Bioaccessibility and bioavailability of biofortified food and food products: Current evidence

Abstract Biofortification increases micronutrient content in staple crops through conventional breeding, agronomic methods, or genetic engineering. Bioaccessibility is a prerequisite for a nutrient to fulfill a biological function, e.g., to be bioavailable. The objective of this systematic review is to examine the bioavailability (and bioaccessibility as a proxy via in vitro and animal models) of the target micronutrients enriched in conventionally biofortified crops that have undergone post-harvest storage and/or processing, which has not been systematically reviewed previously, to our knowledge. We searched for articles indexed in MEDLINE, Agricola, AgEcon, and Center for Agriculture and Biosciences International databases, organizational websites, and hand-searched studies’ reference lists to identify 18 studies reporting on bioaccessibility and 58 studies on bioavailability. Conventionally bred biofortified crops overall had higher bioaccessibility and bioavailability than their conventional counterparts, which generally provide more absorbed micronutrient on a fixed ration basis. However, these estimates depended on exact cultivar, processing method, context (crop measured alone or as part of a composite meal), and experimental method used. Measuring bioaccessibility and bioavailability of target micronutrients in biofortified and conventional foods is critical to optimize nutrient availability and absorption, ultimately to improve programs targeting micronutrient deficiency.


Introduction
Micronutrient deficiencies continue to affect approximately two billion people worldwide, particularly women and children living in low-to-middle-income countries (LMICs) (Bailey, West, and Black 2015;Bhutta 1998).Many populations in LMICs rely on staple carbohydrates and energy-rich crops such as wheat, maize, rice, sweet potato, pearl millet, cassava, and legumes such as common beans.These crops are a staple of the daily diet, but do not generally provide an adequate array of micronutrients to meet human needs, due to both inherently low micronutrient content as well as limited bioaccessibility (the amount or degree to which micronutrients are released after digestion and available for absorption) and/or bioavailability (the amount of target micronutrient absorbed by the body).Processing (for example, grinding or cooking), which is necessary before consuming these crops, can improve bioavailability but may also result in micronutrient losses, depending on the type of processing and micronutrient (Hotz and McClafferty 2007).
Biofortification is a promising solution to reduce micronutrient deficiencies by increasing micronutrient intake through widely consumed staple crops.Biofortification is defined as the process of increasing the concentrations and bioavailability of essential nutrients in a staple crop by conventional crossbreeding, the focus of our review-in addition to agronomic practices, and/or genetic engineering-and is a potential approach to combat micronutrient deficiencies at the population level.Globally, staple carbohydrates and energy-rich crops such as wheat, sorghum, maize, rice, sweet potato, Irish potato, pearl millet, lentils, beans, cowpeas, banana/plantain, and cassava are most commonly biofortified with provitamin A carotenoids, iron, and/or zinc in Africa, Asia, and Latin America (HarvestPlus 2021).Currently, it is estimated that 48 million people from smallholder farming households are benefitting from conventionally bred biofortified crops.HarvestPlus estimates that more than 1 billion people could benefit from regularly eating nutritious biofortified crops and foods by 2030 (HarvestPlus 2021).
Despite the widespread implementation of biofortified crops, there is limited recent systematic consolidation of evidence on bioaccessibility and bioavailability of micronutrients in conventionally biofortified crops, particularly after processing.Bioaccessibility is defined as the quantity of a compound that is released from its matrix in the gastrointestinal tract, becoming available for absorption (e.g., enters the blood stream) (Charis 2017).This term includes digestive transformations of foods into material ready for assimilation, the absorption/assimilation into intestinal epithelium cells as well as presystemic, intestinal and hepatic metabolism (Charis 2017).Bioavailability expresses the fraction of ingested nutrient or bioactive compound that reaches the systemic circulation and is ultimately utilized.Bioaccessibility and bioavailability therefore measure different stages of nutrient assimilation, and only bioavailability estimates actual nutrient absorption and utilization (Charis 2017).It is important to understand the bioaccessibility and bioavailability of micronutrients in biofortified crops as these factors potentially limit the efficacy of biofortified crops to improve micronutrient deficiency (La Frano et al. 2014;Bechoff and Dhuique-Mayer 2017).
The bioaccessibility and bioavailability of conventionally bred biofortified crops has not been previously systematically reviewed to our knowledge.As such, the objective of this systematic review is to examine the bioaccessibility and bioavailability of the target micronutrients enriched in conventionally bred biofortified crops in varied-use settings including farmer and urban households or commercially, and/or that have undergone post-harvest storage and/or processing, such as milling or cooking.

Methods
We registered the protocol for this review on PROSPERO, the international prospective register of systematic reviews of the University of York and the National Institute for Health Research, on June 11, 2021 (Huey et al. 2021).

Participants or models used
Any human studies were considered eligible for inclusion.In addition, studies measuring the micronutrient content released after digestion and/or absorption of said micronutrients from a conventionally biofortified crop or food product using in-vitro gastrointestinal digestion simulation such as Caco-2 cell uptake or via animal models were included in this review (see Supplementary material).Studies that did not use one of the above methods for micronutrient bioavailability assessment or studies that examined soil micronutrient bioavailability were considered ineligible.

Interventions
We included studies examining biofortified crops and food products, including those that have undergone processing post-harvest, that were delivered as crops only or in the form of food products (as defined by study authors).Crops included those biofortified by conventional plant breeding approaches.
We did not include interventions utilizing agronomic biofortification methods, genetic engineering-based biofortification methods, or animal-based biofortified foods such as dairy products or meat from animals that consumed biofortified feed.

Comparators
Comparators included either (A) a non-biofortified version of the same crop (i.e., control) or food product made using non-biofortified crops; or (B) crops and food products fortified with the same micronutrient.

Outcomes
1. Bioaccessibility of iron, zinc, or carotenoids in biofortified crops and the impact of factors such as storage, processing (for example, milling and cooking), freezing, meal composition, etc. 2. Bioavailability of iron, zinc, or carotenoids in biofortified crops and the impact of factors such as storage, processing (for example, milling and cooking), freezing, meal composition, etc.

Literature searches
We originally aimed to conduct a set of four reviews on biofortification and thus designed our search strategy to accommodate the topics examined by all four reviews.We performed a search of relevant literature databases including: MEDLINE (PubMed), AGRICOLA, AgEcon, CABI Abstracts (Web of Science) and organizational websites (e.g., Harvest Plus, CGIAR and partners).
As a preliminary assessment of the literature on biofortification, we conducted a broad search in MEDLINE (PubMed) on March 29, 2021, using the following key terms: [tiab].This resulted in 1434 records.After screening these records and ascertaining key words to use for increasing the sensitivity of the search, we conducted searches in additional databases, using broader or narrower searching depending on the topic focus of the database.These, including the original MEDLINE search, are summarized in Table 1.
We also hand-searched organization websites.The results are included in Table 2.
We also identified 1146 potential citations outside of the original search during the screening process.These included studies that were: cited in review papers but did not include variations of the term "biofortification" in their abstracts; not indexed in any of the literature databases described above and were thus missed by the original search; published after the search was run, which we identified from journals' table of contents alert feeds.Some of the latter included full-text versions of conference abstracts that were found and included in the original screening pool.

Data screening and extraction
SLH, NHM, EMK, and AB screened all records for eligibility, first at the title/abstract level and subsequently at the full text screening level.SLH, NHM, EMK, and AB used a subset of articles to enhance consistency amongst review authors.
SLH, NHM, EMK, and AB extracted data for each identified study, according to the following: study level details including authors or research group, study year, and location; method details including randomization scheme, population, how crops were biofortified, method for ascertaining bioaccessibility and/or bioavailability; and outcomes of the study.We used PlotDigitizer software (https://plotdigitizer. sourceforge.net/)to extract datapoints from graphs.

Data synthesis and analysis
We synthesized the data in a narrative format, pooling data where appropriate (i.e., multiple studies comparing the same crop cultivar, processing method, and nutrient of interest) to generate a summary estimate of the bioaccessibility or bioavailability of a particular micronutrient in a crop cultivar.

Results
For all 4 review topics, we found a total of 5131 records (Figure 1).Ultimately, we found overall 307 studies records across the 4 review topics outlined previously.
For the current review, we identified 18 studies reporting on bioaccessibility and 58 studies on bioavailability, of which 9 reported both bioaccessibility and bioavailability, resulting in 67 studies total.Eighteen of the 58 bioavailability studies involved humans.
We have reported on bioaccessibility studies and on the bioavailability studies done in human populations with some form of biofortified crop processing here in the main text.Bioaccessibility of nutrients from raw crops (cassava, cowpeas) is shown in Supplementary material (Table S1) given their limited applicability to actual consumption patterns in humans.
Bioavailability studies done in animals or in in vitro assays are reported in Supplementary material (Table S2 and S3, respectively) in order to narrow the review focus to results based in human models and thus, closer to real-world applicability.

Bioaccessibility
Bioaccessibility of micronutrients across various biofortified crops was measured using in vitro digestions to measure mineral solubility, carotenoid content in supernatant or micelle fraction, and ME of carotenoid uptake into micelles (Supplementary Figure S1).The cultivars examined and bioaccessibility outcomes measured are shown by processing method in Table 3.Studies that analyzed raw cassava and cowpeas are shown in Table S1 (Supplementary material).We did not identify any studies measuring bioaccessibility in vivo.All varieties are biofortified unless noted.

Summary of findings
There was substantial heterogeneity in cultivar types used, the types of processing done, and the outcomes measured.Main findings are summarized below, given in mean ± SD unless otherwise noted.Individual results may be found in Supplementary material Tables S4 and S5.

Cassava
Carotenoid uptake by micelles was measured by all studies in the form of micellization efficiency (ME) % or the amount of carotenoid(s) in the micellar phase.Only biofortified crops were examined (no conventional cassava crops, which are usually white and therefore for all practical purposes devoid of carotenoids, were included for comparison).Oral, gastric, and intestinal digestions were simulated, and oil (either soybean or canola) was added.
After boiling, the highest bioaccessibility of all-trans BC (ATBC) was shown for cultivar Clone 03-15 with an ME of 41.2%, while cultivar Clone B-14-11 had the lowest ME at 9.4% out of a total of 8 specific cultivars and an additional 3 pooled cultivars (specific variety was not described (Thakkar et al. 2009)).The mean ATBC ME across the cultivars was 24.9% (Figure 2A).
For cultivar BRS Jari, ATBC ME was 38.8% (Berni et al. 2014).Another study showed that BRS Jari contained 77.1 ± 8.3 µg bioaccessible ATBC per 100 g in the micellar fraction after boiling (Gomes et al. 2013).Cultivar BRS Jari was also examined in terms of total carotenoid content (TCC) or beta carotene (BC) bioaccessibility after boiling: TCC ME was 6.0 ± 0.2%, resulting in 153 ± 11.5 µg TCC per 100 g micellar fraction (Gomes et al. 2013).BRS Jari BC ME was 5.8 ± 1.0% (Gomes et al. 2013), which is lower than the data reported in another study for five unspecified cultivars: 30 ± 2% (Thakkar et al. 2007).In the latter study, data from 10 boiled cassava cultivars were also pooled to determine the amount of BC incorporated into the micellar fraction, which ranged from 0.3 to 2.3 µg BC/g by cultivar; however, results by cultivar were not distinguished (Thakkar et al. 2007).
After frying, the highest bioaccessibility of ATBC was shown for cultivar IAC 576-70 with an ME of 57.6%, and the lowest for BRS Jari at 32.4% out of 8 tested cultivars; on average, cultivars showed an ME of 43.2% (Berni et al. 2014; Figure 2B).One study measured BC ME in the BRS Jari cultivar, which resulted in 14.4 ± 2.4 ME% after frying.Another study found that BRS Jari contained 157.5 ± 22.4 µg ATBC per 100 g in the micellar fraction after frying (Gomes et al. 2013).Only cultivar BRS Jari was examined in terms of total carotenoid content (TCC) bioaccessibility after frying: TCC ME was 14.1 ± 2.25%, resulting in 300 ± 39.9 µg TCC per 100 g micellar fraction (Gomes et al. 2013).
After drying at 40 °C for 2 days and milling into flour, beta carotene equivalents (BCE) and TCC ME % were examined across 10 varieties of cassava (Figure 3A, B, blue bars) (Aragon et al. 2018).Variety GM 5194-5 demonstrated the highest BCE ME at mean ± SEM 14.7 ± 2.0% while variety GM 4571-3 yielded the highest TCC ME at 16.5 ± 2.1%.Flour milled from SM 3765-15, which was not examined prior to boiling, MEs were lower at 3.73 ± 0.7% for BCE and 4.7 ± 0.9% for TCC.Adding a 2-day fermentation step increased ME% for some cultivars while decreasing the ME% for others (Figure 3A, B, orange bars).Fermentation had the most beneficial impact on both BCE and TCC ME% for GM 4571-3, SM 3774-21, GM 5194-13, SM 3757-75, and SM 3762-15.Fermented flour from SM 3765-15, which was not examined prior to boiling, showed ME for BCE at 7.5 ± 0.5% for BCE and 8.9 ± 0.4% for TCC.Mixing hot water with fermented flour to create eba porridge from variety SM 3765-15 found MEs for BCE at 6.5 ± 2.4% and TCC at 7.7 ± 2.3%.
Differences in ME% for BCE and TCC for variety SM 3767-84 are shown in Figure 3C.Boiled cassava flour made from SM 3767-84 yielded MEs of 14.8 ± 2.4% for BCE, and 18.4 ± 3.8% for TCC.Boiling fermented flour made from SM 3767-84 resulted in negligible decreases in BCE and TCC (0.06% and 0.6%, respectively) resulting in MEs of 6.3 ± 1.1% and 7.7 ± 1.2%, respectively.Finally, mixing fermented flour with hot water (to make eba) resulted in MEs at 4.5 ± 1.2% for BCE and 5.3 ± 1.2% for TCC.
From these results, only SM 3767-84 was examined in detail comparing several processing steps.Interestingly, unfermented and boiled cassava flour resulted in the highest ME for both BCE and TCC, and adding fermentation appeared to result in lower ME%.This suggests that skipping the fermentation step, at least for SM 3767-84, will result in greater BCE and TCC incorporation into micelles and increase the potential for more BCE and TCC to be absorbed by the body.However, given the results comparing fermentation vs no fermentation for flour among the other cultivars, the ME% findings for SM 3767-84 may not be applicable across all cultivars.
Finally, three unspecified biofortified cassava cultivars were prepared into gari and fufu, which were analyzed for ATBC ME (%) (Thakkar et al. 2009).Mean ± SEM ATBC ME% for gari was approximately 30.3 ± 0.9%, and for fufu, 13.4 ± 0.4%, showing that boiled cassava or cassava prepared into gari may be more suitable for bioaccessibility than cassava prepared into fufu, perhaps due to its extended fermentation time (Hotz and Gibson 2007).
Several types of processing methods were utilized.After simulated oral, gastric, and intestinal digestion, the effects of various slicing and heating methods were examined for an OSP variety not specified (Figure 4A) (Bengtsson, Alminger, and Svanberg 2009).As expected, adding a fat source improved ATBC ME%, even when the percentage of ATBC in the supernatant was similar across processing methods.The best method for increased ATBC ME% was cutting OSP in slices, blanching at a low temperature and then boiling, and adding 2.5% fat.The lowest ME% was found when cutting OSP into cylinders, microwave heating, without fat.
Comparing ATBC ME% across varieties Ejumula, Beauregard, and a commercial OSP baby food after gastric and intestinal simulated digestion (Figure 4B), one study examined making foods such as chapati, mandazi, and porridge (Bengtsson, Alminger, and Svanberg 2009) while other studies compared cooking methods in making various versions of baby food (Dhuique-Mayer et al. 2018;Failla, Thakkar, and Kim 2009).Making chapati from Ejumula had the highest ATBC ME%, and blanching, sterilizing, adding oil and egg yolk powder yielded the highest ATBC ME% of the processing methods done on Beauregard.However, conclusions from these findings are limited given that the cultivar and processing techniques were unique and therefore not comparable.

Pumpkin
One study examined the bioaccessibility of BCE in 5 biofortified pumpkin (Curcubita moschata Duch) genotypes by measuring BCE in the micellar fraction after oral, gastric, and intestinal digestions (Ribeiro et al. 2015; Figure 5).Results are shown in Mean ± SEM.The findings showed the highest BCE in the micellar fraction for genotype Duch 58, across all three preparation methods: boiling, boiling with sugar, and steaming.The other four genotypes were notably lower in BCE but varied across the preparation methods.For Duch 346, Duch 13, and Duch 12, boiling resulted in the highest estimates.For Duch 129, all three preparation methods resulted in similar BCE content between 284 and 314 µg/g dry weight.Adding sugar appeared to reduce BCE content in some genotypes.

Carrots
One study examined the ME% of BC in meals made with biofortified carrots and other vegetables (Schmaelzle et al. 2014).Interestingly, the meal containing carrot mixed with banana had the highest ME at 26%, though variability was wide (Figure 6).

Beans
Iron and zinc solubility (%) were measured in one variety of biofortified beans, BRS Sublime, a carioca variety, after simulated oral, gastric, and intestinal digestion (Figure 7) from one study (Brigide et al. 2019).Out of the four preparation methods involving cooking, freeze drying with or without a soaking step resulted in the highest zinc solubilities at a 69.8 ± 0.19 to 76.8 ± 0.4%, while soaking and freeze-drying or oven-drying resulted in the highest iron solubilities of 16.5 ± 0.03% to 17.9 ± 0.3%.More studies are needed to draw larger conclusions; however, these results suggest that laboratory sample preparation style of biofortified beans may need to be considered depending on which micronutrient deficiency is of more concern.For example, if zinc deficiency is high, freeze-drying beans may be a preferred method; if iron deficiency is high, then soaking beans may be a step that should not be omitted from preparation.

Cowpeas
Two studies examined iron and zinc solubility in biofortified and conventional cowpeas (Coelho et al. 2021;Sant' Ana et al. 2019).The first study examined iron and zinc bioaccessibility in four cowpea varieties after cooking (in water at 100 °C for about 1.5 hours) and simulated digestion utilizing oral, gastric, and intestinal phases (Coelho et al. 2021).The conventional crop, BRS Guariba had the highest zinc solubility (Figure 8A).
Another study examined the bioaccessibility of iron after pressure-cooking with or without a previous germination step, plus gastric and intestinal simulated digestion (Sant' Ana et al. 2019), relative to the total iron content in the sample (Figure 8B).Germinating pearl millet prior to pressure-cooking appeared to result in increased iron    solubility in both the conventional BRS Nova (10.3 ± 0.3 mg/kg iron, compared to pressure cooking only iron solubility concentrations of 8.8 ± 0.3 mg/kg in the supernatant of digesta) and biofortified BRS Tumucumaque (12.7 ± 0.4 mg/kg iron, compared to pressure cooking only iron solubility concentrations of 12.7 ± 0.4 mg/kg in the supernatant of digesta) cowpeas.

Andean potato
Bioaccessibility of iron was examined in 10 varieties of Andean white potato after boiling, peeling, and pureeing and subjecting them to oral, gastric, and intestinal simulated digestion (Figure 9) (Andre et al. 2015).Varieties were similar in iron solubility concentrations, ranging from 15.8 to 28.5 µg/g dry weight, showing that 63.7 to 79% of the total iron in the boiled potato tissue was bioaccessible.

Bioavailability
Bioavailability of micronutrients across various biofortified crops was measured in humans, animal models, and Caco-2 cell models or with other methods such as mineral dialysis.Human studies are described in Table 4, while animal and experimental studies are described in Table S2, and Table S3 (Supplementary material), respectively.Studies that analyzed raw cowpeas and beans are shown in Table S1 (Supplementary material).

Summary of findings
In human studies, there was substantial heterogeneity in the types of processing done to the various cultivars, and the outcomes measured.To our knowledge, the bioavailability of micronutrients in several biofortified crops, including carrots, cowpeas, pumpkin, and combinations of crops, have not yet been tested in humans.

Cassava
The bioavailability of biofortified and conventional cassava porridge or gari with or without oil was measured in humans, specifically American women, in two studies (La Frano et al. 2013;Zhu et al. 2015).Procedures in both crossover-design studies were similar: participants consumed a diet low in vitamin A and carotenoids for four days and then nutrient-controlled meals for the next 3 days before the test day.On the test day, overnight-fasted participants consumed one of three randomly selected cassava, porridge, or gari preparations.Baseline and postprandial blood samples were collected before consumption of the meal and the following 9.5 hours.From these blood samples, the triacylglycerol-rich lipoprotein (TRL) plasma layer, also referred to as the TAG-rich plasma layer, was isolated and tested for BC and retinyl palmitate at each timepoint.After a two-week washout period, these procedures were repeated until all three cassava preparations had been tested.
In the first study, 12 women consumed either porridge made with biofortified cassava cultivar GM 905-69 (2.02 mg BC per 100 g), alone or mixed with peanut or rapeseed oil just before serving (La Frano et al. 2013).In the second study, 8 women consumed gari made from either a locally produced white cassava plus red palm oil (RPO, containing 1 mg BC) or the biofortified TMS 07/0593 cultivar (3.21 ± SEM 0.26 BC) without RPO (Zhu et al. 2015).Over the next 9.5 hours, the postprandial change in TAG-rich plasma BC (Figure 10A) and in retinyl palmitate (Figure 10B) concentrations were measured.White cassava gari fortified with RPO resulted in the largest elevation in BC within 2 hours, and remained higher across the postprandial period compared to the three biofortified varieties.TMS 07/0593 consumption initially elevated BC concentrations, but by 9.5 hours BC concentrations had decreased to baseline levels.GM 905-69 with or without edible oil resulted in differences in initial elevation of BC at 2 hours, but ultimately both   porridge versions reached similar levels of BC in blood at the 9.5-hour mark.Factors that should be considered in comparing these cassava preparations include that red palm oil-fortified white cassava had a slightly different ratio of carotenes than biofortified varieties; and that emulsification of carotenoids in fat (porridge with red palm oil, which inherently contains carotenoids) may increase their bioavailability, versus simply adding a fat source just before consumption (porridges with other edible oils added).
On the other hand, biofortified cassava porridge with oil resulted in the largest increase in retinyl palmitate concentrations in the TAG-rich plasma layer in blood compared to the three other preparations (Figure 10B).Both the GM 905-69 porridge and the TMS 07/0593 gari did not contain oil, and the former sustained higher concentrations of retinyl palmitate over the 9.5-hour postprandial period.Interestingly, there was higher BC content in the TMS 07/0593 (20.2 µg/g) (La Frano et al. 2013) than in the GM 905-69 gari (5.54 µg/g) (Zhu et al. 2015).This could be due to the differences in preparation, as gari contains a fermentation step while porridge does not.A previous study observed that increasing the number of fermentation days from 1 to 4 was paralleled by decreasing amounts of BC in gari (Abiodun, Ayano, and Amanyunose 2020), which may result in less retinyl palmitate converted.

Maize
The bioavailability of zinc and provitamin A has been studied in children and women in four studies, including one in children 1-5 years of age (Chomba et al. 2015), one in children 5-7 years of age (Tanumihardjo et al. 2019), one in men and women ~23 years of age (Titcomb et al. 2018), and one in women 18-30 years of age (Li et al. 2010).Each study examined different biofortified maize cultivars, preparations, and outcomes; these are summarized below.
The fractional absorption of zinc (FAZ), total absorbed zinc, and plasma zinc were examined in Zambian children 1-5 years of age (n=55, mean age: 29 months) who were randomized to and fed either: (1) experimental high-kernel zinc (biofortified) hybrid maize, zinc content: 34 µg/g; (2) control whole grain maize, zinc content: 21 µg/g, or (3) zinc-fortified maize, zinc content: 60 µg/g, ground into flour and prepared as phala and nshima porridge test meals (Chomba et al. 2015).Specifically, on the test day, children consumed 3 test meals made from their assigned maize for breakfast, lunch, and dinner.Halfway through each meal, a standardized amount of 70 Zn oral stable isotope solution was also consumed.Between lunch and dinner, 67 Zn was infused into the antecubital vein, after which children consumed a banana and had blood samples collected.After dinner, families were instructed to collect morning and afternoon urine samples for four days commencing on the fourth day after the test day.FAZ was determined via dual isotope tracer technique based on isotopic enrichments in urine from orally and IV administered isotopes, while total absorbed zinc was calculated by multiplying the total intake of zinc in the test meals by the FAZ.FAZ was significantly higher in the control maize group (mean ± SD: 0.28 ± 0.10) compared to the fortified maize group (0.20 ± 0.07), as expected since FAZ is inversely related to zinc intake.Further, the quantity of zinc absorbed was greater for the biofortified group (1.1 ± 0.5 mg/day) and fortified group (1.2 ± 0.4 mg/day) than the control group (0.6 ± 0.2 mg/day).Unexpectedly, higher mean plasma zinc was observed in the fortified maize group (58 ± 13 µg/dL) compared to the biofortified (59 ± 13 µg/dL) and control (58 ± 13 µg/dL) groups, despite having no differences in the exchangeable zinc pool (EZP), defined as the estimated size of the combined pools of zinc that exchange with zinc in plasma (Krebs et al. 2003).
In a randomized crossover trial among nine adults (men and women, 23.4 ± 2.3 years) in the US, the bioavailability of BC cryptoxanthin was compared between orange whole grain biofortified maize (BC: 500 ± 18 µg/two muffins), orange refined grain biofortified maize (BC: 500 ± 31 µg/two muffins), and white refined grain biofortified maize (BC: 10 ± 2.0 µg/ two muffins) (Titcomb et al. 2018).Participants followed a low beta-cryptoxanthin diet for 4 days, and then a stricter low carotenoid diet for 3 days as a washout to minimize serum carotenoids.Next, each type of grain was consumed in the form of muffins for 12 days, followed by a 1-week washout.Blood was collected at several timepoints during feeding and washout periods.Serum carotenoid area-under-the-curve (AUC), determined in µmol * L −1 * d increased by 131% [from baseline (0.11 ± 0.07 µmol/L) to endline, Day 12 (0.26 ± 0.08 µmol/L] and by 108% (0.14 ± 0.05 to 0.29 ± 0.12 µmol/L) in the whole grain and refined grain orange maize groups, respectively, which did not differ in concentration by endpoint.No differences between baseline and endpoint were seen in the white refined grain muffins group (~28 µmol/L).
Finally, in another study based in the US, bioavailability of BC and retinyl palmitate formation was measured in six women 18-30 years of age who consumed unfermented porridge made from either (DeExp × CI.7) × BC Orange hybrid maize (BCE: 495 ± 2 µg/250 g) or a white conventional maize, with a BC or vitamin A reference dose (Li et al. 2010).Similar to the previous study, participants followed a low vitamin A/ carotenoid diet for 3 days, followed by a stricter standardized low-carotenoid diet for another 3 days.Participants then fasted overnight and consumed the test porridge on the morning of the 7th day, after which blood samples were collected across a 9-hour period.The AUC from 0-9 hours was calculated for both retinyl palmitate formation as well as BC in the plasma triacylglycerol-rich lipoprotein layer, specifically in the chylomicron and large very low-density lipoprotein.Retinyl palmitate AUC values in the triacylglycerol-rich lipoprotein fraction were 80.9 ± 31.3 nmol * h/µmol BC reference dose and 80.3 ± 24.9 nmol * h/µmol vitamin A reference dose in the white maize with BC reference dose group and vitamin A reference dose group, respectively, as compared to 26.0 ± 10.2 nmol * h/µmol ATBC equivalents in the biofortified maize group.BC AUC values were 15.2 ± 8.8 nmol * h/µmol BC reference dose in the white maize with a BC reference dose, as compared to 4.3 ± 4.2 nmol * h/µmol ATBC equivalents in the biofortified group.Vitamin A equivalence values (the quantity of BC that has vitamin A activity equivalent to 1 µg of retinol) were calculated from the AUC values.Consumption of white maize porridge with a BC reference dose resulted in equivalence values of 2.34 ± 1.61 µg of trans b-carotene equivalents ingested/µg retinol formed, compared to 6.48 ± 3.51 µg of trans b-carotene equivalents ingested/µg retinol formed in the biofortified porridge group.
Taken together, provitamin A from biofortified maize is shown to be bioavailable by measurements of serum retinol, serum carotenoids, carotenoids and retinyl palmitate in the triacylglycerol layer and total body vitamin A stores.Zinc from maize was as bioavailable as fortified maize, demonstrating its utility in contributing to zinc status.

Biofortified potato
Studies measured the micronutrient bioavailability in iron-biofortified orange sweet potato (OSP) high in provitamin A content, regular OSP high in provitamin A content, iron-biofortified purple potato, or regular yellow potato in women (Jongstra et al. 2020), canned orange sweet potato in men (Haskell et al. 2004) and biofortified OSP or white potato in children (van Jaarsveld et al. 2005).
After consuming steamed and mashed iron-biofortified potato as part of a mixed meal, iron bioavailability was measured among iron-depleted Malawian women (n = 24) and iron-depleted Peruvian women (n = 35) participating in a randomized cross-over trial (Jongstra et al. 2020).In Malawi, iron-biofortified potato varieties included MUSG15052-2 (15.8 ppm Fe), compared to the regular OSP varietal, Irene (7.7 ppm Fe) (Figure 11A).In Peru, an iron-biofortified purple potato, CIP306417.79(5.6 ppm Fe) and yellow potato, Peruanita (3.0 ppm Fe), were assessed (Figure 11B).Iron-biofortified OSP resulted in a total iron absorption of 0.26 mg Fe/day, 0.12 mg more than regular OSP, though fractional iron absorption was similar.The iron-biofortified purple potato had a higher fractional iron absorption of 28.4%, compared to regular yellow potato at 13.3%; the difference in fractional iron absorption resulted in similar total iron absorptions of about 0.46 mg per day.These differences may be due to the iron-biofortified purple potato's higher polyphenol content, which can inhibit iron absorption.
In a study among Bangladeshi men (Haskell et al. 2004), the bioavailability of provitamin A from canned OSP, sauteed in corn oil with spices (2.25 mg (375 µg RE) as ATBC, was measured from plasma BC, and total vitamin A pool size measured as the plasma isotopic ratio ([2H4]retinol: unlabeled retinol).There were no differences in total vitamin A pool size between the canned OSP and control group, which contained white potato and cauliflower, after 60 days of feeding.However, plasma BC was higher in the intervention group than the control group (0.28 ± 0.05 µmol/L vs. 0.04 ± 0.01 µmol/L, respectively), though measuring plasma carotenoids does not account for preexisting carotenoid content in blood.
Overall, in humans, few studies have measured nutrient bioavailability from biofortified potatoes and sweet potatoes.These studies indicate that polyphenol content must be taken into account to achieve adequate iron absorption.Additionally, methods to measure provitamin A bioavailability from orange sweet potato need to distinguish between newly absorbed carotenoids and preexisting carotenoids.

Rice
One randomized cross-over study measured zinc bioavailability from either zinc-biofortified IR-68144 or conventional BR-28 parboiled, milled, and washed rice among Bangladeshi children (n=42) 36-59 months of age (Islam et al. 2013).FAZ (%) was measured using the dual-isotope tracer ratio technique, and total absorbed zinc (mg/day) was calculated as the product of the total dietary intake of zinc and FAZ.Children consumed a total of 3.83 and 4.83 mg zinc per day from the biofortified and conventional varieties, respectively, as part of a mixed diet.The mean FAZ from conventional rice (25.1%) was greater than the zinc-biofortified rice (20.1%).The mean total absorbed zinc was similar, at 0.96 ± 0.16 and 0.97 ± 0.18 mg/day from the conventional and biofortified rice respectively.This was unexpected given the 12.5 mg increase in zinc content in the biofortified rice compared to conventional rice, but washing the rice resulted in zinc losses resulting in a smaller net difference in zinc content between the varieties.More studies assessing bioavailability of micronutrients from rice are needed.

Beans
Several studies examined the bioavailability of biofortified or conventional beans; however, processing methods as well as techniques to measure bioavailability varied.In men and women aged 20-45 years, the isotopic ratio absorption of 58 Fe (added to the beans) and 57 Fe (reference dose) of cooked BRS Pontal (biofortified) or BRS Estilo (conventional) was similar (0.407 ± 0.04 vs. 0.409 ± 0.04, respectively) (Junqueira-Franco et al. 2018).Hemoglobin concentration was also examined, but the baseline iron status of the participants was not measured, limiting the utility of hemoglobin as a bioavailability parameter.
In several studies among women, biofortified beans consumed as part of a meal containing potatoes or rice provided more bioavailable iron than conventional beans, specifically 431 µg total iron absorbed per meal compared to 278 µg total iron per meal (Petry et al. 2016).In another study, carioca beans (an unspecified biofortified varietal and G4825, a conventional type) were included in a cooked composite meal, with or without partial dephytinization to reduce phytate content (Petry et al. 2014).Reducing phytates, an inhibitor of iron absorption, resulted in higher iron absorption in both bean types.Biofortified beans with their native phytate content resulted in 406 mg iron absorbed, 19% higher than for control beans.However, 50% and 95% dephytinization resulted in 599 and 746 mg iron absorption, respectively, which was 37% to 51% higher than conventional beans.Another study examined an iron-biofortified black bean, MIB 465, compared to SER16, a conventional red bean, among women with low iron status (Petry et al. 2012).These beans had similar PP levels and a PA:iron molar ratio, and were fed with potatoes or rice in multiple meals.Comparing to the conventional beans, biofortified beans had lower fractional iron absorption (6.30% vs 3.80%, respectively), resulting in similar total iron absorbed (225 µg/ meal vs. 234 µg/meal, respectively).Finally, another study found that the fractional iron absorption from whole SER 16 beans that were soaked for 4.5 h at 4 °C, boiled for 40 min was greater than that from dehulled-and subsequently, lower polyphenol content-SER 16 beans that underwent similar processing (Petry et al. 2010).
One study examined n = 57 non-anemic children 2-5 years of age who consumed either iron-and zinc-biofortified Pontal carioca beans or conventional Perola beans for 18 weeks (Vaz-Tostes et al. 2016).Erythrocyte zinc (µg/g Hb) was not significantly greater in the Pontal beans (59.4 ± 17.1 µg/g hemoglobin) vs. the Perola beans (38.8 ± 7.7 µg/g hemoglobin), but the change in erythrocyte zinc was significantly greater in the Pontal group (5.9 ± 18.3 µg/g hemoglobin) compared to the Perola group (−8.7 ± 12.4 µg/g hemoglobin), where baseline erythrocyte zinc was 53.5 ± 13.8 µg/g hemoglobin and 47.6 ± 12.9 µg/g hemoglobin, respectively.It is unclear whether children were randomized to each group, and whether the baseline erythrocyte zinc was significantly higher in the biofortified group, resulting in a null inter-group differences.
In sum, phytate and polyphenol content can reduce iron absorption, but the components of the full meal may impact these results.Randomized trials among children are required to understand the iron or zinc bioavailability from beans.

Pearl millet
The bioavailability of iron and zinc was examined in n = 20 women 17-35 years of age (including iron-deficiency with or without anemia) (Cercamondi et al. 2013) and n = 44 iron-deficient children at around 2-3 years of age (Kodkany et al. 2013a;2013b).Studies used stable isotopes ( 57 Fe, 58 Fe, 67 Zn, 70 Zn) to calculate total and fractional absorption of iron and zinc after consumption of test meals containing iron and zinc biofortified pearl millet milled into millet meal.In women, studied varieties included biofortified ICTP 8203 (8.8 ± 0.3 mg/100 g) and conventional DG-9444 (2.5 ± 0.1 mg/100 g).Fractional and total iron absorptions were presented in geometric mean (95% CI).Fractional iron absorption was similar between the two varieties [7.5% (5.6, 10.1%) and 7.5% (5.7, 10.0%), respectively), resulting in greater total iron absorption in women consuming ICTP 8203 [1.13 (0.83, 1.52) mg/day] compared to 0.53 (0.40, 0.70) mg/day for those consuming DG-9444.In children, iron and zinc absorption from meals-including sheera, upma, roti-made with biofortified (124 ± 7.7 µg iron/g grain, 84.1 ± 4.9 µg zinc/g grain) or conventional pearl millet (46.5 ± 5.0 µg iron/g grain, 43.7 ± 5.2 µg zinc/g grain) ground into flour (cultivars unspecified).Fractional iron absorption was (mean ± SD) 0.09 ± 0.08% and 0.06 ± 0.04% for the biofortified and conventional pearl millet, respectively, while the total iron absorption was 0.7 ± 0.5 mg/day and 0.2 ± 0.2 mg/ day, respectively.FAZ for the biofortified and conventional pearl millet in the children was 0.17 ± 0.08% and 0.2 ± 0.04%, respectively, resulting in a total zinc absorption of 1 ± 0.5 mg/ day and 0.7 ± 0.2 mg/day.While it is unclear which cultivar was used in this study, possible reasons for the lower iron absorption among children are differences in the population under study (children vs. women) and number of days of iron consumption (5 days per each varietal for the women, 1 day per each varietal for the children).

Discussion
In this review, we examined the bioaccessibility and bioavailability of conventionally-biofortified (biofortification via conventional crossbreeding) staple crops, focusing on studies conducted in humans.We found that crops biofortified to have higher amounts of micronutrient(s) generally had greater bioaccessibility and bioavailability of the targeted micronutrients relative to conventional crops or biofortified crops with lower inherent amounts of the micronutrient.Notably, we found that bioaccessibility and bioavailability varied considerably by genotype or cultivar, processing method, and whether the biofortified crop was analyzed alone or in combination with other biofortified or non-biofortified crops, or other ingredients-for example, cassava porridge with oil resulted in greater provitamin A bioavailability compared to without oil (La Frano et al. 2013), suggesting including a fat source with provitamin A-biofortified crops would be potentially of interest.These data also highlight the need to conduct further research to determine the best combination of these factors to achieve the highest bioaccessibility and bioavailability.
This review also found several gaps.While we found 67 studies reporting on bioaccessibility (18 studies) and/or bioavailability (58 studies), and most of these also reported micronutrient concentration in each processed crop or food product, only nine of these studies reported both the bioaccessible quantity and the absorbable, bioavailable quantity.A fuller picture, particularly in cell and animal studies, including all three components-static micronutrient content in the food meant for consumption, bioaccessibility of that amount, and finally the bioavailable amount-will allow researchers and crop developers to understand (1) how much of each targeted micronutrient actually is digested and becomes available for potential absorption, and (2) how much of that bioaccessible quantity actually is absorbed.This will reveal points at which crops could be further improved.For example, if only a small amount of the bioaccessible portion is actually absorbed, investigation into improving that absorption, such as adjusting anti-nutrient content, may be needed.On the other hand, perhaps nearly all of the bioaccessible portion is also bioavailable-in which case, breeding the crop to have even greater amounts of static micronutrient content may improve micronutrient absorption simply by increasing the amount of static micronutrient available.
Human studies were of primary interest, but out of the 18 human studies included, we did not identify any measuring micronutrient bioavailability for some biofortified crop types including rice, carrots, pumpkin, cowpeas, and meals based on combinations of crops.Further, most studies focused on women of reproductive age and children under 5, with a dearth of studies on older children and adolescents.Only two studies included men.For cassava, more studies in children are needed.For rice, only 1 human study was found in children, and for wheat, again only 1 human study in women was found.More studies for these crops in women and other populations are needed.
Methodologically, isolating the triacylglycerol-rich lipoprotein plasma fraction for measuring provitamin A bioavailability in maize and OSP is required, as the included studies' methods (such as measuring plasma beta carotene) may not take into account preexisting blood carotenoids (La Frano et al. 2014).Future issues to investigate are the role of the gut microbiome, as well as how consuming probiotics may also be a factor in bioaccessibility and bioavailability.

Conclusions
In conclusion, traditionally bred biofortified crops generally have higher bioaccessibility and bioavailability than their conventional counterparts.However, these estimates depend on the exact cultivar and baseline micronutrient content prior to processing, processing or cooking method, whether a biofortified crop is measured individually or in combination with other crops, and experimental model used.Measuring both bioaccessibility and bioavailability of target micronutrients in a given biofortified food-and particularly, combinations of either multiple biofortified crops or biofortified foods paired with other non-biofortified foods, considering biofortified crops are staple foods -will be critical to optimize nutrient availability and absorption, ultimately improving micronutrient intake and contributing to reducing micronutrient deficiencies.

Contributions of authors
V.M.F., M.N.N.M., E.M. and A.M.N. from GAIN were involved in conceptualization, interpretation of the results, and reviews and edits of the manuscript, but were not directly involved in conducting the search, deciding on study/report eligibility, data extraction, and statistical analyses.

Figure 3 .
Figure 3. (a) Beta carotene equivalents (BCe) and (B) total carotenoid content micellization efficiency (tCC Me) (%) after oven-drying and milling cassava cultivars into flour (unfermented, blue bars; fermented, orange bars).differences in BCe (blue bars) and tCC (orange bars) Me% for one varietal, sM 3767-84 after various processing steps, are shown in C. estimates reflect means of an unreported number of replicates and error bars represent sd.

Figure 4 .
Figure 4. (a) all-trans beta-carotene (atBC) in supernatant (orange bars) and after microfiltration of supernatant to ascertain micellization efficiency % (blue bars) in orange sweet potato.estimates reflect means of 3 replicates and error bars represent sd.Cultivar tested was not specified.(B) all-trans-beta carotene (atBC) in supernatant (orange bars) and after microfiltration to ascertain micellization efficiency % (blue bars) in orange sweet potato (osP).estimates reflect means of 9 replicates (mandazi, chapati, porridge, boiled + pureed) or 3 replicates (all other methods listed) and error bars represent sd.nr, not reported.

Figure 5 .
Figure 5. Beta-carotene equivalents (BCe) in the micellar fraction of duch genotype biofortified pumpkin after boiling (gray bars), boiling with sugar (orange bars), or steaming (blue bars).estimates reflect means of 4 replicates and error bars represent sd.

Figure 6 .
Figure 6.Beta-carotene (BC) micellization efficiency (Me) % in the micellar fraction of biofortified carrots either alone or in combination with other crops.estimates reflect means of 3 replicates and error bars represent sd.

Figure 7 .
Figure 7. Zinc solubility (orange bars) and iron solubility (blue bars) in Brs sublime (carioca) beans.estimates reflect means of 3 replicates and error bars represent sd.

Figure 8 .
Figure 8.(a) iron and zinc solubility in cooked biofortified (Brs arace, Brs tumucumaque, Brs Xiquexique) and conventional (Brs Guariba) cowpeas.estimates of iron and zinc solubility (%) reflect means of an unreported number of replicates; no variability estimate was reported.(B) iron and zinc solubility in biofortified (Brs tumucumaque) and conventional (Brs nova) cowpeas.estimates reflect means of 3 replicates; no variability estimate was reported.

Figure 9 .
Figure 9. iron solubility in biofortified andean potato genotypes.estimates reflect means of 3 replicates; no variability estimate was reported.values shown are the iron solubility concentrations found in the supernatant of digesta.dw, dry weight.

Figure 10 .
Figure 10.(a) Postprandial beta-carotene (BC) concentrations in the triacy gfisldlglycerol (taG)-rich lipoprotein layer in women after consuming biofortified cassava porridge with or without rapeseed or peanut oil (la Frano et al. 2013), biofortified gari without red palm oil (rPo), or white cassava gari with rPo (Zhu et al. 2015).(B) Postprandial retinyl palmitate concentrations in the triacylglycerol (taG)-rich lipoprotein layer in women after consuming biofortified cassava porridge with or without rapeseed or peanut oil (la Frano et al. 2013), biofortified gari without red palm oil (rPo), or white cassava gari with rPo (Zhu et al. 2015).

Table 1 .
search strategy across included databases.

Table 3 .
Characteristics of included studies for bioaccessibility.

Table 4 .
Characteristics of included studies for bioavailability: human studies.