Linoleic acid (95%) was purchased from Anqing biological engineering Co., Ltd. (Anhui, China). Lipozyme 435 (formerly named Novozym^® 435 before 2010, declared activity 10000 PLU/g) was obtained from Novozymes A/S (Bagsvaerd, Denmark). Sodium hydroxide and glycerol were purchased from Fuyu chemical Co. Ltd. (Tianjin, China). Acetone and methanol were brought from MREDA technology Inc. (USA). All chemicals were of pure, analytical grade, and all solvents used were of HPLC grade.

Glycerol (150.0 g) and sodium hydroxide (5.25 g, 3.5 wt% based on the glycerol) were added to a 250 mL three-neck flask equipped with a condenser. The reaction was performed under nitrogen to prevent the dehydration of glycerol to acrolein. The mixture was reacted for 3.5 h at 250 °C and 400 rpm. After reaction, it was cooled down to room temperature, and the crude oligoglycerols were purified using molecular distillation (MD). The glycerol was removed and the heavy phase was collected under the following conditions: evaporator temperature 120 °C, condenser temperature 50 °C, evaporator vacuum pressure 5.0 Pa, and roller speed 300 rpm. The high polyglycerol (n≥4) removed from the heavy phase of MD has been collected, and the light phase containing oligoglycerols was obtained using the following conditions: evaporator temperature 170 °C, condenser temperature 60 °C, evaporator vacuum pressure 1.0 Pa, and roller speed 300 rpm.

The composition of oligoglycerols was analyzed by gas chromatography (GC) according to a modified method described by Xiao et al. (2012). The GC analysis was carried out on an Agilent GC 7820A instrument (Agilent Technologies Inc., America), equipped with a flame ionization detector (FID) and a capillary column of AT-Wax (15 m×0.32 mm, 0.25 µm). The injector temperature was 280 °C and was used in the split mode with a split ratio of 20:1. FID was kept at 280 °C and the flow rate of the carrier gas (N₂) was kept at 30 mL·min⁻¹. In each measurement, 1 µL of sample was injected. The oven temperature was programmed from 120 °C to 260 °C as follows: rising from 120 °C to 165 °C at a rate of 60 °C·min⁻¹ and held for 4 min, to 230 °C at a rate of 60 °C·min⁻¹ and held for 2 min, to 260 °C at a rate of 10 °C·min⁻¹ and held for 3 min. The percentage content of each oligoglycerol was calculated assuming the same response factor and the results are shown in Table 1.

Enzymatic esterification was carried out in a solvent-free system. Lipozyme 435, oligoglycerols and linoleic acid were added to a 25 mL airtight flask and reacted at 300 rpm under vacuum conditions (4 kPa) to remove water and avoid the oxidation of the reagents. The effects of different molar ratio of oligoglycerol to linoleic acid (2:1∼1:2), enzyme dosage (0.5∼3.0 wt% of the total substrate mass), reaction temperature (60∼120 °C), reaction time (2∼6 h), and water addition (0∼12 wt% of the total substrate mass) on the EE were studied.

At the end of each reaction, 8 mL of reaction mixture were taken and centrifuged at 3000 rpm for 10 min to be separated into three layers. The immobilized enzyme was present in the middle level. It can be isolated by filtration and used repeatedly. The upper oil phase was titrated with a 0.1 M NaOH standard solution using phenolphthalein as the endpoint indicator to obtain the acid value. The EE of the reaction was calculated according to the following formula: EE ( % ) = A V 0 - A V A V 0 × 100

Where AV ₀ is the acid value of linoleic acid, and AV is the acid value of the upper oil phase. All determinations were carried out in duplicate and the average values were reported.

To achieve maximal EE, a three-level three-factor Box-Behnken design was adopted to optimize the reaction conditions. The factors studied were enzyme dosage (E_d, wt%), the molar ratio of substrate (S_r, mol/mol) and the reaction time (T_i, h). The response studied was the efficiency of esterification (EE,%). The ranges of settings for factors were based on the results of single factor experiments and shown in Table 2. The initial reaction temperature was fixed at 90 °C for all experiments. All of the experiments were performed in triplicate and the average values were reported. Design Expert software (version 7.0, Stat-ease, Inc.) was used for the analysis of variance (ANOVA) and regression analysis of the experimental data.

The scaling-up synthesis of OGEs was performed in a 2 L pilot-scale BCR (Figure 1) manufactured by Handway Technology Co., Ltd. (Foshan, China) and the Department of Food Science and Engineering, Jinan University (Guangzhou, China). Oligoglycerols (300.00 g), linoleic acid (267.85 g) and Lipozyme 435 (11.36 g) were added to the reactor vessel and reacted under the optimum conditions (as obtained from section 2.5). N₂ gas was bubbled through the sieve at the bottom of the reactor where it entered into the reaction system and formed a boiling-like mixture. With this gas bubbling, the heterogeneous reaction system was vigorously agitated. Therefore the immiscible reactants and immobilized enzymes were efficiently mixed by the N₂ bubbling to enhance the reaction rate. The water formed by the reaction can be efficiently removed by the N₂ flow (2.5 L·min⁻¹) and collected by the condenser. The removal of water was beneficial to shift the thermodynamic equilibrium to the esterification. The N₂ gas pumped through the reactor was recycled to save on its consumption. After the reaction, the liquid reaction mixture was collected through the sieve. The immobilized enzyme was retained in the reactor, washed by anhydrous ethanol, and used for the next batch reaction under the same conditions. The product collected was centrifuged and separated into two layers. The upper layer was the OGEs phase which was analyzed by HPLC and ESI-MS.

A distinct advantage of the BCR is the lower mechanical shear force which allows for maintaining a considerably longer half-life of enzyme. In this study, 10 consecutive batches of OGEs production were conducted under the optimized conditions to assess the reusability of Lipozyme 435. The relative activity of enzyme was defined as follows: Relative activity ( % ) = EE of each batch EE of the 1st batch × 100

High performance liquid chromatography (HPLC) and ESI-MS were employed to analyze the composition of the esterification products. The HPLC system was composed of a Shimadzu HPLC system (LC-20AD pump, SIL-20A auto injector) equipped with a 150 mm ×4.6 mm Diamonsil C18(2) column (DIKMA, Beijing, China) and a Sedex 55 evaporative light scattering detector (ELSD) (SEDERE, France). The ELSD was set to 70 °C with a carrier air flow rate of 1.7 L·min⁻¹. The column temperature was set to 40 °C. A mixture of acetonitrile/isopropyl alcohol (55/45, v/v) was delivered as the mobile phase at a flow rate of 1.0 mL·min⁻¹. The samples were dissolved in isopropyl alcohol with a concentration of 5.0 mg·mL⁻¹. About 1 mL of this sample solution was filtered through a nylon membrane filter (0.22 µm), transferred to a 1.5 mL sample vial, and analyzed by HPLC. The injection volume of each analysis was 0.5 µL. Each sample was analyzed twice and the average value was reported. The content of each ester compound was estimated by calculating the peak areas obtained from the ELSD chromatogram assuming the same response factors.

The MS spectrum was recorded with an AB 4000 QTRAP mass spectrometer (USA) equipped with an ESI source. The samples were dissolved in methanol and were delivered to the ESI source with a syringe pump at a flow rate of 10 µL·min⁻¹. The operating parameters were as follows: ions spray voltage 5500 V, capillary voltage 40 V, nebulizing gas pressure 30 psi, and capillary temperature 250 °C. Mass spectra within a mass range of 100∼2000 m/z were acquired in the positive ion mode. The compositions of OGEs were identified by measuring their m/z values.

The effect of reaction temperature on the EE was studied at the range of 60∼120 °C for 5 h with an oligoglycerol/linoleic acid molar ratio at 1.5:1 and an enzyme dosage of 3 wt%. As seen from the Figure 2a, The EE soared from 76.6% at 60 °C to 95.0% at 90 °C and plateaued between 90 °C and 110 °C, suggesting that the reaction temperature had a dramatic influence on the esterification reaction. This may be explained by the effect of temperature on the enzyme activity and the affinity of the enzyme for substrate (Eom et al., 2010). The appropriately elevated reaction temperature was very effective in improving the mass transfer by increasing the activity of the enzyme and the diffusion of the substrate. However, the EE declined rapidly above 110 °C, indicating that thermal deactivation of the enzyme had taken place. This deactivation was attributed to the destroying of the spatial conformation and structure of the enzyme at high temperatures.

According to the literature (Duan et al., 2013, Pan et al., 2013), the optimal reaction temperature of Lipozyme 435 is between 60∼70 °C, and above 70 °C the stability of the enzyme is reduced as a result of changing the secondary and tertiary structure of the protein. Few data are found in the literature regarding the reaction temperature at an excess of 80 °C in enzyme catalyzed reactions. Nevertheless, here the reaction was performed above 90 °C, and high EEs (>95%) were achieved. The result indicated that Lipozyme 435 is a thermostable enzyme and can react at above 90 °C. This can be explained by the fact that the immobilization of the lipase makes the enzyme conformation stable and hence enhances its thermal stability. On the other hand, due to the high viscosity of oligoglycerol, it was necessary to raise the reaction temperature in order to reduce the viscosity of the reaction system and to increase the solubility of oligoglycerol in the fatty-acid phase. However, with the increase in temperature, the possibility of water-stripping from the enzyme surface was increased. So the optimal reaction temperature was selected at 90 °C for subsequent experiments.

To determine the optimal reaction time, esterification reactions with different reaction times (2∼6 h) were conducted at 90 °C and an enzyme dosage of 3 wt% with an oligoglycerol/linoleic acid molar ratio of 1.5:1. As shown in Figure 2b, the EE increased from 91.90% at 2 h to 95.51% at 3 h, and it did not improve significantly afterwards, indicating that the equilibrium state had been reached.

Compared to the data reported by Pan et al. (2013), although here the reaction was carried out at a lower enzyme dosage (3 wt% vs. 6.1 wt%), the equilibrium was reached in a shorter time period (3 h vs. 24 h) and the EE obtained was even higher (95.51% vs. 92.4%). Several reasons could explain this: (a) the esterification reaction here was carried out in a solvent-free system. This increased substrate concentrations and avoided the enzyme deactivation caused by organic solvents (e.g. methanol) which were beneficial for achieving a fast reaction rate (Shimada et al., 2002). (b) A higher reaction temperature (90 °C) was adopted in this study. This decreased the viscosity of the reaction mixture, improved the diffusion, and hence enhanced the accessibility of the reagents to the active-site of the enzyme. (c) It has been reported that Lipozyme 435 is more selective to fatty acids with a certain chain length and degree of unsaturation (Zou and Akoh, 2013), thus it might be possible that the enzyme could show a higher activity to linoleic acid. (d) The use of vacuum conditions to avoid linoleic acid oxidation also removed the water formed by the reaction, which drove the equilibrium to form the ester products.

To investigate the effect of the enzyme dosage on the EE, experiments with various enzyme dosage (0.5∼3 wt%) were performed for 3 h at 90 °C with an oligoglycerol/linoleic acid molar ratio of 1.5:1 (Figure 2c). It can be seen that the EE soared from 67.29% to 92.69% when the enzyme dosage was increased from 0.5 wt% to 1 wt%, and it plateaued at 95.85% with the enzyme dosage of 2 wt%.

Higher enzyme dosages enhanced the effective contact area between enzyme and substrate molecules and provided abundant active sites as a result of increasing the probability of substrate enzyme collision, which were conducive to the reaction occurring (Zhong et al., 2013). However, the contacting area between the enzyme and substrates reached saturation when an excessive amount of enzyme was added. It led to the incomplete exposure of the lipase to the substrates at the active site. So a further increase in the enzyme dosage had little effect on the EE. Thus, 2 wt% was chosen as the optimal enzyme dosage in this esterification reaction.

The esterification of oligoglycerol with different molar ratios of linoleic acid (2:1∼1:2) has been carried out for 3 h at 90 °C at the enzyme dosage of 2 wt%. It can be seen from Fig 2d that the maximum EE (95.87%) was obtained at a molar ratio of oligoglycerol/linoleic acid of 1.5:1.

In previous studies, the effect of glycerol/fatty acid molar ratio on acylglycerol synthesis has been reported. According to Byun et al. (2007), a large excess of glycerol could improve the yield of monoacylglycerol (MAG) (68 wt%) by the esterification of fish oil fatty acids with glycerol, because the esterification equilibrium was displaced toward product formation. Brady et al. (1988) pointed out that increasing the concentration of glycerol can greatly improve enzyme stability. However, in this study, the EE was overall decreased with the increasing amount of oligoglycerol. This is possibly due to the higher viscosity of oligoglycerol compared with that of glycerol. High viscosity of the reaction mixture prevented the diffusion of substrates and hindered the removal of water formed by the esterification. As will be discussed in the next section, high amounts of water had a negative effect on EE. On the other hand, high amounts of linoleic acid also decreased EE. This might be explained by the acidification of the micro-aqueous interface of the enzyme which would inhibit the enzyme activity (Chowdary et al., 2000). Previous investigations have reported that free fatty acids had a strong inhibition effect on enzyme activity when the substrate acid concentration rose above a critical level (Duan et al., 2013). Therefore, in this research the optimal oligoglycerol/linoleic acid molar ratio was 1.5:1.

To explore the effect of water content (0–12 wt% of the substrate) on the EE, experiments were carried out under the optimum conditions established from the above discussion (90 °C, 3 h, oligoglycerol/fatty acid molar ratio 1.5:1, enzyme dosage 2 wt%). As displayed in Figure 2e, the EE had a tendency to decline with increased water content.

Theoretically, the appropriate amount of water is essential to maintaining the enzyme activity (Wang et al., 2009). However, as Lipozyme 435 is a type of immobilized lipase that already has a certain content of water, and it can maintain activity without the addition of extra water (Zhong et al., 2013). On the contrary, the addition of extra water reduced the enzyme concentration and affected the hydration of the enzyme directly, or indirectly, by changing the nature of lipase support materials (Freitas et al., 2007), which resulted in a decrease in enzyme activity. On the other hand, esterification is the reverse reaction of hydrolysis. Excessive water accumulated during the progression of the reaction is favorable for hydrolysis rather than esterification. Therefore the removal of water shifted the equilibrium toward the formation of desired products. So in this study, the vacuum condition was applied to remove water and ensure a high EE.

Based on the above individual studies, RSM using the Box-Behnken design (Table 2) was applied to obtain the optimal reaction conditions. The EE obtained from all the experiments are given in Table 3. Model fitting and ANOVA results (Table 4) show that the quadratic polynomial model was suitable (F-value =84.9 at p<0.001, R²=0.9909) for representing the actual relationship of reaction parameters. The insignificant lack of fit test (F-value =0.26) with p>0.05 also indicated that the model represented the experimental data well. The coefficient of determination (R²) and adjusted-R² were 0.9909 and 0.9793, respectively, suggesting that the polynomial model used was accurate. Multiple regression coefficients obtained by the method of least squares to predict a second-order polynomial model for the EE are presented in Table 4. The corresponding p values of coefficients suggested that T_i (reaction time), S_r (substrate molar ratio), E_d (the amount of enzyme), T_i×E_d (substrate molar ratio×the amount of enzyme), S_r² (square of substrate molar ratio), E_d ² (square of the enzyme dosage) were significant variables with p<0.05. The remaining terms had no significant effects on EE. Furthermore, from Table 4 it can be concluded that EE was influenced in the following order of significance: E_d>S_r>T_i. The second-order polynomial model for the EE (Y,%) is shown in the following equation:

Y=95.17+0.31A+1.25B+1.33C-0.075AB-0.16AC -0.62BC-0.13A ²-0.168B²-0.44C ²

Where Y is the EE; A, B, and C are the independent variables. Here A is T_i, B is S_r, and C is E_d.

The optimal conditions for the synthesis of OGEs by Lipozyme 435 were predicted by Design-Expert software. When esterification was catalyzed by Lipozyme 435 under the optimal conditions (90 °C, 4.52 h, enzyme dosage 2 wt%, oligoglycerol/linoleic acid =1.59 in moles), the predicted EE based on the model was 96.15%. The small deviation between the optimal conditions and those local optimal ranges from contour plots (Figure 3) is possibly due to the interaction effects of various experimental factors. To further verify the reliability of the regression model, three independent experiments were performed at the predicted optimum conditions, and the actual EE was 95.82±0.22%, proving the high accuracy of the RSM model used.

According to Mollenhauer et al. (2010), for a heterogeneous reaction system with high viscosity, a BCR is more suitable than a conventional stirred tank or a fixed bed reactor. In this study, a BCR as shown in Figure 1 was employed to scale-up the synthesis of OGEs, and the bubbling of N₂ gas was used as a means of agitation for the reaction mixture. The optimal reaction conditions obtained by RSM were applied to BCR. As expected, the EE achieved was higher than that of a conventional magnetic stirred reactor (97.23% vs. 95.82%). In the operational stability test, Lipozyme 435 still maintained 98.2% of the relative activity after 10 consecutive batch reactions (Figure 4). So in a high viscosity reaction system, agitation by N₂ bubbling can give higher reaction efficiency than by magnetic stirring. That may be because the bubbling of N₂ ensured a convective flow field and created an effective interaction in the multiple-phase reaction system to provide a high mass-transferring rate. The excellent operational stability of Lipozyme 435 also benefitted from the gentle agitation and the protective atmosphere provided by N₂ bubbling. In comparison, magnetic stirring often mechanically destroys the structure of the enzyme carrier and hence reduces the activity (Mollenhauer et al., 2010). Therefore, production in BCR is a preferred way to achieve a higher yield of OGEs.

The OGE product obtained under optimal conditions was analyzed by HPLC. As reported by Cassel et al. (2001), the retention time in the reversed-phase HPLC mode increased as more fatty acid chains were attached to polyglycerol. The four groups of peaks (A∼D in Figure 5) represent the mono-, di-, tri-, and tetraesters of oligoglycerols according to the reversed-phase HPLC theory and the polarity of products (Andersen et al., 2003). By integrating the peak areas in the ELSD chromatogram, it was found that the OGE product contained 56.92% monoesters, 34.94% diesters, 7.74% triesters, and 0.39% tetraesters of oligoglycerols.

To obtain more details about the OGE product, an ESI-MS spectrum recorded in the positive ion mode was used to analyze the product obtained under optimal conditions (Figure 6). The spectrum contains a wide range of peaks corresponding to the characteristic ions of OGEs. Over 10 OGE species were observed, including glycerol moieties ranging from n =2∼4. Based on the ion intensities, the OGE species with n =2 and 3 were the most abundant, which was in agreement with the result of the oligoglycerol composition analyzed by GC. Due to the small vapor pressure difference between PG3 and PG4, after distillation there was still a little PG4 (4.41 wt%) remaining in the final products. Thus, OGE species with n =4 were also found with low ion intensity. Besides the OGE species as sodium and ammonium adducts, their water adducts and dehydrated forms were also detected in the spectrum (Figure 6). According to Orfanakis et al. (2013), the dehydrated species were not formed as the result of the ESI process. It is worth mentioning that the multiple water adducts of PG4-tetraester ([M+Na+4H₂O]⁺, m/z =1476.6) were also observed. It indicated that the intermolecular hydrogen bonding was formed between the water molecules and hydroxyl groups due to the high density of the hydroxyl groups on the PG4-tetraester. The water molecules may come from the esterification or from the immobilized enzyme (as stripped in the carrier material).

Linoleic acid appears to react completely as its signal cannot be observed from the spectra. The mono-, di- and triester of oligoglycerol ions all showed high intensities in the spectrum. The intensity of monoesters was greater than the other ions of OGEs. According to the literature (Byun et al., 2007, Liu et al., 2007), the formation of triacylglycerol (TAG) was much less than that of MAG and diacylglycerol (DAG) when sn-1,3-specific lipases were used for the esterification of glycerol with fatty acids. The small amount of TAG might result from the acyl migration to the 2-position in 1,3-DAG, which requires higher amounts of energy to overcome the steric hindrance for further esterification (Liu et al., 2007). This isomerization reaction was considered to be the confined step of TAG synthesis. However, here the amount of triesters was significant. This may be due to the fact that Lipozyme 435 is a nonspecific lipase, thus the acyl donor can be esterified at any available hydroxyl groups in the oligoglycerol and shows higher activities in direct esterification. On the other hand, PG4-tetraester ([PG4+4A+Na]⁺, m/z =1404.7, A =Linoleic acid) and PG4-pentaester ions ([PG4+5A+H]⁺, m/z =1662.1) were detected with low intensities, while none of the PG2- and PG3-tetraester ions was observed in the spectra. It seems that the esterification occurred more easily between linoleic acid and PG4 than PG2 or PG3. This is possibly because PG4 has more hydroxyl groups so that the reaction with PG4 has a less prominent steric hindrance issue. However, it was clear that esterification was incomplete as none of the fully esterified OGE ions was detected.