Bioaccessibility of Fifteen Elements from Dried Fruits by the BARGE (Bioaccessibility Research Group of Europe) Unified Bioaccessibility Method (UBM) and Multivariate Statistical Analysis

Abstract The BARGE (Bioaccessibility Research Group of Europe) unified bioaccessibility method (UBM) was applied to assess the bioaccessibility of essential and toxic elements (B, Al, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Rb, Sr, Mo, and Cd) in 11 dried fruit samples, two certified reference materials, and British Geological Survey guidance material (BGS) 102. The total, UBM gastric, and gastro-intestinal (GI) phase concentrations of elements were determined by inductively coupled plasma—mass spectrometry. The accuracy of the methods was verified using BGS 102 for the BARGE UBM and SRM 1573a for total concentrations. The order for mean bioaccessible fractions (BF %) in gastric phase for the dried fruit was Rb (101) > Sr (79) > Mn (59) > Cr (46) > Ni (37) > Fe (36) = Zn (36) > Cu (33) > Co (32) > Al (24) > V (22) > B (19) > Cd (18) > Mo (8) > As (6). Total and bioaccessible concentrations for Mn, Rb, and Sr are significantly and positively correlated with each other. Multivariate statistical analysis was used to classify bioaccessible elements and samples in gastric and GI phases.


Introduction
Dried fruits are concentrated food with lower moisture content than their fresh counterparts and are prepared by various techniques.They are mainly composed of carbon, hydrogen, and oxygen (CHO, 61.3-72.8%).The predominant components of dried fruits are carbohydrates.The basic sugars in all dried fruits are glucose and fructose.However, sucrose is most abundant in dates and peaches (Alasalvar, Salvad o, and Ros 2020).Apricots, apples, dates, currants, peaches, figs, prunes, pears, and raisins are "traditional" dried fruits (Hern andez-Alonso et al. 2017).They are a valuable source of antioxidants, macro and micronutrients, and vitamins.Moreover, they have a low content of total fat, saturated fatty acids (<1%), and protein (0.17-4.08%) (Jeszka-Skowron and Czarczy nska-Go sli nska 2020).However, the total concentrations of elements in food provide little information on their nutritional value and toxic effects (Arpadjan et al. 2013).Therefore, it is important to know more than the total concentration of a ICP-MS was used to determine the total concentrations and bioaccessible fractions.ICP-MS offers low detection limits and wide linear dynamic ranges.The relationships between total and bioaccessible element concentrations in dried fruits were investigated in detail using principal component analysis (PCA) and hierarchical cluster analysis (HCA).

Instrumentation
An ICP-MS equipped with an autosampler, Agilent 7500a, was used for the determination of 15 elements.The studied isotopes were 11 B, 27 Al, 51 V, 53 Cr, 55 Mn, 57 Fe, 59 Co, 60 Ni, 63 Cu, 66 Zn, 75 As, 85 Rb, 88 Sr, 95 Mo, and 111 Cd.The components of the ICP-MS were a Babington nebulizer, nickel cones, a peristaltic pump, quarts double pass spray chamber, a glass torch, nickel sampler, and skimmer cones.Operating conditions for the ICP-MS were radio frequency power, 1290 W; sample depth, 7.9 mm; torch-H, À0.5 mm; torch-V, 0.7 mm; carrier gas flow rate, 1.01 L min À1 ; auxiliary gas flow rate, 0.9 L min À1 ; plasma gas flow rate, 15 L min À1 ; nebulizer pump, 0.12 rps; and spray chamber temperature, 2 C. High purity argon gas was used to form the plasma.The performance of the ICP-MS instrument was checked daily.The pulse to analog factor was performed on the day of analysis.An Agilent ICP-MS solution containing 10 lg L À1 Ce, Co, Li, Tl, and Y was used for tuning.A certified reference material was analyzed before samples to verify the accuracy of the measurements.
Acid digestion was performed using a Berghof MWS-4 microwave system with closed 100 mL Teflon vessels.A WTW pH 3110 pH meter (Fisher Scientific) was used for all pH measurements, it was calibrated daily.A N€ uve FN 400 model oven was used to dry samples.A Memmert model thermostatically controlled water bath with a shaking table (Memmert GmbH þ Co. KG), a NF 400 model centrifuge, and Clifton model water bath (Nickel Electro) were employed during the in vitro gastric and GI extractions.

Reagents and solutions
The reagents were of analytical reagent grade.All solutions were prepared using ultrahigh purity (UHP) water (18.2MX cm, Millipore).A high purity ICP-MS multielement standard solution (Merck) of 10 mg L À1 was used for the preparation of calibration curves with eight points including blanks from 0 to 50 mg L À1 for B, Al, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Rb, Sr, Mo, and Cd.The determination coefficients for the calibration relationships were higher than 0.99.

Microwave digestion
A total of 0.20 g of each dried fruit in triplicate were placed into the microwave digestion vessels.A total of 2 mL of concentrated nitric acid and 5 mL of concentrated hydrogen peroxide were added.The vessels were closed, heated in the microwave oven with a four-step procedure, and cooled to room temperature for approximately 20 min.The vessels were opened carefully in a fume hood.After cooling, each digest was transferred into a 25 mL volumetric flask and diluted to volume with UHP water.The elements in the resulting solution were determined by ICP-MS.Blank and certified reference material analysis were performed by the same protocol.

In vitro GI procedure
The in vitro GI digestion, based on the BARGE UBM methodology (Wragg, Cave, Taylor et al. 2011;Tokalıo glu et al. 2014Tokalıo glu et al. , 2020)), was used for assessing the bioaccessibility of elements in the dried fruit, certified reference materials (SRM 1573a Tomato Leaves and SRM 1547 Peach Leaves), and BGS Guidance Material 102.All procedures for the gastric and GI fractions were performed in quadruplicate accompanied by blanks (n ¼ 3).All simulated gastric and intestinal fluids were prepared one day before the bioaccessibility experiments, stored at 4 C, and heated to 37 ± 2 C in a water bath prior to use.The pH values of saliva, gastric, duodenal, and bile fluids were adjusted with concentrated HCl or 1 mol L À1 NaOH (Table S1).The final pH of the mixed saliva and gastric phase and the final pH of the mixed GI fluid was measured.

Gastric and GI extracts
In the UBM, 0.300 g of sample were weighed into a centrifuge tube.A 4.5 mL simulated saliva fluid (pH 6.5 ± 0.5) was added by micropipette.The tubes were shaken manually for 10 s, 6.75 mL of gastric fluid (pH 1.1 ± 0.1) were added, and the pH was adjusted to 1.20 ± 0.05.The sample tubes were placed in the thermostatic water bath with end-overend shaking at 37 ± 2 C and 160 rpm for 1 h.Next, the pH of the suspension was measured.If the pH from 1.2 to 1.7, the gastric phase extract was centrifuged at 4500 rpm for 15 min.1.0 mL of supernatant was diluted to 5.0 mL with 0.1 mol L À1 HNO 3 .
A total of 13.5 mL of duodenal fluid (pH 7.4 ± 0.2) and 4.5 mL of bile fluid (pH 8.0 ± 0.2) were added to each gastric phase extract to obtain the GI extracts.The tubes were manually shaken for 10 s and the pH was adjusted to 6.3 ± 0.5 with 6 mol L À1 NaOH.The samples were incubated in the thermostatic water bath with end-over-end shaking at 37 ± 2 C and 160 rpm for 4 h.If the pH was 6.3 ± 0.5, the suspensions were centrifuged for 15 min at 4500 rpm.A 1.0 mL portion of the supernatant was combined with 4.0 mL of 0.1 mol L À1 nitric acid.Both extraction phases were stored at < 8 C prior to analysis by ICP-MS.

Determination of bioaccessibility (%)
The bioaccessible fraction (BF %) was calculated by BF (%) ¼ BMC/TMC Â 100% where BMC is the bioaccessible metal concentration (mg kg À1 or ng g À1 ) in the gastric or GI extracts and TMC (mg kg À1 or ng g À1 ) is the total metal concentration of the element in the dried fruit.

Statistical analysis
Correlation analysis, PCA, and HCA were performed using IBM SPSS Statistics version 21.The relationships between total element concentrations and their bioaccessible fractions were analyzed by Pearson's correlation coefficient.PCA and HCA were employed to classify elements and samples according to the bioaccessible concentrations in the gastric and GI phases.Eigenvalue criteria exceeding 1.00 were used to determine the number of factors and Varimax as a rotation method.HCA with Ward's method and a z score transformation before cluster analysis (CA) were applied.A squared Euclidean distance was used for distance measurements.

Analytical figures of merit
The certified reference material, 1573a Tomato Leaves, was analyzed to characterize the reliability of the results.The procedure was performed in triplicate for each sample and blank.The results are summarized in Table 1.
The mean recoveries for the elements were from 84% to 120%.The measured results are in good agreement with the certified values.The limits of detection (LOD) and limits of quantification (LOQ) for the total digestion (n ¼ 10) and the UBM procedures (n ¼ 14) were calculated by 3S blank /b and 10S blank /b, respectively, where S blank is the standard deviation of the blank signal and b is the slope of the calibration curve.The LODs for total digestion were from 0.08 to 2.96 mg L À1 .The LODs for gastric and GI extracts were from 0.04 to 5.02 mg L À1 and 0.02 to 2.60 mg L À1 , respectively.The intraday (n ¼ 3) and inter-day (n ¼ 9) precision as the relative standard deviation(%) were characterized for SRM 1573a with total digestion.These values were from 0.26% to 1.34% and from 0.40% to 3.74%, respectively.The analytical figures of merit are summarized in Table S2.

Bioaccessibility of the trace elements
The bioaccessibility factors (BF %) for each element in the gastric and GI extracts for the dried fruit are summarized in Tables 2-4, and S3.The highest bioaccessibility (%) values for elements in all samples were obtained for Rb, Sr, and Mn in the gastric and GI phases.The decreasing mean BF (%) order in gastric phases for all elements is: Rb      6).The highest mean BF (%) in the GI phase was obtained for Rb (89), Sr (82), and Mn (54).The BF (%) values for Mn, Rb, and Sr in all samples were similar in the gastric and GI phases.The BF (%) in the gastric phase for the other elements was higher than in the GI phases.
The low pH of the gastric phase may considerably increase the soluble fraction of the metals, leading to higher bioaccessibility values.In the gastric phase, enzymes (such as pepsin) release most metal ions.Pepsin, which is more effective in acidic conditions, breaks down proteins and therefore it aids in the sample dissolution (Intawongse and Dean 2006;Tokalıo glu et al. 2014;Pelfrêne et al. 2015).Compounds such as ascorbic acid and citric acid facilitate the bioaccessibility of elements, whereas the compounds, such as oxalates, fibers, polyphenols, and phytates inhibit their bioaccessibility (Kulkarni et al. 2007).Ascorbic acid is abundant in strawberries, while present in moderate concentrations in other berries (de Souza et al. 2014).
The BF (%) values of the elements change with the components of the dried fruit (Tokalıo glu et al. 2014).Dried fruits contain bioactive compounds that include flavonoids, phenolic acids, proanthocyanidins, stilbenes, and chalcones/dihydrochalcones.For example, raisins contain the highest number of phenolic acids.Anthocyanins have been only reported in cranberries, dates, figs, peaches, and raisins (de Souza et al. 2014;Alasalvar, Salvad o, and Ros 2020).The possible mechanisms affecting bioaccessibility of the metals in these samples may involve metal chelates with polyphenols at the physiological pH or the metal ions such as Cd 2þ and Zn 2þ which have similar properties that may compete for the binding sites of enzymes.A high concentration of one element may influence the release and the absorption of others.For example, high Fe, Mn, and Zn concentrations may influence the bioaccessibility of Cu which is consistent with the results in the study.In addition, the elemental species in solution may also influence the absorption.For example, Fe 2þ is more easily absorbed than Fe 3þ (Tokalıo glu et al. 2014;de Souza et al. 2021).
The BF (%) of elements in the dried fruit in the gastric phase indicated greater values for Rb, Sr, and Mn.The Rb and Sr bioaccessibility (%) in food has been infrequently studied.The higher bioaccessibility for Rb may be because it is a monovalent ion.Similar results have been reported for Na þ , K þ , and Li þ (Bertin et al. 2016;Santos et al. 2018).Lower bioaccessibility (%) for Sr than Rb may be due to the competition by other monovalent ions such as K þ (Kulkarni et al. 2007).The high BF (%) values for Mn (46-74%) are in good agreement with the reported values in fruit (Khouzam, Pohl, and Lobinski 2011;Pereira et al. 2018), nutritional supplements (Tokalıo glu et al. 2014), seaweed (Intawongse Kongchouy, and Dean 2018), pulses (Santos et al. 2018), and linseed and sesame (Souza et al. 2018).Mn readily binds to proteins, and this interaction is broken in the GI tract by hydrolysis of proteins via the action of pepsin and pancreatin, leading to the cleavage of peptide bonds (Martins et al. 2020).
The BARGE UBM was employed for SRM 1573a Tomato Leaves (Table S5), SRM 1547 Peach Leaves (Table S6), and BGS Guidance Material 102 (Table S7).The BF (%) values for SRM 1573a in the gastric phase were from 1.8% for Al to 82% for Rb, whereas the BF (%) in the GI phase was from 2% for Al to 113% for Cu.The highest BF (%) for SRM 1547 was 78% for Mn, 99% for Rb, 124% for As and 85% for Cd in the gastric phase and 121% for Cu, 112% for Rb, and 87% for Mo in GI phase.The BGS 102 bioaccessibility factors (%) in the gastric phase were 23% for Mn, 8% for Cr, and 0.4% for Fe.The literature values were 23% for Mn, 12% for Cr, and 0.4% for Fe (Tokalıo glu et al. 2014), in good agreement with the results from this study.
The accuracy of the BARGE UBM for extractable As and Pb was verified using BGS 102 guidance material which has certified values for bioaccessible As and Pb.The concentrations of As and Pb in BGS 102 were determined by ICP-MS with the UBM method.The As concentration in the GI phase was 4.2 ± 1.6 mg kg À1 compared to the certified value of 5.4 ± 2.4 mg kg À1 .Pb in the gastric phase was 13.3 ± 0.6 mg kg À1 compared to the certified value of 13 ± 6 mg kg À1 .These results were in agreement with the certified values for BGS 102 and also with the results of Hamilton et al. (2015) and Tokalıo glu et al. (2014) of 15.3 ± 2.97 and 10 ± 1 mg kg À1 for Pb in the gastric phase and 3.3 ± 0.4 mg kg À1 and 3.9 ± 0.1 mg kg À1 for As in the GI phase, respectively.

Correlation analysis
Pearson's correlation coefficients between the total metal concentrations and bioaccessible concentrations in the gastric and GI phases were investigated in dried fruit.The significant relationships between total and bioaccessible elements or between bioaccessible elements provide important information.The positive correlations between metal pairs show a common origin.The differences in correlations may be related to the food matrices and elements.
Total and bioaccessible concentrations for Mn, Rb, and Sr are significantly (at 99% confidence level) and positively correlated with each other (Figure 1).The bioaccessibility of the other elements was not significantly correlated (p < .05)with their total concentrations.A positive and significant correlation (p < .05) between bioaccessibility of   Mn with the total Mn concentration was observed by Leufroy et al. (2012) in seafood.
The results of correlation analysis are also consistent with the PCA and HCA results.

Multivariate statistical analysis
Principal component analysis PCA was applied to the average concentrations of elements for gastric and GI phases in dried fruits by applying varimax rotation of Kaiser Normalization (Tables 5-7).The goal was to identify the relationships between bioaccessible elements in these phases.
The PCA results for the in-vitro gastric extraction indicate that there were four eigenvalues higher than 1.00 and four PCs explain 82.0% of the total variance.The loadings !0.45 are shown in bold in Table 5.The first component accounts for 25.7% of the total variance.It shows high loadings for Cu (0.84), Co (0.80), Cr (0.72), and Zn (0.60) and with weak loadings for B (0.45) and Ni (0.53).The second component, loaded primarily by Mn (086), Sr (0.81), and Zn (0.76), accounts for 21.5% of the total variance.The third component, which has high loading values for Al (0.97) and Fe (0.94), explains 19.6% of the total variance.Rubidium (0.91) and Ni (0.70), which appear in the fourth factor, explain 15.0% of the total variance.
For in-vitro GI, the first five eigenvalues higher than > 1 account for 88.3% of variances for all data.The first component accounted for 19.9%, the second for 19.1%, the third for 17.4%, the fourth for 16.7%, and the fifth for 15.3% of the total variation.The loadings !0.62 are shown in bold in Table 6.The first component contains Rb and Ni with high loading values of 0.94 and 0.89, respectively.The second component is loaded primarily by Sr (0.93) and Mn (0.84).The third component is loaded by Cr and Fe which have 0.82 and 0.81 loading values, respectively.PC4 is primarily dominated by Al (0.90) and Cu (0.83) and PC5 includes loading for Zn (0.82), B (0.70), and Co (0.62).
The PCA for the in-vitro gastric and GI phases extracted six principal components with eigenvalues exceeding 1.The PCs explain 94.0% of the total variance.The loadings greater than or equal to 0.60 are shown in bold in Table 7. PC1 contributed 18.0% to the total variance and showed a close association with CrG, CrGI, CoG, CoGI, and FeGI.PC2 accounts for 17.6% of the total variance and includes AlG, AlGI, FeG, and CuGI.PC3 contributes 17.1% to the total variance and is characterized by SrG, SrGI, MnG, and MnGI.PC4 includes RbG, RbGI, NiG, and NiGI and accounts for 16.4% of the total variance and PC5 is loaded by ZnG, ZnGI, and CuG and accounts for 14.7% of the total variance.PC6 contributes 10.0% to the total variance and is primarily dominated by BG and BGI.The relationships among the bioaccessible elements based on the PCA loadings are shown in Figure 2a-c.

Cluster analysis
CA was used to classify samples into groups with similar properties.The smallest distance indicates the closest relationship; therefore, those objects are considered to belong to the same group (Tokalıo glu et al. 2019).In this study, the variables were standardized by means of z-scores.The measurement of the similarities in the samples was based upon the squared Euclidean distance.Hierarchical clustering by Ward's method was performed using this data set.The results of the HCA in gastric and GI phase for variables as a dendrogram are shown in Figure 2d and e, respectively.Figure 2f shows the results for the fruit samples.
The results of CA were fairly close in agreement with those obtained by PCA and correlation analysis.CA was used to classify the dried fruit (Figure 2f) and shows the presence of two clusters.The samples in the same cluster indicate similarity in their properties.Cluster 1 contains samples 6 (black raisin), 8 (ground strawberry), 7 (blueberry), 10 (sekerpare apricot), 11 (sun-dried apricot), and 9 (Malatya apricot).Cluster II contains samples 3 (fig), 4 (date), 2 (plum), 1 (mulberry), and 5 (yellow raisin).Date, fig, plum, and raisins have the highest concentration of total phenolics and anthocyanins (Alasalvar and Shahidi 2013).

Conclusions
The bioaccessibility of B, Al, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Rb, Sr, Mo, and Cd was investigated in dried fruit by the BARGE UBM method with ICP-MS analysis.This protocol is effective for determining the trace element concentrations that dissolve in gastric and GI solutions.The highest element concentrations (mg kg À1 ) for total digestion were B (37.3) > Al (16.3) > Fe (11.9) > Rb (10.8).The highest mean gastric BF (%) in all samples was for Rb (101), Sr (79), and Mn (59).The highest mean BF (%) in the GI phase was for Rb (89), Sr (82), and Mn (54).The most bioaccessible elements were observed in S4 (dates), S9 (Malatya apricots), and S11 (sun-dried apricots).The BFs (%) of elements in the gastric and GI phases were dependent upon the dried fruit matrix and element.The total and bioaccessible concentrations of Cd and As in dried fruits were lower than their permissible limits.The correlation analysis shows that total and bioaccessible concentrations for Mn, Rb, and Sr were significantly (p < .01)and positively correlated with each other.In addition, significant relationships were supported by PCA and HCA.CA classified the samples into the dried fruit type.

Disclosure statement
No potential conflict of interest was reported by the authors.

Figure 1 .
Figure 1.Correlations between total and bioaccessible Mn, Rb, and Sr concentrations.

Funding
Dr. S ¸erife Tokalıo glu is grateful for the financial support obtained from the Scientific and Technological Research Council of Turkey (T € UB _ ITAK) (Project No: 119Z748) and Erciyes University Scientific Research Projects Unit (Project No: FYL-21-10803).
b Information value.

Table 2 .
Total (n ¼ 3) and in vitro BARGE UBM (n ¼ 4) concentrations of B, Al, Mn, and Fe in dried fruit as mean ± standard deviation.

Table 3 .
Total and in vitro BARGE UBM concentrations of Cu, Zn, Rb, and Sr in dried fruit as mean ± standard deviation.

Table 4 .
Total and in vitro BARGE UBM concentrations of Co, Ni, V, and Cr in dried fruit as mean ± standard deviation.

Table 5 .
Varimax rotated loadings for the dried fruit in the gastric phase.

Table 6 .
Varimax rotated loadings for the dried fruit in the gastrointestinal phase.

Table 7 .
Varimax rotated loadings for the dried fruit in the gastric and gastrintestinal phase.