Profiling the varietal antioxidative contents and macrochemical composition in Australian faba beans (Vicia faba L.)

There is growing interest in pulses such as faba bean for the development of foods with enhanced nutrition, functionality, and health benefits. In this study, seed material from 10 faba bean varieties, grown in replicated field trials in South Australia over consecutive seasons (2016 and 2017), were analysed for ferric reducing antioxidant potential, total phenolics, and total monomeric anthocyanins. Differences in the macrochemical composition of varieties was investigated using attenuated total reflectance mid‐infrared spectroscopy. The mean ferric reducing antioxidant potential of the varieties ranged from 237 to 531 mg trolox equivalents 100 g−1; the total phenolics from 258 to 571 mg gallic acid equivalents 100 g−1; and the total monomeric anthocyanins from 12.7 to 21.0 mg cyanidin‐3‐glucoside equivalents 100 g−1. Statistically significant variances in all three measures were found between varieties. Attenuated total reflectance Fourier transformed mid‐infrared spectroscopy was found to provide a rapid assessment of the phytochemical composition of the samples. Partial least squares discriminant analysis was able to classify samples by growing year with reasonable accuracy (>87%). There is significant variation in the antioxidant, phenolic, and anthocyanin contents between Australian faba bean varieties. Mid‐infrared spectroscopy may prove to be a valuable screening tool for breeders and researchers in the future.

Over the past 35 years, pulse breeders have focussed on improving Australian faba bean varieties, primarily selecting for disease resistance and elevated yield (Siddique et al., 2000). However, new varieties specifically selected for these characteristics alone may display a concomitant, albeit unintentional, alteration in their nutritional and antioxidative properties as observed in other crops (Wrigley, Matakovsky, Melnik, Pascual, & Romanov, 2019). Recent work highlighted variation in nutritional and antinutritional properties in commercial Australian faba bean varieties . However, similar profiling on phytochemical properties such as antioxidant compounds is lacking.    (Valente et al., 2018;Valente et al., 2019), Africa (Chaieb, González, López-Mesas, Bouslama, & Valiente, 2011), and South America (Baginsky et al., 2013).
In this study, we analysed the antioxidative, phenolic, and anthocyanin content of 10 faba bean varieties, grown in replicated field trials, at two sites in South Australia and over two consecutive growing seasons. We also explored the use of mid-infrared (MIR) spectroscopy for non-invasive analysis of faba beans, as this technology has shown promise in other crops such as wheat and mungbeans Johnson, Collins, Skylas, & Naiker, 2019).

| Faba bean samples
Seed material of 10 faba bean varieties included in this study are listed in Table 1 and are described further in Skylas et al. (2019).
These varieties constitute the bulk of domestic production for this crop (Pulse Australia, 2016). Varieties were grown at two locations in South Australia (Charlick and Freeling), over consecutive seasons (2016 and 2017). The four environments are designated herein as 16Char, 17Char, 16Free, and 17Free. Growing details, including rainfall conditions and trial yield data have been previously reported . Two field replicates were analysed (in duplicate) for each of the 2016 sites, and three field replicates were analysed (in duplicate) for each of the 2017 sites. Seed samples were impact milled to produce whole seed flour (falling No. grinder, 0.8 mm screen), and moisture content were determined as previously described .

| Reagents
All reagents used were of analytical grade. Methanol was purchased from Fisher Scientific Australia. Hydrochloric acid and sodium carbonate were purchased from Chem Supply. All other reagents were purchased from Sigma-Aldrich Australia. Unless otherwise specified, all dilutions and assay preparations were made using Milli-Q water. All solutions were stored at 4 C until usage.

| Extraction
Extracts were prepared in duplicate by combining approximately 0.5 g of faba bean flour with 8 ml of 90% v/v aqueous methanol, vortexing for 10 s, and mixing for 60 min using an end-over-end shaker (Ratek RM4) operating at 50 rpm. After centrifugation at 1,000 g for 10 min (Heraeus Multifuge; Thermo Fisher Scientific), the supernatant was collected. To extract any remaining phytochemicals, the extraction process was repeated with another 8 ml of 90% methanol added to the pellet and end-over-end mixing for 20 min. The combined supernatant was made up to 20 ml volume with 90% methanol. Extracts were stored in the dark at 4 C until required for analysis.

| Total phenolics
Total phenolics (TP) were determined through a modification of the Folin-Ciocalteu method developed by Singleton and Rossi (1965).
First, 2 ml of a 1:10 aqueous dilution of Folin-Ciocalteu reagent was combined with 400 μl of sample extract. The samples were incubated at room temperature in darkness for 10 min before 2 ml of 7.5% w/v aqueous sodium carbonate was added. They were then vortexed for 10 s, incubated at 40 C for 30 min in a covered water bath, and vortexed for another 10 s. From the absorbance at 760 nm, the TP concentration was derived as a function of the equivalent absorbance of gallic acid in the range 20 to 120 mg L −1 (R 2 = .9968). Results were expressed as milligrams of gallic acid equivalents (GAE) per 100 g of oven dry sample weight (mg GAE/100 g).

| Ferric reducing antioxidant power
As a measure of total antioxidant capacity, the ferric reducing antioxidant power (FRAP) assay developed by Benzie and Strain (1996) was performed on the samples. FRAP reagent was prepared by combining 300-mM acetate buffer at pH 3.56, 20-mM aqueous ferric chloride, and 10-mM TPTZ (made in 40-mM hydrochloric acid) in the ratio 10:1:1. The FRAP reagent, ferric chloride, and TPTZ solutions were prepared fresh each day. First, 3 ml of FRAP reagent, pre-equilibrated at 37 C, was combined with 100 μl of pre-equilibrated sample and vortexed for 10 s. The samples were incubated in a covered water bath at 37 C for 4 min, vortexed for 10 s, and their absorbances read at 593 nm. The FRAP derived was a function of the equivalent absorbance of trolox in ethanol solution in the range 10-175 mg L −1 (R 2 = .9999). Results were expressed as milligrams of trolox equivalents per 100 g of oven dry sample weight (mg TXE/100 g).

| Total monomeric anthocyanins
The total monomeric anthocyanins were determined using a minor modification of the pH differential method described by Giusti and Wrolstad (2001). Buffer solutions consisting of 0.025-M aqueous potassium chloride and 0.4-M aqueous sodium acetate were prepared and adjusted to pH 1 and 4.5, respectively, using 32% hydrochloric acid.

| Attenuated total reflectance MIR spectroscopy
A Bruker Alpha Fourier transformed infrared spectrophotometer (Bruker Optics Gmbh, Ettlingen, Germany) fitted with a platinum diamond attenuated total reflectance (ATR) single reflection module was used for the MIR analysis. Homogenous faba bean flour was used to cover the reflection module and pressure applied to achieve uniform contact between the ATR interface and flour. Air was used as a reference background; the background measurement was performed every 10 samples. Cross contamination of samples was minimised by cleaning and drying the platform with isopropyl alcohol and laboratory Kimwipes ® between samples (Gordon et al., 2019).
MIR spectra between 4,000 and 400 cm −1 were recorded using the OPUS software version 7.5 (Bruker Optics Gmbh, Ettlingen, Germany) as the average of 24 scans at a resolution of 4 cm −1 . Five replicates were performed on each sample.

| Statistical analysis
Statistical tests were performed in IBM SPSS. As a relatively high number of replicates were included in each test and all data were reasonably normally distributed, parametric testing was used throughout.
MIR spectra were analysed with The Unscrambler X software version 10.5 (Camo ASA, Oslo, Norway). Following previous work on barley (Gordon et al., 2019) and mungbeans (Johnson, Collins, Power, et al., 2019), the spectra were preprocessed to the second derivative using a Savitzky-Golay algorithm at a polynomial number of 2 and a smoothing window of 41 points (Savitzky & Golay, 1964). Using the second derivative removes spectral variations in the baseline and slope (Savitzky & Golay, 1964), minimising differences due to noncompositional variables such as the pressure and contact with the reflection module. Principal component analysis (PCA) and partial least squares (PLS) regression were performed in The Unscrambler X on the second derivative of the MIR spectra.

| Ferric reducing antioxidant potential
The average ferric reducing antioxidant potential determined for the varieties is shown in

| TP contents
The TP contents followed a similar trend to the FRAP values (Table 3).
There was a linear correlation between FRAP values and TP contents

| Total monomeric anthocyanin contents
There was no correlation with the mean total monomeric anthocyanin content of samples (Table 4) and FRAP or TP contents (p > .05 for both), although there was a negative correlation with moisture content (r 100 = −0.202, p < .05).
There was a significant variation in the mean anthocyanin content when analysed by variety (one-way ANOVA; F 9,90 = 3.726, p = .001).
Statistical differences are shown from annotations reported in

| Principal component analysis of moisture, antioxidant, phenolic, and anthocyanin contents
In order to further explore the variation in the moisture, antioxidant, phenolics, and anthocyanin contents, a principal component analysis was conducted on the data for these three measurements. As these parameters varied considerably in terms of their absolute values, each datapoint was weighted by dividing itself by the overall standard deviation for that measurement.
The first two principal components (PCs) explained 78% of the total variation observed. Across PC1, broad separation was observed between PBA Rana and the remainder of the varieties (Figure 1).
Examination of the loadings associated with this PC indicated that PC1 scores were positively correlated with increased FRAP and TPs, confirming that levels of these compounds in PBA Rana were noticeably elevated compared with other varieties. PBA Samira also had higher scores along PC1 than the remainder of samples, indicating that its FRAP and TP levels were higher, albeit not as distinct as those of PBA Rana.
Separation along the PC2 was generally less clear. Both PBA Rana and Fiord were largely associated with positive scores along this axis, indicating above average anthocyanin levels and lower moisture contents. The negative PC2 scores observed for PBA Zahra indicated higher moisture contents and lower anthocyanin levels, as previously observed (Table 4). PBA Samira, which had the lowest mean anthocyanin levels of all faba bean varieties (Table 4), also was largely associated with negative PC2 values.
To further visualise the general relationship between the chemical composition of the varieties obtained through PCA, a cluster analysis was performed on the mean moisture, antioxidant, phenolic, and anthocyanin contents (Figure 2). This confirmed that the composition PBA Rana was highly distinct from the remaining varieties, whereas the composition of PBA Samira was moderately different. The hierarchical cluster analysis also suggested that based on their chemical composition, two general groups could be made of the remaining eight varieties, one comprising Nura, Doza, and PBA Zahra and the other comprising Fiord, Fiesta VF, PBA Nasma, PBA Warda, and Farah.
T A B L E 3 Average total phenolic content (mg GAE 100 g −1 DW) of faba bean varieties Note. Samples with the same letter in the last column were not statistically different at α = 0.05 according to post-hoc Tukey testing. Abbreviation: DW, dry weight. The different letters correspond to different statistical groups.

| ATR-MIR analysis
There was a visible difference in the amplitude of the MIR absorbance between 2016 and 2017 samples, with the 2017 samples showing greater absorbance overall. However, the location of spectral peaks was virtually identical between the 2 years. There was little visible difference in the average spectra between the Charlick and Freeling sites, although slightly larger peak at 2,990-2,950 cm −1 was observed in the Freeling samples. There was also little visible difference in the average spectra of the 10 faba bean varieties, although varietal differences in the size of the 2,990-2,950 cm −1 peak were noted ( Figure 3). The main spectral peaks observed were attributed to a range of constituent compounds, including water, compositional polysaccharides, and protein (Table S1).
Prior to further analysis, the individual spectra were preprocessed to the second derivative to remove any differences in the absorbance amplitudes resulting from variation in the level of contact between the sample and the reflection module. This successfully removed any baseline amplitude variation while amplifying the differences in peak positions, shapes, and relative amplitudes.
To further explore the spectral variation, principal component analysis was conducted on the second derivative of the MIR spectra. F I G U R E 2 Hierarchical cluster analysis of the mean moisture, ferric reducing antioxidant potential, total phenlics, and total monomeric anthocyanin contents of the 10 faba bean varieties. The cluster analysis used Ward's method and the squared Euclidean distance F I G U R E 1 PC analysis of four chemical parameters (moisture, ferric reducing antioxidant potential, total phenlics, and total monomeric anthocyanin content), with the samples separated by variety. The loadings for the first two PCs of the PC analysis are also indicated. PC, principal component

| ATR-MIR Analysis-PLS discriminant analysis
PLS discriminant analysis (PLS-DA) has previously been highlighted as a powerful tool for the discrimination and authentication of grains of different origins (Gordon et al., 2019). Hence, PLS-DA was performed on the second derivative of the faba bean spectra. The growing year was able to be correctly classified in 87.4% of the samples (Table S2), indicating that this technique may be suitable for some authentication purposes when applied to faba bean flour. However, the successful classification rate by growing site was much lower, with an average of 60.2% of samples correctly assigned to their growing site (Table S2).
Further refinement of the PLS regression, perhaps through the isolation and selection of the most relevant waveband, is required before this technique can be utilised for authentication of growing site in faba bean flour.
Overall, the MIR analysis was found to provide valuable information on the chemical composition of the faba bean varieties. The preliminary results presented here are quite promising, indicating that F I G U R E 3 Average attenuated total reflectance mid-infrared spectra of the faba bean samples, before and after spectral processing through the use of the second derivative agronomic aspects such as the year of growth can be determined from the MIR spectra with reasonable certainty. With the development of more sophisticated methods of data analysis, the usefulness of this technology can only increase.

| CONCLUSION
The 10 varieties of Australian faba beans tested show considerable variation in their anthocyanin, phenolic, and antioxidant contents. In particular, PBA Rana contains much higher content of total phenolics and antioxidants than all other varieties tested. Varietal differences in chemical composition were highlighted through principal component analysis and hierarchical cluster analysis. MIR spectra obtained from the faba bean flour provided insight into the phytochemical composition and variation between the varieties. Classification using PLS-DA allowed the growing year to be successfully predicted from the MIR over 87% of the time; however, prediction of growing location was less successful (mean accuracy 59%).