β-Cryptoxanthin and zeaxanthin are highly bioavailable from whole-grain and refined biofortified orange maize in humans with optimal vitamin A status: a randomized, crossover, placebo-controlled trial

Abstract

Background

Biofortification of staple crops with β-carotene is a strategy to reduce vitamin A deficiency, and several varieties are available in some African countries. β-Cryptoxanthin (BCX)-enhanced maize is currently in field trials. To our knowledge, maize BCX bioavailability has not been assessed in humans. Serum retinol ¹³C content and xanthophyll concentrations are proposed effectiveness biomarkers for biofortified maize adoption.

Objective

We determined the relative difference in BCX and zeaxanthin bioavailability from whole-grain and refined BCX-biofortified maize during chronic feeding compared with white maize and evaluated short-term changes in ¹³C-abundance in serum retinol.

Design

After a 7-d washout, 9 adults (mean ± SD age: 23.4 ± 2.3 y; 5 men) were provided with muffins made from BCX-enhanced whole-grain orange maize (WGOM), refined orange maize (ROM), or refined white maize (RWM) for 12 d in a randomized, blinded, crossover study followed by a 7-d washout. Blood was drawn on days 0, 3, 6, 9, 12, 15, and 19. Carotenoid areas under the curve (AUCs) were compared by using a fixed-effects model. ¹³C-Abundance in serum retinol was determined by using gas chromatography/combustion/isotope-ratio mass spectrometry on days 0, 12, and 19. Vitamin A status was determined by ¹³C-retinol isotope dilution postintervention.

Results

The serum BCX AUC was significantly higher for WGOM (1.70 ± 0.63 μmol ⋅ L⁻¹ ⋅ d) and ROM (1.66 ± 1.08 μmol ⋅ L⁻¹ ⋅ d) than for RWM (−0.06 ± 0.13 μmol ⋅ L⁻¹ ⋅ d; P < 0.003). A greater increase occurred in serum BCX from WGOM muffins (131%) than from ROM muffins (108%) (P ≤ 0.003). Zeaxanthin AUCs were higher for WGOM (0.94 ± 0.33) and ROM (0.96 ± 0.47) than for RWM (0.05 ± 0.12 μmol ⋅ L⁻¹ ⋅ d; P < 0.003). The intervention did not affect predose serum retinol ¹³C-abundance. Vitamin A status was within an optimal range (defined as 0.1–0.7 μmol/g liver).

Conclusions

BCX and zeaxanthin were highly bioavailable from BCX-biofortified maize. The adoption of BCX maize could positively affect consumers’ BCX and zeaxanthin intakes and associated health benefits. This trial is registered at www.clinicaltrials.gov as NCT02800408.

INTRODUCTION

Vitamin A deficiency affects millions of children <5 y of age (1). Interventions have included high-dose Vitamin A supplementation, fortification of targeted foods, and biofortification of staple and horticultural crops with provitamin A carotenoids, usually focusing on β-carotene (BC). Porridge prepared with high-BC maize was as efficacious as supplementation with 400 μg retinyl palmitate/d (2) and improved serum BC concentrations in Zambian children (3). Similarly, orange-fleshed sweet potato improved the vitamin A status of South African schoolchildren (4). BC from biofortified cassava mixed with oil modestly improved serum retinol concentrations in Kenyan children (5).

Maize (Zea mays) is an important staple crop with an estimated 900 million people who consume it (6), providing >50% of calories in some countries (7). Maize is processed in many ways, including dry milling, wet milling, and nixtamalization (i.e., cooking with lime) (8–10). Animal studies showed excellent BC bioefficacy [i.e., 3.6 and 3.7 μg BC equivalents to 1 μg retinol for whole and refined grain, respectively (8)]; however, milling may affect provitamin A carotenoid bioavailability from biofortified maize in humans. Up to 90% of carotenoids in yellow maize is found in the endosperm, which is maintained after refining (9). β-Cryptoxanthin (BCX), another provitamin A carotenoid, is found naturally in maize. The theoretical yield of vitamin A from BCX is half the potential yield from BC. However, BCX-biofortified maize feeding showed better bioefficacy than BC supplements in gerbils (11) and improved vitamin A status in chickens (12). Furthermore, BCX has higher apparent bioavailability than BC in humans (13), making it viable for biofortification.

It is well known that there is a difference in ¹³C-natural abundance between C₄- and C₃-photosynthesizing plants (14). Research in gerbils (15) and Zambian children (16) showed that long-term consumption of biofortified maize (a C₄ plant) increased the natural abundance of ¹³C in retinol compared with dietary regimens containing provitamin A from carrots (a C₃ plant) or white maize, respectively. Variation in ¹³C-abundance of serum retinol is due to intake from C₃- and C₄-plant sources, supplements, and fortificants (16, 17). The combination of serum retinol ¹³C-abundance shifts and changes in xanthophyll concentrations (e.g., zeaxanthin) has been proposed as a method to measure the effectiveness of biofortified maize consumption (16).

To our knowledge, high-BCX-biofortified maize, which is currently in field trials, has not been evaluated in humans. The primary objectives of this study were to determine the relative difference in BCX and zeaxanthin bioavailability from whole-grain and refined BCX-biofortified maize during chronic feeding periods in adults compared with white maize as a negative control and to evaluate the stability of and variation in the baseline measure of ¹³C-abundance in serum retinol among individuals with short-term maize feeding, which changed during long-term feeding studies (15, 16). Calculations were performed that used either individual- or group-mean baseline natural abundance in the ¹³C-retinol isotope dilution (¹³C-RID) test to evaluate changes in values when representative baseline blood draws are used in population evaluations.

METHODS

Subjects

All of the procedures involving human subjects were approved by the University of Wisconsin's Health Sciences Human Subjects Institutional Review Board (2015–1607). This study was registered as NIH clinical trial number NCT02800408 and protocol PEP-1508 from the primary sponsor. Subjects were recruited through flyers posted on the university's campus followed by phone screening by TJT and MS (Figure 1). Inclusion criteria included nonsmoking healthy young adults, aged 20–28 y with a BMI (kg/m²) of 19–26, with no known lipid malabsorption concerns. Participants were excluded if they were enrolled in another feeding study, were unable to consume the muffins at the research facility, had known schedule conflicts at the time of recruitment, or were unwilling to discontinue personal supplement use, in addition to any obvious indicators of poor health. Enrollment and treatment assignment occurred in July 2016 by TJT, and written informed consent was obtained from 10 healthy, nonsmoking adults (5 men and 5 women) aged 21–28 y (Table 1). This is the first study of this design to evaluate BCX; 5 subjects of each sex allowed us to have a high probability (∼90% power) of detecting a difference in carotenoid concentration between males and females that is ∼2 times as large as the within-sex SD. In previous studies that provided carrots containing either α- and β-carotene or the xanthophyll lutein, we recruited and reported on 9–10 women and men (18, 19), which provided adequate power to determine the difference between the treatment carrot and placebo white carrot muffin groups.

FIGURE 1.

CONSORT flow diagram of subject recruitment procedures and reasons for exclusions and loss to follow-up. CONSORT, Consolidated Standards of Reporting Trials.

TABLE 1.

Characteristics and baseline serum concentrations of subjects enrolled in a randomized, crossover orange maize feeding study¹

Characteristic	Value
Age, y	23.4 ± 2.3
Weight, kg	76.2 ± 11.8
BMI, kg/m2	23.8 ± 1.62
Body fat, %	17.0 ± 7.9
LDL cholesterol, mmol/L	2.47 ± 0.71
HDL cholesterol, mmol/L	1.50 ± 0.34
Total cholesterol, mmol/L	4.39 ± 0.96
Total triglycerides, mmol/L	0.91 ± 0.42
α-Carotene, μmol/L	0.13 ± 0.13
β-Carotene, μmol/L	0.33 ± 0.27
β-Cryptoxanthin, μmol/L	0.12 ± 0.07
Lutein, μmol/L	0.19 ± 0.06
Lycopene, μmol/L	0.17 ± 0.10
Zeaxanthin, μmol/L	0.04 ± 0.02
Serum retinol, μmol/L	2.43 ± 0.64
Total body retinol stores,2 μmol	954 ± 560
Total liver reserves,2 μmol/g liver	0.42 ± 0.22

¹Values are means ± SDs; n = 9 (4 women, 5 men). Baseline samples were obtained after a 1-wk washout, immediately before the treatment began.

²Vitamin A status was determined by retinol isotope dilution after the study was completed with the use of single baseline and 14-d blood draws at the end of the third-treatment washout period.

Percentage body fat at the beginning of each study arm was determined by air displacement plethysmography (BodPod version 2.30; Life Measurement, Inc.). Blood lipid concentrations were determined for each subject before beginning the study in typical clinical fashion by a contract laboratory (Meriter Laboratories) (Table 1).

Study design

The design was a randomized, single-blinded, 3 × 3, three-sequence crossover with treatment groups composed of 2 muffins from whole-grain orange maize (WGOM), refined orange maize (ROM), and refined white maize (RWM) as a negative control (Figure 2). After enrollment, a statistician not involved with the conduct of the study individually randomly assigned participants to 1 of 3 treatment sequences, blocked by sequence (3–4 subjects/sequence). The subjects remained in the study for a total of 120 d, including 1-wk washout periods before starting the study and before each treatment phase. One week before the first intervention, subjects followed the low-BCX diet (LBD; diets detailed below) for 4 d and then the stricter low-carotenoid diet (LCD) for 3 d as a washout to minimize serum carotenoids. Baseline blood draws were collected after the 1-wk washout period, after which participants immediately started treatments. Each treatment phase lasted 12 d (days 0–11), with a 1-wk washout after the 12-d blood sample while the subjects adhered to the LCD. In order to minimize carryover effects (18, 19), subjects were allowed to return to their regular diet for 2 wk in between phases. Each treatment day, subjects were provided with a breakfast consisting of the treatment muffins, tea or coffee, apple juice, yogurt, and coconut oil spread or strawberry jam, if desired. During each phase of the study, morning blood samples were collected after an 8-h fast on selected days to quantify serum carotenoids during muffin feeding and for 1 wk after treatment cessation. Blood samples were obtained on days 0, 3, 6, 9, 12, 15, and 19 during each treatment phase. Extra blood was drawn on days 0, 12, and 19 to quantify ¹³C-abundance. Blood was allowed to clot at room temperature and centrifuged at 2800 × g for 15 min at 4°C, and serum was stored at −80°C until analysis.

FIGURE 2.

Timeline (A) and randomization order (B) of a 120-d crossover study in which each subject (n = 9) received each of the 3 treatments once. Subjects were individually randomly assigned to sequence group A, B, or C (n = 3/group) to follow 3 maize treatments preceded and followed by washout periods. A statistician not associated with the study performed the randomization by subject. LBD, low-β-cryptoxanthin diet; LCD, low-carotenoid diet; ROM, refined orange maize treatment; RWM, refined white maize treatment; WGOM, whole-grain orange maize treatment.

Carotenoid-restricted and vitamin A–controlled diets

Three diets were developed for the subjects to follow during various times of the study. Subjects followed the LBD during the first 4 d of each washout period. This diet eliminated foods containing significant amounts of BCX (yellow-corn products, oranges, other citrus fruit, and red bell peppers). For the remainder of the phase, 22 d, the subjects were required to follow the LCD, which had all of the restrictions of the LBD but also eliminated additional foods that were high in other carotenoids, such as red tomato products, orange carrots, and mangoes. Alcohol consumption was also restricted (2 standard drinks/d for men and 1/d for women) 2 d before beginning each phase and during the active treatment periods. Compliance was monitored by providing subjects with forms to record fruit and vegetable intake, which were collected weekly and reviewed. The controlled vitamin A diet was followed after completion of the third phase of the study in order to estimate total body stores (TBSs) of vitamin A using ¹³C-RID as described below. This diet had all of the restrictions of the LCD as well as limitations on liver and fortified cereals, which contain preformed vitamin A, as recommended for individual vitamin A status assessment (20, 21).

Maize and muffins

High-BCX experimental maize varieties were conventionally bred at the International Maize and Wheat Improvement Center in Mexico as part of its HarvestPlus maize provitamin A biofortification project (22). For this study, maize was grown during the spring-summer cycle in 2015. Ears were harvested, dried, shelled, and stored at −20°C until being shipped to the University of Wisconsin–Madison. Biofortified and white maize kernels and flours were stored at −30°C for the duration of the study. Kernels (10–12% moisture) were ground by a Meadows 8-inch stone-burr mill in Lone Rock, Wisconsin. Mill stones were set 1.6 mm apart. All ground material was manually sifted through a 1.7-mm sieve. All material that passed through this sieve was considered “whole-grain” flour. For refined maize, ground maize was passed through a 0.6-mm sieve; all material that passed through this sieve was considered “refined” flour. RWM flour was also sifted by using a 0.6-mm sieve. Triplicate samples of the ROM, WGOM, and RWM flours were analyzed for carotenoid content before each treatment phase to monitor retention during storage.

WGOM and ROM recipes were developed to contain 250 μg BCX/muffin after baking for a total consumption of 500 μg/d from 2 muffins. Recipes for each food item were identical for different treatments, except for the maize flour type. The white maize recipe included yellow food coloring to match the appearance of orange maize muffins for blinding purposes. Recipes were standardized to ensure consistency by ingredient weight measured with a digital scale (American Weigh Scales LB-3000). Estimated macronutrient content was calculated by using the USDA Food Composition Database (Table 2). Muffins were made ≤3 d before consumption (except in one case of a participant who left town for >3 d who carried the muffins with her). Muffins were frozen at −30°C until the night before use, at which time they were moved to a refrigerator. In the morning, muffins were placed in a low-temperature oven (120°C) for 10–20 min. Treatment compliance was monitored by muffin weight consumed each day. Muffins were weighed after reheating, before dispensing to subjects, and then any remaining crumbs were weighed after consumption and subtracted from the total.

TABLE 2.

Estimated macronutrient and carotenoid intakes from 2 muffins as daily treatments in a randomized crossover study in young adults

	Maize
	Whole-grain orange	Refined orange	Refined white
Energy,1 kcal	810	800	800
Carbohydrate, g	132	135	135
Protein, g	14	13	13
Fat, g	252	23	23
β-Cryptoxanthin,3 μg	500 ± 184	500 ± 31	10 ± 2.0
β-Carotene, μg	380 ± 30	320 ± 16	13 ± 2.3
Zeaxanthin, μg	870 ± 149	780 ± 146	13 ± 2.2
Lutein, μg	230 ± 37	200 ± 37	12 ± 3.2

¹Energy and macronutrient content was estimated from the USDA Food-Composition Database.

²Whole-grain milling retains most of the germ where the oil is contained.

³Carotenoid profiles were determined by HPLC. Two muffins were equalized to deliver 500 μg β-cryptoxanthin/d.

⁴Mean ± SD (all such values).

Muffin carotenoid analysis

WGOM flour was mixed with RWM flour so that the BCX concentration would match that of the ROM muffins after baking. All maize and muffins were analyzed with the same HPLC procedure (23). Ethanol (6 mL with 0.1% butylated hydroxytoluene) was added to a 0.5-g sample, mixed with a vortex for 20 s, and heated at 85°C for 5 min. Samples were saponified with 500 μL 80:20 KOH:H₂O (weight:volume) at 85°C for 10 min with periodic mixing. Samples were placed in ice, and 3 mL cold deionized water was added along with 200 μL β-apo-8′-carotenal (Sigma-Aldrich) in methanol to calculate extraction efficiency. Carotenoids were extracted 3 times with 3 mL hexane. Extracts were combined, dried under nitrogen, redissolved in 500 μL 50:50 (volume:volume) dichloroethane:methanol; and 50 μL was injected into the HPLC (23) with a modified gradient. The 52-min linear gradient was run at 1 mL/min starting with 70% solvent A and transitioning to 44% A over 26 min. The flow then transitioned to 20% solvent A over 4 min and held for 10 min. The gradient was reversed in 2 min to 70% solvent A and re-equilibrated for 10 min. All of the sample analyses were performed under gold fluorescent lights to prevent photo-oxidation and isomerization of the carotenoids.

Serum carotenoid analysis

Serum aliquots were stored for ≤3 mo before analysis. To 200 μL serum, 100 μL β-apo-8′-carotenal dissolved in methanol as an internal standard was added along with 300 μL ethanol (0.1% butylated hydroxytoluene) to denature proteins. Samples were mixed by vortex for 30 s, and carotenoids were extracted 3 times with 500 μL hexane. The hexane layers were pooled, dried under nitrogen, and redissolved in 100 μL 50:50 (volume:volume) dichloroethane:methanol; and 50 μL was injected into the HPLC system (23). All samples from the same individual for each treatment were analyzed in the same day. Serum concentrations were calculated from standard curves generated for each carotenoid using HPLC-purified standards and the specific extinction coefficients (2592 for BC, 2800 for  α-carotene, 2550 for lutein, 2386 for BCX, 2348 for zeaxanthin, 3450 for lycopene, and 1845 for retinol) at 450 and 325 nm for the carotenoids and retinol, respectively (24, 25). The extraction efficiencies ranged from 85% to 105%. Interassay CVs for BCX and zeaxanthin were 3.4% and 11%, respectively, with the use of aliquots of stock serum analyzed several times throughout the HPLC analyses.

TBSs and liver concentrations of retinol

The subjects’ vitamin A statuses were determined by using the ¹³C-RID test and estimating TBSs and total liver reserves (TLRs) with the equations described below at the end of the study. Analysis with gas chromatography/combustion/isotope-ratio mass spectrometry (GCCIRMS) allows sensitive assays of the abundance of ¹³C in serum retinol (26). After a baseline blood draw (which was the day-19 sample in the third phase of the muffin study), 2.0 μmol [14,15]¹³C₂-retinyl acetate dissolved in soybean oil was delivered orally to each participant using a positive-displacement pipette and immediately followed by a high-fat–containing snack to facilitate absorption. After a 14-d mixing period, during which subjects consumed the controlled vitamin A diet, a second and final blood sample was taken. Serum retinol was purified by HPLC, and ¹³C:total C of serum retinol was determined by GCCIRMS (2).

In order to use the ¹³C-RID test in conjunction with the mass balance equation (20), the tracer-to-tracee ratio (TTR) must be determined with the use of the following calculation derived by Cobelli et al. (27) [with noted changes in abreviations adopted when using the ¹³C-RID test (20)]:

(1)

where a(t) (F_c) is the ¹³C-abundance of serum retinol 14 d postdose, a_N (F_b) is the baseline natural ¹³C-abundance in serum retinol, and a_I (F_a) is ¹³C-abundance in the retinol dose (the number of labeled ¹³C atoms plus natural abundance in the remaining carbon atoms, which was experimentally determined to be 0.11).

Biological factors and assumptions relating to the ¹³C-RID test are built into the mass balance equation as coefficients (20). F_b (natural abundance) was measured directly for each subject at day 0 or estimated by using the group mean, as was used in Thai (28) and Ghanaian (29) children to minimize the number of blood draws required. The parameter F_c was directly measured by GCCIRMS at day 14. With the use of Equation 1 above, the TTR for each individual was calculated from the directly measured day-14 postdose ¹³C-abundance measurement as F_c and either their own predose abundance or the mean of all 9 subjects’ predose measurements as F_b. The TTR was also calculated from F_c and all baseline measures of natural abundance during the muffin feeding (n = 9/subject) to determine variation. TBSs (micromoles) were calculated as previously described (20) with the use of the mass balance equation, as follows:

(2)

where a is the dose of ¹³C₂-retinyl acetate in micromoles. The following assumptions were used: 90% dose absorption, equal serum and liver ¹³C enrichment before dosing and after the mixing period because subjects were fasting and maintained on a low–vitamin A diet (20, 21), and correction for dose catabolism according to a half-life of 140 d (20).

TLRs (micromoles per gram) were then calculated from TBSs using the following formula:

(3)

where BW is body weight (kilograms), liver fraction of BW was estimated as 2.4% in healthy adults, and 80% of TBSs was assumed to be in the liver storage pool in adults with adequate vitamin A status (20).

Statistical analysis

Statistical analyses were performed with SAS software (version 9.4; SAS Institute). The serum concentrations of each carotenoid of interest were corrected by subtracting the baseline concentration from the post-treatment value and then determining the increase or decrease at each time point and plotted against time. The AUCs for days 0–19 (AUC_0–19d) were calculated for all carotenoids by trapezoidal approximation as in previous studies (18, 19). Initially, a mixed-effects PROC MIXED model included a random effect of BMI and fixed-effects terms for treatment sequence, sex, type of muffin, and the interaction between sequence and treatment. Finally, nonsignificant variables were removed from the model to contain only sequence, treatment, and their interaction. Least-square means with the Tukey-Kramer method were used to make multiple group comparisons. A piecewise regression was used to evaluate whether a plateau occurred in serum carotenoid response.

For natural ¹³C-abundance, subject values at each time were compared by 1-factor ANOVA. The effect of time, treatment, and time-by-treatment with randomization for treatment sequence and subject was determined on the entire set of natural ¹³C-abundance values, expressed as atom percent (At%), by repeated-measures ANOVA, and compared by the Tukey-Kramer method. Mean and individual baseline TLR values were compared by 1-sample Student's t test with Bonferroni correction for multiple comparisons. Significance was defined as P ≤ 0.05 for all statistical evaluations.

RESULTS

Subject baseline characteristics and compliance

Baseline serum carotenoid concentrations, vitamin A status, anthropometric characteristics, and a lipid panel were collected (Table 1). Two subjects, 1 man and 1 woman, had total cholesterol concentrations >2000 mg/L. One woman left the study due to time constraints after completion of the first phase; her data are not included. In general, compliance with the LCD was high, with no major deviations noted in daily food records or evident from serum analysis for the xanthophyll carotenoids. All of the subjects consumed >95% of the muffins during all treatments, except for one subject who consumed 89% of the WGOM treatment. Overall, treatment intakes were ≥98% for the 3 respective treatment groups.

Muffin carotenoid concentrations

The BCX retention after baking from the milled maize was ≥90% (30), which is higher than or similar to the BCX retention of other provitamin A–biofortified maize varieties and cooking methods (31, 32). Muffin BCX and α-carotene concentrations remained stable throughout the study. BC, zeaxanthin, and lutein concentrations decreased by ∼30% from the beginning of the study to the third phase, which is evident in the variation in the concentrations (Table 2).

Serum carotenoid concentrations and AUC analyses

The AUCs_0–19d for all carotenoids quantified were determined in μmol ⋅ L⁻¹ ⋅ d (Table 3). The increase in serum BCX from baseline (0.11 ± 0.07 μmol/L) to day 12 (0.26 ± 0.08 μmol/L) was 131% for the WGOM treatment (P < 0.0001) and 108% during the ROM treatment (0.14 ± 0.05 to 0.29 ± 0.12 μmol/L) (P = 0.003). The absolute day-12 serum BCX concentrations did not differ between WGOM (0.26 ± 0.08 μmol/L) and ROM (0.29 ± 0.12 μmol/L) muffin groups (P = 0.5). Serum BCX concentration during the RWM treatment, which was essentially devoid of BCX, did not substantially change from baseline (0.14 ± 0.07 μmol/L) to day 12 (0.10 ± 0.05 μmol/L) (P = 0.16). The BCX AUC_0–19d for the WGOM and ROM treatments did not differ (Figure 3A). WGOM and ROM treatment AUCs were significantly higher than that for the RWM negative control (P < 0.003). Sequence and treatment-by-sequence interaction did not affect the BCX AUC_0–19d (P = 0.34 and 0.41, respectively). Piecewise regression fit to both total WGOM and ROM treatment groups indicated that BCX concentrations had not reached a plateau stage in either treatment, indicating that a steady state had not yet occurred with an intake of 500 μg/d for 12 d.

TABLE 3.

Serum carotenoid AUC_0–19d (determined by trapezoidal approximation) corrected for baseline concentrations at time 0 for each treatment arm¹

	AUC0–19d, μmol ⋅ L−1 ⋅ d
	WGOM	ROM	RWM
β-Cryptoxanthin2	1.70 ± 0.63a	1.66 ± 1.08a	−0.06 ± 0.13b
Zeaxanthin2	0.94 ± 0.33a	0.96 ± 0.47a	0.05 ± 0.12b
β-Carotene	−0.09 ± 0.08	−0.06 ± 0.17	−0.05 ± 0.09
ɑ-Carotene	−0.02 ± 0.13	−0.03 ± 0.05	0.01 ± 0.16
Lutein	0.04 ± 0.17	0.08 ± 0.17	−0.04 ± 0.04
Lycopene	−0.06 ± 0.11	0.08 ± 0.35	−0.05 ± 0.03

¹Values are means ± SDs. All subjects completed each treatment; n = 9 (4 women, 5 men). Treatments were as follows: WGOM (500 μg BCX/d), ROM (500 μg BCX/d), or RWM (10 μg BCX/d) muffins, which were consumed for 12 d followed by 7 d of withdrawal. AUC_0–19d, AUCs for days 0–19; BCX, β-cryptoxanthin; ROM, refined orange maize; RWM, refined white maize; WGOM, whole-grain orange maize.

²Values in rows with different superscript letters are significantly different (least-square means test with the Tukey-Kramer method, P < 0.003).

FIGURE 3.

Serum β-cryptoxanthin (A) and zeaxanthin (B) concentrations after correction for baseline concentration in 9 adults who consumed WGOM muffins (500 μg β-cryptoxanthin and 870 μg zeaxanthin/d; ⬤), ROM muffins (500 μg β-cryptoxanthin and 780 μg zeaxanthin/d; ▴), or RWM muffins (negative control, ∼0 μg/d; ▪) for 12 d followed by a 7-d treatment withdrawal. Values are means ± SDs. All of the subjects completed the 3 treatments. AUC analyses with the use of trapezoidal approximation showed that the effects of the orange maize treatments were significantly different from those in the negative control group (P < 0.003 for both; least-square means test with the Tukey-Kramer method). (Note: the concentration scales are different for each carotenoid.) ROM, refined orange maize treatment; RWM, refined white maize treatment; WGOM, whole-grain orange maize treatment.

The AUC_0–19d for zeaxanthin followed a similar response to that of BCX (Figure 3B), but concentration changes were smaller in magnitude. Similar to BCX, zeaxanthin WGOM and ROM treatment AUCs were significantly higher than that for the RWM negative control (P < 0.003). The baseline to day 12 increases in serum zeaxanthin concentrations in the WGOM and ROM treatments were 178% (P < 0.001) and 176% (P = 0.001), respectively, and there was a decrease of 15% for the RWM negative control. The absolute day-12 serum zeaxanthin concentration did not differ between WGOM (0.10 ± 0.03 μmol/L) and ROM (0.12 ± 0.06 μmol/L) treatment groups (P = 0.28). Piecewise regression for zeaxanthin concentrations showed that treatment with WGOM muffins reached a steady-state plateau by day 9. The piecewise regression was significant (P < 0.001), but it did not appear to explain more variation than a linear regression. A steady state for zeaxanthin did not yet occur for ROM treatment by day 12, meaning that a higher concentration may have occurred with longer feeding.

Serum BC decreased by 0.1 μmol/L in all 3 treatment groups by 6 d, resulting in negative BC AUCs_0–19d (Table 3). Lutein concentrations remained relatively stable during the treatment phases. Although AUCs_0–19d were positive for both WGOM and ROM treatments and negative for the RWM treatment (Table 3), the difference between lutein responses was not significant. Serum lycopene and α-carotene concentrations were used to measure compliance with the LCD because neither carotenoid is found in significant amounts in current varieties of biofortified maize. Lycopene concentrations decreased by 0.05 to 0.1 μmol/L in 6 d and did not differ by sequence or muffin type. The AUC_0–19d for lycopene was positive for the ROM group, but not different from the WGOM and RWM AUCs_0–19d (P ≤ 0.32) (Table 3). The BCX AUC_0–19d was 22.5 times larger than that of the lycopene response from ROM. The positive lycopene AUC_0–19d is largely explained by one subject's high value during the ROM treatment arm, likely caused by reported ketchup consumption. Even clearer evidence of diet compliance was the continued decreases in α-carotene during the study treatments. Serum α-carotene consistently decreased during each treatment, with the largest decrease of 0.05 μmol/L by 6 d, which was not surprising considering that the muffins contained <0.4 μg α-carotene/100 g. The α-carotene AUCs_0–19d were close to zero for all treatments (Table 3).

Serum retinol, TBSs, and liver retinol concentrations

Serum retinol concentrations were not affected by any treatment (P > 0.05) but were higher during the second arm of the study (2.46 ± 0.43 μmol/L) compared with arm 1 (2.17 ± 0.44 μmol/L; P = 0.001) and arm 3 (2.27 ± 0.36 μmol/L; P = 0.04). Mean individual serum retinol concentrations (2.32 ± 0.41 μmol/L; n = 9 measurements/subject) measured during the analysis of natural ¹³C-abundance varied during the 4-mo study, with a mean CV of 17% (Table 4), although no values were <0.7 μmol/L [cutoff for deficiency (33)]. Serum retinol concentrations did not correlate with serum BCX concentrations throughout the study (P = 0.58). The TBSs of these adults were 954 ± 560 μmol (using the blood draw at day 19 after arm 3 as baseline and 14-d postdose values in Equation 2). Estimated TLRs based on a liver weight of 2.4% of body weight were 0.42 ± 0.22 μmol/g, which is considered optimal (33). One subject had a TBS of 2150 μmol (125% higher than the mean) with an estimated TLR of 0.74 μmol/g, which is considered high (33). In consultation with this subject after determining this value, he consumed whole milk and multiple eggs on a daily basis, both excellent sources of preformed retinyl esters.

TABLE 4.

Serum retinol, TBSs, and TLRs calculated by using each of 9 baseline natural abundance measurements (F_b in the mass balance equation for retinol isotope dilution) for subjects enrolled in a biofortified maize intervention study¹

	Serum retinol		TBSs		TLRs
Subject	μmol/L	CV, %	μmol	CV, %	μmol/g	CV, %
1	2.19 ± 0.48	22	951 ± 20	2.12	0.423 ± 0.009	2.12
2	2.28 ± 0.39	17	743 ± 9	1.2	0.303 ± 0.004	1.2
3	2.70 ± 0.26	9.7	2190 ± 120	5.3	0.753 ± 0.040	5.3
4	2.10 ± 0.34	17	508 ± 7	1.4	0.266 ± 0.004	1.4
5	2.13 ± 0.61	29	1050 ± 10	1.3	0.417 ± 0.006	1.3
6	2.57 ± 0.44	17	1130 ± 30	2.9	0.591 ± 0.017	2.9
7	2.27 ± 0.28	12	1270 ± 40	3.5	0.690 ± 0.024	3.5
8	2.10 ± 0.39	18	397 ± 3	0.71	0.177 ± 0.001	0.71
9	2.59 ± 0.34	13	363 ± 11	3.0	0.144 ± 0.004	3.0
Mean	2.32 ± 0.413	17 ± 5.6	956 ± 443	2.4 ± 1.5	0.418 ± 0.0173	2.4 ± 1.5

¹Values are means ± SDs unless otherwise indicated; n = 9. TBS, total body store; TLR, total liver reserve.

²CVs are the same because TLRs were estimated from TBSs.

³Values are means of means ± mean SDs (square root of mean variance = ).

¹³C-Abundance in subjects before, during, and after treatment

The amount of retinol activity equivalents ingested during the WGOM or ROM muffin treatments was 2.2 μmol, estimated by using the Institute of Medicine's conversion factors for a mixed diet of 24 μg BCX or 12 μg BC equivalent to 1 μg of retinol (34). This amount is 0.23% of the mean TBS (954 μmol) in these young adults.

The mean ¹³C-abundances of the entire group composed of 9 subjects (3 subjects undergoing each of 3 treatments at any time point during the study) were not different from each other (mean At% = 1.0764 ± 0.0005; P = 0.83) (Figure 4A). Time (P = 0.62) and treatment (P = 0.18) did not affect ¹³C-abundance, but the treatment-by-time interaction approached significance (P = 0.09) (Figure 4B). However, comparison of individual subject mean ¹³C-abundance values from all time points (days 0, 12, and 19 from each of the 3 treatments) showed that individual subjects had significantly different ¹³C-natural abundance (P = 0.0002) despite having the same total amount of additional maize-derived provitamin A carotenoid added to their low–provitamin A regimens over all 3 treatment periods (Figure 4C). The significant finding of the variation among individual natural abundance values (Figure 4C) and the nonsignificant finding when using the group as a whole (Figure 4A) underscores the importance of determining natural abundance measurements for individual vitamin A status assessment for either the individuals themselves or a carefully selected representative sample of the cohort being evaluated (28, 29). This finding allows latitude when assessing the vitamin A status of population groups for evaluation purposes.

FIGURE 4.

¹³C-Natural abundance determined by gas chromatography-combustion-isotope ratio mass spectrometry in serum retinol expressed as At% by time (A), treatment-by-time (B), and subject (C). Values are means ± SDs. (D) TLRs of subjects calculated by using mean (white bars) or individual (black bars) baseline serum retinol ¹³C-natural abundance. TLRs calculated by using the mean baseline are shown as means ± SDs. Mean-derived and individual-baseline TLRs did not differ between subjects (1-sample t test, Bonferroni corrected P < 0.05). Values with different lowercase letters are significantly different by the Tukey-Kramer method (P < 0.05) after ANOVA. At%, atom percent of ¹³C in total carbon; d, day; TLR, total liver vitamin A reserve; Trt, treatment.

Calculated TLRs of retinol from individual or mean baseline values

TLRs determined by the individual's TTR were considered to be the hypothetical true value, and each true value was compared with the TLRs calculated from the mean F_b TTR. None of the mean-baseline TLR values differed from the individual-baseline TLR values (Bonferroni-corrected P > 0.05), and the difference between the 2 was not different from zero (−0.01 ± 0.03; P = 0.1) (Figure 4D), further supporting the use of a nonintervention group for evaluation studies. Finally, TLRs were calculated by using the 9 measures of the individual's ¹³C-natural abundance as F_bs with the postdose F_c, and the mean CV was 2.4% ± 1.5% (Table 4). The difference between the standard protocol (20) mean TBS value (calculated by using a baseline at time of dose; 954 μmol; Table 1) and that of all 9 baseline values (956 μmol; Table 4) was only 2 μmol (P = 0.93).

DISCUSSION

Muffins made from whole-grain and refined high-BCX maize significantly increased serum BCX concentrations. Maize milling did not affect responses, which confirms findings in animals (8) and in vitro models (35). Other health benefits associated with consuming whole grains include minerals (8), fiber, vitamins, and phytochemicals that are removed by milling. BCX was highly bioavailable from both flours, which is favorable for future products formulated with biofortified orange maize targeted to improve vitamin A status. This experimental variety of BCX maize was selectively bred for enhanced BCX, in part due to better stability of BCX during storage and processing (32, 36) and higher bioavailability in animals (11, 37) and humans (13, 38). High-BC varieties are currently being bred at a faster pace than BCX, but large genetic diversity exists for BCX in maize, which makes the breeding strategy feasible (39). Future product considerations should include high-BCX varieties, not only to improve vitamin A status but also for the other beneficial health effects associated with BCX (e.g., cancer prevention, bone health) (40).

Previous carrot muffin studies found carryover effects when feeding lutein from yellow carrots (19), α- and BC from orange carrots (41), and lycopene from red carrots (18) because subjects transitioned directly to the next treatment. BCX-response curves showed a continuous increase during 12 d of feeding that did not return to baseline by day 19, 7 d after the treatment. The washout period was based on previous experience (19) and the fact that less BCX (500 μg/d) than lutein (1.7 mg/d) was consumed. Nonetheless, baseline BCX concentrations were not different because subjects were allowed to resume their regular eating patterns for 2 wk in between treatments; thus, carryover effects did not occur and muffin sequence did not affect outcomes.

Although the mean lutein AUCs_0–19d for both orange maize treatments were positive and the RWM group was negative, interindividual variations resulted in no overall difference. In contrast, significant treatment effects were obtained with lutein-containing yellow carrots (19), which had 8.5 times more lutein than the ∼200 μg/d from the muffins in this study, which maintained baseline concentrations. This lutein amount is similar to that found in supplements, such as Centrum Silver multivitamin for adults at 250 μg/tablet (http://www.centrum.com/).

Zeaxanthin is concentrated in the macula of the eye and may protect humans against macular degeneration (42). Zeaxanthin is not common in many foods, with the exception of yellow and orange maize, egg yolks, persimmons, and some leafy greens (43). Zeaxanthin intake from BCX-biofortified maize increased serum concentrations, confirming its use as a biomarker of prolonged biofortified maize intake (16). Including orange maize in maize-based foods could increase zeaxanthin consumption. Two hydroxylases, P450 cytochrome 97A and hydroxylase B, hydroxylate β-ionone rings leading to loss of provitamin A activity (44). When selecting for BCX, increased β-ring hydroxylase activity causes flux toward zeaxanthin and zeinoxanthin, resulting in a positive correlation between BCX and zeaxanthin (45). This high-BCX maize (5.0 μg/g) had nearly double the amount of zeaxanthin (8.8 μg/g), but the zeaxanthin response was relatively half that of BCX, confirming superior BCX bioavailability in humans (13).

Serum BC concentrations decreased and did not differ between groups. The amount of BC ingested from the orange maize muffins was ∼350 μg/d. Most of the absorbed BC was likely converted to meet daily retinol needs due to the long-term restricted dietary requirements. Subjects were not allowed to consume multivitamins, although they could consume dairy products, which are sources of preformed vitamin A. We assessed vitamin A status by using ¹³C-RID in case we had a nonresponder phenotype among them, which could partially be explained by low TBSs. Although we had one subject with high TBSs confirmed with previous diet history, the mean TLR of the group was optimal, and all subjects responded to the BCX treatments.

The difference between baseline subject ¹³C-natural abundance F_b values was not surprising because retinol stores build up over time from combinations of C₃ and C₄ foods, supplements, fortificants, and other biological factors. The lowest At% value was from a vegetarian who likely sourced most of her vitamin A from C₃ plants. Shifts in natural abundance due to changes in the proportion of C₃ and C₄ foods consumed over longer periods support the application of using these measurements for program effectiveness in populations targeted to consume biofortified C₄ plants (16, 46) or adherence to increased C₃ vegetable intake regimens (17). Communities tend to consume similar foods with unique natural enrichments of ¹³C (2, 16).

Determining either individual or representative baseline values in evaluation studies is important. To show the extremes, F_b values from gerbils fed exclusively orange maize (C₄) or carrots (C₃) for 62 d resulted in different At% values (i.e., 1.083% and 1.071%, respectively) (15). When applying these extreme F_b values, the adjusted mean TLRs would be estimated as 1.8 ± 3.4 μmol/g and 0.32 ± 0.14 μmol/g, respectively. The value from C₃F_b is similar to the actual value (i.e., 0.42 ± 0.22 μmol/g), but that from C₄F_b reclassifies the group from having optimal vitamin A status to hypervitaminotic TLRs. Thus, the use of a single constant natural abundance is not appropriate in studies using ¹³C-RID tests and should be measured in a nonintervened, nondosed, or predose group from the same community. The lack of change in mean baseline F_bs over several months provides evidence that it may be appropriate to take a baseline blood sample at a reasonable time before a study. An expert recommendation is that in countries implementing high-dose vitamin A supplements (47), vitamin A assessment should occur in children ≤10 y (48). Having a simplified retinol isotope dilution procedure to monitor population subgroups is important when >1 intervention is occurring at the same time (49).

The lack of effect on ¹³C-natural abundance could be due to the generally high vitamin A status of the subjects. During 12 d of either WGOM or ROM feeding, the subjects consumed ∼2 μmol retinol activity equivalents. Despite enrichment of maize for ¹³C at an At% of 1.093 (17), the serum ¹³C-abundance did not change significantly when diluting this amount into TBSs of 370–2150 μmol. Recent studies reported significant natural abundance differences between gerbils when tested after 2 mo of being fed very diverse feeds (15) and between Zambian children fed either biofortified orange or white maize for 6 d/wk for a total of 42 d (16). These data indicate that natural abundance will vary with longer dietary modification, highlighting the need to determine F_b in intervention and effectiveness studies.

The clinical relevance of the ¹³C-RID findings was the applicability of the mean baseline as a good estimation of natural abundance in all subjects. This suggests that it is possible to limit the blood draws required in a population evaluation (for preintervention TLR calculations only). The subject with the highest TBSs showed a much larger difference between mean- and individual-baseline TLRs. If that subject's baseline is removed from the mean to provide a less-biased value, the difference between that subject's 2 TLRs becomes significant, indicating the need for further research in individuals with high or hypervitaminotic TLRs. Overall, these data suggest that the complexity and cost of retinol isotope dilution in the field and laboratory could be reduced by decreasing the stringency of the time requirements for predose baseline blood draws, and by limiting the number of draws by forming a mean baseline rather than taking an individual day 0 draw. The increased accessible field-friendliness of this technique follows the recommendations laid out for the future of the vitamin A–labeled isotope dilution (termed VALID) test (50).

One limitation of this research is that a BCX oil-based supplement in a placebo muffin, such as that used in the lutein carrot study (19), was not included. However, BCX supplements are not currently commercially available and analytical standards are cost-prohibitive. We would have needed ∼100 mg purified BCX for inclusion as a positive control. Although BCX is a vitamin A precursor, BCX and zeaxanthin likely have additional health benefits. BCX is linked to bone health (40) and zeaxanthin to eye health (42). Future long-term interventions with biofortified maize or BCX and zeaxanthin supplements should measure markers of bone and eye health.

Notes

The analytical aspects of this work were presented at Experimental Biology 2017.

Supported by PepsiCo, Inc.; HarvestPlus breeding contract 5404, which was partially funded by the Bill and Melinda Gates Foundation (no. OPP27510); and Global Health Funds at the University of Wisconsin-Madison. HarvestPlus (www.harvestplus.org) is a global alliance of agriculture and nutrition research institutions working to increase the micronutrient density of staple food crops through biofortification. The views expressed do not necessarily reflect those of PepsiCo, Inc., or of HarvestPlus. None of the organizations involved in supporting this work had a role in the analysis or interpretation of the data.

TJT and JS are joint first authors.

Abbreviations used:

At%

atom percent of ¹³C in total carbon

BC

β-carotene

BCX

β-cryptoxanthin

¹³C-RID

¹³C-retinol isotope dilution

GCCIRMS

gas chromatography/combustion/isotope-ratio mass spectrometry

LBD

low-β-cryptoxanthin diet

LCD

low-carotenoid diet

ROM

refined orange maize

RWM

refined white maize

TBS

total body store

TLR

total liver reserve

TTR

tracer to tracee ratio

WGOM

whole-grain orange maize