Seeds from the oranges of the Dörtyol variety (Citrus sinensis L. Osbeck) were provided by the Anadolu Etap Penkon Co. Inc. (Mersin, Turkey) from the 2013–2014 processing season. Around 50 kg of washed and dried seeds were kept in a deep-freezer (–20 ^oC) until cold pressing. All chemicals used were of analytical grade and purchased from Merck Co. (Darmstadt, Germany) and Sigma Chem. Co. (St. Louis, USA). Fatty acid, tocopherol and sterol standards were purchased from Supelco (Bellefonte, PA, USA), Nu-Check (Elysian, MN, USA) and Sigma Chem. Co.

The seed dimensions were measured by a digital caliper (CD-15CP, Mitutoyo Ltd., Andover, UK) on 100 randomly selected seeds. Similarly, to determine the 1000-seed weight, 10 randomly selected seeds were weighed (Sartorius ED224S, Sartorius, Germany), and multiplied by 100. Both measurements were repeated four times. The skin: flesh ratio was determined by removing the seed coats and weighing skin and flesh separately from 10 seeds, four times. A Minolta colorimeter (CR-400, Osaka, Japan) calibrated previously against white tile was used to read the seed color. Ohaus MB45 moisture analyzer (Ohaus, Switzerland) and AquaLab 4TE (Decagon Inc. USA) were used to measure the moisture content (%) and water activity (a_w) values of the orange seeds, respectively. Similarly, seed oil content using method 920.39 (AOAC, 2005); seed protein using the Kjeldahl method of Aa 5–38 (AOCS, 1997), and seed total ash using the Ba 5a-49 method (AOCS, 1997) were measured.

The control group (regular seed roasting) was prepared by heating seeds for 30 min at 150 ^oC in an oven (Inoksan FPE 110, Bursa, Turkey) with frequent agitation. The treatment group (microwave roasting) was achieved by heating seeds at 360 Watt for a total of 30 min, applying a 3-min pause mode of action every three minutes. Then, the seed moisture level was determined and adjusted to 11% by water conditioning (Aydeniz et al., 2014) before cold pressing. A laboratory scale cold-press machine (single head, 2 hp, 1.5 kW power, 12 kg seed/h capacity, Koçmaksan, ESM 3710, İzmir, Turkey) was used to cold press the orange seeds. The operation conditions were 30 rpm screw rotation speed, 10 mm exit die and 40 ^oC maximum oil exit temperature. The cold pressings were repeated twice for each treatment group. The cold pressed oils were first centrifuged (Sigma 2–16 K, Postfach, Germany) at 6797 xg for 10 min, then placed in amber-colored glasses and flushed with nitrogen before capping. Until and during the analyses, the oil samples were kept in the fridge. The press meal exiting the cold press machine was first ground (Retch Grindomix GM 300, Germany), and then placed in refrigerator bags and kept at −20 °C until analysis.

The orange seeds and press meal gained after cold pressing can be observed in Figure. 1. The proximate composition (moisture, ash, remaining oil, and protein contents) and color were measured with the same methods described above for the seeds. The press meal was then deep-frozen for further studies.

The specific gravity of the oil was measured by the AOCS Cc 10c-95 method (AOCS, 1997) at 25 ^oC. The Abbe 5 (Bellingham and Stanley, UK) refractometer was used (25 ^oC) to read the refractive indices of the oil samples. Apparent oil viscosities were measured at 25 ºC with the Brookfield Viscometer (model DV II+Pro with Rheocalc software, Brookfield Eng. Lab., Inc., MA, USA) and a no:18 spindle rotating at 30 rpm. The oil turbidities were measured at 25^oC with a Hach 2100 AN Turbidimeter (USA). Instrumental colors were read with a Minolta Colormeter CR-400 (Minolta Camera Co., Osaka, Japan). The oil sediment contents were assessed gravimetrically (Dündar Emir et al., 2015). The free fatty acidity and acid values of the oil samples were measured by Ca 5a-40 and Cd 3d-63 methods (AOCS, 1997). Likewise, peroxide value, p-anisidine value and iodine number were measured according to the methods Cd 8-53, Cd 18–90, and Cd 1–25 (AOCS, 1997), respectively. The saponification numbers were determined according to Tl 1a-64 (AOCS, 1997) and unsaponifiable matter following ISO 3596 (ISO, 2000) methods. Total phenolic contents were measured according to the Folin-Ciocalteu reagent technique, and the antioxidant capacities of the phenolic extracts were measured by the Trolox equivalent antioxidant capacity (TEAC) technique according to Aydeniz and Yılmaz (2012).

The fatty acid analysis of the orange seed oils was completed by preparation of the methyl esters (Ce 2–66 method, AOCS, 1997) and quantification by a Gas Chromatograph (Agilent Technologies 7890B, Palo Alto, CA, USA) equipped with a flame ionization detector (FID) (Agilent Technologies, Palo Alto, CA, USA), and HP 88 capillary column (100 m × 0.25 mm ID ×0.2 μm film thickness, J&W Scientific Co, CA, USA). Conditions of the analysis were; oven temperature: 120 °C for 1 min, 175 °C (10 °C/min) for 10 min, 210 °C (5 °C/min) for 5 min and 230 °C (5 °C/min) for 5 min; 1 μL injection volume, 1:50 injector split ratio, 2 ml/min flow rate. The carrier gas was hydrogen and the temperatures of the injector and detector were 250 and 280 °C, respectively. Fatty acid quantification was achieved with co-chromatography of the FAME standard mixture (37-components, C4-C24, Supelco, Bellefonte, PA, USA).

The sterol composition of oil samples was determined by the ISO 12228 method (ISO, 1999). The sterol fraction of the oils was first separated on TLC after saponification. Then the sterols were analyzed with a Gas Chromatograph-FID (Agilent Technologies 7890B, Palo Alto, CA, USA) and DB5 capillary column (30 m × 0.25 mm ID × 0.1 μm film thickness, J &W Scientific Co, CA, USA). The oven temperature and thermal programming were 60 °C for 2 min, 60–220 °C (40 °C/min) for 1 min, 220–310 °C (5 °C/min) for 30 min. The injection volume was 1 μL, with a 1:100 injector split ratio, 0.8 ml/min flow rate, 30 ml/min hydrogen carrier gas flow, and 290 and 300 °C injector and detector temperatures, respectively. The relative retention times of available standards (cholesterol, brassicasterol, stigmasterol, β-sitosterol) were used for sample identification and quantification.

The amount of α-tocopherol was determined with reverse-phase HPLC (Shimadzu Corporation, Kyoto, Japan) equipped with a LC-20AT HPLC pump, DGU-20A5R degasser, CTQ-10ASVP column oven, and RF-20A diode array detector. The method of Grilo Câmara et al. (2014) was followed with minor modifications. Each of the 0.15 g oil samples was dissolved in 3 ml dichloromethane, and from this stock around 20 μl of sample were injected into the Inertsil ODS-3 column (250 mm x 4.6 mm x 5 μm, GL Sciences Inc., Japan) by an autosampler (SIL-20AHT). The methanol: water (98:2 v/v) mobile phase was eluted with a 1.6 ml/min isocratic flow rate. The detector excitation and emission wavelengths were 290 nm and 330 nm. The α-tocopherol standard was used for quantification (Merck, Darmstadt, Germany).

The melting and crystallization temperatures, and enthalpies of the orange seed oil samples were assessed by a Perkin-Elmer 4000 Series Differential Scanning Calorimeter (Groningen, The Netherlands). The previously calibrated (with Indium and Zinc) instrument was purged with nitrogen at a 50 ml/min flow for half an hour to clean the sample holder. Then around 5–10 mg oil sample were weighed and sealed hermetically in aluminum pans, and put in the instrument. The temperature program applied was heating from room temperature to 110 ^oC by 10 ^oC/min then cooling to –70 ^oC by 10 ^oC/min and holding at that temperature for 3 min for full crystallization, and finally heating to 50 ºC by 5 ºC/min. The thermograms were used to calculate the parameters with the Pyris 1 Manager software (Yılmaz et al., 2015b).

The oxidative induction time (OIT) was also measured with the DSC. Samples were prepared as previously indicated and exposed to heating from 30 ^oC to 130 ^oC at 20 ^oC/min under constant 50 ml/min nitrogen flushing. Then an isothermal temperature programming at 130 °C with 50 ml/min purified oxygen (99.8%) was applied to the samples until curve departure for OIT calculation. The Pyris 1 Manager software was used to calculate the OIT values (Yılmaz et al., 2015b).

Oil production from the orange seeds by the cold pressing technique were repeated twice, and for each oil sample produced, the analyses were conducted at least in duplicate or in triplicate. Comparisons of the treatments were completed by one-way ANOVA and Tukey’s tests by Minitab ver. 16.1.1 (Minitab, 2010) and SPSS package (SPSS, 1994) statistics programs. There was a 95% minimum confidence level for all statistical analyses.

The main physicochemical properties of the orange seeds used in this study (Figure. 1) are presented in Table 1. These are the raw material properties before the seeds were cold pressed. Seed dimensions, 1000-seed weight, skin: flesh ratio and seed color could depend on the seed variety as well as agricultural practices. These data are just to define the material used in this study. There is no similar data in the literature to compare. The seed moisture content and water activity at the beginning were determined, but before cold pressing, the moisture content of the seeds was arranged to 11% by water conditioning, as explained. The orange seeds used in this study included around 40.75% oil, 19.22% protein and 1.58% ash. In an early study (Habib et al., 1986), orange seeds were shown to contain 17.40% protein, 45.46% lipid, 34.13% carbohydrate and 2.95% ash. Similarly, El-Adawy et al. (1999a) reported the protein, crude lipid and ash contents of raw orange seed as 17.01, 42.59, and 3.12%, respectively. Our results concur with theirs. In another study (Anwar et al., 2008), four different citrus species, mitha (Citrus limetta), grapefruit (Citrus paradisi), mussami (Citrus sinensis), and kinnow (Citrus reticulata) were investigated. They indicated 27.00–36.54% oil, 3.90–9.51% protein and 4.60–5.60% ash in the samples. Differences with our samples could be due to seed variety. In a more recent study (Saloua et al., 2009), 33.89% protein, 32.75% oil, 20.76% carbohydrate and 6.72% ash were reported for orange seeds. The proximate composition data in this study is in accordance with the literature.

After cold pressing of the orange seeds, the press meal (cake) was collected and analyzed (Table 2). There were some differences between the two treatment groups (regularly roasted and cold-pressed control sample and microwave roasted and cold-pressed sample) for water activity, oil and protein content, and color a* values. Obviously, the oil remaining in the meal was significantly higher in the control sample, while for the protein the contrary result was seen. Clearly, microwave roasted seeds infiltrated their oil with higher yields during pressing. A higher percentage of protein was observed in the group treated with microwave, an expected result because decreasing of the percentage of one of the components (oil) will generate an increase in the other components. Higher a* values in the microwave roasted meal means a more reddish color, as can be observed from Figure. 1. Clearly the color values of the seeds (Table 1) and meal (Table 2) are fairly different, since after pressing the wholesomeness of the material is lost and some color reactions may occur. Orange seed meal data show that this material can be valorized for different purposes, since it includes substantial amounts of protein and remaining oil.

Common physicochemical properties, total phenolic contents and antioxidant capacity values of the two cold pressed orange seed oils are presented in Table 3. The oil yield values calculated over the total oil content of the seeds indicate that seed microwave treatment enhanced cold-press oil yield significantly (62.99% versus 52.93% in the control sample). Oil yield is an important parameter in terms of economic gain in cold pressing, but oil properties would also be very important in terms of oil quality. There were no differences in specific gravity, refractive index or viscosity values for the two samples, while oil turbidities were significantly different. Seed microwave treatment caused the oil to be more turbid. While the color parameters, L and b* values were not different, the a* value was higher in the microwave treated sample. This means that the microwave treated sample is a little more reddish. Around 6.65–7.57% sediment content was found in the oil sample with no difference between the treatments. The density of four species of citrus seed oils were reported as 0.920–0.941 (Anwar et al., 2008). El-Adawy et al. (1999a) reported a refractive index of 1.4684, a density of 0.914 and viscosity of 0.007 Pa.s for orange seed oil. Our findings concur with the literature.

For both oil samples, there were no significant differences for acid value, peroxide value, iodine or saponification numbers, or unsaponifiable matters (Table 3). The free fatty acidity of the control sample (0.27%) was significantly lower than that of the microwave treated sample (0.34%), and both samples had fairly low acid values (0.54 and 0.69 mg KOH/g oil). Both oils could be fairly acceptable, since the Turkish codex for named vegetable oils (Codex, 2012) permits up to 4.0 mg KOH/g oil acid value for cold-pressed and virgin oils. In one study (Habib et al., 1986), the acid values of 0.21, 0.65, 1.20, and 0.90 mg KOH/g oil were published for orange, mandarin, lime and grapefruit seed oils, respectively. Likewise, acid values ranging from 0.673 to 1.120 mg KOH/g oil were reported for four other citrus seed oils (El-Adawy et al., 1999a). Our data generally concur. The peroxide values (PV) and p-anisidine values (p-AV) of the oils indicate the oxidation status of samples. The codex (Codex, 2012) permits a maximum of 15 meq O₂/kg oil peroxide value for virgin oils. Hence, the microwave treated sample exceeds this limit by a little (15.49 meq O₂/kg oil), and might require further treatment. Clearly, seed microwave roasting enhances the PV compared to regular roasting. The PV of 6.37 meq O₂/kg oil (El-Adawy et al., 1999a), and PV of 2.33 meq O₂/kg oil and p-AV of 1.86 (Saloua et al., 2009) were reported for orange seed oils. Clearly the reported values are significantly lower than our samples, but it must also be pointed out that in our study we cold pressed the seed oils at laboratory pilot scale production rather than laboratory scale solvent extraction. It was stated that the oxidation status of an oil would depend on its chemical nature as well as its production, processing and storage conditions, and age (O’Brien, 2003). In our previous study (Aydeniz et al., 2014), the PVs of regularly roasted safflower seed oil (3.85 meq O₂/kg oil) was found significantly lower than that of the microwave treated counter sample (9.40 meq O₂/kg oil), and this finding agrees with the measured PVs in this study. Although seed pre-treatment with microwave could enhance oil yield, it also causes the PV to increase.

There were no significant differences in the orange seed oils for the iodine and saponification numbers, or unsaponifiable matter contents. The iodine and saponification numbers for the control and microwave treated oil samples ranged between 117.95–119.33 g I/100 g oil, and 204.08–201.06 mg KOH/g oil, respectively. El-Adawy et al. (1999a) reported 102.57 g I/100 g oil as the iodine number, and 190.2 mg KOH/g oil as the saponification number for orange seed oil. In another study, Anwar et al., (2008) reported iodine values ranging (g I/100 g oil) from 99.85–110.00; saponification number (mg KOH/g oil) from 180.90–198.85; unsaponifiable matter (%) from 0.31–0.50 for four different citrus species. Saloua et al. (2009) reported iodine and saponification numbers and unsaponifiable matter with the same units as 141.43, 174.57, and 1.57 for orange seed oil, respectively. Obviously, our iodine and saponification numbers were a little higher than those reported in the literature.

While the total phenolic contents (4071 and 5607 μg GA/100 g oil) of the two oil samples were not statistically different, their antioxidant capacity values (12431 and 16512 μmol Trolox/100 g oil) were statistically significantly different. Moreover, it is clear that the antioxidant capacity/phenolic content (ac/pc) ratios of both seed oils (3.053 and 2.944) were not very different. The total phenolic content (as μg GAE/100 g oil) of orange seed oil was reported as 115288 (Malacrida et al., 2012). This value is significantly higher than we measured in our samples. Furthermore, Malacrida et al. (2012) reported 47.6% quenching ability for the DPPH radical of orange seed oil, and indicated that the highest antioxidant capacity was in orange seed oil compared with lemon and tangerine seed oils. Unfortunately, there is no other data available in the literature to compare with our findings. We reported (Aydeniz et al., 2014) 2855.70 and 4079.30 μg GA/100 g oil for total phenolics, and 315.32 and 538.26 μmol Trolox/g oil for the antioxidant capacity of regularly roasted and microwave roasted safflower seed oil samples. Clearly, cold-pressed orange seed oils are very high phenolic-containing and high antioxidant capacity-exhibiting samples. Hence, this oil could be added to other oils or foods as antioxidant and/or phenolic containing supplements.

The fatty acid, sterol and α-tocopherol contents of the orange seed oil samples are summarized in Table 4. There was no difference between the control and microwave roasted samples for the fatty acid composition. Six fatty acids were quantified, and the most abundant ones were linoleic acid (around 36%), palmitic acid (around 26%) and oleic acid (around 25%). The amounts of linolenic and palmitoleic acids were around 4% and 0.5%, respectively. There are similar but differently ranged numbers in the literature for different orange seed varieties. Saidani et al. (2004) reported 34.40% palmitic, 3.40% stearic, 14.90% oleic, 40.30% linoleic and 4.0% linolenic acids, as the major fatty acids for sweet orange seeds. Clearly the amounts of oleic and linolenic acid are fairly different from ours. Matthaus and Özcan (2012) reported the fatty acid composition of orange seeds originating from Turkey as 26.6% palmitic, 4.9% stearic, 22.4% oleic, 1.3% vaccenic, 39.9% linoleic, 3.7% linolenic and 0.4% arachidic acids. Likewise, Malacrida et al. (2012) gave a very similar fatty acid composition data for orange seed oils. Hence, our data mostly concur with the data of Matthaus and Özcan (2012) and Malacrida et al. (2012).

The sterol composition (%) of the orange seed oil samples is also presented in Table 4. There were fifteen different sterols with the highest ratios of β-sitosterol (around 78%), campesterol (around 9–10%), stigmasterol (around 3–3.5%) and Δ-5-avenasterol (around 1.5–5%). Furthermore, cholesterol, brassicasterol, 24-methylene cholesterol, campestanol, Δ-7-campesterol, Δ-5,23-stigmastadienol, chlerosterol, sitostanol, Δ-5,24-stigmastadienol, Δ-7-stigmastenol and Δ-7-avenasterol were quantified at less than 1.0% concentrations. Most sterols were not different between the the two samples, while some small differences existed for only several small quantity sterols between the two samples (Table 4). Clearly, seed pre-treatment had no extensive effect on sterol composition other than some small but statistically significant differences. In one study (Saloua et al., 2009), the sterol composition of orange seed oil was reported as around 81% β-sitosterol, 7.4% campesterol, 4.2% stigmasterol, 4.1% lupeol and 3.2% Δ-5-avenasterol. These values turn out to be very similar to ours, except for lupeol, which we have not quantified. Matthaus and Özcan (2012) reported the major sterol composition of orange seed oil as 77.9% β-sitosterol, 8.0% campesterol, 4.3% stigmasterol and 4.1% Δ-5-avenesterol and others. This data is also concurring with our findings. Hence it could be said that seed pre-treatments had no effect on sterol composition, and the sterol composition data are in good agreement with existing literature.

We have quantified only the α-tocopherol contents of the samples (Table 4), and found no significant difference (283.40 and 256.65 mg/kg oil). Saloua et al. (2009) reported 18.92, 10.80, 6.02 and 6.29 mg/100 g oil of α-, γ-, β-, and δ-tocopherols, respectively. Although we have only measured the α-tocopherol content, the α-tocopherol contents of our samples are higher than the reported amount. Another study (Matthaus and Özcan, 2012) showed the tocopherol composition of orange seed oil as 9.3 and 0.5 mg/100 g of α-, γ-tocopherol, and 0.2 mg/100 g of γ-tocotrienol. Again, the α-tocopherol contents measured in our samples (283.40 and 256.65 mg/kg) are higher than the reported 93 mg/kg level.

The crystallization and melting onset temperature, peak temperatures and enthalpies of the two orange seed oil samples are shown in Table 5. There is no significant difference between the two samples. Both samples started to crystallize at around 1.53–1.90 ^oC, and the crystallization peak was reached at around 4.63–5.12 ^oC, respectively. Since the samples cooled down to −70 ^oC for full crystallization, and then heated gradually, it was possible to observe the melting onset and peak temperatures. The oils started to melt at around −26.16 - (−25.79) ^oC, and four different melting peak temperatures were observed at around −23.78 - (−23.04) ^oC, −2.36 - (−2.35) ^oC, −0.47 - (−0.63) ^oC, and 5.97–6.42 ^oC, respectively. This means that there were four main types of triglyceride forms with distinctly different melting ranges. While crystallization enthalpy was around 30.56–31.20 J/g, the melting enthalpy was around 56.81–59.41 J/g. As far as we searched, there is no data in the literature about the thermal properties of orange seed oils; hence, this data would be very useful. Clearly this oil is mostly similar to other unsaturated oils, and stays liquid at refrigerator temperature. This condition could be important in terms of oil storage and transport procedures and preparations of food and other product formulations.

The oxidation induction time (OIT) was also measured (Table 5). The OIT of the microwave roasted sample was higher than the control sample. Accordingly, it could be assumed that there were more antioxidant compounds present in the microwave roasted sample. In fact, it could be observed from Table 3 that the total phenolic content and measured antioxidant capacity value of the microwave roasted sample is higher than that of the control sample. Hence, it might be possible that phenolics and other antioxidant compounds might better leach into the microwave roasted oil. This could be accepted as another benefit of microwave treatment of the seeds prior to cold pressing. There is no data in the literature for orange seed oil to compare, and this data can also be an important contribution.