Impact of ionic liquid properties on selective enrichment of glycerides in direct lipase-catalyzed esterification

Abstract

The direct lipase-catalyzed esterification of oleic acid and glycerol was studied in 23 different ionic liquids (ILs) in order to deduce the effects of solvent properties such as viscosity, polarity (E^T_N), hydrophobicity (log P) and Kamlet–Taft parameters (α, β and π*) on selective enrichment of glycerides. E^T_N and Kamlet–Taft parameters are most likely the determining factors for glyceride selectivity. ILs with high E^T_N and Kamlet–Taft values might form a hydrogen-bond with monoacylglycerols (MAG) and resulted in high MAG content. ILs with higher log P and lower β resulted in higher FFA conversion and ILs with longer alkyl chains lead to higher contents of DAG and TAG. Viscosity may reduce the reaction rate which could decrease the FFA conversion and change glyceride selectivity. Pearson's correlation analysis showed that the MAG content had a significantly positive correlation with E^T_N value (R = 0.4860, P < 0.05) and Kamlet–Taft parameters (R > 0). log P and hydrogen bond basicity (β) have relatively high positive correlation coefficient (R = 0.6065) and negative coefficient (R = −0.3878) with FFA conversion respectively. [NTF₂]⁻ based ILs were more suitable for the rules of polarity. In summary, selective enrichment of glycerides was not simply influenced by any single property, but by multiple properties collaboratively. However, higher polar ILs could always result in higher polar reaction products.

---

1 Introduction

Glycerides, as important food processing ingredients, can be divided into monoacylglycerols (MAG), diacylglycerols (DAG) and triacylglycerols (TAG) according to the number of fatty acids on glyceryl backbone. The lipase-catalyzed syntheses of glycerides have been extensively studied in conventional solvents^1–3 or solvent-free systems.^4–6 These approaches have exhibited comparable advantages to chemical methods because of their mild reaction conditions, high selectivity of biocatalysts and low energy consumption.^7,8 However, there has existed a great challenge in this area. Esterification, hydrolysis and glycerolysis are the three essential processes in industrial manufacture of glycerides. They are all multistep reactions and high acyl donor conversion and high selectivity are two desired goals. Therefore, the application of an efficient system to shift the reaction to generate higher yields of single product should be concerned. Traditional organic solvent engineering have been reported as methods to improve this concern.^9,10 Take esterification reaction for example, thermodynamic equilibrium of overall esterification reaction can be described by eqn (1)–(3) in Fig. 1A, the equations of the equilibrium constants can also be described by the eqn (4)–(6) in Fig. 1B. K is constant at a given temperature, while the thermodynamic activity of a component depends on its nature and specific reaction system. Change of the thermodynamic activity of substrate governs the equilibrium shift. In hydrophilic solvents, interaction between hydrophilic MAG and solvents is stronger than that between hydrophobic TAG or DAG and solvents, which reduces the thermodynamic activity and activity coefficient of MAG. Therefore, the equilibrium will be shifted to the formation of MAG. However, these efforts only achieved a limited success because of the volatility and toxicity of organic solvents and its pollution to environment.

Fig. 1 (A) Thermodynamic equilibrium of the esterification reaction. (B) The equations of the equilibrium constants. K₁, K₂ and K₃ are the equilibrium constants; a_i, χ_i and γ_i represent the thermodynamic activity, mole fraction and activity coefficient of component i, respectively. K_1χ, K_2χ and K_3χ, as well as K_1γ, K_2γ and K_3γ are the corresponding constants denoted by mole fractions and activity coefficients.

ILs are defined as organic salts which are liquid at ambient temperature. Increasing interests in ILs arise primarily from the progressive concerns for environmental problems. As molten salts, ILs have been extensively used as a potential alternative to highly volatile, toxic, hazardous and flammable organic solvent.^11–13 The unique properties of ILs, such as low vapor pressures, high chemical and thermal stability, multiple solvation interactions with organic and inorganic compounds, make these compounds popular as a medium for chemical and biochemical reactions.¹⁴ Also, ILs have other important advantages. They not only possess adjustable solubility, protective effects on enzymes for stability enhancement and to be recoverable and recyclable, but also have positive effects on the specificity of enzymes or on the shift of reaction equilibrium.^15–17 The most interesting nature of ILs for some scientists may be their tunable properties such as polarity, hydrogen-bond basicity, anion nucleophilicity, ion's kosmotropicity and viscosity by selectivity of different cations, anions and substituents. Thus, ILs are real engineered solvents which could provide opportunities for many biocatalytic processes to control reaction progress or to optimize reaction towards the desired products.^18–20 This has been demonstrated successfully in enzymatic glycerolysis reactions,^21–23 where the MAG selectivity or DAG selectivity was increased greatly with CPMA·MS or TOMA·Tf₂N/Ammoeng 102 as reaction medium and Novozym 435 as catalyst. Although these selected ILs are adoptable for lipase-catalyzed glycerolysis procedures, the effects of IL properties on the catalytic behaviors of enzymes are not clear. The influence of the solvent properties on enzymatic activity and stability has been studied in ionic liquids.^24–26 However, the relationship between selectivity of glycerides and solvent properties in this area has been rarely investigated. Thus, it is significant to understand clearly the relationship to select specific ILs for selective enrichment of MAG, DAG and TAG respectively.

In this work, the direct lipase-catalyzed esterification of oleic acid and glycerol was studied in 23 different ionic liquids (ILs) in order to deduce the effects of solvent properties such as viscosity, polarity (E^T_N), hydrophobicity (log P) and Kamlet–Taft parameters (α, β and π*) on selective enrichment of glycerides. Pearson's correlation analysis was used to study the correlation between IL properties and product selectivity or conversion. It was also expected that some specific ionic liquids could be selected in favor of enrichment of MAG, DAG or TAG, respectively.

2 Materials and methods

2.1 Enzyme and chemicals

Novozym 435 (C. antarctica lipase immobilized on macroporous polyacrylate resin beads), bead size: 0.3–0.9 mm; bulk density: 430 kg m⁻³; activity: 7000 PLU g⁻¹, PLU is the ester synthesis activity expressed in “propyl laurate units” was received as gift from Novozymes (Beijing, China). The water content of Novozym 435 was approximately 3% (w/w). [BMIM][BF₄], [HMIM][BF₄], [OMIM][BF₄], [BMIM][N(CN)₂], [HMIM][N(CN)₂], [BMIM][PF₆], [HMIM][PF₆], [OMIM][PF₆], [C₁₀MIM][PF₆], [BMIM][Tf₂N], [HMIM][Tf₂N], [OMIM][Tf₂N], [BMMIM][Tf₂N], [HMMIM][Tf₂N], [OMMIM][Tf₂N], [MeO(CH₂)₂MIM][PF₆], [HO(CH₂)₂MIM][Tf₂N], [MeO(CH₂)₂MIM][NTf₂], [C₁₄MIM][NTf₂], [C₁₂MIM][NTf₂], [Oct₃MeN][NTf₂], [B₃C₁₄PH₂₉][NTf₂] and [1-Oct-3-MePy][NTf₂] were purchased from Chengjie Chemical Co. Ltd., Shanghai, China. Table 1 shows the structures of the 23 ionic liquids studied in this work. 1,3-Diolein (99%), 1,2-diolein (97%), 1-oleoylglycerol (99%), 2-oleoylglycerol (95%), triolein (95%) and oleic acid (90%) were purchased from Sigma-Aldrich Chemical Co. Ltd. (Shanghai, China). Glycerol was purchased from Sinopharm Chemical Reagent Co. Ltd. (Shanghai, China). Hexane and isopropanol used for HPLC analysis were chromatographically pure. All other organic solvents were purchased commercially and were of analytical grade.

Table 1 Structure of the 23 ionic liquids studied in this article

Alkyl-methylimidazolium cation
Abbreviation	R1	R2	R3	Anion	Cation
[C14MIM]NTf2	CH3	C14H29
[C12MIM]NTf2	CH3	C12H25
[OMMIM]NTf2	CH3	C8H17	CH3
[BMMIM]NTf2	CH3	C4H9	CH3
[HMMIM]NTf2	CH3	C6H13	CH3
[OMIM]NTf2	CH3	C8H17
[BMIM]NTf2	CH3	C4H9
[HMIM]NTf2	CH3	C6H13
[MeO(CH2)2MIM]NTf2	CH3	CH2OC2H4
[HO(CH2)2MIM]NTf2	CH3	C2H4OH
[MeO(CH2)2MIM]PF6	CH3	CH2OC2H4
[C10MIM]PF6	CH3	C10H21
[OMIM]PF6	CH3	C8H17
[HMIM]PF6	CH3	C6H13
[BMIM]PF6	CH3	C4H9
[OMIM]BF4	CH3	C8H17
[HMIM]BF4	CH3	C6H13
[BMIM]BF4	CH3	C4H9
[BMIM]N(CN)2	CH3	C4H9
[HMIM]N(CN)2	CH3	C6H13

---

1-Alkylpyridinium cations
Abbreviation	R1	R2	Anion	Cation
[1-Oct-3-MePy]NTf2	C8H17	CH3

---

Alkylammonium cation and Alkylphosphonium cation
Abbreviation	R1	R2	R3	R4	Anion	Cation
[Oct3MeN]NTf2	C8H17	C8H17	C8H17	CH3
[B3C14PH29]NTf2	C4H9	C4H9	C4H9	C14H29

---

2.2 Enzymatic esterification in ionic liquids

The enzymatic esterification reaction was conducted in 25 ml beakers with their jackets connected to water bath. 5 mmol of oleic acid (1.412 g), 5 mmol glycerol (0.46 g), 30% (w/w, relative to total reactants) ionic liquids, 4% activated 4 Å molecular sieves (w/w, relative to total reactants) and 4% (w/w, relative to total reactants) Novozym 435 were added into reactors and mixed thoroughly by magnetic stirrer (450 rpm) at 60 °C. After 8 hours, 1 ml reaction mixtures were withdrawn from the reactor and mixed with 4 ml solvent mixture of diethyl ether and hexane (1 : 4) and be centrifuged at 10 000 rpm for 5 min. After centrifugation, the mixture was divided into two phases, the upper phase were collected for solvent evaporation. The sample was ready for HPLC analysis and free fatty acids determination. The concentration of glycerides was determined by HPLC. Content of free fatty acids was determined by NaOH titration according to the standard method.²⁷ The conversion rate of fatty acid was defined as the proportion of esterified fatty acid to initial amount of fatty acid amount.

2.3 HPLC analysis of the enzymatic reaction product

For the separation and quantification of esterification products including MAG, DAG and TAG, HPLC-ELSD (Evaporative Light Scattering Detector) using a Waters 1525 liquid chromatographic system (Waters Corp., Milford, MA, USA) equipped with a Li Chrospher Si column (250 mm and 4.6 mm, 5 lm particle size, Sigma-Aldrich Corp. K. K., Tokyo, Japan) was chosen and the products were eluted with a binary gradient of solvent A (99 : 1, hexane/isopropanol) and solvent B (1 : 1 : 0.01, isopropanol/hexane/acetic acid) at 1.0 ml min⁻¹. 12 μl of the reaction mixture were withdrawn and dissolved in 25 ml of solvent B. The mixture was then filtered through a microfilter (0.45 μm) to be analyzed by HPLC. The course of the gradient was as followed: solvent A was decreased from 100% to 80% over 10 min and decreased further to 70% from 10 to 14 min. At last, solvent A was held at 100% from 15 min to 20 min. The total run time was 20 min. Triolein (TAG), 1,3-diolein (1,3-DAG), 1,2-diolein (1,2-DAG), 1-oleoylglycerol (1-MAG) and 2-oleoylglycerol (2-MAG) were used as external standards. The concentration of MAG, DAG and TAG was respectively calculated from their calibration curve constructed from known concentrations (0.00714 to 0.2 mg ml⁻¹ for MAG, 0.0075 mg ml⁻¹ to 0.2 mg ml⁻¹ for DAG, 0.0076 to 0.2 mg ml⁻¹ for TAG) of the corresponding standards. Their relative contents were calculated based on grams. All the above experiments were conducted in triplicate and the data presented in this work were an average of the obtained values.

2.4 The determination of viscosity

Viscosity was determined by this study using the Cannon-Fenske Routine (CFR) viscometer at 25 °C.

2.5 Statistical analysis

All statistical data were reported as mean values of independent experiments ± standard deviations (SD). Pearson's correlation analysis was done using SPSS Version 20.0 in Windows 10.

3 Result and discussion

3.1 Esterification of glycerol with free fatty acids by Novozym 435 in ILs

3.1.1 Effect of anion type. The free fatty acids (FFA) conversion and glyceride relative contents of esterification in ILs containing 14 different cations and 7 different anions were shown in Table 2. Anion selections had a significant influence on the FFA conversion and values from 9.22% to 90.89% were obtained. Four kinds of ILs containing [PF₆]⁻, [Tf₂N]⁻, [BF₄]⁻ and [N(CN)₂]⁻ have been usually chosen as the solvents for some lipase-catalyzed reactions.^24,28 According to Table 2, higher FFA conversions were observed in [PF₆]⁻ and [Tf₂N]⁻ based ILs than in ILs [BF₄]⁻ and [N(CN)₂]⁻ based ILs. This could be due to the higher activities of Novozym 435 in [PF₆]⁻, [Tf₂N]⁻ based ILs than [BF₄]⁻ and [N(CN)₂]⁻ based ILs. ILs with anions containing [BF₄]⁻ and [N(CN)₂]⁻ have more nucleophilic properties than those containing[PF₆]⁻ and [Tf₂N]⁻. Thus, they could cause a loss of protein secondary structure and subsequent loss of activity.²⁹

Table 2 Glyceride relative content and FFA conversion in ILs at 60 °C by direct lipase-catalyzed esterification, and correlations with physical properties of solvents^a

Number	Ionic liquids	MAG (wt%)	DAG (wt%)	TAG (wt%)	Conversion (%)	E^TN	α	β	π	Viscosity (25 °C)/cP	log P
a All reactions procedure was as follows: 5 mmol of oleic acid and glycerol at 1 : 1 molar ratios were blended with 30% (w/w, relative to total reactants) ionic liquids using 4% (w/w, relative to total reactants) Novozym 435 as catalyst at 60 °C for 8 h; (a) data from ref. 35; (b) ref. 36; (c) ref. 30; (d) ref. 37; (e) ref. 38; (f) ref. 39; (g) ref. 40; (h) ref. 41; (i) ref. 42; (j) ref. 43; (k) ref. 44; (l) ref. 45; (m) ref. 46; (n) calculated from octanol–water partition coefficient (KOW) in ref. 34; (o) ref. 24.
1	[Oct3MeN]NTf2	13.7 ± 0.8	78.2 ± 4.8	8.1 ± 1.2	86.9 ± 1.4	0.469^a	0.33^a	0.23^a	0.87^a	491
2	[B3C14PH29]NTf2	24.5 ± 1.5	67.6 ± 0.9	7.9 ± 1.1	90.9 ± 0.9
3	[C14MIM]NTf2	31.5 ± 1.1	58.4 ± 2.8	10.2 ± 1.4	80.4 ± 1.37
4	[C12MIM]NTf2	28.7 ± 2.1	62.1 ± 1.6	9 ± 1.0	87.8 ± 1.4
5	[OMMIM]NTf2	30.1 ± 2.6	62.7 ± 1.9	7.2 ± 1.7	83.7 ± 2.1	0.525^b				150
6	[BMMIM]NTf2	32.1 ± 2.3	64.6 ± 2.5	3.4 ± 0.9	81.5 ± 1.5	0.552^b	0.38^c	0.24^c	1.01^c	93
7	[1-Oct-3-MePy]NTf2	29.6 ± 2.6	64.2 ± 1.3	6.1 ± 1.4	87.8 ± 1.2	0.565^d	0.46^d	0.28^d	0.97^d
8	[HMMIM]NTf2	33.6 ± 1.3	63.2 ± 1.6	3.2 ± 0.9	83.3 ± 1.1	0.574^d	0.45^d	0.26^d	0.99^d		0.13–0.25^n
9	[OMIM]NTf2	29.1 ± 1.3	64.5 ± 2.4	6.4 ± 0.8	89.2 ± 2.7	0.627^d	0.6^d	0.28^d	0.97^d	119	0.80–1.05^n
10	[BMIM]NTf2	33.7 ± 0.7	63.3 ± 1.6	3.0 ± 1.0	83.6 ± 1.9	0.645^e	0.62^c	0.25^e	0.9^a	55	−0.96 to −0.21^n or 0.11 ± 0.01°
11	[HMIM]NTf2	34.1 ± 2.0	65.4 ± 1.3	0.5 ± 0.2	83.1 ± 1.2	0.651^d	0.65^d	0.25^d	0.98^d	87	0.15–0.22^n
12	[MeO(CH2)2MIM]NTf2	35.5 ± 1.0	63.1 ± 0.9	1.4 ± 1.2	80.5 ± 2.3	0.723^f				54
13	[HO(CH2)2MIM]NTf2	50.5 ± 3.8	48.7 ± 4.3	0.9 ± 0.6	63.4 ± 3.3	0.929^g	1.17^g	0.34^g	1.03^g	91
14	[MeO(CH2)2MIM]PF6	27.4 ± 0.9	63.9 ± 1.3	8.7 ± 0.9	82.0 ± 1.4					283.6
15	C10MIM PF6	21.8 ± 1.2	72.8 ± 1.7	3.4 ± 0.8	82.3 ± 1.8					419
16	[OMIM]PF6	27.3 ± 2.0	68.8 ± 2.6	3.9 ± 1.1	73.9 ± 2.2	0.633^b	0.58^h	0.46^h	0.88^h	680
17	[HMIM]PF6	24.1 ± 2.6	70.0 ± 3.4	5.9 ± 1.9	82.2 ± 2.4	0.660^i	0.57^h	0.58^h	1.02^i	560
18	[BMIM]PF6	20.7 ± 1.8	66.9 ± 3.0	12.5 ± 2.1	87.6 ± 3.1	0.669^c	0.637^c	0.19^j	1.04^j	380	−1.66^n
19	[OMIM]BF4	56.6 ± 1.9	34.4 ± 2.9	9.0 ± 3.0	9.2 ± 1.6	0.650^i	0.62^i	0.41^i	0.98^i	440	−1.34 ± 0.09°
20	[HMIM]BF4	38.3 ± 2.1	57.0 ± 1.6	4.7 ± 0.9	24.9 ± 3.1					310
21	[BMIM]BF4	28.9 ± 0.8	69.1 ± 1.1	2.0 ± 0.5	36.3 ± 1.6	0.670^k	0.61^k	0.39^i	1.04^k	98	−2.51 ± 0.04°
22	[BMIM]N(CN)2	57.3 ± 2.4	35.0 ± 3.8	7.7 ± 2.7	58.7 ± 4.0	0.639^l	0.54^j	0.71^m	1.05^j	32	−2.32 ± 0.02°
23	[HMIM]N(CN)2	37.9 ± 1.7	56.4 ± 3.5	5.7 ± 1.2	56.0 ± 2.3	0.630^l	0.51^l		1.05^l

---

As for glyceride relative contents, product profiles in these ILs based on different anions were exactly opposite. Higher MAG content was produced in [BF₄]⁻ and [N(CN)₂]⁻ based ILs, especially in [BMIM][N(CN)₂] (57.31% MAG content). While [PF₆]⁻ and [Tf₂N]⁻ based ILs produced more DAG content. It indicated that [BF₄]⁻ and [N(CN)₂]⁻ based ILs had a better MAG selectivity, [PF₆]⁻ and [Tf₂N]⁻ had a better DAG selectivity.

3.1.2 Effect of cation type. [PF₆]⁻ and [Tf₂N]⁻ based ILs resulted in higher FFA conversion. Especially most of [Tf₂N]⁻ based ILs are liquid at room temperature compared with [PF₆]⁻ based ILs. Therefore, 13 kinds of cations of [Tf₂N]⁻ based ILs were chosen to studied the effects of cation on lipase-catalyzed esterification. Most of them resulted in very high FFA conversion (65–90%). The ILs with long alkyl chains resulted in higher FFA conversion, such as [Oct₃MeN]NTf₂, [C₁₂MIM]NTf₂, [B₃PC₁₄H₂₉]NTf₂, which also produced high DAG content. Additionally, relative high MAG content was produced in [MeO(CH₂)₂MIM]NTf₂ (35.45% MAG content) and [HO(CH₂)₂MIM]NTf₂ (50.47% MAG content). Surely all these results could depend on some properties of ILs. Thus, the impact of each IL property on the selectivity of glycerides and FFA conversion was investigated and the results were shown in the following paragraphs.

3.2 Effect of polarity

E^T_N, as one of the popular solvatochromic polarity scales can be used to represent ionic liquids' polarity.³⁰ Product selectivity and FFA conversion were plotted against E^T_N of ILs in Fig. 2. With the increase of E^T_N value, MAG content increased while DAG and TAG contents and FFA conversion had a decreasing trend in most of ionic liquids. This rule was even more obvious in [NTF₂]⁻ based ILs. As already mentioned, MAG content increased with the decreasing alkyl chain length of cations. ILs with long alkyl chains resulted in higher DAG and TAG contents with 80–90% FFA conversion. Data from Table 2 indicated ILs with shorter alkyl chain length had higher E^T_N value. This result was consistent with the conclusion in organic solvents.¹⁰ It could be essentially attributed to the solvent polarity effect on the thermodynamic equilibria of the three consecutive reactions (Fig. 1A). However, many exceptions have been observed. For [BF₄]⁻ and [N(CN)₂]⁻ based ILs, they had relatively lower E^T_N value but resulted in more MAG contents. For [PF₆]⁻ based ILs, MAG content increased while DAG and TAG contents and FFA conversion had a decreasing trend with the decrease of E^T_N value. Thus, it is important to be aware that the E^T_N value is not a universal scale for solvent effect due to the complexity of enzyme–solvent and products–solvent interactions. It probably suggested that many other solvent properties can influence the esterification reaction.

Fig. 2 Glyceride relative content and FFA conversion vs. E^T_N parameter of different ionic liquids: (A) MAG; (B) DAG; (C) TAG; (D) FFA conversion (plot from data in Table 2, and the solvent numbers in the graph are consistent with those in Table 2).

3.3 Effect of Kamlet–Taft values

As mentioned in Table 2, Kamlet–Taft parameters might also influence enzymatic reaction results, which include α, β and π*. Three Kamlet–Taft parameters quantify hydrogen bond donating ability (acidity), hydrogen bond accepting ability (basicity) and dipolarity/polarizability, respectively.³¹ The results from Table 2 suggested that ILs displayed similar π* values, which indicated that the π* value didn't have the correlation with product selectivity and conversion. The effect of α value on product selectivity and FFA conversion suggested the similar trend compared with polarity. The α value is largely determined by the structure of the cation. [HO(CH₂)₂MIM]NTf₂ with the higher α value shows the higher MAG selectivity, which is an O–H hydrogen bond donor because of the hydroxyl group. Thus, the hydroxyl group in the [HO(CH₂)₂MIM]NTf₂ molecule could form hydrogen bonds with the hydroxyl group in the glycerol moiety of MAG, as suggested in Fig. 6A, which made the hydroxyl group of MAG unreactive. The formation of hydrogen bonds would greatly reduce the thermodynamic activity and activity coefficient of MAG, which shifted the equilibrium to the formation of MAG. In fact, there is a relationship between E^T_N value and α value. Two general regression equations have been established as the following,³²

α = 0.0649E_T(30) − 2.03 − 0.27π*

ILs of this work have similar π* values, which suggests that the change of π* does not contribute much to the change of α. Therefore, α has a positive correlation with E^T_N, which could explain why the effect of α value on the glyceride selectivity have a similar trend with E^T_N.

As Fig. 3 suggested, the β value had an obvious positive correlation with MAG selectivity. For example, [BMIM]N(CN)₂ with higher β value and lower α value also resulted in higher MAG content, which could be due to the possibilities of the formation of hydrogen bonds between MAG and anion moiety of ILs according to Fig. 6B. However, the β value had an approximate negative correlation with DAG selectivity and FFA conversion. The β values are in a decreasing order of [BMIM]N(CN)₂ > [BMIM]BF₄ > [BMIM]Tf₂N > BMIM]PF₆. The respective FFA conversions in these ILs are 58.65%, 36.27%, 83.62% and 87.57% (Table 2). Zhao's research reported that the H-bond basicity could have a considerable impact on the enzyme stabilization in some ILs.²⁴ ILs with high hydrogen-bonding basicity were mostly water-miscible, which resulted in a strong interactions between the enzyme and solvent molecules. If such interactions were unfavorable for active sites and/or strong enough to disrupt the protein structures, the enzyme activity would decrease in these hydrophilic media. Higher TAG content and FFA conversion were obtained in [BMIM]PF₆, which could result from its lower β value (0.19). Higher enzyme activity resulted in more FFA conversion, which made more substrates to be transformed to TAG.

Fig. 3 Glyceride relative content and FFA conversion vs. β parameter of different ionic liquids: (A) MAG; (B) DAG; (C) TAG; (D) FFA conversion (plot from data in Table 2, and the solvent numbers in the graph are consistent with those in Table 2).

3.4 Effect of log P

Hydrophobicity could be regarded as a narrower concept of ‘polarity’. However, the former places more emphasis on miscibility with water. The hydrophobicity could be quantified by the log P value, which is related to the partition coefficient of ILs between 1-octanol and water. Joel et al. concluded that free lipase (Candida rugosa) was only active in hydrophobic ILs [BMIM]PF₆, but inactive in other hydrophilic ILs including [BMIM]BF₄, [BMIM]NO₃ and [BMIM]CF₃COO.²⁵

As shown in Fig. 4, the FFA conversion has an obvious positive correlation with log P. However, the correlation between glyceride selectivity and log P was not observed. As was reported by Zhao,²⁴ the lipase activities were generally much lower in water-miscible ILs (such as [BF₄]⁻, [N(CN)₂]⁻, [NO₃]⁻, [OAc]⁻, etc.) than in water-immiscible ones ([PF₆]⁻ and [Tf₂N]⁻), which was consistent with our results. [BF₄]⁻ and [N(CN)₂]⁻ based ILs with lower log P resulted in lower FFA conversion, which could be due to the loss of lipase activity. Similar phenomenon was also found in conventional organic solvents.³³ Enzymes were more active in solvents with a larger log P (>3) (such as hexane, log P = 3.5) than lower log P (such as dimethylsulfoxide, log P = −1.3). Unfortunately, the log P values of these lipophilic ILs with long alkyl chains are not available for comparative discussion. Previous studies have reported hydrophobicity exhibited an increasing trend as alkyl chain length increased.³⁴ This was attributed to greater van der Waals interactions between the IL alkyl chain and longer chain alcohols. Thus, some ILs with longer alkyl chains, such as [B₃C₁₄PH₂₉]NTf₂, [Oct₃MeN]NTf₂, [C₁₄MIM]NTf₂ and [C₁₂MIM]NTf₂ could possess higher log P values, which made them result in higher lipase activity. On the other hand, oleic acid with a long alkyl chain had lower mutual solubility with glycerol. The hydrophobic moiety of the alkyl group of ionic liquids provided a compatible medium for both oleic acid and glycerol, making them soluble in this IL synchronously, which might be the explanation of why higher FFA conversion, DAG and TAG content can be achieved in these kinds of ILs. It also indicated that ILs with longer alkyl chain are good for DAG and TAG selectivity.

Fig. 4 Glyceride relative content and FFA conversion vs. log P parameter of different ionic liquids: (A) MAG; (B) DAG; (C) TAG; (D) FFA conversion (plot from data in Table 2, and the solvent numbers in the graph are consistent with those in Table 2).

3.5 Effect of viscosity

As shown in Fig. 5, the influence of viscosity on the glyceride selectivity and FFA conversion was investigated. Obvious correlation could not be observed between the glyceride selectivity and solvent viscosity for all ILs. However, different rules were obtained in different anions based ILs. For [Tf₂N]⁻ based ILs, the viscosity increased with the increase of cation alkyl chain length, which results in higher FFA conversion, TAG and DAG contents and lower MAG content, but the trend was not obvious. However, for [PF₆]⁻ based ionic liquids the results showed the opposite trend. Higher FFA conversion, TAG and DAG contents and lower MAG content were observed in [PF₆]⁻ based ILs with lower viscosity, the same trend was also found in [BF₄]⁻ based ILs.

Fig. 5 Glyceride relative content and FFA conversion vs. viscosity parameter of different ionic liquids: (A) MAG; (B) DAG; (C) TAG; (D) FFA conversion (plot from data in Table 2, and the solvent numbers in the graph are consistent with those in Table 2).

According to Table 2, IL viscosity was highly dependent on the anion types. [PF₆]⁻ and [BF₄]⁻ based ILs (100–700 cP) have much higher viscosity than [Tf₂N]⁻ and [N(CN)₂]⁻ (20–100 cP). It indicated that viscosity is the main factor affecting glyceride selectivity and conversion in high viscous ILs, such as [PF₆]⁻ and [BF₄]⁻ based ILs. Higher viscosity resulted in stronger mass transfer limitation, affecting the reaction rate and conversion. Esterification of glycerol with free fatty acids was multi-step reaction, MAG were produced firstly, and then transformed to DAG and TAG. In [PF₆]⁻ or [BF₄]⁻ based ILs, with the decrease of conversion, most substrates transformed to MAG and it would be more difficult to transform to DAG and TAG, which resulted in higher MAG content in higher viscous ILs.

3.6 Pearson's correlation analysis

The correlation between IL properties and the reaction results was further studied by Pearson's correlation analysis. The corresponding Pearson's correlation coefficients for all ILs were shown in the form of histogram in Fig. 7 and only the Pearson's correlation coefficients for [Tf₂N]⁻ based ILs were plotted in Fig. 8. Pearson's correlation coefficients represent a linear correlation between the two variables. As shown in Fig. 7, the correlation coefficients were relatively low for all of ILs, which indicated that the results of reaction were not simply influenced by any single property, but by multiple properties collaboratively. MAG content was positively correlated to E^T_N value and Kamlet–Taft parameters (R > 0). To be noted, MAG content had a significantly positive correlation with E^T_N value (R = 0.4860, P < 0.05). While opposite results were obtained for DAG (R = −0.3789), TAG content (R = −0.3834) and FFA conversion (R = 0.2725). log P and hydrogen bond basicity (β) were observed to have main effects on the FFA conversion, because relatively higher positive correlation coefficient (R = 0.6065) and negative coefficient (R = −0.3878) were obtained respectively. According to Fig. 8, the results indicated that [NTF₂]⁻ based ILs were more suitable for the rules of polarity. E^T_N value was significantly positively correlated with MAG content (R = 0.883, P < 0.01). The log P had a significantly positive correlation with TAG content (R = 0.8395) and FFA conversion (R = 0.992, P < 0.01) and a negative correlation with MAG (R = 0.992).

Fig. 6 Proposed mechanisms for the decrease of the thermodynamic activity of monoglycerides.

Fig. 7 Pearson's correlation coefficients between properties and the results of reaction for all kinds of ILs: (A) MAG; (B) DAG; (C) TAG; (D) FFA conversion (*means p < 0.05, **means p < 0.01).

Fig. 8 Pearson's correlation coefficients between properties and the results of reaction for [Tf₂N]⁻ abased ILs: (A) MAG; (B) DAG; (C) TAG; (D) FFA conversion (*means p < 0.05, **means p < 0.01).

4 Conclusions

In summary, the results of this work allowed a better understanding of the impact of IL properties on selective enrichment of glycerides in direct lipase-catalyzed esterification. E^T_N and Kamlet–Taft parameters are most likely the determining factors for glyceride selectivity. ILs with high E^T_N and Kamlet–Taft values might form the hydrogen-bond with MAG and resulted in high MAG content. ILs with higher log P and lower β resulted in higher FFA conversion and ILs with longer alkyl chains lead to higher contents of DAG and TAG. Viscosity may reduce the reaction rate which could decrease the FFA conversion and change glyceride selectivity. All in all, selective enrichment of glycerides was not simply influenced by any single property, but by multiple properties collaboratively. However, higher polar ILs could always result in higher polar reaction products such as MAG.

According to the conclusion, specific ILs could be selected for the production of MAG, DAG and TAG respectively by enzymatic esterification. [HO(CH₂)₂MIM]NTf₂ and [MeO(CH₂)₂MIM]NTf₂ indicated selective enrichment of higher polar products such as MAG. ILs with longer alkyl chain, lower β value and lower viscosity, such as [Oct₃MeN]NTf₂, [B₃C₁₄PH₂₉]NTf₂, [BMIM][PF₆] and [C₁₂MIM]NTf₂, were good for selective enrichment of lower polar products such as DAG and TAG.

Acknowledgements

Support for this work was provided by the National Natural Science Foundation of PR China (31601433) and Jiangsu Provincial Natural Science Foundation (BK20140149).

References

1. Z.-Q. Duan, W. Du and D.-H. Liu, Process Biochem., 2010, 45, 1923–1927.
2. N. Zhong, L. Li, X. Xu, L. Cheong, B. Li, S. Hu and X. Zhao, J. Am. Oil Chem. Soc., 2009, 86, 783–789.
3. D. M. Cetina, G. I. Giraldo and C. E. Orrego, J. Mol. Catal. B: Enzym., 2011, 72, 13–19.
4. N. Zhong, Z. Gui, L. Xu, J. Huang, K. Hu, Y. Gao, X. Zhang, Z. Xu, J. Su and B. Li, Lipids Health Dis., 2013, 12, 1.
5. H. Ghamgui, N. Miled, A. Rebaï, M. Karra-chaâbouni and Y. Gargouri, Enzyme Microb. Technol., 2006, 39, 717–723.
6. Y. J. Zhao, J. F. Liu, L. Deng, F. Wang and T. W. Tan, J. Mol. Catal. B: Enzym., 2011, 72, 157–162.
7. Z. Guo and Y. Sun, Biotechnol. Prog., 2004, 20, 619–622.
8. Y. Watanabe, Y. Yamauchi-Sato, T. Nagao, T. Yamamoto, K. Ogita and Y. Shimada, J. Mol. Catal. B: Enzym., 2004, 27, 249–254.
9. A. E. Janseen, A. Van der Padt, H. M. Van Sonsbeek and K. Van't Riet, Biotechnol. Bioeng., 1993, 41, 95–103.
10. J. C. Bellot, L. Choisnard, E. Castillo and A. Marty, Enzyme Microb. Technol., 2001, 28, 362–369.
11. U. Kragl, M. Eckstein and N. Kaftzik, Curr. Opin. Biotechnol., 2002, 13, 565–571.
12. M. Moniruzzaman, K. Nakashima, N. Kamiya and M. Goto, Biochem. Eng. J., 2010, 48, 295–314.
13. Z. Yang and W. Pan, Enzyme Microb. Technol., 2005, 37, 19–28.
14. S. Mallakpour and M. Dinari, in Green Solvents II, Springer, 2012, pp. 1–32.
15. H. Zhao, J. Chem. Technol. Biotechnol., 2010, 85, 891–907.
16. H. Zhao, G. A. Baker, Z. Song, O. Olubajo, T. Crittle and D. Peters, Green Chem., 2008, 10, 696–705.
17. S. Sun and X. Chen, Biochem. Eng. J., 2015, 97, 25–31.
18. Z. Guo, B. Q. Chen, R. L. Murillo, T. W. Tan and X. B. Xu, Org. Biomol. Chem., 2006, 4, 2772–2776.
19. Y. Abe, K. Kude, S. Hayase, M. Kawatsura, K. Tsunashima and T. Itoh, J. Mol. Catal. B: Enzym., 2008, 51, 81–85.
20. Y. Tsukada, K. Iwamoto, H. Furutani, Y. Matsushita, Y. Abe, K. Matsumoto, K. Monda, S. Hayase, M. Kawatsura and T. Itoh, Tetrahedron Lett., 2006, 47, 1801–1804.
21. Z. Guo and X. B. Xu, Org. Biomol. Chem., 2005, 3, 2615–2619.
22. D. Kahveci, Z. Guo, B. Özçelik and X. Xu, Food Chem., 2010, 119, 880–885.
23. D. Kahveci, Z. Guo, B. Özçelik and X. Xu, Process Biochem., 2009, 44, 1358–1365.
24. H. Zhao, G. A. Baker, Z. Song, O. Olubajo, L. Zanders and S. M. Campbell, J. Mol. Catal. B: Enzym., 2009, 57, 149–157.
25. J. L. Kaar, A. M. Jesionowski, J. A. Berberich, R. Moulton and A. J. Russell, J. Am. Chem. Soc., 2003, 125, 4125–4131.
26. T. De Diego, P. Lozano, M. A. Abad, K. Steffensky, M. Vaultier and J. L. Iborra, J. Biotechnol., 2009, 140, 234–241.
27. GB/T 5530–2005/ISO 660:1996 Animal and vegetable fats and oils: determination of acid value and acidity.
28. Y. Hu, Z. Guo, B.-M. Lue and X. Xu, J. Agric. Food Chem., 2009, 57, 3845–3852.
29. Q. Y. Pan, L. P. Yang and X. H. Meng, J. Am. Oil Chem. Soc., 2013, 90, 501–509.
30. L. Crowhurst, P. R. Mawdsley, J. M. Perez-Arlandis, P. A. Salter and T. Welton, Phys. Chem. Chem. Phys., 2003, 5, 2790–2794.
31. M. J. Kamlet and R. Taft, J. Am. Chem. Soc., 1976, 98, 377–383.
32. J.-M. Lee, S. Ruckes and J. M. Prausnitz, J. Phys. Chem. B, 2008, 112, 1473–1476.
33. J. Lu, K. Nie, F. Wang and T. Tan, Bioresour. Technol., 2008, 99, 6070–6074.
34. L. Ropel, L. S. Belvèze, S. N. Aki, M. A. Stadtherr and J. F. Brennecke, Green Chem., 2005, 7, 83–90.
35. S. Coleman, R. Byrne, S. Minkovska and D. Diamond, Phys. Chem. Chem. Phys., 2009, 11, 5608–5614.
36. M. J. Muldoon, C. M. Gordon and I. R. Dunkin, J. Chem. Soc., Perkin Trans. 2, 2001, 433–435.
37. B. R. Mellein, S. N. Aki, R. L. Ladewski and J. F. Brennecke, J. Phys. Chem. B, 2007, 111, 131–138.
38. C. Chiappe and D. Pieraccini, J. Phys. Chem. A, 2006, 110, 4937–4941.
39. S. V. Dzyuba and R. A. Bartsch, Tetrahedron Lett., 2002, 43, 4657–4659.
40. S. G. Zhang, X. J. Qi, X. Y. Ma, L. J. Lu and Y. Q. Deng, J. Phys. Chem. B, 2010, 114, 3912–3920.
41. S. K. R. D. Rogers Robin, ACS Symp. Ser., 2002, 270.
42. A. Jelicic, N. García, H.-G. Löhmannsröben and S. Beuermann, Macromolecules, 2009, 42, 8801–8808.
43. M. Ab Rani, A. Brant, L. Crowhurst, A. Dolan, M. Lui, N. Hassan, J. Hallett, P. Hunt, H. Niedermeyer and J. Perez-Arlandis, Phys. Chem. Chem. Phys., 2011, 13, 16831–16840.
44. H. Tokuda, S. Tsuzuki, M. A. B. H. Susan, K. Hayamizu and M. Watanabe, J. Phys. Chem. B, 2006, 110, 19593–19600.
45. Y. Yoshida, O. Baba and G. Saito, J. Phys. Chem. B, 2007, 111, 4742–4749.
46. C. Chiappe, C. S. Pomelli and S. Rajamani, J. Phys. Chem. B, 2011, 115, 9653–9661.

---