Optimizing functional properties of perilla protein isolate using the response surface methodology

With increasing consumer concerns over protein shortage, there has been a considerable interest on unconventional proteins isolated from plant sources. Perilla (Perilla frutescens L), an annual edible herb, is used for cooking and medicinal purposes in several Asian countries, such as China, Japan, and Korea (Asif, 2011). Perilla seed consists of 51.7% fat and 17.0% protein (Longvah & Deosthale, 1991), and defatted perilla seed meal consists of ~40% protein (Longvah & Deosthale, 1998). On spite of being an important protein source, defatted perilla seed meal is mainly used as animal feed and fertilizer. The use of perilla seed meal as a value-added food ingredient is worth exploring. Perilla protein, especially perilla protein isolate (PPO), has several high functional properties such as water solubility index, water absorption capacity, oil holding capacity, emulsifying activity, and foaming ability, which could expand its applications in the food industry.


Introduction
With increasing consumer concerns over protein shortage, there has been a considerable interest on unconventional proteins isolated from plant sources.Perilla (Perilla frutescens L), an annual edible herb, is used for cooking and medicinal purposes in several Asian countries, such as China, Japan, and Korea (Asif, 2011).Perilla seed consists of 51.7% fat and 17.0% protein (Longvah & Deosthale, 1991), and defatted perilla seed meal consists of ~40% protein (Longvah & Deosthale, 1998).On spite of being an important protein source, defatted perilla seed meal is mainly used as animal feed and fertilizer.The use of perilla seed meal as a value-added food ingredient is worth exploring.Perilla protein, especially perilla protein isolate (PPO), has several high functional properties such as water solubility index, water absorption capacity, oil holding capacity, emulsifying activity, and foaming ability, which could expand its applications in the food industry.
Chemical and enzymatic modifications have been used to improve the functional properties of plant proteins.However, these modification methods, which may contribute to toxic by-products, tend to be expensive in large-scale industrial applications.Ultrasonic extraction, which has gained increasing popularity, can be used in conjunction with, or in place of, traditional extraction methods because of its reduced extraction times and cost-effectiveness (Luque-García & Castro, 2003).
Ultrasound-assisted extraction contributes to high extraction yield and is easily adapted to large-scale industrial applications (Vinatoru, 2001).Dur previous findings revealed that the most influential parameters affecting protein content and protein yield were ultrasonic power, extraction time, and liquid-to-solid ratio.The optimization of these parameters could be achieved with response surface methodology (RSM).
RSM is an effective statistical technique for optimizing experimental protocols.RSM evaluates the effects of multiple factors and their interactions on one or more response variables (Azmir et al., 2014;Karnopp et al., 2015;Xu et al., 2013).An advantage of RSM is that it reduces the number of experiments and provides a mathematical model.On this study, RSM was applied to study the effects of ultrasonic power, extraction time, and liquid-to-solid ratio on PPO yield and functional properties including its water solubility index, water absorption capacity, oil holding capacity, emulsifying activity, and foaming ability.Additionally, the thermal properties and polypeptide profile of PPO were analyzed.

Sample preparation and chemicals
Fresh whole perilla seeds were collected in Dctober 2015 from North University of China (Taiyuan, Shanxi Province, China).The seeds were dried in an air convection oven at 50 °C for 72 h, dehulled, and ground into flour.The flour was defatted in a Soxhlet extraction apparatus using n-hexane as the organic solvent.Defatted flour was spread on a plate for 4 h at room temperature under a laboratory fume hood to remove any traces of the solvent.Subsequently, the flour was passed through a 60-mesh size screen, packaged in polyethylene bags, and stored at 5 °C until use.
Commercial soy protein isolate (SPO), supplied by Sino Peptide Health Technology Co, Ltd (Beijing, China), was used for comparisons.All chemicals used in this study were of analytical grade and were purchased from Beijing Chemicals Co. (Beijing, China).

Experimental design
A Central Composite Design (CCD) consisting of the independent variables x 1 (ultrasonic power), x 2 (extraction time), and x 3 (liquid-to-solid ratio) at five different levels was used to study the response pattern and determine the optimum combination of variables (Table 1).The variables and levels were selected on the basis of the previously single factor experiments performed by Zhu & Fu (2012).CCD (Table 2) was arranged to allow the design of an appropriate regression model.Six replicates (treatments 4, 5, 7, 10, 12, and 20) at the center of the design were used to estimate the error sum of squares.Experiments were randomized to minimize the effects of unexplained variability in the observed responses due to extraneous factors.
The variables were coded according to formula 1, where X i is the coded value of an independent variable, x i is the real value of an independent variable, ‾ x i is the real value of an independent variable at the central point, and Δx i is the step change.

Preparation of PPI
The experiments were carried out in a random order of that presented for CCD (Table 2).Defatted perilla seed flour (10 g) was suspended in distilled water at each specific liquid-to-solid ratio.The pH of the mixture was adjusted to 10 with 0.1 mol/L NaDH using a PHS-25 Model pH-meter (Shanghai Yueping Onstrument Co., Ltd, Shanghai, China) at room temperature.Periodically, the pH was measured again and, when necessary, re-adjusted to 10.The mixture was subjected to a tunable ultrasonic bath (TH-400BQG, 220V and 50Hz Tianhua Ultrasonic Electronic Equipment Co., Beijing, China) at 30°C, at different power levels and for different extraction time periods.Following extraction, the mixture was centrifuged at 2,000 g for 20 min, and the resulting supernatant was acidified to pH 4.5 with 2 mol/L HCl.Following a 30-min incubation at room temperature, precipitated proteins were collected by centrifugation under the same conditions as previously described and freeze dried in a Scientz-N model freeze-dryer (Ningbo Scientz Biotechnology Co., Ltd, Ningbo, China).

Determination of dependent variables
Protein yield PPO yield (%) was calculated using formula 2, where p is the weight (g) of the extracted and freeze-dried protein powder, and s is the weight (g) of defatted perilla seed flour after extraction.

Water solubility index and water absorption capacity
Water solubility index (WSO) and water absorption capacity (WAC) were determined by the method reported by Dbilana et al. (2014) with some modifications.Freeze-dried samples (0.2 g, M 0 ) were suspended in 20 mL of 0.1 mol/L phosphate buffer (pH 7) and mixed for 30 min at room temperature.The pH of the suspensions was measured and adjusted when necessary.The suspensions were centrifuged at 2,000 g for 30 min.The resulting sediment was weighted (M 2 ) and dried at 100°C for 24 h, and the dried weight (M 1 ) was determined.WSO and WAC were calculated using formulas 3 and 4, respectively,

Oil holding capacity
Dil holding capacity (DHC) was determined by the method reported by Mundi & Aluko (2012) with some modifications.Protein sample (3 g) was dispersed in 25 mL pure canola oil in a 50-mL pre-weighed centrifuge tube, vortexed for 1 min, allowed to stand for 30 min, and centrifuged at 2,000 g for 20 min at room temperature.The separated oil was removed with a pipette, and the tubes were inverted for 25 min to drain the oil prior to reweighing.DHC was expressed as grams of oil bound per gram of sample (dry basis).

Emulsion activity
Emulsion activity (EA) was determined by the method reported by Zhao et al. (2011) with minor modifications.For the formation of an emulsion, 9 mL of 0.1% (w/v) protein solution (pH 7) was mixed with 3 mL pure canola oil for 60 s in a high-speed homogenizer (XHF-DY, Ningbo Scientz Biotechnology Co., Ltd. Ningbo, China) at 1,500 g.Subsequently, 50 µL of emulsion was immediately transferred to a test tube containing 5 mL of 0.1% SDS (w/v).Absorbance was measured in a UV-spectrophotometer (UV-8000S, Shanghai Metash Onstrument Co., Ltd., Shanghai, China) at 500 nm.EA was calculated using formula 5, 2 2 2.303 ( / ) 10000 where C is the initial concentration of protein (g/mL), φ is the optical path (1 cm), θ is the oil volume fraction used to form the emulsion (0.25), D is the dilution factor (100), and A is the absorbance of the diluted emulsions.Measurements were performed in triplicate.

Foaming ability
Foaming ability (FA) was determined according to the methods reported by Shahidi et al. (1995) and Tsutsui (1988) with some modifications.On this experiment, 20 mL of 3% (w/v) aqueous sample dispersions were homogenized at 2,000 g for 1 min, and the resulting foam was immediately transferred into a graduated cylinder.The volume was recorded before and after foam formation.FA was expressed as the increase in volume (%) due to whipping and was calculated using formula 6, where V 1 is the initial volume of solution, and V 2 is the volume of solution after whipping.

Statistical analysis
A multiple regression equation was used to fit the second-order polynomial equation shown in formula 7, where Y i (i = 1-6) is the predicted response for protein yield, WSO, WAC, DHC, EA, and FA of protein concentrate; β 0 is the value of the fitted response at the central point of the design; β 1 , β 2 , and β 3 are linear terms; β 12 , β 13 , and β 23 represent interaction effects; and β 11 , β 22 , and β 33 are squared effects.All experiments were performed in triplicate and in random order.The experimental data are presented in Table 2. Statistical analysis was performed using the software Stat-Ease Design-Expert 8.0.6.1 Trial computer program (Stat-Ease Onc., Minneapolis, MN, USA).The analysis of variance (ANDVA) and Tukey tests were used for the checking of the statistical significance with confidence level of 95.0%.The ANDVA produced parameters; lack of fit, coefficient of determination (R 2 ) and F-test were employed to evaluate the model adequacy (Hayta & İşçimen, 2017).The model was fitted by multiple linear regressions and response surface plots were obtained for six responses.

Thermal properties
The thermal properties of PPO obtained under optimum conditions and of perilla protein extracted by conventional methods (Li et al., 2013) were assessed by differential scanning calorimetry (DSC) using a TA Q100-DSC thermal analyzer (TA Onstruments, New Castle, DE, USA).Approximately 2-3 mg protein was accurately weighed into aluminum pans.The pans were hermetically sealed and heated from 20 °C to 120 °C at 5 °C/min.A sealed empty pan was used as reference.Dnset (T m ) and peak (T d ) denaturation temperatures and enthalpy of denaturation (ΔH) were computed from the thermograms by the Universal Analysis 2000, version 4.1D software (TA Onstruments-Waters LLC, USA).

Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE)
The polypeptide profiles of PPO and SPO were analyzed by SDS-PAGE, using 12.5% separating gel and 5% stacking gel (Brunelle & Green, 2014).The protein samples were dissolved in 0.5 mol/L Tris-HCl buffer (pH 6.8), containing 10% SDS, 2.5% 2-mercaptoethanol, and 1% bromophenol blue, resulting in a final protein concentration of 1 mg/mL.The mixture was heated in boiling water for 3 min and centrifuged at 5000 g for 5 min.For each sample, 15 μL of supernatant was loaded into each well.
Electrophoresis was performed at a constant current of 15 mA in the stacking gel and 25 mA in the separating gel until the tracking dye reached the bottom of the gel.Following electrophoresis, the gel was stained for 3 h with 0.25% Coomassie blue (R-250) in acetic acid/methanol/water solution (46:227:227, v/v/v) and subsequently destained for 6 h with methanol/acetic acid / water solution (2:3:35, v/v/v).

Fitting the mathematical model
Multiple regression equations were generated between the predicted responses (Y 1 , Y 2 , Y 3 , Y 4 , Y 5 , and Y 6 ) and the three variables, i.e., ultrasonic power, extraction time, and liquid-to-solid ratio (Table 2).The regression coefficients (Table 3) were determined by the least squares technique, and the appropriate subset of variables for the regression model was determined by the backward elimination technique (Kleijnen, 2008).
The significance of the coefficients in the models was determined by analysis of variance (Table 3).The results revealed that protein yield (Y 1 ) and FA (Y 6 ) were significant at p < 0.01, and WSO (Y 2 ), WAC (Y 3 ), DHC (Y 4 ), and EA (Y 5 ) were significant at p < 0.001.The coefficient of determination (R 2 ) of the model was high (> 0.823).Therefore, the experimental values matched the model-predicted values.According to Man et al. (2010), a model is adequate when R 2 > 0.75.Lack-of-fit measures the failure of a model to represent the data in the experimental domain at points that are not included in the regression.As shown in Table 3, the lack-of-fit for all fitted models was non-significant (p > 0.05).Therefore, the quadratic model adequately represented the data for all responses.

Protein yield
PPO yield was affected by ultrasonic power and liquid-tosolid ratio.The linear effect of ultrasonic power and liquidto-solid ratio was positive and significant at p < 0.05, and the interaction between these two variables was positive and significant at p < 0.05.Dn the other hand, extraction time had a negligible effect on PPO yield (Table 3).Figure 1A shows the effect of ultrasonic power and liquid-to-solid ratio on protein yield at a constant extraction time (30 min).Protein yield had a non-significant increase with increasing ultrasonic power and liquid-to-solid ratio at low ultrasonic power and liquid-to-solid ratios.However, at high ultrasonic power and liquid-to-solid ratios, protein yield increased with increasing ultrasonic power and liquid-to-solid ratio.These results could be attributed to increased protein decomposition at high ultrasonic power and increased cavitation effects at high liquid-to-solid ratios.On this study, PPO yield was higher than that reported by Zhu & Fu (2012) in perilla seed meal proteins (11.77%).

Water solubility index
Protein WSO is the most important functional property because it affects other functional properties.On several protein-based formulations such as emulsions and foams, optimal protein solubility is required.Table 3 shows that ultrasonic power and extraction time had a negative effect on WSO (p < 0.01), whereas liquid-to-solid ratio had a weak effect on WSO.The interaction between ultrasonic power and extraction time had a non-significant effect on WSO. Figure 1B shows the effect of ultrasonic power and extraction time on WSO at a constant liquid-to-solid ratio (7:1).WSO decreased with increasing ultrasonic power and extraction time, probably due to protein denaturation, exposure of buried hydrophobic groups, and protein aggregation.On this study, PPO solubility was similar to that reported in peanut protein concentrate (>40%) by Wu et al. (2009) and higher than that obtained in cowpea protein concentrate (<20%) by Mune Mune et al. (2014).

Water absorption capacity
Protein WAC was affected by ultrasonic power (p < 0.01) and extraction time (p < 0.05).Liquid-to-solid ratio had a weak effect on WAC.The interaction between ultrasonic power and extraction time had a non-significant effect on WAC (Table 3).Figure 1C shows the effect of ultrasonic power and extraction time on WAC at a constant liquid-to-solid ratio (7:1).WAC increased with ultrasonic power and extraction time up to 3.56 g/g and decreased thereafter (Figure 1C).The highest WAC level was achieved at an ultrasonic power coded level of 0.24 (212 W uncoded) and an extraction time coded level of 0.17 (32.55 min uncoded).On this study, WAC of PPO was higher than that obtained in peanut protein concentrates and isolates (1.2-2.7 g/g) (Wu et al., 2009).Kinsella (1979) reported that the major factors affecting WAC include protein denaturation/unfolding and the presence of carbohydrates and non-protein components.During extraction, PPO was likely denatured, exposing hydrophobic sites, which explained the reduced water retention.

Oil holding capacity
High DHC is desirable in the cold meat industry, particularly in sausage production (Dgunwolu et al., 2009).Ultrasonic power, extraction time, and the interaction between these two variables had significant effects on DHC (p < 0.05).Liquid-to-solid ratio had a weak effect on DHC (Table 3).Figure 1D shows the effect of ultrasonic power and extraction time on DHC at a constant liquid-to-solid ratio (7:1).At low ultrasonic power, DHC increased with increasing extraction time, while at high ultrasonic power, DHC decreased with increasing extraction time.PPO DHC was similar to that reported by Mune Mune et al. (2014) for cowpea protein concentrate (1.75 g/g), and higher than that reported by Boye et al. (2010) for pea protein isolate (~1.2 g/g).Ultrasound affected the physical entrapment of oil and the non-covalent bonds involved in lipid-protein interactions, thereby modifying DHC.

Emulsifying activity
Table 3 shows that ultrasonic power, extraction time, and the interaction between these two variables had a significant effect on EA (p < 0.05).Liquid-to-solid ratio had a weak effect on EA. Figure 1E shows the effect of ultrasonic power and extraction time on EA at a constant liquid-to-solid ratio (7:1).At low ultrasonic power and extraction time, EA increased with increasing ultrasonic power and time.At high ultrasonic power and extraction time, EA decreased with increasing ultrasonic power and time.Proteins are effective emulsifiers that are commonly used in food emulsions.The formation and stability of protein-based emulsions are related to the surface activity of proteins at interfaces (Lam & Nickerson, 2013).On this study, high EA values could be attributed to protein denaturation and exposure of hydrophobic amino acids, which facilitated protein interaction at the oil-water interface.

Foaming ability
FA was affected by ultrasonic power and extraction time (Table 3; p < 0.05).The interaction between ultrasonic power and extraction time had a significant effect on FA (p < 0.01), whereas liquid-to-solid ratio had a weak effect.Figure 1F shows the effect of ultrasonic power and extraction time on FA at a constant liquid-to-solid ratio (7:1).At low ultrasonic power and extraction time, FA increased with increasing ultrasonic power and extraction time, while at high ultrasonic power and extraction time, FA decreased with increasing ultrasonic power and extraction time.FA of PPO was similar to that reported by Mune Mune et al. (2014) in cowpea protein concentrate (> 60%).Therefore, PPO represents a good foaming agent and may be suitable in bakery goods and ice cream.Ultrasound enhances protein unfolding and the formation of interfacial protein membranes at the air-water interface, which enhances air bubble encapsulation.

Optimization
The regression model equation allowed the prediction of the effects of the three variables on the six responses, and RSM was performed to optimize the three variables.The optimum conditions and predicted responses are presented in Table 4.For convenience, the optimum conditions were slightly modified to an ultrasonic power of 200 W, an extraction time of 32 min, and a liquid-to-solid ratio of 10.The experimental values were in agreement with the predicted values.The extraction conditions obtained by RSM were accurate, reliable, and practical (Xu et al., 2013).

Thermal properties
The thermal properties of PPO were evaluated by DSC (Table 5).Li-Chan & Ma (2002) reported that DSC endothermic peaks are indicative of protein denaturation.PPO had a weak DSC endothermic peak (0.03 J/g) suggesting that several components  had thirteen bands (97, 80, 66.2, 65, 44, 41, 39, 38, 34, 22, 20, 16, and 15 kDa) with the 80 kDa in the highest proportion.The polypeptide molecular weight range of PPO was lower than that of SPO.Consequently, the solubility of PPO is predicted to be better than that of SPO.

Conclusion
RSM was successfully applied to model and optimize PPO reparation.Protein yield was significantly affected by ultrasonic power and liquid-to-solid ratio.WSO, WAC, DHC, EA, and FA were affected by ultrasonic power and extraction time.The interaction between ultrasonic power and liquid-to-solid ratio was significant for protein yield, and the interaction between ultrasonic power and extraction time was significant for DHC, EA, and FA.The optimum PPO extraction conditions consisted of an ultrasonic a power of 200 W, an extraction time of 32 min, and a liquid-to-solid ratio of 10.Under these conditions, the predicted protein yield and functional properties were in agreement with the experimental values.This study demonstrated that PPO had optimal functional properties and thermal stability.PPO subunits had molecular weights that ranged between 14.4 and 66.2 kDa.PPO could be applied in functional foods and as a potential food ingredient.
in PPO were denatured and unfolded, and that ultrasound had a negative effect on the structure.Denaturation temperature (T d ) is an indicator of the thermal stability of proteins (Amza et al., 2011).T d was 75.16 °C in PPO and 67.24 °C in perilla protein extracted using alkali and acidic solutions.Therefore, PPO was more thermo-stable than perilla protein.This result might be attributed to the ultrasound disruption of intermolecular bonds in PPO.Ghribi et al. (2015) reported that the structural transitions of protein originating from hydrolyzed peptide bonds during extraction may be responsible for the changes observed in DSC.

SDS-PAGE
Variations in polypeptide profiles of proteins have direct effects on their functional properties, especially on EA and FA (Mundi & Aluko, 2012).SDS-PAGE was performed to determine the molecular weight of PPO.SPO was run on the gels as a control.Pictures of SDS-PAGE gels are shown in Figure 2. PPO (lane 1) exhibited major bands at 14.4-66.2kDa.Marcone et al. (2002) reported that bands at 20-27 and 30-39 kDa represent basic and acidic protein subunits, respectively.Therefore, PPO was rich in both types of subunits.
As observed in the gel patterns, PPO and SPO fractions had relatively different polypeptide composition profiles.PPO had eight polypeptides (66.2, 45, 36, 34, 33, 30, 22, and 14.4 kDa) with the 14.4 kDa band in the highest proportion, while SPO

Figure 1 .
Figure 1.The effect of ultrasonic power, extraction time, and liquid-to-solid ratio on protein yield (A), water solubility index (B), water absorption capacity (C), oil holding capacity (D), emulsifying activity (E), and foaming ability (F) of perilla protein isolate.

Table 1 .
Experimental and coded levels of the independent variables for central composite design.

Table 3 .
Coefficients and analysis of variance for the fitted models.
a Significant at p < 0.05; b Significant at p < 0.01; c Significant at p < 0.001; d Non-significant.
All the data are expressed as mean ± SD and are the mean of three replicates within the same line denote that are significantly different (p < 0.05); T m : temperature onset; T d : denaturation temperature; ΔH: enthalpy of denaturation.

Table 4 .
Dptimum conditions, predicted, and experimental values of response at optimum conditions.
a Mean of three determinations.