3.38 0.24 81.0 6.36 0.25
b,c
Formula 3 983 5.70 0.33 8.71 3.76 0.28 137 6.60 0.29
b,d
Formula 4 526 3.71 0.18 6.00 2.95 0.20 179 7.16 0.33
WMP 323 5.28 0.25 na na na 141 6.07 0.26
a b c d
Wet-blended; Dry blended; retinyl acetate; retinyl palmitate; na = not applicable
732
733
734
735
736
737 Table 2: Same-day (r) and between-day (iR) precision estimates from ten duplicate between-day
738 analyses
739
Vitamin A Vitamin E β-Carotene
A: Intermediate Precision
Mean RSDiR Mean RSDiR Mean RSDiR
HorRat HorRat HorRat
ug/100 g % mg/100 g % ug/100 g %
Formula 1 542 5.26 0.42 7.54 4.94 0.59 134 8.20 0.54
Formula 2 359 6.29 0.48 4.60 5.89 0.66 87.7 6.49 0.40
B: Repeatability
Mean RSDr Mean RSDr Mean RSDr
HorRat HorRat HorRat
ug/100 g % mg/100 g % ug/100 g %
Formula 1 542 3.67 0.30 7.54 3.08 0.37 134 5.76 0.38
Formula 2 359 3.83 0.29 4.60 3.68 0.41 87.7 4.25 0.26
740
741
742
743
744 Table 3: Recovery of vitamin A esters, vitamin E acetate and β-carotene (provitamin A)
745 spiked into nutritional formula or milk powders
746
749
750
751 Table 4; Vitamin A and E ester results compared with certified reference materials
752 (CRM) from National Institute of Standards and Technology (NIST)
753
754
Observeda Certified Range
Retinol Palmitateb
NIST 1849 1631 (35.5) 1640 1510 – 1770
NIST 1849a 788 (19.2) 768 745 – 791
α-Tocopheryl Acetatec
757
758
759 Highlights for Review
760
761 We developed a fast method for vitamin A, E and carotene additives in dairy powders.
764 We extended the method to include carotenes with apo-carotenoate internal standard.
765 The method has potential to improve existing AOAC official method 2012.10
766
767
768