Experiment sets and the corresponding oil yield obtained are given in Table 1. The ANOVA analysis of the quadratic regression models for oil yield showed that the model was significant (p < 0.01) with an F-value of 89.63 (Table 2). The p-value of the “lack of fit” of the model was 0.6932, which implied that the “lack of fit” was not significant relative to the error [19]. The coefficient of determination (R²) and adjusted coefficient of determination (Adj. R²) were 0.9878 and 0.9767, respectively, which indicated that the accuracy of the polynomial model was adequate. The second-order polynomial model was expressed by the following quadratic equation:

where Y represents the oil yield, X_1, X₂ and X₃ are the levels of temperature, time and pressure, respectively. It was observed that the factor with the largest effect on oil yield was linear extraction time (p < 0.01), followed by extraction pressure (p < 0.01), and the interaction between time and pressure (p < 0.01). The interaction between temperature and time also had a significant effect on oil yield (p < 0.05), whereas, the linear extraction temperature and the interaction of temperature and pressure showed no statistically significant (p > 0.05) impact on oil yield. All of the second-order terms of extraction parameters had negative effects on oil yield.

Response surfaces can be illustrated on three-dimensional plots by presenting the response as a function of two factors and keeping the other constant. The influence of extraction time and temperature on oil yield at a fixed pressure of 30 MPa is shown in (Figure 1a). Yield decreased slightly with increasing extraction temperature when extraction time was relatively shorter (around 50 min), but increased with longer extraction time. The effect of extraction pressure and temperature on oil yield at a fixed extraction time of 70 min is presented in (Figure 1b). It can be found that temperature showed a slight effect on oil yield regardless of extraction pressure, the influence of pressure on oil yield at a fixed temperature was observed to be similar with that of time (Figure 1a). The influence of pressure and time at a temperature of 40 °C is shown in (Figure 1c). A linear increase in yield with the increase of pressure at low level of time, while an obvious quadratic effect on yield of oil was observed when extraction time was longer (around 90 min). The effect of time on oil yield at fixed pressures was similar to that of pressure.

It is well-known that temperature has a dual effect on SC-CO₂ extraction. Generally, higher temperature could accelerate mass transfer and improve the extraction yield [20]. However, increasing temperature would reduce the solvent density and decrease the yield at pressures in the critical range [21]. Consequently, solubility of the solute at a constant pressure was likely to depend on whether solvent density or solute vapor pressure was higher [22]. In our study, extraction temperature showed a minor effect on yield, an almost constant oil yield was observed with increasing extraction temperature (Figure 1). A similar phenomenon was also observed in SC-CO₂ extraction of chia seed oil [13]. That may be explained by the fact that a dynamic balance was maintained between solvent density and vapor pressure under the experimental conditions. Extraction time and pressure were found to be main parameters that influenced yield of tea seed oil by SC-CO₂ extraction in our study. Increasing pressure caused an increase in supercritical CO₂ density, resulting in an enhanced solubility of solutes. However, high pressure is not preferred due to increased repulsing solute-solvent interactions resulting from highly compressed CO₂ at high pressure, which potentially induced complex extraction and difficult analysis. Despite four parameters, including miscibility and threshold, the pressure at which the solute reaches its maximum solubility, the fractionation pressure range, and a knowledge of the physical properties of the solute, were suggested to be critical to understand the solute behavior in supercritical media; the underlying mechanisms of various performance of SC-CO₂ extraction under different conditions have not been clearly illustrated yet.

Based on the above model, the optimal values for temperature, time and pressure were predicted to be 45 °C, 89.7 min and 32 MPa, respectively, to give a maximum yield of 29.5%. Validation experiments carried out under these optimal conditions found the yield to be 29.2 ± 0.6%, which was not significantly different from the predicted value (p > 0.05). These results indicated that the quadratic model was reliable.

Tea seed (Jiukeng variety) was collected from the Panban tea garden (Zhejiang, China) in December 2010. Seeds were de-hulled and washed with water, then air-dried at ambient temperature. The dried seeds were milled into a powder by a pulverizer (DFY-500, Linda Machinery Co., Ltd., Zhejiang, China) and passed through a 20 mesh sieve. The resulting flour was oven-dried at 80 °C for 12 h, and then stored in a refrigerator at 4 °C until needed. Moisture content in the seed powder was 4.9% after oven-drying.

Carbon dioxide (99.99%) was obtained from Zhejiang Gas Co., Ltd. (Zhejiang, China). The Supelco fatty acid methyl ester (FAME) mix and 1,1-diphenyl-2-picrylhydrazyl (DPPH) were purchased from Sigma Aldrich (Munich, Germany). Methanol and hexane, of gas chromatography grade, were purchased from Tianjin Shield Co. Ltd. (Tianjin, China). All other chemicals and solvents were of analytical grade.

The SC-CO₂ extraction was carried out using a supercritical fluid extractor (Spe-ed^TM SFE-2, Applied Separations Inc., Allentown, PA, USA). For each experiment, a 10 g sample of tea seed powder was loaded into a 50 mL thick-walled stainless steel cylindrical extractor vessel filled with defatted cotton. Liquefied CO₂ was pumped into the extraction vessel at 2 L/min to a given pressure. The temperature inside the vessel was raised to desired ones and maintained by a heating jacket encasing the vessel. When the relevant pressure and temperature were reached, the extraction was started. Extracts were separated from the CO₂ phase and collected. The oil samples were then weighed and yield obtained as follows:

Conventional organic solvent extraction (SE) was carried out using petroleum ether (30–60 °C) to evaluate the extraction capacity and quality of oil using SC-CO₂. A 10 g sample of milled tea seed was weighed and placed in a Soxhlet apparatus using 300 mL petroleum ether, and then continuously extracted for 8 h. After extraction, the solvent was evaporated at 60 °C under a nitrogen stream. Experiments were conducted in triplicate.

Experimental design, data analysis and quadratic model building were carried out using Design-Expert software (Trial Version 7.1.6, Stat-Ease Inc., Minneapolis, MN, USA). This software determined the optimum levels of the three independent variables of temperature (X₁), time (X₂) and pressure (X₃) for maximum seed oil yield (Y). Each parameter was set to one of three levels; for temperature, the levels were 35, 40 and 45 °C; durations of extraction were 50, 70 and 90 min, and pressure levels were 25, 30 and 35 MPa (Table 4). A total of twenty experiments were performed. The experimental design was based on a central composite design (CCD) that consisted of eight (2³) factorial points. The six replicates for the center point were used to estimate the experimental error. All experiments were carried out in triplicate, and in a random order to minimize the effects of extraneous factors. A quadratic polynomial regression model was used to predict responses as follows:

where Y represents the response variable, β₀ is a constant and β_i, β_ii, and β_ij are the coefficients for linearity, square and interaction, respectively. X_i and X_j are the levels of the independent variables. The model was built based on confidence levels of 95%.

FAMEs of extracted oils were prepared using a fatty acid methylation method [28]. The FAMEs were identified by gas chromatography (GC) using an Agilent 6890N GC-FID (Agilent Technologies, Folsom, CA, USA) equipped with a fused silica capillary column (30 mm × 0.25 mm i.d). Thickness of the polyethylglycol-coated film was 0.32 μm. A solution of 1 μL of hexane containing the methylated sample was injected with a split ratio of 100:1. The inlet temperature was 220 °C and the detector temperature was 260 °C. The initial oven temperature was 170 °C and was then increased to 240 °C at a rate of 5 °C/min, and held at 240 °C for 5 min. Nitrogen was used as the carrier gas at a flow rate of 1.4 mL/min. The FAMEs were identified and quantified by comparison with standard FAME mix standards.

The antioxidant activity of tea seed oils extracted using both methods was determined using a DPPH assay, based on the method proposed by Brand-Williams et al. [29]. A 2 mL sample of the extracted oil at various concentrations (10–160 mg/mL) in DMSO was added to 2 mL of 0.005% (w/v) ethanolic DPPH solution. The decrease in absorbance of DPPH at 517 nm was measured by a UV/Vis spectrophotometer (HP8453, Hewlett Packard, USA) after incubation for 30 min at 30 °C in the dark. The radical-scavenging ability of the tested samples was calculated according to the following formula:

where A_cont and A_sample were defined as absorbance of the control and extracted oils, respectively.

The results were statistically evaluated by analysis of variance (ANOVA) followed by a Tukey’s range test. Statistical tests were carried out using the SAS system for Windows (Version 9.1, SAS Institute Inc., Cary, NC, USA), and p < 0.05 was regarded as statistically significant.