Bioactive compounds and physical attributes of Cornus mas genotypes through multivariate approaches

Seed-propagated cornelian cherry genotypes in Mesudiye town of Ordu province constituted the material of the present study. A total of 100 fruits were taken from each genotype. Collected fruits were placed into plastic boxes and transported to the laboratory in cooling thermos bottles.

Experiments were conducted in 5 replicates with 20 fruits in each replicate. Stones of fruits were taken out with a stainless-steel knife and resultant flesh was homogenized in a blender. Homogenized fruit samples were placed into falcon tubes (about 50 g) and preserved at −20°C until bioactive compound analyses.

Fruit 1.1-diphenyl-2-picryl-hydraziyl (DPPH) antioxidant activity was determined with the aid of modified method of Brand-Williams et al., (1995). For analysis, 0.26 mM DPPH solution was prepared. Then, 100 μL fruit extract was supplemented with 2,900 μL ethyl alcohol and 1 mL DPPH solution. The resultant mixture was vortexed for 30 min and kept at dark. Sample absorbance readings were performed in a spectrophotometer at 517 nm wavelength. Results were expressed in μmol Trolox equivalent (TE) · kg⁻¹ mmol of fresh weight.

Sample total flavonoid content was determined with the aid of the method recommended in Chang et al., (2002). Fruit extracts (1,000 μL) were supplemented with 3.3 mL methanol, then with 0.1 mL 10% AlCl₃·6H₂O and 0.1 mL of 1 M potassium acetate (CH₃COOK). Spectrophotometer readings were performed at 415 nm wavelength and total flavonoid contents were expressed in quercetin equivalent (mg QE · kg⁻¹) of fresh weight.

Folin-Ciocalteu's method was used to determine total phenolics of the samples. Initially, 500 μL fresh fruit extract was supplemented with 4.2 mL distilled water, then with 100 μL Folin-Ciocalteu reagent and 2% sodium carbonate (Na₂CO₃). The resultant solution was incubated for 2 h. Following the incubation, spectrophotometer readings were performed at 760 nm wavelength and the results were expressed in gallic acid equivalent (mg GAE · kg⁻¹) of fresh weight (Beyhan et al., 2010).

Samples were named between G1 and G18 for 18 genotypes (Figure 1). To determine the size and the shape attributes of each genotype, 75 cornelian cherry samples were placed over a lightbox in 5 × 15 matrix arrangement. Nikon D90 model camera was used to take the images of fruits at horizontal and vertical orientations. Imaging was performed from standard height of 50 cm.

Figure 1

Physical characteristics of cornelian cherry fruits were determined with the aid of the image processing method. SigmaScan Pro v.5.0 software was used to determine the size and shape parameters. Length (L), width (W), thickness (T), projection area (PA), equivalent diameter (ED) and perimeter (P) measurements of cornelian cherry samples imaged at horizontal and vertical orientations were directly made with the aid of the software (Figure 2). The other size and shape parameters of cornelian cherry fruits and equations used to determine these parameters are provided in Table 1.

Figure 2

Elliptic Fourier analysis (EFA) includes the stages of (i) formation of the points defining the closed contour of two-dimensional geometry, (ii) determination of x and y coordinates of each point over the contour and (iii) transformation of coordinates into a mathematical function and obtaining function coefficients defining the shape (Sayıncı, 2016). Function coefficients are generated based on the number of harmonics and each harmonic generates four Fourier coefficients (a_n, b_n, c_n and d_n). The a_n and b_n coefficients correspond to x coordinates and c_n and d_n coefficients correspond to y coordinates of the contour (Neto et al., 2006; Özkan-Koca, 2012).

SHAPE (version 1.03) software was used to compare the geometry of cornelian cherry genotypes identified through the EF method (Iwata and Ukai, 2002). For analysis in software, fruit images were converted into 24-bit *bmp files. EFA and statistical assessments were performed with the use of four sub-modules; (i) ‘ChainCoder’ module was used for image processing and shape contour generation, (ii) ‘Chc2Nef’ module was used to normalize contour codes and to get elliptic Fourier descriptors, (iii) ‘PrinComp’ module was used for principal component analysis (PCA) of descriptors and to generate component scores and (iv) ‘PrinPrint’ module was used to visualise shape variations of fruit image contours. Analyses were conducted over 20 harmonics.

Nutritional attributes of cornelian genotypes were subjected to analysis of variance at 95% significance level by using one-way ANOVA procedure and statistically significant differences among the genotypes were revealed with the aid of Duncan's multiple comparison test. Statistical analyses were performed with the use of SPSS v.23.0 software.

One-way ANOVA was used to test the differences in size and shape parameters of cornelian cherry genotypes. Significant means were compared with the use of Tukey's multiple comparison test at 95% significance level. PCA was performed to elucidate the relationships between size and shape parameters of the cornelian cherry genotypes. Eigen statistics were provided for principle components presenting the relationships between the genotypes. Factor loads were used to present the differences between the genotypes over a scatter chart.

Shape contours of cornelian cherry images were normalised to get elliptic Fourier scores presenting common variance. The components explaining the variance between these scores were visualised and differences between the genotypes were determined with the aid of multivariate analysis of variance (MANOVA). In the significance test of variance, ‘Wilks's Lambda’ and ‘Pillai Trace’ statistics were used. Pair-wise comparisons were made with the aid of Bonferroni-correction test (Bonferroni corrected P values). Statistical analyses were conducted with the use of PAST v.3.25 software (Hammer et al., 2001).

For a visual comparison of differences between the genotypes, PC scores obtained from elliptic Fourier descriptors were used and these scores were subjected to discriminant analysis. At the end of this analysis, centroid coordinates of each genotype were determined, and visual similarities of the genotypes were presented over a scatter chart.

There were significant differences in antioxidant activity, total flavonoids and total phenolics of the seed-propagated cornelian cherry genotypes (p < 0.05) (Table 2).

The greatest antioxidant activity was obtained from G13 (152.420 mmol TE · kg⁻¹), G14 (152.180 mmol TE · kg⁻¹) and G11 (151.943 mmol TE · kg⁻¹), which were all placed into the same statistical group. The lowest antioxidant activity was obtained from G10 (55.062 mmol TE · kg⁻¹) and G18 (59.670 mmol TE · kg⁻¹) genotypes that have yellow fruits.

Present findings on antioxidant capacity of seed-propagated cornelian cherry genotypes were found higher than the values reported in previous studies. Dragovic-Uzelac et al. (2007) conducted a study in Croatia and reported DPPH-free radical scavenging activity of fruit extracts of two cornelian cherry genotypes as between 33.41 and 39.89 mmol Trolox · kg⁻¹, Cosmulescu et al. (2019) conducted a study in Romania and reported the antioxidant activity of 6 genotypes as between 1.24 and 2.71 μmol Trolox 100 g⁻¹; Klymenko et al. (2019) in Ukraine, reported antioxidant activity of 20 genotypes as between 5.94 and 16.56 μmol Trolox · g⁻¹. On the other hand, Kucharska et al. (2011) conducted a study in Poland on cornelian cherry gene sources and reported the greatest antioxidant activity for Dublany cultivar (20.72 μmol Trolox · g⁻¹) and the lowest value for Juliusz cultivar (10.85 μmol Trolox · g⁻¹).

Total flavonoids of the present genotypes varied between 286.40 mg QE · kg⁻¹ (G10) and 2,882.80 mg QE · kg⁻¹ (G7) and a large variation was observed in total flavonoid contents of the genotypes.

Hassanpour et al. (2011) reported total flavonoids of cornelian cherry fruits in Iran as between 321.27 and 669.00 mg QE · 100 g⁻¹; Cosmulescu et al. (2019) reported total flavonoids of 6 cornelian cherry genotypes collected from different regions of Romania in the range of 12.14–64.48 mg QE · 100 g⁻¹. Present findings on total flavonoids were lower than the values of Hassanpour et al. (2011) and greater than the values of Cosmulescu et al. (2019).

Total phenolics of the present genotypes varied between 2,644.80 mg GAE · kg⁻¹ (G2) and 12,959.00 mg GAE · kg⁻¹ (G14).

In previous studies conducted with cornelian cherry fruits, Tural and Koca (2008) reported total phenolics of 24 genotypes selected from Black Sea region of Turkey as between 2.81 and 5.79 mg GAE · g⁻¹; Yılmaz et al. (2009) reported total phenolics of 16 genotypes collected from Western Black Sea and Inner Anatolia of Turkey range from 26.59 to 74.83 mg GAE · g⁻¹; Hassanpour et al. (2011) reported total phenolics of cornelian cherry genotypes grown in Iran varied between 1,097.19 and 2,695.75 mg GAE · 100 g⁻¹; Dragovic-Uzelac et al. (2007) reported total phenolics in Croatia as between 2,095 and 3,055 mg GAE · kg⁻¹. Present findings on total phenolics were greater than the values of Tural and Koca (2008), Yılmaz et al. (2009) and Dragovic-Uzelac et al. (2007) and lower than the values of Hassanpour et al. (2011).

Considering the bioactive components of the genotypes, it was observed that yellow fruits generally had lower antioxidant activity and total phenolics; there were significant variations among the genotypes and present values were mostly greater than earlier reports. Although the differences are thought to be mainly caused by the variation of the examined varieties, differences in climate and soil characteristics, the geographical situation of the area where the cultivation is made, the type and time of harvest, the storage or processing of the crop, the method or periodic differences of the applied cultural processes, crop load, tree ages and fruit ripening levels cause significant differences on the final form and the amount of the phytochemical composition (Benvenuti et al., 2004; Yılmaz et al., 2009; Ercisli et al., 2011; Gündüz et al., 2013; Ersoy et al., 2018b; Sochorova et al., 2019; Szot et al., 2019b; Bolat and Ikinci, 2020; Mertoglu et al., 2020).

The greatest fruit lengths were obtained from G9 and G14 genotypes whereas the greatest geometric mean diameters and surface areas were observed in G3, G4, G9 and G18 genotypes (Table 3). The lowest surface areas were observed in G2, G5 and G8 genotypes. On the other hand, G12 and G18 genotypes had the closest geometry to sphere, thus had the lowest shape index values.

The greatest projected area, equivalent diameter and perimeter values at horizontal orientation were observed in G9 and G14 genotypes, whereas the lowest values were observed in G2 and G8 genotypes (Table 4). The greatest values at vertical orientation were observed in G18 genotype and the lowest values were observed in G14 genotype.

Some shape parameters were calculated with the use of dimensional attributes of cornelian cherry fruit and the results are given in Table 5. According to elongation averages at horizontal orientation, the difference between fruit length and widths were greater in G5, G9, G14 and G17 genotypes than the others. Therefore, relevant genotypes had more ellipsoidal form than the other genotypes. Coating ratio of these genotypes to a circular sieve opening was also lower in these genotypes than the others. The fruits with a full circle geometry had a shape factor of 1. At horizontal orientation, genotypes G1, G3 and G15 had the closest geometry to a circle. According to the elongation values at vertical orientation, all genotypes, except for G18, had close geometry to a circle and were able to largely cover a circular opening. The greatest shape factors at vertical orientation were observed in G2, G3, G8 and G14 genotypes.

PCA results are provided in Table 6. Two principle components (PC1 and PC2) were able to explain 92.56% of total variation in size and shape parameters of cornelian cherry genotypes. The variance elements explained by PC1 included size and shape parameters of cornelian cherry fruits. PC2 directly explained the variance in size parameters.

Based on the size and the shape parameters, factor scores identified for each genotype were presented in a scatter plot (Figure 3). Based on variance elements, the genotypes different in terms of size and shape parameters were indicated within a coloured frame. While the genotypes G2 and G8 with a negative factor score on PC2 had the smallest size, the genotypes G4 and G9 had greater size data than the others. The genotype G14 with the lowest sphericity was placed on the left side of PC1 axis and apart from the others. The genotype G5 exhibited a difference in elongation parameter.

Figure 3

When the physical attributes of 18 cornelian cherry genotypes were tested with PCA, it was observed that two principle components were able to explain a large portion of the variance in size and shape parameters. The physical characteristics of the genotypes G2, G4, G5, G8, G9 and G14, placed apart from the other genotypes on scatter plot, were remarkably different from the others.

Two principle components identified based on shape descriptors explained 86.65% of total variance in cornelian cherry genotypes (Figure 4). According to PC1, the greatest shape difference (explained variance of 80.05%) was attributed to genotypes with oval and circular geometry. Based on average shape geometry, it can be stated that cornelian cherry genotypes had an oval geometry. However, based on ±2 SD variations, it is possible to mention about pointed or spherical forms. According to PC2, shape differences were slightly (explained variance of 6.60%) attributed to the drop-like image of the fruits. According to elliptic Fourier analysis results, it can be stated that cornelian cherry genotypes had an oval appearance in shape geometry. However, there were some genotypes with a shape geometry close to a spherical form. Such a finding is especially significant for the geometry of sieve openings in size-based separation.

Figure 4

Results of multivariate variance analysis conducted to put forth the shape differences between cornelian cherry genotypes with the aid of EF method are provided in Table 7. The first two functions obtained from the discriminant analysis clearly put forth the shape differences of cornelian cherry genotypes and the entire variance was explained by these discriminant functions. Accordingly, the first function explained 78.6% and the second function explained 21.4% of the variation. According to Wilks’ Lambda and Pillai Trace statistics, there were significant shape differences between the genotypes. Pair-wise comparisons were made in the Hotelling test and the genotypes with insignificant differences between them were indicated in coloured format. In shape assessments based on fruit contours, there were not any genotypes similar to G12 genotype.

Among the cornelian cherry genotypes, the groups with similar shapes were presented in coloured fashion in the scatter plot (Figure 5). Accordingly, 18 cornelian cherry genotypes were divided into 5 main shape groups. The first group included the genotypes G5, G9, G14 and G17; the second group included genotypes G4, G7, G8, G11 and G18; the third group included genotypes G2 and G13; the fourth group included genotypes G1, G3, G6, G10, G15 and G16; the fifth group included genotype G12.

Figure 5

In the present study, differences in physical characteristics of cornelian cherry genotypes were determined with PCA. Physical characteristics included size parameters and some shape parameters calculated accordingly. Elliptic Fourier method allowed comparisons based only on the shape, without considering size parameters. When common assessments were made based on two methods, it was observed that the genotypes G5, G9 and G14 with different size parameters have a similar appearance in shape.