Casein phosphopeptides (CPPs) result from the enzymatic hydrolysis of casein, the predominant protein found in cow's milk. CPPs contain clusters of phosphoserine residues, which can effectively bind calcium, and inhibit formation of insoluble calcium phosphates (Narva et al., 2003). Pure CPPs have been shown to promote calcium absorption in in vitro assays using HT-29 cells and Caco-2 cells (Ferraretto et al., 2003; Cosentino et al., 2010) and in vivo. Erba et al. (2001) who studied the intestinal calcium absorption in rats found that the absorption from CaCl₂ solutions decreased by 90% when in the presence of phosphate (Ca:Pi molar ratio of 1:1), but decreased by only 40% from Ca-CPP at the same Ca:Pi molar ratio.

On the other hand, Drago and Valencia (2004) who used an in vitro dialyzability assay to measure bioaccessibility from infant formulas found no increase in calcium dialyzability with increasing casein concentration, perhaps due to an incomplete casein proteolysis. Kennefick and Cashman (2000) similarly found no effect of three different casein phosphopeptide preparations on calcium dialyzability. A human study performed on nine Finnish postmenopausal women who received milk and milk enriched with CPPs, found no differences in serum calcium between the two groups. According to the authors, a stimulatory effect of CPPs on calcium absorption might have been observed had the subjects been vitamin-D deficient (Narva et al., 2003). Likewise, a calcium lactate drink supplemented with CPPs led to lower fractional absorption of calcium in adults than the unsupplemented kind (P = 0.015). Thus, there appears to be conflicting results on the effects of CPPs both in in vivo and in in vitro experiments, and more experiments are needed to clarify their role in mineral bioavailability.

Components in plant foods like phytate can form insoluble complexes with calcium, thereby reducing its bioavailability. Kennefick and Cashman (2000) reported that phytate had a more pronounced negative effect on calcium solubility than oxalate, wheat fibre-extract, barley fibre-extract, and casein. Liang et al. (2010), who used an in vitro solubility assay to compare rice-based foods from China, found that the high level of phytate in the brown rice (ranging from 14.9 to 19.4 mg of phytic acid/gram of rice) resulted in the lowest calcium solubility (12%) among all the rice foods tested. Brown rice germination, a process that results in phytate hydrolysis (Schlemmer et al., 2009), increased calcium solubility from 12% to 18%. Not surprisingly, the calcium solubility of white rice, which was produced by milling and polishing of the brown rice to remove the outer layer, increased with respect to brown rice (16.2% vs. 12%). Rice noodles, which are soaked and fermented prior to noodle making, had a percent calcium solubility ranging from 33.7% to 38.2% probably as a result of the low levels of phytic acid present (ranging from 0.0 to 4.1 mg of phytic acid/gram of rice noodles).

Phytate's inhibitory role on calcium absorption is significant only when the phytate to calcium molar ratio is above a certain value; below that value, the inhibitory effect is trivial. According to Frontela et al. (2009), the cut-off value is a molar ratio of 0.24. The authors, who used an in vitro digestion/Caco-2 cell uptake model to compare three different commercial cereals sold in Spain, found that calcium uptake was higher from infant cereals which had been dephytinized. However, results were significant (P < 0.05) only for the infant cereal which contained the highest phytate to calcium molar ratio. The other infant cereals tested had a phytate to calcium molar ratio ≤ 0.18 (Frontela et al., 2009).

Using dialyzability assays, Kamchan et al. (2004) found that vegetables containing the highest in vitro dialyzability for calcium (20–39%) corresponded to the ones that contained the lowest levels of phytate, fiber, and oxalate (e.g., kale, celery, collard, Chinese cabbage, and soybean sprouts). On the other hand, low dialyzable calcium (2–7%) corresponded to samples with high levels of oxalate and phytate (e.g., amaranth, white, and black sesame seeds).

Soluble fibers may have negative or positive effects on calcium absorption. In some European countries, carbohydrate gums such as alginic acid, guar gum, and locust bean gum are used as thickeners in commercial anti-regurgitation milk formulas for infants with evidence of gastroesophageal reflux (Bass and Chan, 2006). Bosscher et al. (2000) found that the incorporation of locust bean gum into an anti-regurgitation infant formula significantly lowered calcium dialyzability (9.4% ± 0.7%; P < 0.01) in comparison with the corresponding nonthickened formula (13.3% ± 1.2%). According to the authors, locust bean gum appears to affect calcium dialyzability by means of its physical properties to act as a thickening agent, rather than to its chemical ability to form complexes (Bosscher et al., 2003a). In another in vitro study, calcium availability was similarly reduced after supplementation with locust bean gum (11.9%) and high esterified pectin (11.7%), but it increased by 30% after inulin supplementation (Bosscher et al., 2003b). The ability of inulin to enhance calcium absorption has also been shown both in human (Abrams et al., 2005, 2007; Holloway et al., 2007) and animal (Coudray et al., 2005; Raschka and Daniel, 2005) studies.

Maillard reaction products are compounds in foods or beverages that are generated in the presence of heat, amino acids, and reducing sugars. The Maillard reaction induces browning of foods, has an effect on nutritive value, can have toxicological implications (such as the formation of acrylamide), can produce antioxidative components and it has also a large effect on flavor (van Boekel, 2006). Furthermore, Maillard reaction products may affect calcium bioavailability. Seiquer et al. (2010) used an in vitro digestion/solubility assay to compare the effect on calcium of thermally damaged milk, by comparing overheated milk (three cycles of sterilization at 116°C, 16 min) with ultra-high temperature (UHT) milk (150°C, 6 s). Calcium solubility was lower from the overheated milk, which has higher concentrations of Maillard reaction products, than from the UHT milk. The results were validated against rat feeding trials. Feeding rats the diet containing the overheated milk as the main protein source led to significantly lower values of apparent calcium absorption and retention than those found among animals fed the UHT milk diet. On the other hand, Mesías et al. (2009) found no effect of Maillard reaction products on Caco-2 calcium transport. The authors used two diets: a “white diet (WD)” (low in Maillard reaction products) and a “brown diet (BD)” (high in Maillard reaction products). For the preparation of the WD, cooking practices in which the Maillard reaction products develop (i.e., frying, toasting, and roasting) were avoided. The BD was rich in processed foods (breakfast cereals, baked products, chocolate, fried foods, toasted foods, and breaded foods, etc.,) with an evident development of browning and, thus, rich in Maillard reaction products. When 20 male adolescents were fed the two diets using a randomized crossover trial, there were also no differences in bioavailability (% calcium absorption; WD = 40.4%, BD = 38.2%) (Mesías et al., 2009).

Processing conditions were also tested. Viadel et al. (2006) used the Caco-2 cell uptake model to assess the effect of cooking on calcium availability. The bioavailability of calcium from cooked white beans (Phaseolus vulgaris L.) was higher (calcium uptake 18.8%) than from the raw beans (3.6%). Repo-Carrasco-Valencia et al. (2010) showed that boiled kañiwa (Chenopodium pallidicaule), a grain that grows in the Andes, had higher calcium dialyzability values than the raw kañiwa. On the other hand, calcium dialyzability was lower for the roasted and boiled quinoa (Chenopodium quinoa) than in the raw quinoa. According to the authors, cooking might increase the digestibility of the proteins with which calcium is bound, thus increasing the release of the mineral from any protein complexes. On the other hand, boiling might lead to an increase in mineral loss into the water.

Using an in vitro digestion/Caco-2 cell model, Etcheverry et al. (2005a) found no differences in calcium uptake results when human milk fortifiers (i.e., supplements containing protein, energy, minerals and an ample range of vitamins which are added to expressed human milk) were supplemented with three types of calcium salts: calcium glycerosphosphate gluconate, calcium phosphate, and calcium chloride.

Rao et al. (2007) used an in vitro solubility assay to measure calcium bioaccessibility from a commercial calcium-milk protein supplement. The results showed that the calcium present in this supplement was readily released by enzymatic digestion: with increasing pepsin concentration, more mineral was released from the supplement. This was probably a result of the proteolytic role that this enzyme has on the proteins present in this supplement, such as β-lactoglobulin, α-lactalbumin, and lactoferrin. Both β-lactoglobulin and α-lactalbumin have the ability to chelate/bind calcium. Thus, the proteolytic digestion of these proteins might liberate more calcium.

Organic acids might have an enhancing effect on calcium absorption (Pak et al., 1987). Perales et al. (2005) used Caco-2 cells to compare calcium uptake from infant formulas and from fruit juices containing milk and cereals (FMC). The calcium uptake was higher from the FMC samples than from the infant formulas, probably as a result of the presence of citric and malic acids in the juices. Shiowatana et al. (2006) also found an enhancing effect of citric acid on calcium absorption using a continuous flow dialysis system. The authors added organic acids to amaranth leaves and found that the enhancement on calcium dialyzability was most pronounced with the addition of citric acid followed by tartaric, malic, and ascorbic acids. The authors pointed out that the organic acids favorably affected calcium availability in spite of the likely presence of oxalate and phytate in the amaranth leaves. Bernardi et al. (2006) concluded that citric acid addition to a cookie formulation made with seeds of algarrobo (Prosopis alba), a leguminous tree, improved calcium dialyzability.

There are four methods for assessing calcium bioaccessibility and/or bioavailability: solubility, dialyzability, Caco-2 cell uptake, and transport. The Caco-2 cell model is a good model for predicting calcium bioavailability in humans (Cashman, 2003). The cells have features, including calbindin, vitamin D receptors, calcium transport channels, etc., that are essential for the study of vitamin D-mediated intestinal calcium absorption (Fleet et al., 2002). Furthermore, the in vitro digestion/Caco-2 transport method has been validated against human studies. When Mesías et al. (2009) compared two diets with different content of Maillard reaction products, the authors found no differences in calcium bioavailability results when studied in humans or in Caco-2 cells. The recommended method is therefore the in vitro digestion/Caco-2 uptake/transport method.

A comprehensive literature search in PubMed revealed that there are basically three main in vitro methods to determine the bioaccessibility and/or the bioavailability of carotenoids from foods.

An in vitro solubility method for measuring carotenoids has been utilized for the bioaccessibility screening of multiple foods (Hedrén et al., 2002a,b; Mulokozi et al., 2004). The method consists of a digestion method that simulates the human digestive system, followed by an assessment via HPLC of the types and quantity of carotenoids released from the food. Following the intestinal digestion, the samples are centrifuged, and the aqueous portion is extracted with petroleum ether that is then evaporated. The residue, containing the released carotenoids, is dissolved in a mobile phase solvent (consisting of methanol, methyl-t-butyl ether, and water) and filtered through a 0.45 μm pore size cellulose membrane filter and subjected to reverse phase HPLC. This method has been used after minor modifications to study the effects of thermal processing (Lemmens et al., 2011) and particle size (Lemmens et al., 2010) on β-carotene bioaccessibility from carrots.

A modification of this method was introduced by Reboul et al. (2006). What is essentially different in this method is that following the in vitro digestion the samples are ultracentrifuged at very high speeds and the aqueous portion is collected and passed through a 0.22 μm filter, thereby obtaining micelles. Thus, the authors ultimately quantify the carotenoids present in micelles (i.e., micellarized carotenoids) as a measure of bioaccessibility. This method has been used to compare carotenoid bioaccessibility from durum wheat and egg pasta (Werner and Böhm, 2011) and from different varieties and species of citrus fruits (Dhuique-Mayer et al., 2007); and to assess the effect of thermal processing on lycopene bioaccessibility from tomato pulp (Colle et al., 2010), and others vegetables.

The study by Reboul et al. (2006) has been validated against human studies. The in vivo bioaccessibility results were obtained from a study published by Tyssandier et al. (2003). In this study, Tyssandier et al. (2003) measured the percentage of carotenoids recovered in the micellar phase (i.e., micellarized carotenoids) from human duodenum during digestion of a carotenoid rich meal. The meal contained sunflower oil, tomato puree (main source of lycopene), chopped spinach (main source of lutein), and carrot puree (main source of β-carotene). As reported by Reboul et al. (2006), the bioaccessibility values from the in vivo human results were in the same range as those measured after the in vitro digestion model, with the exception of spinach lutein bioaccessibility which was about fivefold higher in in vitro than in in vivo studies.

Results from the solubility assay agree with what is expected to occur in vivo. Cooking, which results in a more efficient release of carotenoids from the food matrix by softening cell structures so that digestive enzymes can work more efficiently, resulted in higher β-carotene release from carrots compared to the uncooked kind (Hedrén et al., 2002a). Homogenization, which represents a mechanical disruption of the tissue, resulted in a sevenfold and an almost fivefold improvement of β-carotene bioaccessibility from the raw and cooked carrot samples, respectively (Hedrén et al., 2002a). Reboul et al. (2006) similarly found that percent β-carotene bioaccessibility increased with the level of processing: 2.5–2.6% from canned or raw carrots, 4.4% from pureed carrots, and 14.1% from carrot juice.

Addition of cooking oil to the carrots increased the percent of β-carotene released from both the raw and cooked carrots, but the results were more significant with homogenized samples (Hedrén et al., 2002a). Addition of oil similarly resulted in higher bioaccessibility values from orange fleshed sweet potatoes (Bengtsson et al., 2009a). Cooking green leafy vegetables (leaves of amaranth (Amaranthus spp.), cowpea (Vigna unguiculata), sweet potato (Ipomoea batatas), pumpkin (Cucurbita moschata), and cassava (Manihot esculenta) in red palm oil instead of sunflower oil, resulted in 1.7–2.5 times as much bioaccessible β-carotene (Hedrén et al., 2002b).

Different cooking methods will affect in dissimilar manner the release of carotenoid from foods. Microwaved orange fleshed sweet potatoes resulted in lower β-carotene release either in the absence or presence of oil (without oil: 23.7%; with oil: 27.5%) than boiling or steaming (without oil: 38–40.7%, with oil: 45%) (Bengtsson et al., 2009a). The authors concluded that the short heating period for the microwaved samples was not sufficient to obtain an adequate breakdown of the sweet potato cell matrix and, subsequently, the release and transfer of β-carotene to the supernatant/micellar fraction was impaired.

It has to be kept in mind that carotenoids are susceptible to destruction by heat. Mulokozi et al. (2004) compared two cooking methods on carotenoid bioaccessibility and retention from diverse African vegetables: a traditional cooking method, which consisted of boiling samples for 20–30 min in the absence of oil, and a modified cooking method, which consisted of reduced boiling times, and thus a potential for reduced carotenoid destruction. Bioaccessibility of β-carotene from the traditional cooking method ranged from 5% to 26% and from 18% to 77% from the modified method. Losses of β-carotene were 14–51% from vegetables prepared via traditional methods and 6–34% when prepared with the modified method. Thus, while cooking will increase carotene release and bioaccessibility from the food matrix, it will also lead to a reduction in carotene concentration, due to destruction of the molecule.

Lycopene and β-carotene appear to be sensitive to digestive conditions. Déat et al. (2009) found there was a 25% loss of lycopene in a simulated gastrointestinal TIM model that measured bioaccessibility from a meal containing red tomatoes and sunflower oil. While lycopene appeared to be stable in the gastric and duodenal compartments, it was in part degraded in the terminal parts of the small intestine. Blanquet-Diot et al. (2009) also showed that lycopene, along with β-carotene, were sensitive to destruction. Recovery percentages of β-carotene were lower for a red tomato-containing meal than from a yellow tomato-containing meal (P < 0.05). On the other hand, zeaxanthin and lutein were stable during in vitro digestion.

Garrett et al. (1999) were basically the pioneers in the development of the Caco-2 method for carotenoid bioavailability. The method relies on an in vitro digestion followed by the addition of the aqueous, filtered portion of the digestate (which would be representative of micellarized carotenoids) to Caco-2 cells. The cells are then harvested in phosphate buffered saline, containing ethanol, and BHT (butylated hydroxytoluene, an antioxidant) and stored at −20°C. On the day of the carotenoid analysis, the carotenoids are extracted from cells with a series of acetone and/or hexane additions. The pooled hexane extract is then evaporated to dryness, reconstituted and analyzed by reverse-phase HPLC.

The method by Garrett et al. (1999) has been used to study the bioavailability of carotenoids from vegetables (Huo et al., 2007), spinach puree (Ferruzzi et al., 2001), and orange fleshed melons (Fleshman et al., 2011), among others. A very similar method was introduced by Liu et al. (2004). In this method the authors measured both bioaccessibility and bioavailability, but they did not ultracentrifuge nor did they filter the samples, thus they did not necessarily add micellarized carotenoids to the Caco-2. The authors found that cooking corn samples enhanced the amount of lutein (0.9 fold) and zeaxanthin (1.2-fold) taken up by the cells compared to the raw grain (Liu et al., 2004).

A concern with this bioavailability method has been the stability of the micellar carotenoids during the incubation time with Caco-2 cells. Some of these bioavailability studies have incubation times as long as 6 (Garrett et al., 2000) or 8 h (Liu et al., 2004). Oxidative reactions might modify and affect the quantity of carotenoids during their exposure to the Caco-2 cells, thus it is important to keep incubation time to a minimum while not affecting the sensitivity of this assay. Garrett et al. (2000) observed that the addition of 500 μmol/L α-tocopherol to the medium might confer protection against oxidation and thus improve the stability of carotenoids.

Interestingly, Biehler et al. (2011) found that the addition of calcium, iron, and zinc significantly reduced both micellarization and Caco-2 uptake of total carotenoids from a spinach meal by up to 55% (Ca) and 90% (Fe, Zn), respectively. The minerals, which had been added at concentrations ranging from 3.8 to 25 mM, can presumably interact with free fatty acids, forming insoluble soaps, and with bile acids, thus compromising carotenoid emulsification. Also, minerals might reduce the size of the micelles, resulting in a marked and significant decrease of carotenoids in the micelles. Bengtsson et al. (2009b) also found that iron inhibits β-carotene uptake by Caco-2 cells, and that an inverse relationship between the beta-carotene uptake and iron concentration in the test solution exists (r² = 0.93, P < 0.05). With the addition of ferrous chloride (30 μM), the beta-carotene uptake was significantly reduced (P < 0.05), on average by 22%.

An extension to the above method involves transport studies in Caco-2 cells in which the cells are grown on Transwell inserts. Only a couple of transport studies have been conducted (O'Sullivan et al., 2008, 2010).

In all of the above methods, carotenoid bioaccessibility can be assessed; however, the Caco-2 method allows the measurement of both bioaccessibility and bioavailability. There are basically two in vitro solubility methods: one that measures soluble carotenoids and one that measures soluble micellarized carotenoids. In the first method there is always the possibility of overestimating the true bioaccessibility of carotenoids, because in the supernatant one is measuring carotenoids which are not micellarized as well as micellarized carotenoids. Micellarized carotenoids are obtained by measuring the fraction of the food carotenoid incorporated into the micelles (obtained from ultracentrifugation and filtration of the aqueous component through a 0.22 μM pore size membrane).

It is important to choose an in vitro method for carotenoid bioaccessibility that includes the extraction and measurement of carotenoids in micelles, the form in which the carotenoids will ultimately be absorbed by the intestinal cells. This is important for various reasons. First, there are compounds in foods that impair the transfer of carotenoids from the food matrix into the micelles, such as sucrose polyester, the structure in Olestra (Weststrate and van het Hof, 1995), fibers such as alginates, cellulose, and pectins (Yonekura and Nagao, 2009) plant sterols and stanols (Yonekura and Nagao, 2007) and divalent cations (Biehler et al., 2011). By the first solubility method, one could never assess this impairment in the carotenoid transfer from the food matrix to the micelle. Second, isomers of the same compound may incorporate into the micelle differently. For example, cis lycopene is more likely to be incorporated into micelles than trans lycopene, resulting in a higher bioavailability from the cis form than from the trans form. This might be as a result of a greater tendency for the trans isomer to form aggregates or due to its slightly lower solubility (Boileau et al., 1999; Failla et al., 2008). A higher micellarization was similarly reported for cis β-carotene than for trans β-carotene (Ferruzzi et al., 2006). This is of importance if different foods contain different amounts or ratios of cis and trans carotenoids. Third, different carotenoids might compete with each other at the level of entry into the micelle (van Het Hof et al., 2000) and different carotenoids might be incorporated into micelles differently. For instance, according to Garrett et al. (1999), the differential transfer of the carotenoids into micelles is dependent on their hydrophilicity. Carotenoids that have been released from the food matrix but are embedded in the very core of the fat droplet will not transfer to the micelle with the same ease as those carotenoids that are associated with the surface of the oil droplet. Thus, carotenoids like lutein are likely to be micellarized to a greater extent than α-carotene and β-carotene (O'Sullivan et al., 2010). Consequently, it is important to follow an in vitro digestion model that uses micelles to measure bioaccessibility.

An important question to ask is whether carotenoid bioaccessibility is a reliable predictor of bioavailability. According to O'Sullivan et al. (2010) and Garrett et al. (2000), this might indeed be the case: the amount of carotenoids present in the plant food and in their respective micelles will reflect the amount accumulated (a measure of uptake) and also secreted (a measure of transport) by Caco-2 cells. Thus, a measure of bioaccessibility might be sufficient as an estimation of how bioavailable the carotenoid is from the food in question.

When studying cellular carotenoid transport it is important to note that the presence of the cytosolic enzyme (β-C 15,15′-oxygenase) responsible for the cleavage of β-carotene into retinoids could affect the amount of carotenoids being released and consequently measured at the basolateral end. This is of no concern, however, when working with the parent line (HTB 37) of Caco-2 cells as this cell line does not produce the enzyme. However, in two clones of Caco-2 cells, PF11 and TC7, β-C 15,15′-oxygenase has been detected (During et al., 1998).

It is very difficult to compare results from different carotenoid in vitro bioavailability studies. As noted previously, one of the most important factors limiting the availability of carotenoids from foods is their release from the food matrix (Parker, 1996). Thus, not only will the species, cultivar, growth conditions, harvest method, storage conditions affect carotenoid levels in the food, the processing conditions will most certainly affect the bioaccessibility data. Added to this is the wide inter- and, even intra-, variations of different research laboratories in preparing the samples for in vitro digestion experiments, making the carotenoid bioavailability results very difficult, and almost impossible, to compare and make sense of.

Another problem one finds when reviewing the literature is the lack of homogeneity among different labs in presenting the data. For the most part, the results of carotenoid bioavailability are expressed as a percentage of the amount taken up by the cells, relative to the total amount of carotenoids in the micelles that are given to the cells. However, some authors express results in terms of the amount of absorbed carotenoids per cell protein. It would be advisable to present the data both as a percentage and as an absolute amount absorbed. A higher percent carotenoid uptake from one test meal versus another does not translate into a higher carotenoid amount taken up by the cells if the test meals have different carotenoid concentrations to begin with, or if the amount of carotenoids in the micelles is different.

Without a doubt, the method that has been used the most, in the past decade, to measure folate bioaccessibility is the dynamic gastrointestinal model (TIM) (Arkbåge et al., 2003; Verwei et al., 2003; Ohrvik and Witthöft, 2008; Öhrvik et al., 2010). It has been used to study both folate and folic acid bioaccessibility from foods like orange juice, breads, milk, and yogurt. Using this model, Verwei et al. (2003) found that folate binding proteins (FBPs) added to milk samples have different binding characteristics for folic acid and for 5-CH₃-H₄-folate. During gastric passage, a large fraction of folic acid remains bound to FBPs, whereas a large fraction of 5-CH₃-H₄-folate dissociates from the FBP, increasing the bioaccessibility of the vitamin. Fortification of milk with 5-CH₃-H₄-folate leads to higher folate bioaccessibility (~70%) than that fortified with folic acid (~60%). The authors attributed this difference to a lower binding affinity of FBP for 5-CH₃-H₄-folate compared with folic acid at the pH range of 5–7.4. A lower binding affinity could result in a higher release or dissociation of the folate compound from the folate-FBP complex during gastric passage and/or through the duodenum.

Arkbåge et al. (2003) also found a more pronounced inhibitory role of FBPs on folic acid than on folate (P < 0.05). In the absence of FPBs, folate bioaccessibility was 82% from yogurt fortified with folic acid and 5-CH₃-H₄-folate (Arkbåge et al., 2003). When FBPs were added, folic acid bioaccessibility decreased to 34% and 5-CH₃-H₄-folate bioaccessibility decreased to 54%. Interestingly, this study also found that FBPs were somewhat resistant to the digestive enzymes in the stomach and small intestine, and this resistance was dependent on the folate form present in yogurt. The FBP stability in yogurt fortified with folic acid (34%) was twice as high as the FBP stability in yogurt fortified with 5-CH₃-H₄-folate (17%). Thus, a relationship between the inhibitory effect of FBP on the bioaccessibility of folic acid and 5-CH₃-H₄-folate, and the FBP stability in folic acid and 5-CH₃-H₄-folate fortified yogurt appears to exist (Arkbåge et al., 2003).

While the TIM method allows the removal of digested material (along with the subsequent determination of folate) at any step of the digestion model, it only measures bioaccessibility, and not absorption. Absorption ultimately depends on the ability of the brush border enzyme glutamate carboxypeptidase II to deconjugate the polyglutamate forms of folate.

In 1998, Seyoum and Selhub incorporated a method in which the susceptibility of food folates to glutamate carboxypeptidase II was studied. In this method, the food was subjected to a peptic digestion at low pH and then incubated with a porcine jejunal brush border membrane extract which contained the hydrolase enzyme. The folate bioavailability index was assessed by comparing the concentration of the monoglutamyl folate in the experimental group to the total folate concentration in the control group as follows: Folate bioavailability index=(M/T)×100 where M is the monoglutamyl folate concentration after treatment and T is the total folate concentration (5-CH₃-H₄-folate) in the control group.

The authors compared the folate bioavailability indices with the indices of bioavailability for the same foods (egg yolk, cow's liver, lettuce, lima beans, orange juice, cabbage, and baker's yeast) reported in human studies (Tamura and Stokstad, 1973; Babu and Srikantia, 1976). The results showed that the two sets of indices have a significant correlation (P = 0.068). Thus, this method measures the potential for food folates to be absorbed.

The main in vitro method which has been used to assess folate bioaccessibility is the dynamic TIM. This model mimics the human digestive system in a way that cannot be replicated by other in vitro systems. Effects like churning, peristaltic movements, flow of saliva, etc., are all replicated and controlled in the TIM. However, this model only measures bioaccessibility, and not absorption. Absorption of dietary folate ultimately depends on the ability of an intestinal enzyme located on the cell surface (called glutamate carboxypeptidase II) to deconjugate the polyglutamate form to the monoglutamate form. Thus, it is important not to rely solely on bioaccessibility results since absorption would ultimately depend on the deconjugation of folate and the effect that certain food components might have on the activity of glutamate carboxypeptidase II. Further studies which incorporate the susceptibility of food folates to the intestinal enzyme (Seyoum and Selhub, 1998) should be conducted.

Several in vitro studies have been performed on the effect of phytate and dephytinization on iron bioavailability from plant-based foods. Afify et al. (2011) used an in vitro digestion/solubility assay to measure iron bioaccessibility from three white sorghum varieties (Sorghum bicolor L.). Iron solubility was 8.02–13.60% for the raw sorghum grains, 14.62–20.75% for the soaked grains, and 16.67–20.63% for the germinated grains. Soaking and germination are processes that activate the endogenous phytase present in the plant material. Soaking may also lead to a phytate reduction through water solubilization and subsequent leaching (from the food) of some phytic acid salts. Interestingly, after soaking and germination the iron content in the seed significantly decreased, which could be attributed to leaching of iron ions into the soaking medium (Afify et al., 2011).

Caco-2 iron uptake from infant cereals was improved after treatment with exogenous phytases (Frontela et al., 2009). Likewise, iron solubility of whole faba bean flours was significantly improved by phytate degradation (Luo et al., 2010). Total dephytinization of dehulled faba bean flour led to an increase in iron solubility, but dephytinization of hull flour had no effect on iron solubility. This is because the hull is rich in fiber and tannins, but has a low content of phytate compared to the dehulled faba bean. Phytate is more localized in the cotyledon of the bean. Treatment with endogenous phytases (achieved by incubating the samples at 55°C in the presence of acetate buffer) significantly decreased (P < 0.05) the total iron content of faba bean flour from 3.52 to 3.15 mg/100 g because of iron leaching into the medium. By contrast, when exogenous phytases were added, the total iron content was apparently less affected, probably because it was complexed with the added proteins (Luo et al., 2010).

Pynaert et al. (2006) compared processed vs. unprocessed complementary foods (CF) in Tanzania. The processed CF consisted of germinated, autoclaved and dried finger millet, kidney beans, roasted peanuts, and mango puree. The same ingredients in identical proportions were used for the unprocessed CF. Iron solubility was higher in the processed samples (19%) than in the unprocessed samples (5%) (P < 0.001). The in vitro solubility results, however, did not agree with a field trial in which no improvement in iron status could be demonstrated in children who were fed the processed food (Mamiro et al., 2004). The reduction in phytates by 34% and improvement in iron solubility to 19% due to processing might not have been enough to compensate for the rather low iron content of the complementary food.

Engle-Stone et al. (2005) studied iron bioavailability from an iron-phytic acid (PA) solution (1 Fe:20 PA molar ratios) with different amounts of AA added to achieve Fe: AA molar ratios of 1:0, 1, 5, 10, 20, 40, and 100. Caco-2 iron uptake from the 1:20 molar ratio of iron to phytic acid decreased Caco-2 cell ferritin formation by 91% in comparison to the control (i.e., Fe without PA). When AA was added (1:20:1 molar ratio of FeCl₃:PA:AA) iron uptake increased by 180% relative to the control (i.e., Fe without PA or AA). Additional AA increased cell ferritin formation, but the effect was maximal at a 1:20:10 molar ratio of FeCl₃:PA:AA. Clearly, the AA was able to partially reverse the effects of phytate inhibition under these conditions.

On the other hand, Beiseigel et al. (2007) found no differences in Caco-2 ferritin formation between two maize varieties, one of which contained more phytate (7% more) than the other. Adding AA to the two maize samples, significantly enhanced iron uptake from 2% to 7%. When Caco-2 values were compared to absorption values obtained from female participants who were fed the maize samples in the presence and absence of AA, the authors found that the Caco-2 model accurately predicted relative iron absorption from the maize meals (Beiseigel et al., 2007).

The bioaccessibility/bioavailability of different iron salts was also studied. Kapsokefalou et al. (2005) reported that iron dialyzability was higher in pasteurized milk samples fortified with iron pyrophosphate, ferrous lactate and ferrous bis-glycinate (P < 0.05) than with ferrous sulfate and ferrous gluconate. However, in commercial pasteurized and UHT milk products, there were no differences in dialyzable iron in products fortified with ferrous lactate or ferrous sulfate. Zhu et al. (2009) also found increased Caco-2 iron uptake from pure ferric pyrophosphate than from any pure iron compounds or chelates. Exposure of iron to pH 2 followed by adjustment to pH 7 markedly decreased FeSO₄ bioavailability but had a smaller effect on bioavailabilities from ferric pyrosphosphate and sodium iron(III) ethylenediaminetetraacetate (NaFeEDTA), suggesting that these chelating agents minimize the effects of pH on iron bioavailability.

Kloots et al. (2004) found that the iron dialyzability was higher in chapatis (a typical Indian bread) prepared from whole-grain wheat flour fortified with NaFeEDTA or SunActive® Fe (ferric pyrophosphate) than those fortified with ferrous sulfate. Iron dialyzability from whole-grain wheat flour baked into chapatis was similar for all added iron sources (ferrous sulfate, ferrous lactate, ferrous fumarate, ferric pyrophosphate, carbonyl iron, electrolytic iron, Ferrochel® amino acid chelate, ferric amino acid chelate taste free [TF], and Lipofer™ which is a complex of ferric pyrophosphate, starch, and lecithin).

The effectiveness of disodium EDTA (Na₂EDTA) on enhancing iron bioaccessibility was studied by Walter et al. (2003). Flour tortillas were fortified with different iron salts in the presence and absence of Na₂EDTA. Iron dialyzability from flour tortillas fortified with reduced iron alone, reduced iron- Na₂EDTA, ferrous fumarate Na₂EDTA and native iron plus Na₂EDTA were 8.8, 15.3, 10.2, and 18.2%, respectively. Native iron from corn-masa flour had a dialyzability of 1.4%, but upon addition of Na₂EDTA, it increased to 18.2%. Like AA, Na₂EDTA may combine with the iron fortificant, thereby enhancing dialyzability, with the advantage that it is stable during storage and processing. The authors conducted human iron absorption studies using the same flour tortillas used in the in vitro solubility studies. The human bioavailability results closely paralleled the ranks obtained in the dialyzability studies. The in vitro dialyzability and in vivo human absorption results were highly correlated (r = 0. 89, P < 0.001).

The variable effects of different proteins on iron bioaccessibility and/or bioavailability was assessed. Bosscher et al. (2001a) found that iron dialyzability was reduced by soluble dietary fiber. However, the inhibitory effect of soluble dietary fiber was more pronounced in casein than in whey-based formulas. Iron dialyzability from casein- and whey-based formulas supplemented with 0.42 g of locust-bean gum/100 mL were 0.32% and 1.45% (P < 0.05), respectively. Drago and Valencia (2004) similarly found a more pronounced inhibitory effect of casein than whey on iron dialyzability.

Hypoallergenic formulas (which were based on protein hydrolysates) resulted in the highest iron dialyzability values, followed by a preterm formula, the followup and soy, an adapted formula, and finally one without lactose (García et al., 1998). No differences, however, were observed in formulas having whey or casein as the main protein fraction.

The addition of milk to fortified fruit beverages containing either iron or iron and zinc had a positive effect on iron uptake by Caco-2 cells. There was a significant (P < 0.05) threefold increase in ferritin formation in samples with milk vs. no-milk added samples. Intact bovine milk proteins may maintain iron in a soluble form in the digestive tract, but inhibit its absorption unless the proteins are hydrolyzed. The increase in iron uptake could have been due to the effect of CPPs formed during gastrointestinal digestion (Cilla et al., 2008).

In the presence of tannic acid (TA), Caco-2 iron uptake was significantly inhibited (98%) in comparison to the control (i.e., Fe without TA). An increase in cellular iron uptake was observed when AA was added at a molar ratio of 1:1:1000 Fe:TA:AA. However, the ferritin formation (i.e., iron uptake) at the 1:1:1000 Fe:TA:AA ratios was only half the ferritin observed for the control (i.e., Fe without TA or AA) (Engle-Stone et al., 2005).

Using the Caco-2 iron uptake assay, Miret et al. (2010) studied different food matrices (water, dough, powdered drink, and chocolate) containing one of the following iron forms: iron sulfate, hemoglobin, or sodium iron chlorophyllin, a water-soluble semisynthetic chlorophyll derivative where the magnesium in the porphyrin ring has been substituted by iron. Iron uptake from hemoglobin was not reduced by the dough but was significantly reduced by the powdered drink and chocolate (a source of polyphenols). This was interesting, since polyphenols are known to inhibit nonheme iron, not heme iron, absorption. According to the authors, polyphenols from wine and tea have been shown to increase pepsin activity, and this could influence the digestion of hemoglobin and the solubility of the released heme. Peptides derived from hemoglobin digestion are known to maintain heme solubility and to allow heme uptake. Extensive digestion of the peptides could decrease heme solubility and consequently, heme-iron bioavailability. Iron uptake from sodium iron chlorophyllin was significantly reduced by the dough and powdered drink but not by chocolate. However, the iron uptake of hemoglobin and sodium iron chlorophyllin was significantly higher than that of FeSO₄.

A handful of studies have been conducted using the in vitro digestion/Caco-2 uptake model to compare white and red common bean (Phaseolus vulgaris L.) (Hu et al., 2006; Laparra et al., 2008; Tako et al., 2009; Tako and Glahn, 2010). All of them showed that Caco-2 iron uptake was lower from the red beans than from the white beans, probably due to the higher presence of polyphenolic compounds in the colored beans (Tako et al., 2009) that included flavonoids such as kaempferol and astragalin (Laparra et al., 2008). Animal trials were conducted with 1-week-old chicks (Gallus gallus) fed white beans and red beans with and without iron for 8 weeks. Following the 8 weeks, divalent metal transporter 1 (DMT1; iron-uptake-transporter), duodenal-cytochrome-B (Dcytb; iron reductase), and ferroportin (iron-exporter) expressions were higher (P < 0.05) in the intestines of the group fed red beans vs. other groups (i.e., groups fed red beans + Fe, white beans, or white beans + Fe). Higher expression of DMT1, Dcytb and ferroportin (as was seen in the red bean group) is indicative of a more iron deficient state (Tako and Glahn, 2010). Iron absorption from white beans was also higher in anemic piglets compared to red beans (14–16% vs. 9–10.5%, P < 0.05) (Tako et al., 2009).

The low cellular iron uptake results are supported by the lower dialyzability of Fe from colored beans (1.5-2.7%) than white beans (12.1-18.8%) (Laparra et al., 2008). Interestingly, there was no significant difference in iron uptake from red and black beans, in spite of differences in iron concentration. The MIB465 sample contained 49.7% more Fe (up to 30 μg g−1 of bean, dw) than DOR500, but both of the black bean (DOR500 and MIB465) genotypes exhibited no significant (P < 0.05) difference in Fe uptake (Laparra et al., 2008).

Beiseigel et al. (2007) found that following an in vitro digestion, Caco-2 cell uptake was higher from cooked great northern beans, which are white in color, than from cooked pinto beans, which are a mottled red color. Caco-2 ferritin values increased when the beans were mixed with orange juice, a source of AA. Human subjects were also fed the cooked beans with and without orange juice. When the in vitro data were compared to the in vivo data, the authors found that the Caco-2 cells inaccurately predicted lower iron bioavailability from pinto beans than from great northern beans, and a lesser enhancing effect of AA with pinto beans than with great northern beans.

Solubility, dialyzability, Caco-2 uptake and/or transport assays have all been used as iron bioaccessibility/bioavailability screening methods. It is important, however, to be cautious about solubility assays. A review by Miller and Berner (1989) concluded that discrepancies do exist between in vitro iron solubility and in vivo iron absorption results, especially when the effects of protein on iron bioavailability are being assessed. On the other hand, the authors stated that iron solubility appears to be a reliable indicator of AA effects on bioavailability. Dialyzability (Walter et al., 2003) and Caco-2 uptake studies (Au and Reddy, 2000; Yun et al., 2004) have been validated against human absorption results. However, a significant drawback to the dialyzability method is that when iron diffuses into the dialysis bag, during the intestinal digestion phase, a significant amount of the iron immediately becomes insoluble at the higher pH (van Campen and Glahn, 1999), which might significantly affect results. The in vitro digestion/Caco-2 uptake model is the recommended bioavailability method for iron, because it is an assay that can provide more information than bioaccessibility studies alone, such as the impact of food components on absorption rate and efficiency, and the possible competition amongst nutrients or between nutrients and food components for the same absorptive site.

None of the methods currently used to assess magnesium bioaccessibility/bioavailability have been validated against human absorption studies, and no method has been used extensively. Thus, there are insufficient data to make a recommendation on the most appropriate bioavailability/bioaccessibility method for this particular nutrient.

The main in vitro bioavailability method that has been used repeatedly to measure the bioaccessibility of polyphenols is in vitro solubility. This method has been used to test the bioaccessibility of various polyphenols in extra virgin olive oil (Dinnella et al., 2007), orange (Gil-Izquierdo et al., 2001, 2003) and pomegranate juices (Pérez-Vicente et al., 2002), broccoli (Vallejo et al., 2004), cocoa liquor (Ortega et al., 2009), and raspberries (McDougall et al., 2005), among other foods. Fazzari et al. (2008) studied the polyphenol bioaccessibility of cherries (Prunus avium L.) of different degree of maturity. The authors found that the percent polyphenol bioaccessibility was higher in immature cherries (i.e., picked 1 week early) than the mature or overmature (i.e., picked 1 week late) cherries. Because immature cherries had a lower concentration of polyphenols, the actual bioavailable amounts of these compounds were lower than for mature and overmature fruit (Fazzari et al., 2008).

McDougall et al. (2005) studied the bioaccessibility of polyphenols from raspberries (Rubus idaeus L., variety Glen Ample) in the presence of different food matrices. Results showed that co-digestion of raspberries with commonly combined foodstuffs such as bread, breakfast cereal, ice cream, and cooked minced beef gave different patterns. Phenol bioaccessibility was slightly decreased by co-digestion with ice cream and cereal, whereas bread had no effect and minced beef caused an increase. Anthocyanin bioaccessibility was either unaffected or increased by co-digestion with the foodstuffs. Thus, anthocyanins may bind to food matrices during digestion, protecting them from degradation and increasing their bioaccessibility (McDougall et al., 2005).

Only one in vitro polyphenol bioavailability study using Caco-2 cells has been conducted. This study assessed the absorption of resveratrol from boiled and roasted peanuts (Chukwumah et al., 2011). Digests of roasted peanuts showed higher resveratrol transport as opposed to boiled peanuts, even though bioaccessibility results were higher for boiled than for roasted peanut, which supports the idea that a higher amount does not necessarily imply higher bioavailability.

It is important to note that Caco-2 cells are able to metabolize some polyphenols. Kern et al. (2003) found that after a 24 h exposure of hydroxycinnamates to differentiated Caco-2 cells, several metabolites were generated including ferullic acid-sulfate, synaptic acids-sulfate, p-coumaric acid-sulfate, and methyl ferulate-sulfate. Similarly, incubation in the presence of diferulates resulted in free acid metabolites. Furthermore, after a 2 h incubation, only 10% of the original methyl ferulate (a hydroxycinnamic acid) was present in the media, disappearing completely by 4 h of incubation. Yi et al. (2006) who added anthocyanins from blueberries to Caco-2 cells grown on Transwell membranes, suggested that anthocyanins can be degraded and demethylated during absorption and transport by the cells.

Caco-2 cells therefore have the capacity to carry out processes like glucuronidation, sulfation and methylation which are normal metabolic processes that polyphenols undergo both in the small intestine and in the liver. Thus, it appears that in order to assess uptake or transport in this human cell line, following the incubation it would be appropriate not only to measure the original polyphenol present but also any possible metabolite/degradation products that might have resulted from it. This is something that would be very challenging to do if one is not aware of all the possible metabolic and degradation products that might arise.

For polyphenols, there is not a substantial amount of evidence as to which method is the most appropriate for measuring bioaccessibility/bioavailability. In general, in vitro methods are somewhat limited for the assessment of polyphenol bioaccessibility/bioavailability due to the active participation of the colon in the digestion and absorption of these compounds. An exception would be certain gastrointestinal models, like TNO's TIM model, which allow the incorporation of colonic fermentation experiments. Certainly, the in vitro solubility method has been utilized more frequently over the past 10 years than the other methods, and is more economical. No method has been validated against human absorption studies. Thus, there are insufficient data to make a recommendation on the most appropriate bioavailability/bioaccessibility method for this particular phytochemical.

Based on an extensive literature search, there are currently two methods for assessing vitamin B₆ bioavailability and/or bioaccessibility: solubility and uptake by Caco-2 cells. No method has been validated against human absorption studies. Thus, there are insufficient data to make a recommendation on the most appropriate bioavailability/bioaccessibility method for this particular nutrient.

Because bioavailability of dietary vitamin B₁₂ is dependent on the complex production and release of proteins from the mouth and stomach (i.e., haptocorrins and intrinsic factor), it is no surprise that there are no in vitro methods for studying vitamin B₁₂ bioavailability from foods. There is one in vitro method which measures bioaccessibility (Miyamoto et al., 2009). In this method, the authors studied the bioaccessibility of vitamin B₁₂ from Korean purple lavers, an edible alga, which appears to contain more vitamin B₁₂ than other edible algae. Bioaccessibility was based on a gastric and intestinal digestion of the sample. Following the intestinal digestion, the samples were centrifuged and the soluble fraction was applied to a Sephadex G-50 fine gel filtration column. The macromolecular and free B₁₂ fractions were estimated with blue dextran and pure B₁₂ by measuring absorbance at 600 and 551 nm, respectively. The results indicated that the dried purple laver could be well digested only under the pH 2.0 conditions, but not under the pH 4.0 and 7.0 conditions. Under the pH 2.0 and 4.0 conditions, about half of the B₁₂ found in the dried purple laver was soluble. Release of B₁₂ from the purple laver was significantly decreased under the pH 7.0 conditions, a pH that serves as a model for severe atrophic gastritis, which prevails in elderly people (Miyamoto et al., 2009).

Unlike other nutrients, bioaccessibility of vitamin B₁₂ does not equal bioavailability due to the complex physiological process involved in B₁₂ absorption. An extensive literature review, did not reveal any in vitro method for measuring vitamin B₁₂ bioavailability from foods except for the one previously discussed. The recommended method is to conduct absorption studies (using either fecal excretion or body retention methods) in human subjects.

The study by Goncalves et al. (2011) is not a “standard” method for measuring vitamin D₃ bioavailability from foods. For starters, the study did not include an in vitro digestion of the samples (mostly because pure vitamin D₃ was used, as opposed to vitamin D₃ from foods), but that could easily be incorporated into the methodology. Bioavailability of vitamin D₃, unlike the bioavailability of other nutrients, is not exclusively dependent on the release of the nutrient from the food matrix (a measure of bioaccessibility) or on the absorption of the vitamin by intestinal cells (a measure of bioavailability). Vitamin D₃ bioavailability is also dependent on the metabolism of the vitamin which includes the conversion of vitamin D₃ into 1,25(OH)₂D₃. Thus, there is insufficient data to make a recommendation.

Two methods have been used to measure vitamin E bioaccessibility: solubility and the TIM developed by TNO. The validation studies performed using solubility assays showed inconsistent results (Granado-Lorencio et al., 2009). The gastrointestinal model has not been used extensively, and has yet to be validated against human absorption data.

Luo et al. (2010) studied the zinc solubility from whole faba bean flour. Solubility was 31.6%, but it increased to 45.4% and 52.3% after the endogenous phytases were activated. Exogenous phytases, on the other hand, did not improve zinc bioaccessibility (35.4%). According to the authors, the added enzyme (which is a protein) might interact with zinc and prevent it from becoming soluble, in spite of the dephytinization. Both treatments (with or without exogenous phytases) reduced the total zinc content of faba bean flour, by 16% and 32% after a short and long incubation period, respectively, probably as a result of leaching into the medium.

Lestienne et al. (2005) evaluated the zinc solubility of whole pearl millet flour. Nondephytinized samples had a zinc solubility of 14.2%. After phytate degradation by endogenous phytases, zinc availability was increased to 23.8% and 27.4% after incubation for 1 and 3.5 h, respectively. However, treatment with exogenous phytases did not improve zinc bioaccessibility which could be due to the added proteins (i.e., enzymes) interacting with the zinc in the millet flour and preventing its solubilization. As noted by the authors, a review by Matsui (2002) reported that zinc bioavailability was not increased by addition of exogenous phytases in a number of animal studies. The authors also stated that some of the hydrolysis products of IP6 (i.e., phytate with 6 phosphate groups), particularly IP5 and possibly IP4 and IP3 (phytate with 3 phosphate groups), participate in the inhibition of zinc availability. Regarding the content of zinc, there were no differences between the treatment with and without exogenous phytases. On the other hand, the longer incubation period with the endogenous phytases significantly reduced zinc content (P < 0.0001) because of gradual zinc leaching into the medium.

An experiment by Bosscher et al. (2001b) done with infant formulas indicated that phytate:Zn molar ratios > 1.5, or [phytate]×[Ca]/[Zn] molar ratios > 200, can negatively affect zinc dialyzability. According to the authors, diets from which Zn availability is low include those that contain high phytate, soyabean-protein products, or have a phytate:Zn molar ratio > 15. Because of the synergistic effects between phytate and high Ca on Zn absorption, the [phytate]×[Ca]/[Zn] molar ratio of the diet is also frequently used to express Zn bioavailability.

Zinc dialyzability was the highest from hypoallergenic infant formula consisting of protein hydrolysates (García et al., 1998). The zinc dialysis percentages (2.2–6.1%) obtained from the soy-based formulas were quite low, probably due to the concentration of phytate. There were no differences in the percentage of dialyzable zinc in formulas having whey or casein as the main protein source. The highest zinc content corresponded to the soy-based formulas. Drago and Valencia (2004) found that zinc dialyzability was adversely affected by casein content in infant formulas; the lowest values were found in formulas with the highest casein-to-whey protein ratio. Binding of a large proportion of zinc to casein may result in the entrapment of zinc in casein curds, which may be incompletely digested in the small intestine, thus rendering a significant proportion of zinc unavailable for absorption. Bosscher et al. (2001a) found that zinc dialyzability was inhibited by soluble dietary fiber. However, the inhibitory effect of soluble fiber on zinc dialyzability was more pronounced in casein than in whey-based formulas. Zinc dialyzability from casein- and whey-based formulas supplemented with 0.42 g of locust-bean gum/100 mL were 3.2 and 5.6% for zinc (P < 0.05), respectively.

The bioavailability of zinc salts was also assessed via in vitro methods. Guillem et al. (2000) fortified milk and soy-based infant formulas with different salts. Dialyzability results from the milk formula were as follows (in decreasing order): oxide > gluconate = chloride = lactate = citrate > acetate, and from the soy-based infant formula were: gluconate > oxide > lactate = chloride = acetate > sulfate > citrate. According to the results obtained and without taking into account other factors that could also influence bioavailability, the choice compounds for zinc supplementation would be oxide and gluconate for milk-based products and gluconate and oxide for soy-based ones. According to the authors, dialyzability values have nothing to do with the water solubilities of the salts used. In fact, the higher dialyzability in milk-based formulas corresponded to one of the zinc compounds with the lowest water solubility (i.e., zinc oxide).

Finger millet was explored as a source of zinc fortification (Tripathi and Platel, 2010). Finger millet flour was fortified with either zinc oxide or zinc stearate. Zinc dialyzability from the fortified flour (with the different zinc salts) increased by 1.5–3 times relative to the unfortified flour. The bioaccessible zinc content in the unfortified finger millet flour was 0.18 mg/100 g, while that in the flours fortified with zinc oxide and zinc stearate was 0.25 and 0.49 mg/100 g, respectively. Thus, zinc stearate seemed to provide more bioaccessible zinc. Inclusion of EDTA along with the zinc salt significantly enhanced the bioaccessibility of zinc from the fortified flours, the increase being threefold. Inclusion of citric acid along with the zinc salt and EDTA during fortification did not have any additional beneficial effect on zinc bioaccessiblity. The [phytate]×[Ca]/[Zn] molar ratio in the finger millet flour was 329.1, which was brought down to 84.1 after fortification (Tripathi and Platel, 2010).

Using an in vitro digestion/Caco-2 cell model, Etcheverry et al. (2005a) found that addition of calcium glycerophosphate/gluconate (CaGPG) increased zinc uptake by Caco-2 cells from human milk fortifiers. Why CaGPG may have an enhancing effect on zinc is not known. Gluconate, present in CaGPG, may have an enhancing effect on zinc absorption.

There is a need for more studies that validate the in vitro methods for measuring zinc bioavailability. The Caco-2 model holds potential for studying zinc bioavailability. Zinc uptake in these cells has been characterized (Etcheverry and Grusak, in preparation), but certainly further studies are needed to assess how they can be used to study zinc bioavailability without the need to heat inactivate proteases (which might affect food nutrients and hence bioavailability). In the meantime, in vitro dialyzability assays might be the most appropriate method to study zinc bioaccessibility. It is the only method thus far that has been validated against human studies.