Nutrient and Total Polyphenol Contents of Dark Green Leafy Vegetables, and Estimation of Their Iron Bioaccessibility Using the In Vitro Digestion/Caco-2 Cell Model

Dark green leafy vegetables (DGLVs) are considered as important sources of iron and vitamin A. However, iron concentration may not indicate bioaccessibility. The objectives of this study were to compare the nutrient content and iron bioaccessibility of five sweet potato cultivars, including three orange-fleshed types, with other commonly consumed DGLVs in Ghana: cocoyam, corchorus, baobab, kenaf and moringa, using the in vitro digestion/Caco-2 cell model. Moringa had the highest numbers of iron absorption enhancers on an “as-would-be-eaten” basis, β-carotene (14169 μg/100 g; p < 0.05) and ascorbic acid (46.30 mg/100 g; p < 0.001), and the best iron bioaccessibility (10.28 ng ferritin/mg protein). Baobab and an orange-fleshed sweet potato with purplish young leaves had a lower iron bioaccessibility (6.51 and 6.76 ng ferritin/mg protein, respectively) compared with that of moringa, although these three greens contained similar (p > 0.05) iron (averaging 4.18 mg/100 g) and β-carotene levels. The ascorbic acid concentration of 25.50 mg/100 g in the cooked baobab did not enhance the iron bioaccessibility. Baobab and the orange-fleshed sweet potato with purplish young leaves contained the highest levels of total polyphenols (1646.75 and 506.95 mg Gallic Acid Equivalents/100 g, respectively; p < 0.001). This suggests that iron bioaccessibility in greens cannot be inferred based on the mineral concentration. Based on the similarity of the iron bioaccessibility of the sweet potato leaves and cocoyam leaf (a widely-promoted “nutritious” DGLV in Ghana), the former greens have an added advantage of increasing the dietary intake of provitamin A.


Introduction
It is generally accepted that dark green leafy vegetables (DGLVs) are important sources of micronutrients such as iron and vitamin A. For example, on the basis of compositional data, DGLVs were reported to contribute about 19-39% of iron and 42-68% of vitamin A [1] in the diets of rural South Africans. However, iron and vitamin A deficiencies are perennial malnutrition problems in developing countries where DGLVs are important food ingredients [2,3]. One of the common food ingredients,

Sample Cultivation and Collection
Five cultivars of sweet potato-three orange-fleshed (Coded OFSP1, OFSP2 and OFSP3), one purple-fleshed (PFSP), and one white-fleshed (WFSP)-and three other DGLVs, namely moringa, corchorus and kenaf (Figures 1 and 2), were nursed in a screen house up to maturity (8 weeks). Each DGLV was cultivated in three replicates, and each replicate contained five pots of the particular green. Baobab and cocoyam were purposively sampled from three different geographical locations. Baobab leaves were collected from trees near settlements from the Upper East, Upper West and North regions, while cocoyam leaves were harvested from farmlands from the Ashanti, East and Brong-Ahafo regions of Ghana. The baobab was not nursed due to a relatively long time for the initiation of vegetative growth. Cocoyam is normally cultivated in the rainforest regions in Ghana and not in northern Ghana.

Sample Cultivation and Collection
Five cultivars of sweet potato-three orange-fleshed (Coded OFSP1, OFSP2 and OFSP3), one purple-fleshed (PFSP), and one white-fleshed (WFSP)-and three other DGLVs, namely moringa, corchorus and kenaf (Figures 1 and 2), were nursed in a screen house up to maturity (8 weeks). Each DGLV was cultivated in three replicates, and each replicate contained five pots of the particular green. Baobab and cocoyam were purposively sampled from three different geographical locations. Baobab leaves were collected from trees near settlements from the Upper East, Upper West and North regions, while cocoyam leaves were harvested from farmlands from the Ashanti, East and Brong-Ahafo regions of Ghana. The baobab was not nursed due to a relatively long time for the initiation of vegetative growth. Cocoyam is normally cultivated in the rainforest regions in Ghana and not in northern Ghana.

Sample Cultivation and Collection
Five cultivars of sweet potato-three orange-fleshed (Coded OFSP1, OFSP2 and OFSP3), one purple-fleshed (PFSP), and one white-fleshed (WFSP)-and three other DGLVs, namely moringa, corchorus and kenaf (Figures 1 and 2), were nursed in a screen house up to maturity (8 weeks). Each DGLV was cultivated in three replicates, and each replicate contained five pots of the particular green. Baobab and cocoyam were purposively sampled from three different geographical locations. Baobab leaves were collected from trees near settlements from the Upper East, Upper West and North regions, while cocoyam leaves were harvested from farmlands from the Ashanti, East and Brong-Ahafo regions of Ghana. The baobab was not nursed due to a relatively long time for the initiation of vegetative growth. Cocoyam is normally cultivated in the rainforest regions in Ghana and not in northern Ghana.

Sample Preparation
The replicates of the DGLVs were separately washed twice under running tap water and rinsed in distilled water; about two handfuls of DGLVs put into a stainless steel cup with 100 mL of distilled water added were covered with aluminium foil and boiled until soft, for between approximately 15 and 20 min. The cooked DGLVs were allowed to cool, and all the contents of the cup were transferred into coded, transparent, low-density polyethylene zip-lock bags, and stored in a freezer at −18 • C for 2 weeks. Prior to storage in the freezer, about 5 g aliquot portions were taken for moisture determination. The frozen samples were then freeze-dried (TK-118 Vacuum Freeze-Dryer, True Ten Industrial Company Limited, Taichung, Taiwan) for 72 h. The samples were then milled (Thomas Scientific, Dayton Electric Manufacturing Company Limited, Niles, IL, USA) and sieved into fine powder using a 60 mm sieve.
Triplicate aliquots of three-letter-coded powdered samples were couriered to the University of Greenwich at Medway, Chatham-Maritime United Kingdom, and Massey University, Palmerston North, New Zealand, from Ghana. The moisture determination of fresh leaves was performed in Ghana.

Moisture and Protein
The moisture contents of freshly harvested leaves and cooked leaves were gravimetrically determined using the forced air oven method (AOAC 925.10). For the milled freeze-dried samples, the vacuum oven protocol (AOAC 926.12), as published in the official methods of analysis of AOAC International [29], was used for the moisture determination.
The concentration of nitrogen in the freeze-dried greens was performed by the Dumas method (AOAC 968.06), and a nitrogen-to-protein conversion factor of 6.25 was used to quantify the amount of protein in the leaves on a fee-for-service basis by Massey University Nutrition Laboratory, Palmerston North, New Zealand.

Mineral Analysis: Calcium, Iron, and Zinc
Approximately 0.50 g of the freeze-dried DGLV samples was microwave-digested using an accelerated reaction system (CEM MARS 5H with XP-1500 vessels) for 20 min at 400 psi and 1200 W. Subsequently, calcium, iron and zinc were quantified using an Inductively Coupled Plasma-Optical Emission Spectrometer (ICP-OES, Perkin-Elmer Optima 4300 DV, Perkin-Elmer, Coventry, UK) using protocols as previously described [30]. A certified reference material (ERMCD281, Sigma-Aldrich, UK) was included and run in parallel with the DGLV samples. The data obtained for all three minerals in the reference material were within 5% of the expected values.

β-Carotene
Other researchers have described the extraction and quantification methods used in this study [31]. Averagely, 0.50 g of the freeze-dried samples of the leaves was used for the extraction. A certified reference material (BCR-485, Sigma-Aldrich now Merck, provided to Sigma-Aldrich from the European Commission Joint Research Centre, Institute for Certified Reference Materials and Measurements, Geel, Belgium) was included in three out of the five batches of extraction carried out on DGLV samples. A mean recovery of 128% was obtained for the β-carotene level for the reference material. Therefore, the values obtained for DGLVs were adjusted for a systematic error of 28%.

Ascorbic Acid
The method for vitamin C determination as published by Lee and Coates [32] was carried out by the Massey University Nutrition Laboratory, Palmerston North, New Zealand, on a fee-for-service basis.

Polyphenols
The Folin-Ciocalteu method described by Isabelle and co-workers [33] was used to quantify the total polyphenols in the samples, as gallic acid equivalents. The Nutrition Laboratory, Massey University, New Zealand Palmerston North, New Zealand, carried out the analysis on a fee-for-service basis.

In Vitro Digestion/Caco-2 Cell Model for Iron Availability
The iron availability from the freeze-dried DGLVs as received from Ghana was assessed using the TC7 Caco-2 cell clone (INSERM U505, Paris, France) from cell passages 42-45 in the in vitro digestion/Caco-2 cell model, as previously described [34], with slight modification. Averagely, 0.5 g rather than 1 g of the sample was weighed for the assessment, as 1 g of the starting material led to a matrix that was too viscous for the multiple mixing and pH adjustments required in this method. Cells were grown in six-well tissue culture plates for the experiments and maintained in DMEM supplemented with 10% v/v foetal bovine serum (FBS). On days 12 and 13, cell media were changed to MEM without FBS, as in the method developed by Glahn [35,36], to ensure low iron media, but optimal expression of Caco-2 cell iron transport proteins [37]. On day 14, foods were subjected to in vitro digestion with a sequential addition of digestive enzymes to mimic exposure to the stomach and small intestine (pepsin at pH 2, followed by bile/pancreatin at pH 7). Digested foods (digestates) and controls, including a blank "No food/added iron" digestate, were then applied to Caco-2 cells through an upper chamber suspended over the plate wells, created using a 15 kD dialysis membrane fitted over a Transwell insert and held in place with a silicon ring. The membrane protected the cells from the digestive enzymes, and also mimicked the gut mucous layer by only allowing soluble iron of a selected size to be available for enterocyte absorption. Cells were treated for 2 h, the digestates were removed, and the cells were returned to the incubator. The cells were harvested for ferritin 24 h after the initiation of the digestive process. Ferritin was measured using a commercial enzyme-linked immunosorbent assay (Spectro ferritin, RAMCO Laboratories Inc., Stafford, TX, USA), and corrected for differing numbers of cell per tissue culture well by measurement of cell protein as an indicator of cell numbers; the cell protein was measured using the Pierce protein bicinchoninic acid assay. Ferritin values were expressed as ng ferritin/mg cell protein.

Statistical Analysis
The compositional data were converted to an "as-would-be-eaten" basis prior to statistical analysis, using the dry matter content obtained for the cooked samples prior to storage in the freezer. The univariate analysis, followed by Tukey's studentised range test with the significance set at p < 0.05, was used for the compositional data. For the in vitro digestion/Caco-2 cell model for iron availability, the data generated were normalised prior to using the general linear model procedure for one-factor analysis, and the results were presented as interval plots of the means with 95% confidence intervals. The Minitab 16.2.2 (Minitab Inc., State College, PA, USA) statistical package was employed for the data analysis.

Compositional Profile
The data in Table 1 is expressed on the as-would-be-eaten basis, with the exception of the moisture value of the freshly harvested leaves. The moisture content of the sweet potato cultivars ranged from 83 to 87 g/100 g, and it was similar to other cultivars cultivated in China [38]. The greens of the sweet potato cultivars were generally not significantly different (p > 0.05) from each other for all the components analysed, with the exception of the total polyphenols. OFSP1, Apomuden, a variety being promoted in Ghana because of the β-carotene content in the storage root [39], had approximately 1.7 times more total polyphenols than the other sweet potato cultivars. The leaves of the sweet potato cultivars were not distinctively superior in the levels of the micronutrients analysed, compared with the other DGLVs. However, OFSP1 contained appreciably higher levels of β-carotene (10,533 µg/100 g) and total polyphenols than the other greens, apart from the β-carotene level in moringa (1.3 times more), and the total polyphenols in baobab, which was about thrice higher. Although the roots of the OFSP cultivars are promoted as a dietary source of vitamin A, moringa leaves actually had the highest β-carotene concentration among the DGLVs investigated. Although the WFSP root is devoid of β-carotene [40], the amount of provitamin A in the leaf was more than that in the greens of OFSP2 and OFSP3.
In contrast, among the commonly consumed DGLVs, only baobab leaves contained the highest amount of calcium (p < 0.001): on average, about four times more. There was no significant difference in the iron concentration (p > 0.05), but the data showed that the iron level in baobab and moringa (4.59 ± 1.28 and 4.55 ± 1.88 mg/100 g, respectively) was higher. Previous data indicated that moringa contained 28.29 ± 0.05 mg/100 g of compositional iron [17], the highest compared with the seven sweet potato varieties in Ghana; the data in this study followed a similar trend.
Three of the DGLVs with notable amounts of ascorbic acid were moringa, baobab, and kenaf. The total polyphenols in baobab was the highest (1646.75 ± 69.44 mg GAE; p < 0.001) among all the DGLVs, including the sweet potato cultivars considered in this study. Moringa had a moderate content of total polyphenols, about one-fifth of that in Baobab (p < 0.05).
The concentration of zinc in the cocoyam leaf was 1.49 mg/100 g, about thrice more than the average of all the other DGLVs (p < 0.001). A similar trend of the zinc data between moringa and the sweet potato cultivars in this study was observed in a previous study in Ghana [17]. Figure 3 shows the crude protein content of all the DGLVs, ranging from 3.62-6.54 g/100 g on the as-would-be-eaten basis. Moringa contained the highest protein (6.54 ± 0.36 g/100 g), and was significantly different (p < 0.05) from the next DGLV, baobab (5.67 ± 0.05 g/100 g), which was followed by two cultivars of sweet potato: OFSP1 (5.37 ± 0.04 g/100 g) and OFSP3 (4.99 ± 0.17 g/100 g). The two DGLVs with the lowest protein levels were WFSP (3.87 ± 0.05 g/100 g) and Cocoyam (3.62 ± 0.17 g/100 g). A trend between the protein data for moringa and the sweet potato cultivars was similar to a previous study in Ghana [17]. OFSP1, Apomuden, a variety being promoted in Ghana because of the β-carotene content in the storage root [39], had approximately 1.7 times more total polyphenols than the other sweet potato cultivars. The leaves of the sweet potato cultivars were not distinctively superior in the levels of the micronutrients analysed, compared with the other DGLVs. However, OFSP1 contained appreciably higher levels of β-carotene (10,533 μg/100 g) and total polyphenols than the other greens, apart from the β-carotene level in moringa (1.3 times more), and the total polyphenols in baobab, which was about thrice higher. Although the roots of the OFSP cultivars are promoted as a dietary source of vitamin A, moringa leaves actually had the highest β-carotene concentration among the DGLVs investigated. Although the WFSP root is devoid of β-carotene [40], the amount of provitamin A in the leaf was more than that in the greens of OFSP2 and OFSP3.
In contrast, among the commonly consumed DGLVs, only baobab leaves contained the highest amount of calcium (p < 0.001): on average, about four times more. There was no significant difference in the iron concentration (p > 0.05), but the data showed that the iron level in baobab and moringa (4.59 ± 1.28 and 4.55 ± 1.88 mg/100 g, respectively) was higher. Previous data indicated that moringa contained 28.29 ± 0.05 mg/100 g of compositional iron [17], the highest compared with the seven sweet potato varieties in Ghana; the data in this study followed a similar trend.
Three of the DGLVs with notable amounts of ascorbic acid were moringa, baobab, and kenaf. The total polyphenols in baobab was the highest (1646.75 ± 69.44 mg GAE; p < 0.001) among all the DGLVs, including the sweet potato cultivars considered in this study. Moringa had a moderate content of total polyphenols, about one-fifth of that in Baobab (p < 0.05).
The concentration of zinc in the cocoyam leaf was 1.49 mg/100 g, about thrice more than the average of all the other DGLVs (p < 0.001). A similar trend of the zinc data between moringa and the sweet potato cultivars in this study was observed in a previous study in Ghana [17]. Figure 3 shows the crude protein content of all the DGLVs, ranging from 3.62-6.54 g/100 g on the as-would-be-eaten basis. Moringa contained the highest protein (6.54 ± 0.36 g/100 g), and was significantly different (p < 0.05) from the next DGLV, baobab (5.67 ± 0.05 g/100 g), which was followed by two cultivars of sweet potato: OFSP1 (5.37 ± 0.04 g/100 g) and OFSP3 (4.99 ± 0.17 g/100 g). The two DGLVs with the lowest protein levels were WFSP (3.87 ± 0.05 g/100 g) and Cocoyam (3.62 ± 0.17 g/100 g). A trend between the protein data for moringa and the sweet potato cultivars was similar to a previous study in Ghana [17].

In Vitro Iron Bioaccessibility Using Caco-2 Cells as a Model
The data representing the in vitro iron bioaccessibility are shown in Figure 4. The overall mean of the iron bioaccessibility was 7.71 ng ferritin/mg protein. Moringa markedly had the best iron bioaccessibility, 10.28 ± 2.73 ng ferritin/mg protein, and was significantly different (p < 0.0001) from all the DGLVs investigated.
The two greens (baobab and OFSP1) that could be ranked first and second in terms of the concentrations of total polyphenols had the lowest iron bioaccessibility using the Caco-2 cell model; their bioaccessibility was below the group mean. Conversely, cocoyam had an iron bioaccessibility at the overall mean, although it contained the lowest concentration of polyphenols. Apart from baobab, moringa and OFSP1, all the other DGLVs had a bioaccessibility similar to that of the overall mean.

In Vitro Iron Bioaccessibility Using Caco-2 Cells as a Model
The data representing the in vitro iron bioaccessibility are shown in Figure 4. The overall mean of the iron bioaccessibility was 7.71 ng ferritin/mg protein. Moringa markedly had the best iron bioaccessibility, 10.28 ± 2.73 ng ferritin/mg protein, and was significantly different (p < 0.0001) from all the DGLVs investigated.
The two greens (baobab and OFSP1) that could be ranked first and second in terms of the concentrations of total polyphenols had the lowest iron bioaccessibility using the Caco-2 cell model; their bioaccessibility was below the group mean. Conversely, cocoyam had an iron bioaccessibility at the overall mean, although it contained the lowest concentration of polyphenols. Apart from baobab, moringa and OFSP1, all the other DGLVs had a bioaccessibility similar to that of the overall mean.  Table 2. Effect of selected components (on an "as-would-be-eaten" basis) in DGLVs on iron bioaccesibility.   To test the relationship between the iron bioaccesibility and some of the components (in the DGLVs investigated), a multiple linear regression was conducted ( Table 2). Although the model explained about 75% of the variation in the iron bioaccesibility, it was the protein and iron levels that showed a marginal but positive effect such that a unit increase could respectively lead to 0.29 and 0.13 ng ferritin/mg protein formations in Caco-2 cells. However, with regard to zinc, an increase in its concentration resulted in a reduction of the ferritin formation by a 0.87 ng ferritin/mg protein. In this study, ascorbic acid and β-carotene (known enhancers of iron absorption), as well as total polyphenols (inhibitors of iron), had almost no effect on the iron bioaccessibility using the in vitro digestion/Caco-2 cell model.

Discussion
OFSP1 was the only sweet potato cultivar with purplish young leaves [41], among the five sweet potato genotypes evaluated in this study. This may have accounted for the highest total polyphenol content of OFSP1, compared to the other sweet potato cultivars. The difference in the iron data for moringa in this study compared to previous work [17] was due to how the data were reported. In the previous study, the result was reported on powdered samples, while in our study, it was on an as-would-be-eaten basis. Nonetheless, the trend of iron levels being the highest in moringa was also confirmed in this study.
Although cocoyam leaf is widely consumed, and promoted in Ghana as a "nutritious" green to improve iron status (anecdotally), on the basis of its composition data, it was highest only in zinc, and lowest in β-carotene and total polyphenols, compared with the OFSP cultivars. Because both the sweet potato leaves and cocoyam had a similar iron bioaccessibility, the sweet potato leaves could be used in culinary preparations, and had an added advantage of increasing the dietary intake of β-carotene, compared to those of cocoyam.
Generally, the level of iron bioaccessibility from the DGLVs was relatively low (6-10 ng ferritin/mg protein) compared with our previous work on complementary food from the same laboratory (12-34 ng ferritin/mg protein) [34]. However, a strong comparison cannot be made between the data from the two studies, as different sample weights were used: 1 g in the previous work, and 0.5 g in the present study. A previous community-based feeding trial using Weanimix, which had an iron bioaccessibility of 17.32 ± 2.84 ng ferritin/mg protein [34], resulted in a poor iron status among older infants in Ghana [42,43]. The lower availability of iron in the greens in this study lends support to the finding of the work on young Burkinabe women, which resulted in no increase in iron absorption after eating Jew's mallow with a thick maize paste [4].
As mentioned earlier, moringa contained the highest number of enhancers of iron absorption: β-carotene [7,8] and ascorbic acid [44]; although in this study their effect were not realised except for the concentration of iron. Additionally, the concentration of total polyphenols in this DGLV was moderate. The composition of nutrients in moringa, compared with the other DGLVs, may have contributed to it having the highest bioaccessibility of iron, as obtained from the in vitro Caco-2 cells model study. Although OFSP1 had significantly similar levels of β-carotene and iron to moringa, and one-third of the total polyphenols of baobab, its iron bioaccessibility was lower than for moringa, indicating that the reported caffeoylquinic acid derivatives in sweet potato leaves [20] may have limited the bioaccessibility of iron. Baobab had the lowest iron bioaccessibility, in spite of being one of the greens that contained the highest amounts of iron and ascorbic acid. This may have been attributed to the high concentration of total polyphenols [20], and not calcium, which is known to inhibit iron absorption [25,45]; and relative to the other DGLVs, suggesting that the polyphenols in baobab may be very inhibitory, even in the presence of endogenous ascorbic acid. However, the amount of calcium in the greens explicitly did not suggest inhibitory effects on iron, as moringa contained the second highest level of this mineral among all the DGLVs investigated, but had a markedly better iron availability. Therefore, predicting iron bioaccessibility based only on compositional data could lead to false conclusions.
The effect of the concentration of plant protein on the iron bioaccessibility cannot be explicitly substantiated in this study. Moringa, having the highest as-would-be-eaten protein, was the green with the highest bioaccessibility. Both baobab and OFSP1, which contained relatively high concentrations of protein compared to the rest of the DGLVs with the exception of Moringa, were those that recorded the lowest bioaccessibility of iron, although not significantly. Thus, from the data in this study, it is difficult to use the protein concentration to predict the iron bioaccessibility, although the regression analysis showed a direct effect. The inverse association between the zinc concentration and the index of iron bioaccessibility could be attributed to the cocoyam leaf, which had the highest zinc concentration and the lowest ferritin formation in the Caco-2 cells.
The major limitations of this study were that phytate and the constituents of the different classes of polyphenols were not quantified. The assay method previously used for phytate determination [46,47] gave very inconsistent results within replicates in this study; possibly the colour of DGLVs interfered with the spectrophotometer readings.

Conclusions
The studied greens varied in terms of calcium, iron and zinc levels. In addition, moringa had the highest levels of β-carotene and ascorbic acid. Baobab had the highest levels of calcium and total polyphenols. Within the limits of this study, iron bioaccessibility is influenced by a complex interplay of several components in DGLVs, including protein, ascorbic acid, β-carotene and total polyphenols. Moringa had the best iron bioaccessibility, and the lowest was found in baobab and one of the orange-fleshed sweet potatoes with purplish young leaves. Estimating iron bioaccessibility in greens based on the mineral concentration may lead to incorrect conclusions. Based on the similarity of the iron bioaccessibility of the sweet potato leaves and cocoyam leaf, the widely promoted "nutritious" DGLVs in Ghana, the former greens have an added advantage of increasing the dietary intake of provitamin A.