The research was carried out in Selçuk University, Faculty of Agriculture, Sarıcalar Research and Application Farm, in 2017. The trial zone is located between 38˚ 05’ northern latitudes and 32˚ 36’ east longitudes and its average height from the sea is 1006 m. In the study, two wild rootstocks Citrullus lanatus var. citroides (A1 and A2), one open-pollinated Lagenaria siceraria (A3) and one Cucurbita maxima Duchesne x Cucurbita moschata Duchesne hybrid (‘TZ-148’) (A4) were used as rootstocks. ‘Crimson Tide’ (CT), a commercial watermelon variety widely grown in Turkey, was grafted on these four rootstocks and production materials were obtained. Ungrafted CT (K) was used as control. A total of 5 trial subjects were used in three replications according to the randomized block design. All cultural practices up to the harvest were applied to all applications equally. Watermelon fruits were harvested on time and after harvest, watermelon seeds from each application were hand collected seperately, washed to remove the fruit pulp residue and dried in an environment free from sunlight at room temperature for a week. The seeds were kept in freezer bags at 4 ºC until oil extraction and analyses.

Seeds (50 samples) were counted and weighed, and the weight was multiplied by 20 to obtain a thousand grain weight (Pradhan, 2010). The width, length and thickness of the seed grains were determined in mm using a caliper.

According to AOCS (1980) with some modifications, approximately 2 g of the milled seeds and kernels (shelled seeds) were extracted using 250 ml of petroleum ether in the Soxhlet device for 4 hours. The solvent was removed from the miscella by a rotary vacuum evaporator (RE100-Pro, Scilogex, CT, USA) set at 40 ºC, and the oil contents of the samples were calculated.

The acid number analysis was determined by making some modifications according to TSI (2003). Firstly, 25 ml of ethyl alcohol-diethyl ether mixture (1:1) were added to 2 g oil sample. The mixture was titrated with a 0.1 N ethanolic NaOH solution with 1 ml phenolphthalein indicator. Results were given as mg NaOH/g oil.

The fatty acid methyl esters of seed oils were prepared according to the method described by Williams (1984) with some modifications. The sample was shaken vigorously in isooctane (0.4 g in 4 ml) and combined with 0.2 ml of 2 N methanolic potassium hydroxide. After the duration in the dark for 6 min, a drop of methyl orange and 0.45 ml of 1 N HCl were added, mixed and centrifuged (2000 rpm/5 min). The supernatants were then analyzed by a GC (Agilent, Santa Clara, CA) equipped with an injector, a capillary column (Innowax, 100 m length, 0.25 mm i.d., 0.20 µm of thickness), and a flame ionization detector (FID). Hydrogen was used as carrier gas. The temperature of the detector and injector was set as 250 ^oC. The initial oven temperature was 180 ^oC and increased to 220 ^oC at 30 ^oC/min.

Color measurements of the seeds and their oils were determined as L* (lightness-darkness), a* (redness-greenness), b* (yellowness-blueness), C* (chroma) and h (hue) values in the CIE color system profile using Minolta colorimeter (CM-5, Minolta, Japan).

To extract the polyphenols and antioxidant effective compounds from seeds, 0.1 g of defatted ground seed obtained at the end of the oil extraction was weighed and 40 ml of 80% methanol solution were added. Homogenization was obtained by an ultra-turrax operation for 2 minutes at 200 rpm. Similarly, 1 g of oil sample was mixed with 2 ml methanol for 10 minutes in a shaker. The supernatants of oil-methanol and seed-methanol obtained from the centrifugation at 4100 rpm for 10 minutes was used for total phenolic compounds and antioxidant activity analyses.

The total phenolic compounds (TPC) of oils and seeds were obtained using the Folin-Ciocalteu method (Singleton and Rossi, 1965), with slight modifications. The Folin-Ciocalteu reagent of 1.25 ml (0.2 N) and 1 ml of sodium carbonate (75 g/l) solution were added to 250 µl of extract. Before determining the absorbance values of samples by spectrophotometer at 765 nm, they were kept in the dark for 2 hours. Results were expressed as mg gallic acid equivalent (GAE) /kg.

The antioxidant activity of watermelon seeds and their oils was determined using the 2, 2-diphenyl-1-picrylhydracyl (DPPH) radical scavenging method, according to an adapted colorimetric procedure (Akbulut et al., 2009). After adding 1.95 ml of DPPH reagent to 50 µl extract, the mixture was kept in the dark for 30 minutes. The absorbance values of the samples were measured by spectrophotometer at 515 nm and antioxidant activity values were calculated in mmol Trolox Equivalent (TE)/kg.

Significant differences among samples were determined through analysis of variance (ANOVA) and Duncan’s multiple range test using SPSS package program version 22.0. Differences were considered significant at p < 0.01. The JMP 13.0 software was used to analyze the important parameters by Principal Component Analysis (PCA). The relation between rootstocks was compared by drawing a loading plot graph using the same software and the relationship between each parameter and another was presented on the score plot graph.

Some physicochemical properties of the seeds obtained from watermelon fruits are given in Tables 1 and 2. According to Table 1, the rootstock had statistically significant effects on thousand grain weight, width-length-thickness dimensions, total phenolic compounds (TPC) and antioxidant activity (AA) values for the seeds. Thousand grain weight results of seeds ranged from 40.70 g to 48.70 g, and watermelons grafted on the bottle gourd rootstock (A3) with the highest thousand grain weight were statistically separated from other samples according to Duncan test results. While the thousand grain weight of the seeds of watermelons grafted on Citroides 1 (A1), Citroides 2 (A2) and TZ148 (A4) rootstocks was significantly decreased, that of watermelon seeds grafted on the bottle gourd (A3) was significantly increased (p < 0.01). Similar to the results of this study, the grafting of Crimson Sweet (CS) watermelon (Citrullus lanatus) steel on three different rootstocks (Cucurbita maxima x Cucurbita moschata ‘NUN-9075’, Lagenaria siceraria ‘Argentario’, and citron watermelon Citrullus amarus Schard ‘PI296341’) had a significant effect on the thousand grain weight of the seeds obtained (Kombo and Sari, 2019).

As seen in Table 1, the width, length and thickness of the watermelon seeds varied from 5.0-5.8 mm, 8.4-9.0 mm and 1.7-1.9 mm, respectively. A decrease in the total amount of phenolic compounds of all watermelon seeds due to grafting was observed. It was determined that the seeds with the highest TPC belonged to the control sample (2340.9 mg GAE/kg) and the lowest to the A3 sample (1140.9 mg GAE/kg). In an analysis of the seeds of Citrullus lanatus var. citroides, Acar et al. (2012) found that the TPC of the seeds were 130 mg GAE/kg, and the reason for the decrease in the TPC of A1 and A2 samples may be due to the low content in the citroides species. Many recent studies (Seidu and Otutu, 2016; Tabiri et al., 2016) have shown that watermelon seeds have much more TPC than those of this study. These differences in TPC may depend on factors such as variety and environmental conditions. The antioxidant activity values of watermelon seeds varied between 3.57 and 6.80 mmol TE/kg. As in the amount of phenolic compounds, a statistically significant decrease (except A2 sample) was observed in the antioxidant activity values of the seeds obtained as a result of grafting. Tabiri et al. (2016) reported the antioxidant activity values of the seeds of “Charleston gray”, “Black diamond” and “Crimson sweet” watermelon varieties as 82.59, 96.63 and 130.29 mmol TE/kg, respectively.

The decrease in color values of the watermelon seeds was found to be significant for L*, a*, b* and h, but not for C* (Table 2). The L*, a*, b*, C* and h values varied between 29.77-37.73, 3.17-5.25, 7.38-12.92, 63.49-67.77 and 8.04-13.92, respectively. If we expanded the results for watermelon seeds, no significant difference was observed only in the C* values after grafting.

The seed oil content, kernel oil content, acid value, TPC and AA values of the samples are presented in Table 3 and the fatty acid profile in Table 4. It has been determined that the differences that occured in these values except for the kernel oil content and TPC results are statistically significant (p < 0.01). The increase observed in the oil efficiency of all the seeds by grafting was found statistically significant. Sample A2 (26.34%) had the highest seed oil content, followed by A1 (25.90%), A4 (25.46%), A3 (24.75%) and lastly K (24.05%). However, there was no similarity between seed oil content and kernel oil content in the samples. Kernel oil content values were listed in descending order as follows: A1, K, A4, A3, A2. The oil contents in the seeds and kernels of Citrullus lanatus var. citroides were reported by Acar et al. (2012) as 26.83 and 52.34%, respectively. In addition, four Lagenaria siceraria cultivars were studied by Essien et al. (2013) and the content in seed oils, which is close to our results, was obtained in the range of 23.0-29.5%. Tabiri et al. (2016) reported that the seed oil contents in “Crimson sweet”, “Charleston gray” and “Black diamond” varieties were 26.50, 26.83 and 27.85%, respectively. In one of the studies, the seed oil content values of the “Crimson” cultivar was 19.23% (de Conto et al., 2011), while in another, in the “Mateera” and “Sugar baby” varieties were 25.53% (46.85% kernel oil content) and 21.93% (38.88% kernel oil content), respectively (Wani et al., 2013). Considering all of this evidence, it seems that oil content values can vary based on the watermelon variety.

The acid values for the samples (4.33 to 5.35 mg NaOH/g) were relatively high compared to that reported for watermelon (Citrullus lanatus L.) seeds (2.37 mg NaOH/g) (Duduyemi et al., 2013). With the effect of grafting, there was a significant decrease in acid value as given in Table 3 (p < 0.01). Among the grafted rootstocks, A2 (4.86 mg NaOH/g) had the highest acid value, while A1 (4.33 mg NaOH/g) had the lowest. Essien et al. (2013) determined the acid values for the seed oils from four gourd cultivars as between 2.1-2.4 mg KOH/g. For this value, the decrease in A3 compared to the control sample may be due to the low acidity of the oil from the gourd seeds. Wani et al. (2013) obtained oil from “Mateera” and “Sugar baby” cultivars and reported that the acid values in mg KOH/g oil were 4.27 and 6.46, respectively. In another investigation, in which cold press was applied, the acid value of the “Crimson” variety seed oil was found to be 9.5 mg KOH/g oil (Rezig et al., 2019). The TPC of the seed oils from grafted (49.25-53.56 mg GAE/kg) and ungrafted (51.86 mg GAE/kg) watermelons was lower than that of Iranian watermelon seed oils (111 mg GAE/kg), as reported by Hashemi et al. (2017). In addition, the TPC of the seed oils from four gourd cultivars was found to be between 80 and 105 mg GAE/kg as reported by Essien et al. (2015). The decrease observed in the antioxidant activity of oils by grafting was found statistically significant (p < 0.01). The control sample (K) (0.52 mmol TE/kg) had the highest antioxidant activity, followed by the A4 (0.49 mmol TE/kg), A3 (0.46 mmol TE/kg), A1 (0.45 mmol TE/kg), A2 (0.44 mmol TE/kg) samples.

The polyunsaturated, monounsaturated and saturated fatty acid profiles of the watermelon seed oils statistically changed with the application of grafting (p<0.01) (Table 4). The fatty acid distribution in the ungrafted control sample (Crimson Tide) was as follows: 63.40% linoleic acid, 16.74% oleic acid, 11.90% palmitic acid and 7.97% stearic acid. When compared to the control sample, an increase in the amount of unsaturated fatty acids (UFAs) in the A1, A2 and A4 samples and of saturated fatty acids (SFAs) in the A3 was observed. The highest linoleic acid content was determined in the seed oils from the citroides rootstock samples as 64.81% (A1) and 64.68% (A2), the lowest oleic acid content was also determined in the same samples as 15.98% and 15.78%, respectively. According to the control sample, the grafting application in citroides samples caused a statistically significant increase in linoleic acid content, with a significant decrease in the oleic acid content. The seed oil from the bottle gourd rootstock sample (A3) had the highest oleic acid content (17.86%), but the lowest linoleic acid (61.88%). No work regarding the effect of grafting on the fatty acid profile of the seed oils from watermelons or the other plants in the Cucurbitaceae family could be found. This situation is also valid for all analyses performed on the seeds and their oils except thousand grain weight. The obtained results for the fatty acid profiles show similarity with the studies performed by Wani et al. (2013), Hashemi et al. (2017), Rezig et al. (2019), and Eke et al. (2021). With the effect of grafting, a significant difference was observed in the color values of watermelon seed oils (Table 5). The increase in L* and h values and the decrease in a*, b* and C* values were statistically significant (p < 0.01). The values for L*, a*, b*, C* and h in the oils were between 87.09-89.08, (-)0.47- (-)1.21, 62.11-80.37, 62.12-80.37 and 90.34-91.12, respectively.