2.2.1. Samples details

under vigorous stirring for 10 min with 20 mL of an acidic aqueous methanol solution (80% methanol, 19% water, 1% formic acid). The resulting heterogeneous mixture was transferred into standard glass sample tubes and centrifuged at 2500 g/min for 10 min. After that, the supernatant was removed and the extraction was repeated. Collected supernatants were

The solid residue from the free phenolic extraction was subjected to alkaline hydrolysis to recover the bound phenolic compounds, according to Mattila et al. [16]. Distilled water (12 mL) and 5 mL of 10 M NaOH were added to the residue and stirred overnight at room temperature. The mixture was acidified to pH = 2 and then extracted three times with 15 mL of a 1:1 (v/v) mixture of cold diethyl ether and ethyl acetate by manually shaking and centrifuging. Organic layers were combined, evaporated to dryness, and dissolved into 2 mL of the aqueous metha-

For HPLC/DAD analyses dry extracts were reconstituted in 3 mL of the extracting solvent and immediately analyzed. Quantitative analyses were carried out on a UltiMate3000 "UHPLC focused" instrument equipped with a binary high pressure pump, a Photodiode Array detector, a Thermostatted Column Compartment and an Automated Sample Injector (Thermo Scientific, Italy). Collected data were processed through a Chromeleon Chromatography Information Management System v. 6.80. Chromatographic runs were all performed using a reverse-phase column (Gemini C18, 250 4.6 mm, 5 μm particle size, Phenomenex, Italy) equipped with a guard column (Gemini C18 4 3.0 mm, 5 μm particle size, Phenomenex, Italy). Wheat polyphenols were eluted with the following gradient of B (formic acid, 2.5% solution in acetonitrile) in A (2.5% solution of formic acid in water): 0 min: 5% B; 10 min: 15% B; 30 min: 25% B; 35 min: 30% B; 50 min: 90% B; then kept for 7 min at 100% B. The solvent flow rate was 1 mL/min and. Quantifications were carried out at 350 nm using orientin (R2 = 0.9999) as external standard; the detector was set at 280 nm to build the calibration curve for vanillic acid (R2 = 0.9997), whilst vitexin (R2 = 0.9999), caffeic acid and ferulic acid were quantified at 330 nm using the corresponding reference substances (R2 = 0.9999 and R2 = 0.9998, respectively). The same reference wavelength was used for the quantification of coumarins against p-coumaric acid

In order to unambiguously identify the chromatographic signals and/or to confirm peak assignments, a series of HPLC/ESI/MS analyses were performed on wheat samples. In this case, variable aliquots (1.0–1.5 mL) of the above mentioned hydro-alcoholic solutions coming from quantitative analyses (see previous paragraph) were transferred into standard laboratory vials and brought to dryness in vacuo with a rotary evaporator (Heidolph Laborota 400). The resulting yellowish residues were then re-dissolved in 500 μL of the original hydroalcoholic solution and submitted to qualitative analyses. The HPLC apparatus used was the same described above, whilst ESI mass spectra were acquired by a Thermo Scientific Exactive Plus Orbitrap MS (Thermo Fisher Scientific, Inc., Milan, Italy), using a heated electrospray ionization (HESI II) interface. Mass spectra were recorded operating in negative ion mode in the m/z

pooled, evaporated to dryness, and then stored at 20C until use.

nol solution to analytical determinations.

26 Rediscovery of Landraces as a Resource for the Future

(R2 = 0.9998). All analyses were carried out in triplicate.

2.1.5. Identification of main components via HPLC/ESI-MS

2.1.4. HPLC/DAD quantitative analyses

For the glumes image analysis, ears of the same ten wheat landraces were reaped, at the time of maximum ripening, in order to include a widest morphological and environmental variability, the wheat ears were collected during three consecutive years (2012, 2013, 2014).

From three to six ears were sampled and from two to four glumes were removed from the spikelets of the ear middle section and from the both sides of each ear. The glumes were stored at room temperature under controlled conditions (20C and 50% RH).

### 2.2.2. Images acquisition

Digital images of glumes samples were acquired using a flatbed scanner (ScanMaker 9800 XL, Microtek Denver, CO), applying the same resolution and scanning area conditions reported in Grillo et al. [4]. As suggested by Venora et al. [17], before digital image capture, the scanner was standardized according to the calibration protocol proposed by Shahin and Symons [18]. Morpho-colorimetric features were only measured for sound intact glumes, rejecting that ones with broken beak or shoulder, distinguishing in right and left side of the ear. A total of 902 wheat glumes were analyzed (Table 1).


Table 1. List of the ten different wheat local varieties studied.

#### 2.2.3. Image processing and analysis

All the images were processed and analyzed using the software package KS-400 V. 3.0 (Carl Zeiss, Vision, Oberkochen, Germany). The same macro used by Grillo et al. [4], specifically developed for the characterization of wheat glumes was applied to perform automatically all the morpho-colorimetric measurements on the glume samples of the present study.

The macro allowed to compute 138 quantitative variables measured for each analyzed left and right glume (Tables 2 and 3). In particular, it was possible to measure 18 parameters descriptive of the glume size and shape and 20 features descriptive of the glume surface color. Afterwards, applying the same procedure reported by Orrù et al. [19], 78 quantitative Elliptic Fourier Descriptors (EFDs) were used to describe the shape of the glume. Finally, the macro was kitted to compute 11 Haralick's descriptors including the relative standard deviations, as reported in Lo Bianco et al. [20].

$$G = \begin{bmatrix} p(1,1) & p(1,2) & \cdots & p(1,N\_{\mathfrak{g}}) \\ p(2,1) & p(2,2) & \cdots & p(2,N\_{\mathfrak{g}}) \\ \vdots & \vdots & \ddots & \vdots \\ p(N\_{\mathfrak{g}},1) & p(N\_{\mathfrak{g}},2) & \cdots & p(N\_{\mathfrak{g}},N\_{\mathfrak{g}}) \end{bmatrix} \tag{1}$$

Feature Equation

where μx, μy, σ<sup>x</sup> and σ<sup>y</sup> are the means and the standard deviations of px and py.

http://dx.doi.org/10.5772/intechopen.79595

29

Phenolic Fingerprinting and Glumes Image Analysis as an Effective Approach for Durum Wheat Landraces…

where x and y are the coordinates (row and column) of an entry in the cooccurrence matrix, and px+y(i) is the probability of co-occurrence matrix

coordinates summing to x+y.

The basis for these features is the gray-level co-occurrence matrix (G in Eq. (1)). This matrix is square with dimension Ng, where Ng is the number of gray levels in the image. Element [i,j] of the matrix is generated by counting the number of times a pixel (p) with value i is adjacent to a pixel with value j and then dividing the entire matrix by the total number of such comparisons made. Each entry is therefore considered to be the probability that a pixel with value i will be found

Table 2. Haralick's descriptors measured as reported in Haralick et al. [21].

Har 1 Angular second moment

Har 4 Sum of square: variance

Har 5 Inverse difference moment

Har 2 Contrast

Har 3 Correlation

Har 6 Sum average

Har 7 Sum variance

Har 8 Sum entropy

Har 9 Entropy

Har 10 Difference variance

Har 11 Difference entropy

adjacent to a pixel of value j.

#### 2.3. Statistics

The data, obtained from chemical and image analysis, were used to build a global database. Statistical elaborations were executed using SPSS software package release 16.0 (SPSS Inc. for Windows, Chicago, Illinois, USA), and the stepwise Linear Discriminant Analysis (LDA) method was applied to identify and discriminate among the investigated wheat samples [23]. This approach is commonly used to classify/identify unknown groups characterized by quantitative and qualitative variables [24–27], finding the combination of predictor variables with the aim of minimizing the within-class distance and maximizing the between-class distance simultaneously, thus achieving maximum class discrimination [28–31]. Then, the stepwise procedure, carried out as explained in [4], identifies and selects the most statistically significant features among the chemical metabolites and the 138 traits measured on each glume. Finally, a cross-validation procedure was applied to verify the performance of the identification system, testing individual unknown cases and classifying them on the basis of all others [32].

All the raw data were standardized before starting any statistical elaboration. Moreover, in order to evaluate the quality of the discriminant functions achieved for each statistical comparison, the Wilks' Lambda, the percentage of explained variance and the canonical correlation between the discriminant functions and the group membership, were computed. The Box's M test was executed to assess the homogeneity of covariance matrices of the features chosen by the stepwise LDA while the analysis of the standardized residuals was performed to verify the homoscedasticity of the variance of the dependent variables used to discriminate among the groups' membership [33]. Kolmogorov-Smirnov's test was performed to compare the empirical distribution of the discriminant functions with the relative cumulative distribution function of the reference probability distribution, while the Levene's test was executed to assess the equality of variances for the used discriminant functions calculated for groups membership [34].

2.2.3. Image processing and analysis

28 Rediscovery of Landraces as a Resource for the Future

reported in Lo Bianco et al. [20].

2.3. Statistics

All the images were processed and analyzed using the software package KS-400 V. 3.0 (Carl Zeiss, Vision, Oberkochen, Germany). The same macro used by Grillo et al. [4], specifically developed for the characterization of wheat glumes was applied to perform automatically all

The macro allowed to compute 138 quantitative variables measured for each analyzed left and right glume (Tables 2 and 3). In particular, it was possible to measure 18 parameters descriptive of the glume size and shape and 20 features descriptive of the glume surface color. Afterwards, applying the same procedure reported by Orrù et al. [19], 78 quantitative Elliptic Fourier Descriptors (EFDs) were used to describe the shape of the glume. Finally, the macro was kitted to compute 11 Haralick's descriptors including the relative standard deviations, as

The data, obtained from chemical and image analysis, were used to build a global database. Statistical elaborations were executed using SPSS software package release 16.0 (SPSS Inc. for Windows, Chicago, Illinois, USA), and the stepwise Linear Discriminant Analysis (LDA) method was applied to identify and discriminate among the investigated wheat samples [23]. This approach is commonly used to classify/identify unknown groups characterized by quantitative and qualitative variables [24–27], finding the combination of predictor variables with the aim of minimizing the within-class distance and maximizing the between-class distance simultaneously, thus achieving maximum class discrimination [28–31]. Then, the stepwise procedure, carried out as explained in [4], identifies and selects the most statistically significant features among the chemical metabolites and the 138 traits measured on each glume. Finally, a cross-validation procedure was applied to verify the performance of the identification system,

testing individual unknown cases and classifying them on the basis of all others [32].

for the used discriminant functions calculated for groups membership [34].

All the raw data were standardized before starting any statistical elaboration. Moreover, in order to evaluate the quality of the discriminant functions achieved for each statistical comparison, the Wilks' Lambda, the percentage of explained variance and the canonical correlation between the discriminant functions and the group membership, were computed. The Box's M test was executed to assess the homogeneity of covariance matrices of the features chosen by the stepwise LDA while the analysis of the standardized residuals was performed to verify the homoscedasticity of the variance of the dependent variables used to discriminate among the groups' membership [33]. Kolmogorov-Smirnov's test was performed to compare the empirical distribution of the discriminant functions with the relative cumulative distribution function of the reference probability distribution, while the Levene's test was executed to assess the equality of variances

ð1Þ

the morpho-colorimetric measurements on the glume samples of the present study.


The basis for these features is the gray-level co-occurrence matrix (G in Eq. (1)). This matrix is square with dimension Ng, where Ng is the number of gray levels in the image. Element [i,j] of the matrix is generated by counting the number of times a pixel (p) with value i is adjacent to a pixel with value j and then dividing the entire matrix by the total number of such comparisons made. Each entry is therefore considered to be the probability that a pixel with value i will be found adjacent to a pixel of value j.

Table 2. Haralick's descriptors measured as reported in Haralick et al. [21].


To graphically highlight the differences among groups, multidimensional plots were drawn

Phenolic Fingerprinting and Glumes Image Analysis as an Effective Approach for Durum Wheat Landraces…

Phenolics are mainly concentrated in the outer layers of kernel and contribute to the wheat flour nutraceutical value owing to their antioxidant, anti-inflammatory and anticancer properties [35]. In literature, ca. 70 different phenolic compounds, including coumarins, phenolic acids, anthocyanins, flavones, isoflavones, proanthocyanidins, stilbenes and lignans, were

Referring to flavones, whose interest has grown enormously due to their putative beneficial effects against atherosclerosis, osteoporosis, diabetes mellitus and certain cancers [36] 5,7,4<sup>0</sup>

sentatives in wheat, where they accumulate as 6-C and/or 8-C-glycosidic conjugates. The 8-Cglucosides of apigenin and luteolin are also known as vitexin and orientin, respectively.

Hydro-alcoholic extracts from wheat grains were exhaustively analyzed by means of HPLC/ DAD and HPLC/ESI-MS. Although the major portion of phenolics in grains exist in the bound form [37], there is a general trend for studying polyphenols in the free form when dealing with chemotaxonomic studies [38, 39]. The chromatograms relating to free phenolics profile of durum wheat grains showed ca. 20 different signals, eluting in the range from 7 to 30 min. Among these, 13 signals were tentatively identified: a preliminary analysis of the UV–VIS (in terms of spectrum shape and absorption maximum, see Table 4) spectra of the peaks revealed the presence of compounds belonging to the chemical subclasses of hydroxycinnamic acids and organic acids; several peaks showing the typical spectrum of apigenin derivatives were

The use of mass spectrometry as detector was helpful in tentatively identifying wheat metabolites (Table 4); peak assignments were further confirmed by comparison with literature data [14, 40, 41] and co-injection with pure reference standards when available (see material and

According to Lo Bianco et al. [11], three hydroxycinnamic acids were identified in wheat grains: caffeic acid (peak 4), ferulic acid (peak 10) and another member of this class (peak 1) for which unfortunately the MS spectrum was not determined. Vanillic acid (peak 3) was identified for its diagnostic UV–VIS and mass spectrum; the assignment was confirmed with co-injection with the corresponding standard. Peaks 2 and 6, showing almost identical UV–VIS spectra (a symmetrical absorption with λ max = 317 nm) were tentatively identified as coumarins; furthermore, peak 2 showed a clear mass spectrum with a pseudomolecular ion at 145.14 m/z (M-H)�. Presence of coumarins in durum wheat has been reported by other authors [14, 40]. The UV–VIS spectrum of peak 5 (λ max = 268, 35 nm) was typical of that of luteolin derivatives; the corresponding mass spectrum exhibited a base peak of 609.52 m/z (pseudomolecular ion) with no signals ascribable to fragments generated by the loss of sugars. The peak corresponding to luteolin aglycone was


http://dx.doi.org/10.5772/intechopen.79595

,40


31

using the first three discriminant functions.

3.1. Phenolic profile in wheat landraces

identified in durum wheat genotypes [14].

trihydroxyflavone (apigenin) and 5,7,3<sup>0</sup>

also detected.

methods).

3. Results and discussion

Table 3. List of morphometric features measured on seeds, excluding the elliptic Fourier descriptors (EFDs) calculated according to Hâruta [22] and the Haralick's descriptors reported in Table 2.

To graphically highlight the differences among groups, multidimensional plots were drawn using the first three discriminant functions.
