Optimization of Solvent Extraction of Lipids from Yarrowia lipolytica towards Industrial Applications

Abstract

Extraction of intracellular lipids of the oleaginous yeast Yarrowia lipolytica has been systematically studied aiming towards a sustainable extraction process for lipid recovery. Selection of suitable industrial (bulk) solvents and extraction parameters that lead to maximization of lipid recovery are significant issues to be addressed, with industrial applications motivating this study. Biomass from fermentation of Yarrowia lipolytica (MUCL 28849) was used in small laboratory tests to assess different solvent mixtures (i.e., methanol/hexane, isopropanol/hexane, and methanol/ethyl acetate), implementing a systematic design of experiments methodology to identify near-optimum values of key extraction variables (i.e., polar/non-polar ratio, vortex time, dry biomass/solvent ratio) in regard to lipid yield (g lipids/g dry biomass). The methanol/hexane mixture exhibited the highest extraction yield in a wide range of experimental conditions, resulting in the following optimum parameters: polar/non-polar ratio 3/5, vortex time 0.75 h, and dry biomass/solvent ratio 40. Extraction tests on a fifty-times-larger scale (in a Soxhlet apparatus employing the optimal extraction parameters) confirmed the optimization outcome by obtaining up to 27.6% lipids per dry biomass (L/DB), compared to 12.1% L/DB with the reference lipid extraction method employing chloroform/methanol. Assessment of lipid composition showed that unsaturated fatty acid recovery was favored by the methanol/hexane solvent. Fatty acid composition was not affected by the increase in Soxhlet reflux cycles, whilst the lipid yield was notably favored.

1. Introduction

Yarrowia lipolytica (Y. lipolytica) is an oleaginous yeast that can produce and accumulate intracellular lipids in high concentrations, usually exceeding 20–40% w/w of its dry biomass [1,2]. This strictly aerobic yeast has been widely studied both due to its oleaginous ability and its robustness to grow in diverse environments regarding pH, temperature, and osmotic stress [3,4,5]. Furthermore, Y. lipolytica can grow on a number of low-cost substrates (both hydrophilic and hydrophobic) such as raw glycerol, fatty acids, oil-containing wastes, and by-product streams [6,7,8,9,10]. The utilization of low-cost by-products or wastes as feedstock is critical for the sustainability of the lipid production process. The composition of the fatty acid methyl esters (FAMEs), derived from yeast lipids after transesterification, is affected by the characteristics of each different strain and the fermentation process parameters (e.g., substrate composition, pH, dissolved oxygen concentration, etc.) [11,12,13]. However, the fatty acid composition in most Y. lipolytica strains greatly resembles the composition of plant oils, which are commonly used for production of first-generation biodiesel. The intracellular lipids of Y. lipolytica are mainly composed of palmitic, stearic, and oleic acid, with low amounts of linoleic and linolenic acid [14]; thus, they can be used for the production of second-generation biodiesel that meets current specification standards [15].

The efficient and economically sustainable extraction of intracellular lipids from Y. lipolytica still remains a significant technological barrier to their commercial exploitation for biodiesel production [14,16]. Lipids are concentrated in specific cell compartments, known as lipid bodies or lipid droplets. The latter are mainly composed of a phospholipid monolayer wall, containing non-polar compounds, such as triacylglycerol and steryl esters in their core [17,18]. Therefore, high lipid extraction yields usually involve a mixture of both polar and non-polar organic solvents to disrupt the complex cell wall structure, to facilitate the transfer of the solvent molecules into the cell compartments, and finally to solubilize the lipids for efficient extraction and separation [16]. Pre-treatment of yeast biomass, including drying to remove water that hinders the solvent’s intrusion efficiency, and cell disruption techniques (i.e., bead beating, freeze-thaw, ultrasonication, alkaline or acid chemical treatment, application of surfactants or enzymes, etc.) can greatly assist the subsequent solvent extraction process [19,20,21,22]. Moreover, appropriate solvent selection is critical for both optimizing the lipid extraction yield and the chemical composition of FAMEs that may affect the quality characteristics of biodiesel after the transesterification process [23,24].

During lipid extraction, wet, or most often dry, yeast biomass is exposed to an eluting extraction solvent, or a mixture thereof, which extracts the lipids from the cellular matrix. Then, the crude lipid phase is separated from the cell debris and the aqueous phase (if present), and the organic solvent/s together with any traces of water are evaporated, prior to measurement of the lipid mass. Although lipid extractability varies for different microorganism species, the most commonly used methods for determining total lipid content in microorganisms are based on using a mixture of chloroform and methanol, as described by Folch et al. [25] and Bligh and Dyer [26]. The combination of chloroform/methanol/water mixture for lipid extraction was first reported by Folch et al. [25], who suspended a sample of homogenized animal tissue into a chloroform/methanol mixture of 2:1 ratio. Polar and non-polar phases were distinct after the addition of water, rendering the crude lipid layer accessible. The latter, being the lower phase, was separated and washed several times with either water or a salt solution to purify and collect the lipids [25]. Two years later, Bligh and Dyer [26] proposed a rapid method for extraction and purification of lipids from biological materials, employing similar steps such as homogenization and solvent synthesis. A second step of adding pure chloroform was included in the assay and emphasis was given to the initially existing water of the sample [26].

The fact that these two methods are the most commonly applied and are used as a standard is evident by their widespread use in several studies of lipid extraction (

Table 1. Summary of studies of lipid extraction from Yarrowia lipolytica (Y. lipolytica) yeast, conducted mainly using chloroform/methanol mixture as solvent. Critical extraction steps and parameters are briefly presented.

Species-Strain	Type of Solvent	Biomass Sample (g)	mg Biomass/mL Solvent	Drying Method	Cell Disruption	Separation	Temperature	Reference
Y. lipolytica MUCL 28849	Chl/MeOH 2:1 Chl/MeOH 1:1 Chl/MeOH 1:2	0.5	N/A	Lyophilization	Shaking for 1 day	Centrifugation	RT	[6]
Y. lipolytica W29 (ATCC20460)	Chl/MeOH 2:1	0.001; WB	1	Dried at 60 °C for 2 nights	Vortex for 1 h	Centrifugation	RT	[29]
Y. lipolytica E26E1	Dimethylcyclohexylamine Ethylbutylamine Dipropylamine Chl/MeOH 2:1	N/A	1	N/A	Mixed at 50 rpm	Centrifugation	RT	[23]
Y. lipolytica po1g	Chl/MeOH 2:1	0.002	4	Dried in oven at 90 °C	Vortex	Centrifugation	RT	[30]
Y. lipolytica IFP29 (ATCC 20460)	Chl/MeOH 1:2	3; WB/ 0.3; DB	29.79	Cold drying under reduced pressure, freeze-drying	Stirring, ultrasonication, bead milling, microwaves	Centrifugation	RT	[20]
Y. lipolytica SKY-7	Chl/MeOH 2:1	3.1 ± 0.2; WB/ 0.162; DB	10.8	N/A	4 h in Water Bath (60 °C, 100 rpm)	Centrifugation	RT	[31]
Y. lipolytica	Hex/Isop 5:3	1	50	Dried under a hot plate	Bead milling	Centrifugation	N/A	[32]
Y. lipolytica (strains ACA-DC 5033 and LFMB Y19)	Chl/MeOH 2:1	0.3	12	N/A	Shaking or acid hydrolysis	Filtration	N/A	[28]
Y. lipolytica modified JMY5289	Chl/MeOH 2:1 n-Hexane EtOH d-limonene p-cymene Isoamyl acetate Butyl acetate Ethyl acetate	0.1	100	Freeze-drying Or wet biomass	High-pressure homogenization, bead milling, and shaking	Centrifugation	N/A	[33]
Y. lipolytica (JMY 5289)	Chl/MeOH 2:1 n-Hexane	0.1	100 or 1/10 ratio	Lyophilization	High-pressure homogenization, bead milling, and shaking	Centrifugation	RT	[34]
Y. lipolytica NCIM 3589	Chl/MeOH 1:1	0.5	25	Dried in oven	Acid treatment and shaking	Filtration	30 °C	[35]
Y. lipolytica	SCCO2 SCCO2/EtOH 9:1	20	N/A	Air dried at 60 °C for 15 h	Milling, rapid decompression, solvent maceration	Centrifugation	40–60 °C	[36]

Abbreviations: RT, room temperature; N/A, not available; Chl, chloroform; MeOH, methanol; Hex, hexane; Isop, isopropanol; EtOH, ethanol; SCCO₂, supercritical CO₂.

). Raut et al. used chloroform/methanol in 2:1 (Folch method) and 1:1 ratios for lipid extraction from Y. lipolytica grown on waste cooking oil [10]. They also tested different cell lysis techniques to investigate their effect on lipid recovery. By merging two methods, i.e., a modified acid hydrolysis [27] and a bead beater, they measured a lipid content approx. 42% w/w, whereas performing acid hydrolysis and homogenization resulted in 24% w/w and 22% w/w, respectively. Sarantou et al. studied two lipid extraction methods for different strains of Y. lipolytica [28]. The first extraction technique tested was a modified Folch method (employing reduced use of organic solvent and increased shaking time), whereas the second technique comprised acidification and boiling prior to adding the solvent. They concluded that the modified Folch method without the acidification step resulted in higher lipid extraction yield, i.e., approx. 3 g/L, while with the acidification step it reached approx. 2.5 g/L. [28]. Meullemiestre et al. studied different ways to maximize lipid extraction using a chloroform/methanol mixture in a 1:2 ratio according to the Bligh and Dyer method. It was shown that freeze-drying under pressure and bead milling resulted in maximum lipid recovery [20].

The main concern related to the aforementioned (commonly employed) extraction methods, is that chloroform is a highly toxic and suspected carcinogenic chemical [37], which hinders its use on an industrial scale [38]. Health, safety, and environmental concerns motivated the research on the use of alternative, i.e., “green”, solvents for lipid extraction from Y. lipolytica. Imatoukene et al. [33] evaluated the performance of some alternative solvents compared to conventional chloroform/methanol and n-hexane with bead milling and high-pressure homogenization as pre-treatment steps. Forty-one solvents were initially selected and subsequently narrowed down based on their solubility properties. The alternative solvents finally assessed included esters (isoamylacetate, ethyl acetate, and n-butyl acetate) and terpenes (d-limonene and p-cymene). Among these solvents, isoamylacetate (together with bead milling preceded by high-pressure homogenization) exhibited the greatest amount of lipid yield (96.7%) compared to the conventional chloroform:methanol (99.6%) and n-hexane (98%) [33]. Furthermore, Yook et al. [23] studied the lipid extraction efficiency of selected switchable hydrophilicity solvents (SHSs) which included N-ethylbutylamine (EB), N-dipropylamine (DP), and N,N-dimethyl-cyclohexyl-amine (DMCHA). Two of them, N-ethylbutylamine and N,N-dimethyl-cyclohexyl-amine, resulted in 13% higher amount of lipid yields (58.6%) compared to the conventional methods [23]. Finally, Breil et al. [39] investigated the effectiveness of several biobased solvents compared to the conventional ones. The biobased solvents that were studied included alcohols (isopropanol and ethanol), esters (ethyl acetate, ethyl lactate, and dimethyl carbonate), terpenes (cymene, pinene, and limonene) and ethers (2-methyltetrahydrofuran and cyclopentyl methyl ether). The lipid yields with all these solvents were in the range of 14–15%, similar to the result obtained for hexane, which was the reference solvent [39]. The use of supercritical CO₂ (SCCO₂) has been also studied for extraction of lipids of different yeast strains [19,40]. Milanesio et al. [36] employed SCCO₂ extraction at 40 °C and 20 MPa with ethanol as a co-solvent for the recovery of lipids from Y. lipolytica cells. The addition of ethanol enhanced the oil solubility and the lipid yield reached values up to 25.4% compared to 13.4% employing Soxhlet extraction with Chl/MeOH 2:1. In all these studies with alternative solvents, the fatty acid profile was not significantly different compared to the one with the conventional solvents.

The use of novel extraction solvents presents significant challenges, especially when considering the upscaling of lipid extraction to an industrial scale. The main concerns include their higher costs compared to conventional solvents and the lack of experience and/or specifically designed equipment for application on a large scale. In the case of SCCO₂, the significantly high operating pressure (e.g., 20 MPa) [36] increases the operating cost, rendering SCCO₂ a rather expensive extraction process. In oilseed industries, continuous solvent extraction is employed as a reliable and efficient process to extract oil from oilseeds with high yields (e.g., >95%), attaining a residual oil concentration as low as <0.5% w/w [41]. Hexane is the solvent universally preferred due to its high oil solubility, narrow boiling point, and moderate energy cost for recovery through evaporation and distillation [42]. Different types of hexane extractors are used; however, the continuous belt extractor type is mainly employed, with a capacity typically varying between 500 and 5000 tn/d of pressed oilseed and approx. 5–10 extraction equilibrium stages [41]. Typical biomass-to-solvent ratio is around 1.0 and the oilseed mean residence time is approx. 2–3 h. A study employing n-hexane as a solvent for lipid extraction from Y. lipolytica with different pre-treatment steps resulted in maximum oil recovery equal to approx 100% for dried and lyophilized biomass, and 80% for wet biomass, compared to the standard chloroform/methanol mixture (2:1 v/v). High-pressure homogenization (approx. 1500 bar and five passes) was employed as pre-treatment for cell disruption [34].

The review of the scientific literature reveals that there are limited data on lipid extraction from Y. lipolytica employing solvents other than the standard chloroform/methanol mixture and few studies report the effectiveness of different mixtures of bulk (commodity) chemicals for their extraction. Moreover, the few available studies focus on alternative solvents that are not attractive for large-scale processes due to high cost or lack of relevant experience and equipment. The aim of this study is to investigate various solvent mixtures of common bulk (commodity) chemicals and optimize the extraction of lipids from Y. lipolytica. Three different mixtures, namely (a) methanol/hexane, (b) isopropanol/hexane, and (c) methanol/ethyl-acetate, were assessed concerning the effect of (a) the polar/non-polar solvent ratio, (b) the extraction time, and (c) the dry biomass/solvent ratio on the extracted lipid yield. Near-optimum conditions were determined and further studied, employing Soxhlet extraction for different numbers of reflux cycles that better resembles continuous industrial solvent extraction processes. Apart from assessing the lipid extraction yield in regard to different extraction solvents, the lipids’ fatty acids composition was analyzed to assess their quality as feedstock for biodiesel production. This work provides useful data on the efficiency of different bulk (commodity) chemicals’ mixtures for the extraction of lipids from Y. lipolytica cells and specifies process parameters that could be employed for the development of a sustainable extraction process.

2. Materials and Methods

2.1. Strain and Culture Media

From Y. lipolytica clades, a wild-type strain, registered as Y. lipolytica MUCL 28849 (BCCM/MUCL (Agro) Industrial Fungi and Yeasts Collection, Louvain-la-Neuve, Belgium), was used for this study. From a glycerol stock (25% v/v pure glycerol) stored at −80 °C, Y. lipolytica was plated on YPG-agar (yeast extract 10 g/L; peptone 20 g/L; glycerol 2% v/v), which was incubated for 24 h in 30 °C prior to use, and then stored at 4 °C. Both precultures and main cultures were carried out in synthetic nitrogen-limited media (SM). A modified version of the latter, described by Egermeier et al. [43], consisted of (g/L): (NH₄)₂SO₄, 3.0; KH₂PO₄, 2.0; Na₂HPO₄ × 2 H₂O, 2.6; MgSO₄ × 7 H₂O, 1.0; CaCl₂ × 2 H₂O, 0.2; FeCl₃, 0.02; (mg/L): Thiamin-HCl, 1.0; H₃BO₃, 0.5; CuSO₄ × 5 H₂O, 0.06; KI, 0.1; MnSO₄ × H₂O, 0.45; ZnSO₄ × 7 H₂O, 0.71; Na₂MoO₄ × 2 H₂O, 0.23. The sole carbon source was crude glycerol (≥90–92% w/w purity), which was obtained directly from a biodiesel production plant (Fytoenergeia/NewEnergy S.A., Serres, Greece). The substrate was freshly prepared and sterilized at 121 °C for 20 min, excluding the mineral stock which contained all the micronutrients. The mineral stock was sterilized through filter sterilization prior to each experiment. Crude glycerol was added at a final concentration of 94.5 g/L reaching a carbon-to-nitrogen ratio of 68. Pre-cultures were performed in 500 mL flasks on a working volume of 100 mL, whereas the main cultures were conducted on a 3-L bioreactor (BioFlo120, Eppendorf, Germany). Preculture was centrifuged (3750× g × 6 min—Heraeus Megafuge 16R, Thermo Fisher Scientific Inc., Waltham, MA, USA) and the pellet was resuspended to 10 mL of sterile deionized water and added to the bioreactor retaining the volume of the main fermentation practically constant.

2.2. Bioreactor Settings and Biomass Preparation

The bioreactor was equipped with a pH electrode (405-DPAS-SC, © METTLER TOLEDO, Greifensee, Switzerland), an optical density probe (OD, Dencytee Unit, Hamilton Bonaduz AG, Bonaduz, Switzerland), and a dissolved oxygen probe (DO, InPro6860i, © METTLER TOLEDO, Greifensee, Switzerland). The working volume of each fermentation was 1.75 L and the temperature 30 °C. The pH was maintained at 6.0 by automated addition of 5 N H₂SO₄ and 5 N NaOH. Aeration was employed through sterile air supply (1.0 vvm) and agitation at 800 rpm, as already described elsewhere [44]. Cell growth was continuously monitored through the OD probe. Correlation with dry biomass was assessed by regular analysis of the dry biomass concentration; 1 mL of culture was sampled and centrifuged (3750× g × 6 min), while the remaining cell pellet was double-washed. The washed pellet was dried at 60 °C until reaching constant weight. Yeast biomass was collected after 60 h of cultivation through centrifugation (3750× g × 6 min). After double-washing, the total wet biomass of the fermentation was weighed and stored at −40 °C until lyophilization. The frozen pellets were freeze-dried under vacuum at a temperature of −55 °C, for 48 h, until complete water sublimation. The lyophilized biomass was stored at room temperature before lipid extraction procedure.

2.3. Lipid Extraction—Control Method

The control method for lipid extraction was based on Tai and Stephanopoulos [29], who employed a modified Folch method [25]. In 100 mg of lyophilized biomass, 10 mL of chloroform/methanol solvent (Chl/MeOH, 1:2 v/v) was added and the samples were placed in an ultrasonication bath (3 × five-minute cycles) [25], and subsequently vortexed for one hour [29]. Then, 1 mL of 0.9% w/v aqueous NaCl solution was added to the samples to form a biphasic mixture, which, after gentle stirring, was centrifuged (3750× g × 6 min). After the first centrifugation, two layers were formed. To the bottom layer, which contained the non-polar solvent, 1 mL of isotonic aqueous NaCl solution 0.9% w/v was added, while to the upper layer, which contained the polar solvent and biomass, 1 mL of the original solvent was added for optimum lipid removal. All samples were vortexed for 1 min and centrifuged again (3750× g × 6 min). After the second centrifugation, the non-polar solvent, which contained the lipids, was collected. The samples were placed in an oven at 70 °C until complete evaporation of chloroform. Samples containing the lipid mass were weighed regularly until constant weight. Finally, the lipids were re-suspended in hexane and stored at −20 °C for subsequent FAME analysis.

2.4. Lipid Composition (FAME Analysis)

Lipid transesterification was performed for the analysis of the fatty acid profile. The lipid transesterification method was based on the method of Tai and Stephanopoulos [29]. The lipids were stored in 2 mL vials after re-suspending in hexane at −20 °C. In each sample, 0.8 mL of a 2% v/v solution of sulfuric acid in methanol were added to initiate the transesterification reaction. The samples were incubated for 2 h at 60 °C and then they were left for partial evaporation in the fume hood. For the extraction of FAMEs, 0.8 mL of hexane was added and the samples were stirred vigorously for 10 min and left for 5–10 min until the separation of the 2 phases was clear [29]. Finally, 0.3 mL of the upper phase of the sample was transferred to glass vials suitable for Gas Chromatographic (GC) analysis. The lipid profile was determined according to Regulation (EC) No 796/2002 [45] on a Gas Chromatograph (Shimadzu GCMS-QP2010) equipped with a capillary column SP-2340 (60 mm × 0.25 mm, 20 μm film thickness) (Supelco, Bellefonte, PA, USA). The statistical significance between the means of the fatty acid composition and the different types of organic solvents was evaluated by using one-way ANOVA for every fatty acid chain length followed by post hoc tests (Bonferroni correction). Any difference with a p-value lower than 0.05 for one-way ANOVA and p-value lower than 0.008 (Bonferroni corrected) was considered as statistically significant. All statistical evaluations were performed by using Microsoft Excel^® software v.15.0.

2.5. Design of Experiments (DoE) for Lipid Extraction Optimization

Statistical experimental design, i.e., a Face-Centered Composite (FCC) design was used to design and perform the lipid extraction experiments. During the lipid extraction optimization experiments, four independent variables (factors) were studied, namely (i) polar/non-polar solvent ratio (g/g), X_A; (ii) vortex time equivalent to extraction time (h), X_B; (iii) dry biomass/solvent ratio (mg/mL), X_C; and (iv) type of solvent, X_D. The first three factors were numerical, while the last one was categorical. The lipid extraction yield, i.e., the ratio of extracted lipids to the dry biomass (Lipids/Dry Biomass—L/DB, g/g) was assigned as the dependent variable (response Y₁).

Table 2. Level of values of the independent variables of the statistical design experiments.

Level of Values	Factor XA Polar/Non-Polar Ratio	Factor XB Time (Vortex) (h)	Factor XC Dry Biomass/Solvent Ratio (mg/mL)
−1	0	0.5	10
0	0.5	1.75	40
+1	1	3	70

shows the levels for the three numerical variables which were selected based on preliminary experimental data and literature. The types of solvents (X_D), used in the present work were (a) methanol/hexane (MeOH/Hex), (b) isopropanol/hexane (Isop/Hex), and (c) methanol/ethyl-acetate (MeOH/AcOEt). The designed FCC indicated three identical strategies, one for each solvent type, proposing a set of 45 runs per solvent type, i.e., 135 runs in total. All experiments were carried out in triplicate and in a randomized order to restrict the effects of variability on the observed response. Analysis of variance (ANOVA) was used to evaluate the statistical significance of the factors (p-value < 0.05 considered significant). Multi-linear regression (MLR) was employed to fit the mathematical model to the experimental data (correlation coefficient, R^2,, and adjusted R² were used as criteria of the fit quality). The adequacy of the used models was evaluated using criteria such as adequate accuracy (with the desired value of the signal-to-noise ratio > 4.0) and the reproducibility of the model (expressed as coefficient of variation (CV) with the desired value < 10%). The optimal areas of the independent variables, i.e., the desired space for all three types of solvents, were determined with the aid of 3D response surface analysis of the independent and dependent variables. The experimental design was carried out using Design Expert^® (v. 13.0 free trial, Stat-Ease Inc., Minneapolis, MN, USA) (Montgomery, 2017; Myers et al., 2016).

2.6. Lipid Extraction for FCC Design Experiments

For the FCC design experiments, a quantity of dry biomass equal to 120 mg was initially weighed in 15 mL test tubes. The appropriate amount of solvents mixture was added (depending on the dry biomass/solvent ratio) and the samples were vortexed for the pre-determined time without any additional pre-treatment step for cell lysis. After vortexing, 1 mL of 0.9% w/v isotonic aqueous NaCl solution was added for better separation of the layers and then the samples were vortexed for 1 min and centrifuged (3750× g × 6 min). After the first centrifugation, the layers were separated. To the upper layer, which contained the non-polar solvent, 1 mL of isotonic aqueous NaCl solution 0.9% w/v was added, while to the bottom layer, which contained the polar solvent and biomass, 1 mL of the original solvents’ mixture was added for optimum lipid removal; all samples were vortexed for 1 min and centrifuged again (3750× g × 6 min). After the second centrifugation, the non-polar solvent layer (containing the lipids) was separated and transferred to a hood for solvent evaporation at room temperature. Finally, the lipid samples were transferred to an oven at 60 °C for complete evaporation of the solvent. Dry lyophilized biomass samples were also extracted, employing the control extraction method, i.e., chloroform/methanol in a ratio of 2:1 v/v (as previously described in Section 2.3). These samples were considered as controls and their results were used for comparison with the results obtained from the three FCC designs.

2.7. Soxhlet Extraction

Experiments of lipid extraction from Y. lipolytica were performed, using a Soxhlet setup (WHM 12391, Witeg Labortechnik GmbH, Wertheim, Germany), after assessing the results of the previously described FCC design optimization experiments. In 6 g of lyophilized biomass, 150 mL of solvent was added, and the extraction process was performed for 5, 10, 15, 20, and 40 reflux cycles. After the extraction process, the samples were firstly placed on a rotary evaporator (Hei-VAP Advantage, Heidolph Instruments GmbH and Co. KG, Schwabach, Germany) at 80 rpm and 50 °C to remove most of the solvent under vacuum and then left in an oven at 60 °C for further solvent evaporation, until achievement of constant weight.

3. Results and Discussion

3.1. Comparison of Different Lipid Extraction Solvents

The results from lipid extraction using the three different solvent mixtures are presented in the frequency histogram plots (Figure 1), showing that the lipid yield (L/DB) varied from 0 to 0.13 g L/g DB. The control extraction method for the same dry biomass sample resulted in approx. 0.10 g L/g DB. The frequency of higher lipid yields was critical for the selection of parameter combinations for further investigation. The highest L/DB values occurred at a considerably higher frequency (Figure 1a) after using MeOH/Hex mixture. Specifically, the most frequent lipid yield values were between 0.10 and 0.11 g L/g DB. On the contrary, for the Isop/Hex mixture, the frequency of L/DB was higher in the range of 0.04–0.07 g L/g DB, while for the MeOH/AcOEt mixture, most frequent values varied from 0.05 to 0.09 g L/g DB. The combination of methanol and hexane, as polar and non-polar solvent, respectively, seems to result in an increased lipid yield. Compared to the standard extraction solvent mixture (i.e., Chl/MeOH), the MeOH/Hex mixture exhibited equal or higher lipid yields for 48.9% of the experiments, whereas for the Isop/Hex and MeOH/AcOEt mixtures, these percentages were 6.7% and 8.9%, respectively. Moreover, the MeOH/Hex mixture resulted in the higher overall lipid yield, reaching 0.13, which is approx. 31% higher than the lipid yield with the standard extraction process. Since values higher than 0.10 L/DB (g/g) were less frequent for the two other solvent mixtures, further study was focused on the MeOH/Hex mixture.

The obtained results from the FCC, testing the MeOH/Hex type, are included in

Table 3. Face-Centered Composite (FCC) design matrix for MeOH/Hex and observed response.

Run	Factor XA Polar/Non-Polar Ratio	Factor XB Time (Vortex) (h)	Factor XC Dry Biomass/Solvent Ratio (mg/mL)	Response Y1 L/DB (g/g)
1	0.5	1.75	40	0.11
2	1	1.75	40	0.11
3	0	3	70	0.04
4	1	3	70	0.11
5	0	3	10	0.06
6	0.5	0.5	40	0.12
7	0	1.75	40	0.05
8	0	0.5	70	0.05
9	0	0.5	10	0.07
10	1	3	10	0.13
11	1	1.75	40	0.10
12	0	0.5	70	0.04
13	1	3	70	0.09
14	1	3	10	0.12
15	1	0.5	10	0.07
16	0.5	1.75	10	0.10
17	0.5	3	40	0.13
18	1	0.5	70	0.08
19	1	0.5	70	0.10
20	0	1.75	40	0.03
21	0	0.5	70	0.03
22	0.5	0.5	40	0.11
23	1	1.75	40	0.10
24	0.5	1.75	10	0.10
25	0.5	1.75	70	0.12
26	0	0.5	10	0.07
27	0.5	1.75	10	0.10
28	0.5	3	40	0.13
29	1	3	10	0.10
30	0	0.5	10	0.06
31	0.5	3	40	0.11
32	0	3	10	0.05
33	0	1.75	40	0.04
34	0.5	1.75	70	0.11
35	0.5	1.75	40	0.11
36	0.5	0.5	40	0.12
37	1	0.5	10	0.07
38	1	0.5	70	0.07
39	1	3	70	0.09
40	0	3	70	0.03
41	0	3	70	0.03
42	0.5	1.75	70	0.11
43	0	3	10	0.05
44	0.5	1.75	40	0.11
45	1	0.5	10	0.09

. Experiments were performed in a randomized order. The corresponding FCC design matrices for Isop/Hex, MeOH/AcOEt, and their observed responses are presented in the supplementary material (Tables S1 and S2).

The experimental data for the response L/DB (for the three types of solvents) were statistically analyzed by analysis of variance, and the significance of the independent variables was evaluated based on their p-value. From the ANOVA results of MeOH/Hex mixture (

Table 4. ANOVA results for the applied FCC experimental design of lipid extraction using MeOH/Hex solvent.

Source	Sum of Squares	Degree of Freedom	Mean Square	F-Value	p-Value
Model	0.0394	9	0.0044	35.20	<0.0001
XA-Polar/non-polar ratio	0.0178	1	0.0178	142.92	<0.0001
XB-Time (Vortex)	0.0005	1	0.0005	3.86	0.0574
XC-Dry biomass/solvent ratio	0.0007	1	0.0007	5.26	0.0280
XAXB	0.0020	1	0.0020	16.23	0.0003
XAXC	0.0004	1	0.0004	3.35	0.0756
XBXC	0.0001	1	0.0001	1.21	0.2794
XA2	0.0134	1	0.0134	107.76	<0.0001
XB2	0.0003	1	0.0003	2.76	0.1057
XC2	0.0003	1	0.0003	2.76	0.1057
Residual	0.0044	35	0.0001
Lack of Fit	0.0018	5	0.0004	4.30	0.0045
Pure Error	0.0025	30	0.0001
Cor Total	0.0437	44

SD = 0.0111, R² = 0.9005, R²_adj = 0.8749, adequate precision = 17.7279.

), it can be seen that the independent variables X_A (polar/non-polar ratio) and X_C (dry biomass/solvent ratio), as well as the interaction between X_A and X_B (vortex time) and the quadratic term X_A², were significant since their p-values were smaller than 0.05. For the Isop/Hex mixture (Table S3), all the independent variables (X_A, X_B, X_C) and the interaction between the time and the dry biomass/solvent ratio were statistically significant. MeOH/AcOEt (Table S4) was significantly affected by a greater number of factors, i.e., all the independent variables (X_A, X_B, X_C), the interaction between X_A and X_C as well as the quadratic models X_A² and X_B².

MLR fitting on the FCC outcome was performed based on the ANOVA results. For both MeOH/Hex and MeOH/AcOEt, the quadratic model had the best fit, while for Isop/Hex, the two-factor integration (2FI) model was used. Adequate accuracy was confirmed through calculation of the signal-to-noise ratio, which was greater than 4.0 for all solvent types. The following MLR model for MeOH/Hex can be used to predict the response values based on any independent variable, into the specified experimental space:

L/DB = +0.062226 + (0.178556 · X_A) − (0.016400 · X_B) + (0.000415 · X_C) + (0.014667 · X_A · X_B) + (0.000278 · X_A · X_C) − (0.000067 · X_B · X_C) − (0.166667 · X_A²) + (0.004267 · X_B²) − (7.40741 · 10⁻⁶ · X_C²),

High correlation coefficient value (R²) illustrates the correlation between estimated and experimental data. All R² values were satisfactory and indicated that the developed models could be used to predict the response (L/DB) with good accuracy, according to the values of the main factors. The correlation graphs of the predicted and the actual values (Figure S1) confirmed the accuracy of each model. Perturbation plots were used to evaluate the effect of an independent variable on the response, where the main factors showing a steep slope or curvature were those affecting responses the most (Figure 2). In the case of MeOH/Hex mixture, the factor with the most significant impact was the polar/non-polar ratio (X_A), compared to the other two independent variables, as obvious from the obtained steep curvature (Figure 2a). For MeOH/Hex, it is evident that there is a range where the response initially increased with the increase in variable X_A and subsequently decreased. In parallel, the vortex time (X_B) and the dry biomass/solvent ratio (X_C) did not affect the response in a similar manner, i.e., the increase in vortex time (X_B) led to a slightly enhanced yield of the extracted lipids (L/DB), whereas the decrease in the solvent volume (X_C) had the opposite effect. According to Figure 2b, for the Isop/Hex mixture, all the variables were significant, which is reflected in their steep slopes. Specifically, the polar/non-polar ratio (X_A) and vortex time (X_B) had a synergistic effect on the response, while the dry biomass/solvent ratio (X_C) had an antagonistic one. Finally, the perturbation plot for the MeOH/AcOEt mixture (Figure 2c) shows that the effect of all factors in relation to the response exhibits steep curvatures. The relation between the independent factors X_A and X_B, and the response, resembles that of factor X_A using MeOH/Hex (Figure 2). Dry-biomass-to-solvent ratio (X_C) displays an antagonistic effect for all three types of solvents.

Response Surface Methodology was used for enhanced visualization of the effects of the independent variables (polar/non-polar ratio, vortex time, and dry biomass/solvent ratio) and their simultaneous interaction on the L/DB response. For this purpose, three-dimensional response surface plots (Figure 3) and two-dimensional contour plots (Figure S2) were designed, keeping constant the dry biomass/solvent ratio (X_C). Since the increase in X_C (i.e., less solvent per yeast biomass) is proportional to the cost-effectiveness of the process, the X_C constant parameter was chosen on the basis of the higher X_C value that would maximize the L/DB optimal area. The X_C value for the MeOH/Hex, Isop/Hex, and MeOH/AcOEt mixtures was chosen to be 40, 10, and 10, respectively. For the aforementioned X_C values, the interactions between the polar/non-polar ratio (X_A), vortex time (X_B), and the response (L/DB) can be readily assessed.

Optimum area for MeOH/Hex mixture was considered where the values of the polar/non-polar ratio were between 0.5 and 0.8 and vortex time was higher than 2.0 h (Figure 3a). In this experimental area, the lipid yield is higher than 0.12 g L/g DB. The plots for the solvent Isop/Hex (Figure 3b) show that the response increases when both independent factors are increasing. This mixture exhibits a less favorable outcome since the area of the highest lipid yields is quite limited. Finally, for the MeOH/AcOEt mixture, when the polar/non-polar ratio is between 0.4 and 0.9, the highest values of the response are observed, reaching a peak at approx. 0.5–0.9 g L/g DB (Figure 3c). In comparison to the vortex time, the polar/non-polar ratio appears to play a more significant role for optimizing the extraction process. It seems that the difference in polarity provides the capability to stimulate an improved separation of polar and non-polar compounds, which leads to enhanced lipid extraction yields. As expected, the use of MeOH/Hex resulted in a high lipid yield (L/DB) in the range of 10 to 70 dry biomass/solvent ratio, whereas for Isop/Hex and MeOH/AcOEt, a higher lipid yield occurred only at dry biomass/solvent ratio of 10. This is a critical finding since high dry biomass/solvent ratio indicates the need for lower quantity of solvents to achieve optimum/high lipid yields.

3.2. Optimization of Independent Variables and Validation Experiments

To determine optimal conditions for every type of solvent, specific targets were imposed on the values of the response. Specifically, for the MeOH/Hex mixture the target for the response was to reach a value of L/DB ≥ 0.100 (10.0%), while for Isop/Hex and MeOH/AcOEt, the target response was set at L/DB ≥ 0.065 (6.5%) and L/DB ≥ 0.090 (9.0%), respectively. These limits were selected based on how frequently these values were observed, based on the results of the experiments presented in the histogram plots of Figure 1. As for the independent variable Χ_A (polar/non-polar ratio), the aim was to reduce the quantity of the polar solvent, while at the same time attaining a satisfactory lipid extraction yield. Taking the aforementioned targets into consideration, the polar/non-polar ratio was set to 0.33, 1.0, and 0.6 for MeOH/Hex, Isop/Hex, and MeOH/AcOEt, respectively. In Figure 4, the overlaying plots depict the area (in yellow) where the set targets are satisfied for the three solvent mixtures.

The next step was to examine the accuracy of the proposed models concerning the prediction of the response value. Selection of the near-optimal conditions was made after investigation of the optimal areas obtained from the response surface and overlay plots (Figure 3 and Figure 4). A decrease in vortex time was desirable to reduce the overall duration of the extraction process, while a simultaneous increase in the dry biomass/solvent ratio corresponds to use of less organic solvent. Therefore, validation experiments were conducted under the near-optimal conditions summarized in Table S5. The acquired experimental and predicted values (Figure 5) exhibited high convergence for all validation points. In particular, the percent convergence for MeOH/Hex was 97.7%, for Isop/Hex 95.0%, and for MeOH/AcOEt 95.8%.

The MeOH/Hex mixture resulted in the highest lipid yield in comparison to Isop/Hex and MeOH/AcOEt (Figure 5), displaying a response value close to the one obtained from the control lipid extraction (0.105). The use of MeOH/AcOEt mixture showed a significantly lower response value compared to the control method, whereas the Isop/Hex was the least effective one, demonstrating a 40% lower lipid yield than the control. It is noted that lipid extraction trials on algal biomass [46] showed a similar trend regarding MeOH/Hex and Isop/Hex performances, with the former organic mixture yielding four times higher lipid recovery in comparison to that of the Isop/Hex solvent mixture. The same study showed that L/DB values—after extraction using MeOH/Hex or MeOH/Chl—were similar, which was also observed in our study, even though yeast and algae species differ, especially concerning the cell wall structure and properties. It is possible that when MeOH/Hex was used, cell wall disruption increases. Another potential contributor to high lipid yields is likely due to the polarity of methanol, which facilitates the disintegration of the lipids from cell membranes and lipoproteins [47], resulting in higher lipids yields.

Table 5. Ratios of the reported maximum lipid yield values of solvent(s) X to the lipid yield values employing the conventional Chl:MeOH solvent mixture.

Type of Solvent (x)	YX/YChl:MeOH	Reference
MeOH/Hex	1.25 a	this study
Isop/Hex	0.99 b
MeOH/AcOEt	1.06 c
Dimethylcyclohexylamine	1.11	[23]
Ethylbutylamine	1.08
Dipropylamine	0.90
n-Hex	0.98	[33]
EtOH	0.47
d-limonene	0.95
p-cymene	0.94
Isoamyl acetate	0.97
Butyl acetate	0.96
AcOEt	0.96
n-Hex	1.0	[34]
EtOH/AcOEt	1.0	[48]
SCCO2	0.86	[36]

Abbreviations: Hex, hexane; MeOH, methanol; Isop, isopropanol; AcOEt, ethyl-acetate; EtOH, ethanol; SCCO₂, supercritical CO₂. ^a Data from Table 3. ^b Data from Table S1. ^c Data from Table S2.

summarizes the findings from relevant studies by comparing the ratios of the maximum reported lipid yield (L/DB) of various solvent(s) with that of the conventional solvent mixture (i.e., Chl/MeOH). It is obvious that the best solvent mixture in this study (i.e., MeOH/Hex) has a higher lipid yield compared to both the conventional solvent mixture of Chl:MeOH as well as to other solvent(s). Compared to other studies employing hexane as the sole solvent (i.e., Imatoukene et al. [33] and Drevillon et al. [34], who report ratios of lipid yields equal to 0.98 and 1.00, respectively), this study reports approx. 25% higher ratio of lipid yield. The addition of MeOH to the hexane extraction solvent seems to favor the lipid recovery efficiency due to the enhanced solvent penetration into the cell and the disruption of the intracellular bonds of lipids with different cell molecules. The efficiency of the MeOH solvent as polar solvent is also obvious by the fact that the MeOH/AcOEt mixture exhibits enhanced lipid yield ratio (i.e., 1.06) compared to the EtOH/AcOEt mixture (i.e., 1.0). MeOH is a smaller molecule and a more polar solvent compared to EtOH, and these properties seem to enhance the aforementioned penetration and disruption efficiency.

The solvent and process parameter combination with the peak performance was found to be the MeOH/Hex mixture in a ratio of 3/5 = 0.6, a ratio of dry biomass/solvent equal to 40, and vortex time of 0.75 h. The selection of the solvent mixture was made on the basis of results from FCC trials that attained the highest lipid yield. Additionally, the value of polar/non-polar ratio was chosen based on the optimal areas identified in the response surface plots of Figure 3. The vortex time was minimized while achieving a high lipid yield, while for the dry biomass/solvent ratio, an average value used in the experiments was selected. Under these conditions, the lipid yield was 0.111 g L/g DB, which is slightly higher than the lipid yield of the control method (i.e., 0.105 g L/g DB). Moreover, the extraction time is 25% less than in the control method (i.e., 0.75 h vs. 1 h) and the required solvent quantity 75% lower (i.e., dry biomass/solvent ratio 40 compared to 10). These optimized conditions were employed in Soxhlet extraction tests to investigate lipid extraction performance on a 50-fold increased volume scale.

3.3. Soxhlet Extraction

The Soxhlet extraction experiments with lyophilized biomass and MeOH/Hex led to results with high lipid yields, as shown in Figure 6. Increasing the reflux cycles results in increased lipid yields, which reached a value of 0.276 g L/g DB after 40 reflux cycles. This yield was obtained without any additional pretreatment such as an agitation step. The lowest lipid yield was observed in the case of five reflux cycles, which was slightly lower than the yield resulting from the control method (i.e., 0.121 g L/g DB). It is very promising that the lipid yields for 5 to 10 reflux cycles (which simulate the extraction equilibrium stages of the continuous industrial process) are comparable to the control method, whereas if the reflux cycles increase further (i.e., at 20 reflux cycles), the lipid yield increases by 70%, compared to the control method.

3.4. Effect of Different Solvents on FAME Composition

Aiming to assess the differences in lipid composition related to the organic solvent mixtures, transesterification of the extracted lipids from the validation experiments was carried out. FAMEs were analyzed through GC-FID chromatography to determine the relative percentage of each fatty acid (FA), calculated by internal normalization of the chromatographic peak area (

Table 6. Fatty acid (FA) composition of the intracellular lipids of Yarrowia lipolytica, after extraction using different types of organic solvents and transesterification. Number of replicates (N = 3).

FAME	MeOH/Hex	Isop/Hex	MeOH/AcOEt	MeOH/Chl
Palmitic C16:0 (%)	16.8 ± 0.3 a	22.7 ± 3.4	22.5 ± 0.8 a	26.4 ± 2.9
Palmitoleic C16:1 (cis-9) (%)	5.3 ± 0.1	4.5 ± 0.9	4.1 ± 0.5	2.7 ± 1.7
Stearic C18:0 (%)	12.7 ± 0.7 ab	17.9 ± 4.4	19.1 ± 1.2 a	17.8 ± 0.6 b
Oleic C18:1 (cis-9) (%)	51.5 ± 1.0 ab	41.8 ± 5.5	42.0 ± 1.2 a	43.6 ± 1.5 b
Linoleic C18:2 (cis-9, 12) (%)	13.7 ± 0.3 a	13.1 ± 1.5	12.3 ± 0.1 a	9.5 ± 1.8

^ab Letters within rows indicate pairs with significant differences on the relative percentage of each fatty acid chain after one-way ANOVA and post hoc tests (level of significance, 0.05). All values are expressed as mean ± S.D.

).

From the results of Table 6 and Figure 7, it is obvious that oleic acid (C18:1) was attaining the highest concentration of the total fatty acid composition, followed by the palmitic (C16:0), stearic (C18:0), and, in some cases, linoleic acid (C18:2). In general, the fatty acid percentages were not considerably influenced by the use of different organic solvents during extraction; however, there was a notable difference when MeOH/Hex was used, i.e., oleic acid recovery was favored in relation to the other solvents’ mixtures by approx. 10%, while palmitoleic (C16:2) and linoleic (C18:2) percentages were slightly enhanced as well. Attaining high percentage of unsaturated fatty acids is critical since it is directly linked to enhanced cold flow properties of biodiesel. Therefore, the mixture of MeOH/Hex is a promising solvent for further research. To distinguish the statistically significant interactions between organic solvents and different fatty acid chains, one-way ANOVA and post hoc tests were conducted. Major differences appeared between MeOH/Hex and MeOH/AcOEt solvent related to palmitic C16:0 (p-value = 0.001), stearic C18:0 (p-value = 0.005), oleic C18:1 (p-value = 0.002), and linoleic C18:2 acid (p-value = 0.008). Furthermore, there was a significant difference between MeOH/Hex and MeOH/Chl solvents concerning stearic C18:0 (p-value = 0.002) and oleic C18:1 acid (p-value = 0.006).

Lipid profile during the Soxhlet extraction experiments was not affected by increasing the reflux cycles (Figure 8). Moreover, traces of two other fatty acids were observed (less than 2%) after scale-up, namely heptadecenoic (C17:1) and lignoceric (C24:0) acid.

4. Conclusions

Lipid extraction was performed from the yeast Y. lipolytica, employing different solvent mixtures (i.e., MeOH/Hex, Isop/Hex, MeOH/AcOEt). Optimization of critical extraction parameters, namely polar/non-polar ratio, vortex time, and dry biomass-to-solvent ratio, was conducted through a DoE approach. A significant amount of data was obtained concerning the effect of extraction parameters and type of solvent mixture on the extraction yield of Y. lipolytica lipids. The identified near-optimum conditions were: MeOH/Hex ratio 3:5, vortex time 0.75 h, and dry biomass/solvent ratio 40, resulting in a lipid yield equal to 0.111 g L/g DB. Applying these optimal process conditions in tests on a much larger scale in Soxhlet apparatus led to encouraging lipid yield values with minor differences in the composition of the fatty acids compared to the reference method. Moreover, the MeOH/Hex mixture seems to favor the extraction of monounsaturated fatty acids, which would result in biodiesel with improved cold flow properties but possibly lower cetane number and higher heating values (HHV).

This work provides useful data and insights for the development of a sustainable process for extraction of Y. lipolytica lipids on a large (i.e., industrial) scale. Hexane is a solvent currently used in the plant oil industry. Similarly, methanol is another bulk (commodity) solvent, widely used for the transesterification process in the biodiesel industry. Considering these attributes, the MeOH/Hex mixture appears to be a promising solvent for industrial applications in a Y. lipolytica lipid extraction plant.