Efficient Conversion of Glycerol to Ethanol by an Engineered Saccharomyces cerevisiae Strain

ABSTRACT

Glycerol is an eco-friendly solvent that enhances plant biomass decomposition via glycerolysis in many pretreatment methods. Nonetheless, inefficient conversion of glycerol to ethanol by natural Saccharomyces cerevisiae limits its use in these processes. In this study, we have developed an efficient glycerol-converting yeast strain by genetically modifying the oxidation of cytosolic NAD (NADH) by an O₂-dependent dynamic shuttle and abolishing both glycerol phosphorylation and biosynthesis in S. cerevisiae strain D452-2, as well as by vigorous expression of whole genes in the dihydroxyacetone (DHA) pathway (Candida utilis glycerol facilitator, Ogataea polymorpha glycerol dehydrogenase, endogenous dihydroxyacetone kinase, and triosephosphate isomerase). The engineered strain showed conversion efficiencies (CE) up to 0.49 g ethanol/g glycerol (98% of theoretical CE), with a production rate of >1 g liter⁻¹ h⁻¹ when glycerol was supplemented in a single fed-batch fermentation in a rich medium. Furthermore, the engineered strain converted a mixture of glycerol and glucose into bioethanol (>86 g/liter) with 92.8% CE. To the best of our knowledge, this is the highest reported titer of bioethanol produced from glycerol and glucose. Notably, we developed a glycerol-utilizing transformant from a parent strain which cannot utilize glycerol as a sole carbon source. The developed strain converted glycerol to ethanol with a productivity of 0.44 g liter⁻¹ h⁻¹ on minimal medium under semiaerobic conditions. Our findings will promote the utilization of glycerol in eco-friendly biorefineries and integrate bioethanol and plant oil industries.

IMPORTANCE With the development of efficient lignocellulosic biorefineries, glycerol has attracted attention as an eco-friendly biomass-derived solvent that can enhance the dissociation of lignin and cell wall polysaccharides during the pretreatment process. Coconversion of glycerol with the sugars released from biomass after glycerolysis increases the resources for ethanol production and lowers the burden of component separation. However, low conversion efficiency from glycerol and sugars limits the industrial application of this process. Therefore, the generation of an efficient glycerol-fermenting yeast will promote the applicability of integrated biorefineries. Hence, metabolic flux control in yeast grown on glycerol will lead to the generation of cell factories that produce chemicals, which will boost biodiesel and bioethanol industries. Additionally, the use of glycerol-fermenting yeast will reduce global warming and generation of agricultural waste, leading to the establishment of a sustainable society.

INTRODUCTION

The demand for ethanol has increased remarkably because of its use in sanitizers in the COVID-19 pandemic, necessitating enhancement of its production via various pathways, including glycerol (C₃H₈O₃) fermentation. In the past decade, glycerol-producing industries, particularly the biodiesel industry, have produced substantial amounts of glycerol (1). Natural Saccharomyces cerevisiae does not ferment glycerol, although glycerol has two additional hydrogen atoms for reduction during its catabolism compared with fermentable sugars (2, 3). In addition, glycerol as a carbon source is poorly utilized, primarily via the glycerol-3-phosphate (G3P) pathway, which is composed of glycerol kinase (ScGUT1) and flavin adenine dinucleotide (FAD)-dependent mitochondrial glycerol-3-phosphate dehydrogenase (ScGUT2) (4). S. cerevisiae prefers biosynthesizing glycerol, as it mitigates osmotic stress and optimizes redox balance (5). In contrast, glycerol catabolism is subject to repression and transcriptional regulation of glucose metabolism via several respiratory factors, as well as ScGUT1 and ScGUT2 (6–8).

Several groups have attempted to ferment glycerol by overexpressing the native oxidative pathway (dihydroxyacetone [DHA] pathway) in S. cerevisiae. In a previous study, overexpression of the endogenous glycerol dehydrogenase (ScGCY1) with dihydroxyacetone kinase (ScDAK1) led to the production of 0.12 g ethanol/g glycerol (g^e/g^g), and the production rate was 0.025 g liter⁻¹ h⁻¹ (9). On the other hand, the importance of glycerol utilization as a carbon source for yeast cell cultivation prompted the assessment of 52 S. cerevisiae strains, which revealed the intraspecies diversity of yeasts, ranging from good glycerol growers to nongrowers. Growth phenotype in these strains is controlled by quantitative gene traits (10). Reports have shown that many genes, such as UBR2, encoding cytoplasmic ubiquitin-protein ligase E3, link GUT1 with growth on glycerol in a synthetic medium in the absence of supplements (11). Heterologous replacement of G3P by the DHA pathway (glycerol dehydrogenase from Ogataea polymorpha [OpGDH] and ScDAK1) and the insertion of the glycerol facilitator from Cyberlindnera jadinii (CjFPS1) resulted in the restoration of growth characteristics similar to those of the parental strain or even higher (12). This replacement in a parent strain capable of growing on glycerol increased ethanol production from 0.17 to 0.34 g^e/g^g in shaking flask cultures supplemented with synthetic medium under a hypoxic atmosphere; the maximum titer reached 15.7 g/liter after 144 h (13). Furthermore, the methylotrophic yeast O. polymorpha has been modified for bioethanol production from glycerol by overexpressing multiple genes involved in either the DHA or G3P pathway, FPS1, pyruvate decarboxylase (PDC) 1, and alcohol dehydrogenase (ADH) 1. Nonetheless, the overall titer was 10.7 g/liter ethanol, with a conversion efficiency of 0.132 g^e/g^g (14). However, there are no previous reports on native or genetically engineered S. cerevisiae strains that efficiently ferment or convert glycerol to ethanol.

Glycerol has been applied to the pretreatment of recalcitrant lignocellulosic biomass for enzymatic saccharification and fermentation. We have previously reported microwave-assisted pretreatments of lignocelluloses in aqueous glycerol (15) and an improvement in the process by using alum as a Lewis acid catalyst (16). However, in this process the separation of the released sugars from glycerol is costly and affects its overall feasibility. The development of an S. cerevisiae strain capable of efficiently fermenting glycerol with glucose will eliminate the burden of solvent separation and create a new system that can simultaneously ferment lignocelluloses and oil plant-derived glycerol produced in biodiesel, vegetative oil, and soap industries. Therefore, we have aimed to model S. cerevisiae by redirecting glycerol traffic to bioethanol production in the presence of glucose via systematic metabolic engineering, which has been outlined in Fig. 1. Here, we report the generation of a hyper-glycerol-converting yeast strain by using genome editing and its potential in producing ethanol from a mixture of glucose and glycerol, aiming at integrating bioethanol and plant oil industries.

FIG 1.

Scenario of a biorefinery integrating bioethanol and plant oil industries using a hyper S. cerevisiae transformant capable of efficiently converting glycerol to ethanol. The highlighted circle within the yeast cell shows the metabolic engineering of glycerol pathways for efficient conversion to ethanol. Herein, NAD⁺, the key cofactor, is replenished by O₂-dependent water-forming NADH oxidase from Lactococcus lactis subsp. lactis Il1403 (LlnoxE) and all the enzymes involved in the DHA pathway are overexpressed. In addition, we knocked out glycerol-3-phosphate dehydrogenase 1 (ScGPD1) and glycerol kinase 1 (ScGUT1) for enhancing the influxes from glycerol to ethanol. The DHA pathway includes the following genes: those for dihydroxyacetone kinases (ScDAK) 1 and 2, triosephosphate isomerase (ScTPI1), glycerol dehydrogenase from Ogataea polymorpha (OpGDH), and glycerol facilitator from Candida utilis (CuFPS1).

RESULTS AND DISCUSSION

Boost in the conversion of glycerol to ethanol by the expression of a glycerol dehydrogenase gene.

To overcome the limitations of glycerol flux in S. cerevisiae, we first overexpressed the OpGDH gene (17) at the URA3 locus of strain D452-2 (Table 1). We used an efficient expression system (18) which consisted of the glyceraldehyde 3-phosphate dehydrogenase promoter (TDH3p) and a mutated DIT1 terminator (DIT1td22). In contrast to the reference D452-2 strain, which lacks GDH activity, 3.5 mU/mg GDH activity was observed in the case of the transformant SK-FGG1 (Table 1), which harbored OpGDH (Table 2). GDH activity in strain SK-FGG1 was around 10 times higher than what was reported previously (19), which could be because of the expression system (18) used in this study. Consequently, glycerol consumption increased from 8% to 58% by overexpression of the OpGDH gene, and the production of ethanol increased from 4.5 g/liter (in the reference strain) to 11.8 g/liter by strain SK-FGG1, i.e., a 2.6-fold increase, reaching the productivity of 0.28 g^e/g^s after 26 h (Fig. 2). Moreover, GDH activity in 79 mM HEPES buffer (pH 7.4) was 3.75-fold higher than that in 167 mM potassium phosphate buffer (pH 7.5). Aßkamp et al. reported that external mitochondrial NADH dehydrogenase (NdeI) is involved in glycerol metabolism in the presence of oxygen as the final electron acceptor (20), although cytosolic NAD⁺ is partially regenerated. We found that under semiaerobic conditions, the NADH/NAD⁺ ratio increased by 45% compared with that for the reference D452-2 strain (Table 2). Therefore, overexpression of OpGDH in S. cerevisiae is the first step in the conversion of glycerol into alcohol. Nonetheless, the overexpression of OpGDH alone in the SK-FGG1 strain was not sufficient for the efficient conversion of glycerol to ethanol (Fig. 2). In contrast, the conversion efficiency of strains overexpressing endogenous glycerol dehydrogenase (ScGCY1) alone or in combination with all enzymes of the DHA pathway was lower than that of the strains overexpressing OpGDH (data not shown).

TABLE 1.

Characteristics of the Saccharomyces cerevisiae strains generated in this study

S. cerevisiae strain	Relevant genotype	Reference
D452-2	MATa leu2 his3 ura3 can1	34
D452-2URA3	D452-2, URA3::TDH3 promoter and DIT1d22 terminator	This study
D452-2URA3-AUR1-C	D452-2URA3, AUR1::AUR1-C	This study
SK-FGG1	D452-2, URA3::TDH3-OpGDH-DIT1d22	This study
SK-FGG2	SK-FGG1, ΔGPD1::TDH3-LlnoxE-DIT1d22	This study
SK-FGG3	SK-FGG2, AUR1::PGKp-CuFPS1-RPL41Bt; PGKp-ScTPI1-PGKt; PGKp-ScDAK2-PGKt; PGKp-ScDAK1-PGKt	This study
SK-FGG4	SK-FGG3, ΔGUT1::TEFp-CuFPS1-CYC1t; TYS1p-OpGDH-ATP15t; TDH3-ScDAK1-DIT1d22; FBA1p-ScTPI1-TDH3t	This study

TABLE 2.

Specific activities of GDH, DAK, and TPI1 and NADH/NAD⁺ ratio with intracellular concentrations in the recombinant strains used in this study^a

Relevant strain	Enzyme activity (U/min/mg) of cell extracted protein(s)			Intracellular concn		NADH/NAD+ ratio
	GDH (mmol)	DAKs (μmol)	TPI1 (mmol)	NADH (μM)	NAD+ (μM)
WT-URA3	−0.6 ± 0.1b	3.9 ± 0.2	15.3 ± 0.2	0.17 ± 0.01	0.36 ± 0.01	0.47 ± 0.01
SK-FGG1	3.5 ± 0.2	NMc	12.6 ± 0.9	0.26 ± 0.01	0.38 ± 0.02	0.68 ± 0.02
SK-FGG2	5.3 ± 0.2	NM	5.4 ± 0.03	0.34 ± 0.02	0.35 ± 0.01	0.99 ± 0.02
SK-FGG3	5.8 ± 0.3	7.2 ± 0.2	5.7 ± 0.2	0.17 ± 0.01	0.52 ± 0.01	0.33 ± 0.02
SK-FGG4	5.9 ± 0.2	18.3 ± 0.3	18.0 ± 0.6	0.11 ± 0.01	0.63 ± 0.01	0.18 ± 0.01

^aError values represent SD (n = 2). GDH, glycerol dehydrogenase; DAK, dihydroxyacetone kinase; TPI1, triosephosphate isomerase.

^bAbsorption at 340 nm decreased over time, which indicates the lack of activity of native glycerol dehydrogenase.

^cNM, not measured.

FIG 2.

Comparison of the time course of glycerol-glucose fermentation between the S. cerevisiae strains used in this study. (A) Glucose consumption (solid lines) and glycerol consumption (dashed lines); (B) ethanol production. Fermentations were run in triplicates under semiaerobic conditions in 200-ml Erlenmeyer flasks with 20 ml liquid medium at 30°C and shaking at 180 rpm. YP medium was supplemented with 15 g/liter glucose and 70 g/liter glycerol. Error bars represent the standard deviations from the mean (SD). Blue, reference ancestral strain; black, SK-FGG1 strain; orange, SK-FGG2 strain; green, SK-FGG3 strain; red, SK-FGG4 strain.

Retrofit NADH oxidation pathway for promoting glycerol oxidation.

Heterologous expression of water-forming NADH oxidase (noxE) in S. cerevisiae has been used to decrease NADH production during the fermentation of glucose to ethanol, 2,3-butanediol, and acetoin (21–25). The recombinant strain SK-FGG1 (Table 1) was transformed to express Lactococcus lactis noxE (LlnoxE) to compensate for the deficiency of NAD⁺ (Fig. 1). The integration of LlnoxE and the replacement of glycerol-3-phosphate dehydrogenase (ScGPD1) in the SK-FGG2 strain (Table 1) further increased the rate of conversion of glycerol to ethanol from 11.8 to 13.3 g/liter after 26 h and then to 14.4 g/liter after 32 h. The expression of LlnoxE in the SK-FGG1 strain, which harbored OpGDH, increased the conversion efficiency by 128% (Fig. 2). The efficiency of strain SK-FGG2 (Table 1) was higher than those of the strains harboring other native oxidizing shuttles, ScGPD and/or Nde (20), indicating the higher efficiency of the NADH oxidation system for glycerol metabolism due to LlnoxE. Surprisingly, GDH activity increased by 51%, which could be due to the deletion of ScGPD1. On the other hand, the activity of triosephosphate isomerase 1 (ScTPI1) decreased by 65%, probably due to the interference by LlNoxE. The effectiveness of oxidation of NADH by LlNoxE in the strain SK-FGG2 (Table 1) was further confirmed by the restoration of the NADH/NAD⁺ ratio (Table 2). Thus, the NADH/NAD⁺ ratio was restored but other limitations in metabolic flow markedly lowered the conversion efficiency of glycerol to ethanol. So, we further controlled glycerol traffic to ethanol by boosting the expression of the DHA pathway.

Overexpression of all genes encoding the enzymes of the DHA pathway.

Considering the conversion efficiency (glycerol to ethanol) of strain SK-FGG2, we hypothesized that other genes in the DHA pathway (ScTPI1, ScDAKs, and ScFPS1) may affect the flow of glycerol to ethanol. Dihydroxyacetone phosphate (DHAP) is a hub for several pathways, including phospholipid and methylglyoxal biosynthesis (26, 27). Overexpression of ScTPI1 increases the metabolic flow of DHAP to glyceraldehyde-3-phosphate (Fig. 1), thereby suppressing flow to other pathways through the bifurcation point, DHAP. Although the pivotal role of ScTPI1 is evident from another study on glycerol production (28), there has been no report of ScTPI1 overexpression aimed at enhancing ethanol fermentation.

The expression of OpGDH may cause the accumulation of dihydroxyacetone (DHA) (29), a cytotoxic metabolite, if the metabolic flow from DHA is not sufficient to pump out the intermediate. Overexpression of ScDAK1 and CjFPS1 may resolve this problem and enhance the rate of conversion of glycerol to ethanol (9, 12, 13). In this study, we strengthened the metabolic flow from DHA to DHAP by overexpressing ScDAK1 and ScDAK2, which have considerably low K_m(DHA-ATP)s. Furthermore, we overexpressed ScTPI1 to increase the metabolic flow from DHAP to glyceraldehyde-3-phosphate. We also expressed the gene encoding glycerol facilitator from Candida utilis (CuFPS1) to strengthen the cellular uptake of glycerol. Thus, we constructed the SK-FGG3 strain by integrating a set of genes (CuFPS1, ScDAK1, ScDAK2, and ScTPI1) (Table 1).

To confirm the construction of the metabolic pathway, activities of the key enzymes—GDH, DAKs, and TPI1—were measured, along with the concentration ratio of NADH and NAD⁺ (Table 2). We found that the activity of ScDAKs in strain SK-FGG3 increased by 185% in comparison to that of the native strain. High GDH activity was also found in strains SK-FGG1, SK-FGG2, and SK-FGG3 as expected, while ScTPI1 activity in these strains decreased in comparison to that of the wild-type strain (Table 2). Thus, an increase in ScDAK activity supports the fact that the influx of glycerol was controlled in the yeast model as per our objective. At the same time, it became evident that influx through ScTPI1 should be strengthened for improving the conversion of glycerol to ethanol. The expression of CuFPS1 is also one of the essential steps for the conversion of glycerol to ethanol, as it enhances glycerol uptake. Klein et al. reported the heterologous expression of different glycerol facilitators from yeast species and showed that superior glycerol consumption considerably enhances the maximal specific growth rate of the microbial strain in the presence of glycerol (30).

Reinforcement of the DHA pathway while abolishing the native G3P route.

Our data for the transformants, SK-FGG1, SK-FGG2, and SK-FGG3 (Table 1), suggested that an increase in the activities of all enzymes involved in the DHA pathway increases ethanol production (Fig. 2), Furthermore, it was found that the activity of ADH in cells grown on a rich medium containing glycerol supplemented with yeast extract/peptone (YP) was 10 times higher than those grown in the presence of glucose (31). A previous study suggests that the availability of ATP for ScDAKs is a limiting factor for the DAH-DHAP pathway in the presence of ScGUT1 (12). Moreover, the consolidation of glycerol catabolism can be achieved only through the oxidative pathway (11, 12). Therefore, we hypothesized that the introduction of another copy of the DHA pathway genes with concomitant knockout of ScGUT1 could enhance the glycerol conversion efficiency. We overexpressed another copy each of CuFPS1, OpGDH, ScDAK1, and ScTPI1 by constructing an M1 module (Fig. 3), which was placed under the control of highly constitutive expression systems (18, 32, 33). The expression system enhanced the activities of ScDAK and ScTPI1 by 256% and 316%, respectively, compared to those in strain SK-FGG3 (Table 2). Interestingly, strain SK-FGG4, which contained the M1 module, showed outstanding conversion efficiencies; the glycerol consumption rate and ethanol productivity reached 2.6 g liter⁻¹ h⁻¹ and 1.38 g liter⁻¹ h⁻¹, respectively, leading to the accumulation of 36 g/liter ethanol. The conversion efficiency was 0.44 g^e/g^s (Fig. 2). These results support the idea that glycerol influx was strengthened through the upregulation of the genes involved in the DHA pathway along with the restoration of NAD⁺. Further improvement can be expected by adding more copies of the genes encoding the enzymes of the DHA pathway.

FIG 3.

Module M1 for replacing ScGUT1 via homologous recombination using multiplex CRISPR Cas 9.

Higher conversion efficiency of glycerol was observed in a fed-batch system using strain SK-FGG4 when around 110 g/liter glycerol was fed to the culture at a high agitation speed, 200 rpm. The conversion efficiency reached 0.49 g^e/g^g (98%), with a production rate of >1 g liter⁻¹ h⁻¹ ethanol after consumption of 82.5 g/liter glycerol within 40 h. During this conversion, 1.1 g/liter acetic acid was produced as a by-product (Table 3, condition A). Thereafter, the conversion efficiency dropped to 97% at the end of cultivation (48 h) (Fig. 4A). Rapid reconsumption of the produced ethanol by the engineered strain SK-FGG4 was observed at the end of the cultivation (Fig. 2B). Importantly, no suppressive effects were observed when 22.5 g/liter glucose was added in addition to glycerol (Table 3, condition B, and Fig. 4B). Furthermore, supplementation of around 100 g/liter glycerol, as the second fed batch, did not change the conversion efficiency after 30 h (Table 3, condition C, and Fig. 4C). To obtain further evidence for the fact that the engineered strain was not substantially repressed by glucose metabolism (5–7), the glucose dosage was doubled to 45 g/liter with the reduction of the initial glycerol concentration by 25% under the condition (Table 3, condition C, and Fig. 4C) to 82 g/liter before starting the second fed-batch cultivation of glycerol (Table 3, condition D, and Fig. 4D). At this stage, economically feasible ethanol productivity and concentration were achieved. The concentration of glycerol and glucose in the conversion system (condition D) was comparable to that of the carbon source from lignocellulosic biomass obtained by glycerolysis pretreatment (16). Thus, the engineered strain SK-FGG4 displayed simultaneous conversion of glycerol and glucose with an accumulation of >86 g/liter bioethanol when the additional fed batch of glycerol was applied, indicating exceptional conversion capacity. The efficiency of ethanol conversion decreased to 92%, but cell density increased by 31% (Table 3, condition D, and Fig. 4D).

TABLE 3.

Semiaerobic fermentation characteristics at maximum efficiency by strain SK-FGG4 at high initial concentrations of glycerol under various conditions^a

Parameter	Fermentation condition and product value
	One fed batchb (−glucose) (condition A)	One fed batchb (+glucose) (condition B)	Two fed batchesc (+glucose) (condition C)	Two fed batchesc (+glucose) (condition D)
Initial glycerol concn (g/liter)	111.0 ± 1.0	108.2 ± 1.0	108.2 ± 1.0	82.8 ± 1.0
Glycerol fed-batch concn (g/liter)	NSd	NS	94.0 ± 1.0	99.0 ± 2.5
Conversion time (h)	40 ± 0.30	40 ± 0.30	96 ± 0.30	96 ± 0.30
Total consumed glycerol (g/liter)	82.5 ± 1.6	75.8 ± 1.4	141.3 ± 2.7	141.7 ± 5.0
Total consumed glucose (g/liter)	NS	22.5 ± 1.2	22.2 ± 0.7	44.7 ± 1.0
Rate of consumption (g liter−1h−1)	2.1 ± 0.2	2.4 ± 0.1	1.7 ± 0.1	1.9 ± 0.1
Ethanol yield (g/liter)	40.4 ± 0.1	48.2 ± 1.2	79.7 ± 1.1	86.5 ± 2.5
Rate of ethanol production (g liter−1 h−1)	1.0 ± 0.01	1.2 ± 0.02	0.83 ± 0.01	0.9 ± 0.02
Efficiency of ethanol production (ge/gs)e	0.49 ± 0.01	0.49 ± 0.01	0.48 ± 0.01	0.46 ± 0.01
Efficient/theoretical (%)	97.0 ± 0.6	98.0 ± 0.5	97.2 ± 0.6	92.8 ± 0.4
Acetic acid accumulation (g/liter)	1.1 ± 0.02	1.1 ± 0.02	1.7 ± 0.1	2.5 ± 0.02
Total conversion (gp/gs)f	0.5 ± 0.01	0.5 ± 0.01	0.5 ± 0.01	0.48 ± 0.01
Final cell density (OD600)	10.6 ± 0.2	11.6 ± 0.2	11.6 ± 0.3	15.2 ± 0.3

^aFermentation was conducted in triplicates in semiaerobic conditions (20 ml culture/200-ml flask). Data represent the averages ± SD. Conditions were as follows: condition A, a single dose of glycerol; condition B, condition A in the presence of glucose; condition C, condition B with second fed batch of glycerol; and condition D, condition C with double glucose concentration and initial glycerol concentration reduced by 25%.

^bInitial fed batch.

^cInitial fed batch followed by 2nd glycerol fed batch after 30 h.

^dNS, not supplemented.

^eg^e/g^s, gram of ethanol per gram of consumed substrate.

^fg^p/g^s, gram of product per gram of substrate.

FIG 4.

Time-course profiles for semiaerobic fermentation of glycerol under different conditions using strain SK-FGG4. Shown are fermentation of one fed batch of glycerol (ca. 110 g/liter) (A) in the presence of (ca. 22 g/liter) glucose (B), fermentation of two fed batches of glycerol (ca. 110 and 100 g/liter) in the presence of (ca. 22 g/liter) glucose (C), and fermentation of two fed batches of glycerol (ca. 82 and 100 g/liter) in the presence of (ca. 45 g/liter) glucose (D). Blue, glycerol consumption; black, glucose consumption; red, ethanol production; orange, cell density (OD₆₀₀); green, acetic acid production. Fermentation was carried out in triplicates in 200-ml Erlenmeyer flasks with 20 ml liquid culture in each flask, with orbital shaking at 200 rpm at 30°C. The data represent the means of three independent experiments, and error bars represent the SD.

Conversion of glycerol to ethanol as a sole carbon source.

To avoid the interference of any nutrients in the complex YP medium, we studied the efficiency of the SK-FGG4 strain for converting glycerol as a sole carbon source in yeast nitrogen base (YNB) minimal medium supplemented with 20 mg/liter leucine and histidine in a single-batch fermentation. The conversion experiments were performed under four different conditions of oxygen availability (anaerobic, microaerobic, semiaerobic, and aerobic), as described in Materials and Methods. No significant growth or production of ethanol was observed under strictly anaerobic conditions (Table 4). Under the microaerobic condition, the strain consumed 32.1 g glycerol after 48 h, with a consumption rate of 0.67 g liter⁻¹ h⁻¹ and a production rate of 0.26 g liter⁻¹ h⁻¹. Acetic acid accumulated at a rate of 0.65 g/liter under this condition, and the efficiency of ethanol conversion was 0.39 g^e/g^g (Table 4 and Fig. 5). Glycerol was consumed more rapidly under semiaerobic conditions, and its consumption rate exceeded 1 g liter⁻¹ h⁻¹; the production rate and concentration of ethanol also increased to 0.44 g liter⁻¹ h⁻¹ and 21 g/liter, respectively. Under this condition, 2.9 g/liter acetic acid was accumulated, and the optical density at 600 nm (OD₆₀₀) of the cells was nearly doubled (Table 4 and Fig. 5). Compared to the case under semiaerobic conditions, ethanol was produced at a slightly lower rate, 0.37 g^e/g^g, with higher accumulation of acetic acid (3.3 g/liter after 48 h), under aerobic conditions (Table 4 and Fig. 5D).

TABLE 4.

Fermentation characteristics of strain SK-FGG4 in the presence of glycerol as the sole carbon source at different oxygen concentrations^a

Parameter	Fermentation conditions and product value
	Strict anaerobic (20:50)b	Microaerobic (20:100)b	Semiaerobic (20:200)b	Aerobic (20:300)b
Fermentation time (h)	48 ± 0.30	48 ± 0.30	48 ± 0.30	48 ± 0.30
Total consumed glycerol (g/liter)	0.3 ± 0.0	32.1 ± 1.8	54.0 ± 6.1	52.5 ± 01.1
Rate of consumption (g liter−1 h−1)	0.0 ± 0.0	0.67 ± 0.04	1.12 ± 0.13	1.1 ± 0.02
Total ethanol produced (g/liter)	0.0 ± 0.0	12.5 ± 0.8	21.0 ± 2.0	19.3 ± 0.4
Rate of ethanol production (g liter−1 h−1)	0.0 ± 0.0	0.26 ± 0.02	0.44 ± 0.05	0.40 ± 0.01
Efficiency of ethanol production (ge/gs)	0.0 ± 0.0	0.39 ± 0.01	0.39 ± 0.01	0.37 ± 0.01
Efficient/theoretical (%)	0.0 ± 0.0	78.0 ± 0.02	78.0 ± 0.2	74.0 ± 0.2
Acetic acid accumulation (g/liter)	0.0 ± 0.0	0.7 ± 0.1	2.90 ± 0.2	3.30 ± 0.04
Total conversions (gp/gs)	0.0 ± 0.0	0.41 ± 0.1	0.44 ± 0.1	0.43 ± 0.1

^aFermentation was performed in triplicate, with each flask containing YNB medium supplemented with 69.0 ± 0.7 g/liter glycerol and 20 mg/liter leucine and histidine. Values represent the averages ± SD.

^bOxygen availability in the flasks was measured as the ratio to yeast cells in the liquid medium; (liquid medium: flask volume) at 30°C with shaking at 200 rpm.

FIG 5.

Time-course profiles of glycerol fermentation (ca. 69 g/liter) as a sole carbon source in YNB minimal medium supplemented with 20 mg/liter leucine and histidine using strain SK-FGG4. Fermentation was conducted at different oxygen levels, which were achieved by varying the ratio between the amount of the liquid culture and the size of the Erlenmeyer flasks. Microaerobic (20 ml culture and 100-ml flask, blue lines), semiaerobic (20 ml culture and 200-ml flask, orange lines), and aerobic (20 ml culture and 300-ml flask, black lines) conditions were used. For the strictly anaerobic condition, N₂ was additionally flushed (20 ml: 50 ml, red lines). (A) Glycerol consumption; (B) ethanol production; (C) OD₆₀₀; (D) acetic acid accumulation. Experiments were carried out in triplicates with orbital shaking at 200 rpm at 30°C. The data represent the means of three independent experiments, and error bars represent the SD.

Growth profile of strain SK-FGG4 on glycerol-supplemented minimal medium.

The growth rate of strain SK-FGG4 was 0.045 OD₆₀₀ unit/h in YNB minimal medium supplemented with 69 g/liter glycerol and 200 mg/liter leucine and histidine (see Fig. S1 in the supplemental material). This result clearly demonstrated that using comprehensive metabolic engineering, the metabolic flow began from glycerol intake because the reference D452-2 strain was derived in part from the S288C strain (Fig. S2) (34–36), which cannot grow in a medium containing glycerol as a sole carbon source (Fig. S1). Swinnen et al. established that strain S288C could not grow in a synthetic medium with glycerol as the sole carbon source without supplements (10). Amino acids play a crucial role in cell growth (37). The biosynthesis of NAD⁺ from tryptophan via the kynurenine pathway is well known (38, 39). In addition, nicotinic acid, nicotinamide, quinolinic acid, and nicotinamide riboside can salvage NAD⁺ biosynthesis. In this context, under anaerobic conditions, S. cerevisiae exhibits auxotrophy for nicotinic acid (38). Although the YNB minimal medium contains 400 μg of nicotinic acid, we did not find a significant growth of SK-FGG4 in the medium containing glycerol as a sole carbon source under strict anaerobic conditions; the OD₆₀₀ increased from 4.3 to 4.5 and then stabilized (Fig. 5C). Therefore, further studies are needed to elucidate the differences between the auxotrophic growth requirements in the glycerol medium and those listed in previous studies that used glucose as a carbon source (38).

Insights on glycerol fermentation using minimal and rich media.

One of the remarkable differences in the conversion of glycerol to ethanol using minimal and rich media is the increased accumulation of acetic acid in the YNB minimal medium (from 14 mg to 53 mg acetic acid/g glycerol) under semiaerobic conditions (Tables 3 and 4). This could be because of a competition between aldehyde dehydrogenase (ScALD) and OpGDH for NAD⁺; however, further studies are required to validate this hypothesis. The inability to ferment glycerol under strict anaerobic conditions (Table 4) indicates the lack of renovation shuttles in the absence of oxygen and oxidizing agents. Hence, the recycling of NADH/NAD⁺ is essential for robust oxidation via the DHA pathway and efficient utilization of glycerol to produce bioethanol or other biobased chemicals. We are currently designing metabolic engineering strategies for glycerol fermentation under anaerobic conditions, using the high reduction ability of glycerol to improve the fermentation efficiency of lignocellulose sugars. In addition, we intend to determine the amino acids that may play a significant role in defined media to reach production level similar to that of complex medium.

In summary, we developed an efficient model that comprehensively controlled glycerol traffic during ethanol production in S. cerevisiae. This model, based on systematic metabolic engineering, included integration of the following elements: (i) robust expression of all genes involved in the DHA pathway; (ii) predominant glycerol oxidation by an oxygen-dependent dynamic of the water-forming NADH oxidase LlNoxE, which controls the reaction stoichiometry with the regeneration of the cofactor NAD⁺; and (iii) elimination of the first step of both glycerol biosynthesis and glycerol catabolism via G3P. Our study provides an innovative metabolic engineering strategy for rerouting glycerol traffic in S. cerevisiae while tracking ethanol production to levels that have not yet been attained within any other safe model organism, either native or genetically engineered (9, 13, 14, 40–42). This strategy represents another pivotal step for fermenting glycerol in several biorefineries, wherein glycerol and saccharified glucose can be coconverted to ethanol after glycerolysis pretreatment of recalcitrant lignocellulosic biomass.

MATERIALS AND METHODS

Strains, primers, cassettes, and plasmid construction.

All the strains used in this study are shown in Table 1 and were derived from the laboratory strain D452-2 (34). The parents of the D452-2 strain are shown in Fig. S2. The plasmids and primers used in this study are listed in Table 5 and Table S1, respectively. The primers were designed based on the sequences available in the S288C Saccharomyces Genome Database. The details of the DNA fragments and cassettes and the construction of plasmids are described below.

TABLE 5.

Plasmids constructed in this study

Plasmid	Relevant genotype	Reference
pPGK-URA3	URA3; PGK promoter and terminator	45
TDH3-DIT1d22-URA3	URA3; TDH3 promoter and mutated DIT1d22 terminator	This study
TDH3-GDH-DIT1d22-URA3	URA3; expresses OpGDH	This study
TDH3-noxE-DIT1d22-URA3	URA3; expresses LlnoxE	This study
pPGK-DAK1-URA3	URA3; expresses ScDAK1	This study
pPGK-DAK2-URA3	URA3; expresses ScDAK2	This study
pPGK-TPI1-URA3	URA3; expresses ScTPI1	This study
pPGK-TPI1-DAK2-URA3	URA3; expresses ScTPI1, ScDAK2	This study
pPGK-TPI1-DAK2-DAK1-URA3	URA3; expresses ScTPI1, ScDAK2, ScDAK1	This study
pAUR101	AUR1-c	TaKaRa Bio
pAUR101-PGK-RPL41B	AUR1-C; PGK promoter-RPL41B terminator	This study
pAUR101-FPS1	AUR1-C; expresses CuFPS1	This study
pAUR101-FPS1-TPI1-DAK2-DAK1	AUR1-C; expresses CuFPS1, ScTPI1, ScDAK2, ScDAK1	This study
pAUR101-M1a	AUR1-C; expresses CuFPS1, OpGDH, ScDAK1, ScTPI1	This study
Multiplex pCAS (Addgene; 60847)	Multiplex; expresses Cas9 and HDV ribozyme-sgRNA	43
Multiplex pCAS/GPD1-1	Multiplex; expresses Cas9 and gRNA to base 135 of ScGPD1	This study
Multiplex pCAS/GPD1-2	Multiplex; expresses Cas9 and gRNA to base 1045 of ScGPD1	This study
Multiplex pCAS/GUT1	Multiplex; expresses Cas9 and gRNA to base 616 of ScGUT1	This study

^aM1, TEFp-CuFPS1-CYC1t; TYS1p-OpGDH-ATP15t; TDH3p-ScDAK1-d22-DIT1t; FBA1p-ScTPI1-TDH3t.

Construction of pTDH3-DIT1_d22-URA3, pTDH3-GDH-DIT1_d22-URA3, and pTDH3-noxE-DIT1_d22-URA3 plasmids.

Initially, D452-2 cells were picked using a toothpick, resuspended in a PCR tube containing 20 μl 30 mM NaOH (Wako, Osaka, Japan), and heated at 95°C for 10 min in a thermal cycler (Astec thermal cycler; GeneAtlas, Japan). Next, a 1-μl volume of the disrupted cells was used as a template for a 50-μl reaction. High-fidelity polymerase KOD-plus neo (Toyobo, Osaka, Japan) was used to amplify the TDH3 promoter with an extra region of 52 bp complementary to the upstream sequence of the start codon of ScGPD1 (fragment 1). The nucleotide sequence for the mutated terminator DIT1_d22 was purchased from Integrated DNA Technology (IDT; Tokyo, Japan) (18) and then amplified to generate an extra region (46 bp) corresponding to the downstream sequence of the stop codon of ScGPD1 (fragment 2). Thereafter, the first fragment was digested using XhoI and NotI (TaKaRa, Shiga, Japan) and the second fragment was digested using NotI and SalI. Then the digested DNA was purified using the FastGene gel/PCR extraction kit (Nippon Genetics, Tokyo, Japan). One-step cloning was used to clone the TDH3 promoter and mutated DIT1_d22 terminator into the XhoI/SalI sites of the pGK-URA3 plasmid to construct the pTDH3-DIT1_d22-URA3 plasmid (Table 5). The nucleotide sequence for the OpGDH gene was purchased from IDT (GenBank accession number XP_018210953.1) (Table S2). NotI/BamHI sites were added to OpGDH via PCR and then cleaved by restriction enzymes to form the pTDH3-GDH-DIT1_d22-URA3 plasmid (Table 5). LlnoxE (accession number AAK04489.1) was prepared as described for OpGDH, and its sequence is shown in Table S2. LlnoxE was then cloned into the pTDH3-DIT1_d22-URA3 plasmid to assemble the pTDH3-noxE-DIT1_d22-URA3 plasmid (Table 5).

Construction of the multiplex pCAS-gRNA-CRISPR system.

The original multiplex pCAS-guide RNA (gRNA) system was a gift from Jamie Cate (43). The online tool CRISPRdirect (https://crispr.dbcls.jp/) was used for the rational design of the CRISPR/Cas target (44). The efficiency of the target design was confirmed using CHOPCHOP (https://chopchop.cbu.uib.no/). Accordingly, a highly specific guide, 20 bp before the protospacer-adjacent motif (PAM), was selected and used to design the primers (Table S1). Instead of a single-step amplification of the entire multiplex plasmid, which could result in mutations, two rounds of PCR were used to produce only universal scaffolds. In the first round, the two fragments were synthesized separately. The first fragment encompassed the upstream region of the gRNA scaffold, while the second one encompassed the downstream region (Fig. S3). After purification of each DNA segment, the overlapping gRNA (20-nucleotide guide sequence) was used to generate overhangs during the second round of PCR (Fig. S3). During the second PCR, KOD-plus neo and pCAS F and R primers were used with 6 pg each DNA segment as a template. Then the unified DNA fragment was cleaved by SmaI/PstI to form SmaI-universal scaffold-PstI. Subsequently, the universal scaffold was cloned into a multiplex plasmid cleaved at SmaI/PstI sites. As a result, a new multiplex pCAS-gRNA plasmid was formed. These steps were repeated for the construction of multiplex pCAS-gRNA plasmids targeting ScGPD1 and ScGUT1 (Table 5). The sequences of the new SmaI-universal scaffold-PstI in the constructed plasmids were confirmed via sequencing.

Construction of pPGK-DAK1-URA3, pPGK-DAK2-URA3, pPGK-TPI1-URA3, pPGK-TPI1-DAK2-URA3, and pPGK-TPI1-DAK2-DAK1-URA3 plasmids.

DAK1, DAK2, and TPI1 amplicons were obtained by using the genomic DNA of the parent strain D452-2 as the template, as described above. The Xhol site of ScDAK2 was disrupted by incorporating a silent mutation before cloning. The ends of the genes were cleaved using restriction enzymes; the restriction sites were present in the supplemented primers (Table S1, section 1). First, each gene was separately cloned in the pPGK-URA3 plasmid (45) under the control of the PGK promoter and terminator (Table 5). After cloning, the sequence of the genes was confirmed using the primers listed in Table S1. To construct the pPGK-TPI1-DAK2-URA3 plasmid, the XhoI/SalI-ScDAK2 cassette was inserted in the SalI site of the pPGK-TPI1-URA3 plasmid after dephosphorylation of the site. Since the previous ligation of SalI with XhoI altered the sequence of the ligation site, we reused the SalI site during the cloning of the XhoI/SalI-ScDAK2 and ScDAK1cassettes by preventing cleavage and detachment of the cloned cassettes, XhoI/SalI-ScTPI1 or ScDAK2, to obtain the plasmids pPGK-TPI1-DAK2-URA3 and pPGK-TPI1-DAK2-DAK1-URA3 (Table 5).

Construction of pAUR101-PGK-RPL41B, pAUR101-FPS1, and pAUR101-FPS1-TPI1-DAK2-DAK1 plasmids.

Candida utilis (NBRC 0988) was obtained from the National Biological Resource Center (NBRC) of the National Institute of Technology and Evaluation (Japan) and was used as a source of the glycerol facilitator gene (CuFPS1). The sequence of CuFPS1 (GenBank accession number BAEL01000108.1), along with its expression system, including the PGK promoter and RPL41B terminator, is shown in Table S2. To construct the pAUR101-PGK-RPL41B plasmid, the pAUR101 vector (TaKaRa, Shiga, Japan) was subjected to one-step cloning of the SmaI-PGK-NotI and NotI-RPL41B-SacI fragments into its SmaI-SacI sites. Thereafter, CuFPS1 was cloned into a dephosphorylated NotI site of pAUR101-PGK-RPL41B to assemble the pAUR101-FPS1 vector. The direction of cloning was confirmed using PCR. Then a detached set of cassettes, ScTPI1, ScDAK2, and ScDAK1, obtained via Xhol-SalI digestion from the previously constructed plasmid pPGK-TPI1-DAK2-DAK1-URA3, was cloned into the dephosphorylated SalI site of the pAUR101-FPS1 vector to assemble the pAUR101-FPS1-TPI1-DAK2-DAK1 plasmid (Table 5).

Construction of the M1 and pAUR101-M1 plasmids.

All fragments constituting module M1 were first amplified separately via PCR (Fig. 3). CuFPS1, OpGDH, and the mutated DIT1_d22 terminator were amplified from their synthetic DNA stocks, while the other fragments were amplified from the genomic DNA of the D452-2 strain. Following the generation of 12 fragments via PCR, the products were electrophoresed on 1% or 2% agarose gels, excised according to the sizes of the fragments, and purified using a gel extraction kit (Nippon Genetics, Tokyo, Japan). The highly purified fragments were used for assembly using the Gibson assembly master mix (New England BioLabs, Tokyo, Japan). Initially, the first three consecutive segments of M1 were joined seamlessly according to the manufacturer’s protocol (Gibson; New England BioLabs); this was immediately followed by PCR amplification of the joined segments, which were then purified again with an agarose gel. These steps were applied for the second, third, and fourth three consecutive segments of the 12 segments of M1. Then every six sequential fragments were assembled via Gibson assembly followed by PCR. Next, the 12 sequential segments were joined the module M1: TEFp-CuFPS1-CYC1t, TYS1p-OpGDH-ATP15t, TDH3p-ScDAK1-DIT1_d22t, and FBA1p-ScTPI1-TDH3t (Fig. 3). The SacI site was added upstream of module M1, while the SmaI site was added downstream for cloning into the pAUR101 vector to form pAUR101-M1 (Table 5). Finally, the vector pAUR101-M1 was transformed into Escherichia coli as described later, and the accurate structure of M1 was confirmed by sequencing module M1 in the plasmid pAUR-M1.

Transformation and recombination of strains.

All constructed plasmids were transformed in E. coli NEB 10-beta via the heat shock method of transformation according to the manufacturer’s instructions (New England BioLabs, Tokyo, Japan). All plasmids were extracted using the QIAprep Spin miniprep kit (Qiagen, Hilden, Germany). All DNA measurements were performed using a BioSpec-nano UV-visible (UV-Vis) spectrophotometer (Shimadzu, Japan). Plasmids and DNA fragments were stored at −20°C. Yeast transformation was performed using the Fast Yeast Transformation kit (G-Biosciences, USA) to integrate the linear pAUR101 vector. BsiWI (New England BioLabs) was used to linearize the AUR1-C gene before transformation, which was integrated in the AUR1 locus of S. cerevisiae via homologous recombination. Positive colonies were selected on aureobasidin A (TaKaRa, Shiga, Japan) and further confirmed via PCR. The same method and kit were used to transform the linearized pPGK-URA3 plasmid, in which URA3 was linearized with NcoI (TaKaRa). The transformants were grown on YNB medium supplemented with 20 mg/liter leucine and histidine; the fast-growing colonies were confirmed using PCR and then screened for glycerol consumption and ethanol production. ScGPD1 with OpGDH were replaced via homologous repair of the double-strand break, which was induced by CRISPR Cas 9 (46). Similarly, ScGUT1 was replaced with the M1 module. The homologous regions were further extended via PCR during double-strand DNA repair. Positive colonies were confirmed via PCR using primers located on the inserted cassettes and regions upstream and downstream of the flanking recombined loci. The recombinant cells were reconfirmed after the loss of the pCAS multiplex plasmids. All the recombinant strains and their genotypes are listed in Table 1.

Preculture and culture conditions.

YPD medium (yeast extract, 10 g; peptone, 20 g; glucose, 20 g per liter) was routinely used to cultivate stocks and maintain the strains listed in Table 1. All liquid cultivation was conducted in Falcon 50-ml conical tubes to enhance the workflow by a decrease in sterilization steps, easy handling, and direct recovery of the cell pellets. The cells were cultured under microaerobic conditions (10 ml culture per tube with 200 rpm of orbital shaking [BioShaker BR-42FL; TAITEC, Japan] and approximately 45° sitting angle) overnight at 30°C. The screw caps of the tubes were closed at a level that permitted CO₂ escape. The cell densities were monitored at 600 nm (OD₆₀₀) after 10-fold dilution using a spectrophotometer (AS ONE, Japan). YPD₁₅G₇₀ medium (15 g/liter glucose and 70 g/liter glycerol) was used to preculture the cells used in this study under the microaerobic conditions described above (Tables 2 to 4 and Fig. 2 to 5). YNB minimal medium supplemented with 20 g/liter glucose, 20 mg/liter leucine, and histidine was used to culture the recombinant strains harboring active URA3 during or after the transformation. YNB medium supplemented with 69 g/liter glycerol, 200 mg/liter leucine, and histidine was used to compare the cultivations of the D452-2 and SK-FGG4 strains (Fig. S1).

Preparation of cell extract.

The OD₆₀₀ (growth) of the recombinant strains ranged from 6 to 9 (Fig. S4). Therefore, proteins were extracted by considering the difference in the cell concentration as described previously (19, 47, 48), with the following modifications: 10-ml cell cultures with an OD₆₀₀ of approximately 5 were harvested via centrifugation at 700 × g for 2 min at 4°C. The cell pellets were washed with 20 ml 100 mM HEPES buffer (pH 7.4), followed by another washing with 1 ml buffer. Then the cell pellets were lysed with approximately 400 mg glass beads in 1 ml HEPES buffer supplemented with 1 mM MgCl₂ and 10 mM 2-mercaptoethanol in a 2-ml Eppendorf tube. Lysis was accomplished by vigorous shaking using a bench vortex with cooling six times on ice at intervals of 30 s. The crude proteins were separated from the glass beads and cell debris via two rounds of centrifugation at 22,300 × g at 4°C for 5 min. The total protein concentration of the supernatant was estimated using Quick Start Bradford (Bio-Rad, USA) and bovine serum albumin standards (Novagen, USA) at 595 nm using an Infinite M200 plate reader (Tecan, Switzerland).

Enzyme assays.

The specific activity of GDH was determined by monitoring the changes in NADH absorbance at 340 nm (UV-2700; UV-Vis spectrophotometer; Shimadzu, Japan) as reported previously (19), with the following modifications. One milliliter 79 mM HEPES buffer (pH 7.4), 10 μl crude extract, and 10 mM NAD⁺ were mixed in a quartz cuvette and incubated for 30 s to complete any side reaction and obtain a stable baseline. The reaction was initiated by adding 100 μl 1 M glycerol and incubating the mixture for 1 min. The activity of ScTPI1 was assayed as described previously (49), with some modifications. The reaction mixture consisted of 100 mM (pH 7.53) triethanolamine hydrochloride (Sigma-Aldrich), 10 mM dihydroxyacetone phosphate (DHAP) hemimagnesium salt (Sigma-Aldrich), 100 μl crude extract, and 5 mM NAD⁺ as a starting point for the reaction, which lasted 400 s. The main changes in absorbance were detected after approximately 100 s. Changes in the absorbance from 200 s to 260 s were used to calculate the activities. Dihydroxyacetone kinase was assayed using a universal kinase activity kit according to the manufacturer’s instructions (EA004; R&D Systems, USA).

Determination of the intracellular concentration of NADH/NAD⁺.

NADH/NAD⁺ was estimated per the instructional manual of the EnzyChrom NAD⁺/NADH assay kit (BioAssay Systems; E2ND-100) as previously described (21). A 20-μl volume of cultivated cells at an OD₆₀₀ of 5 was harvested via centrifugation at 1,600 × g for 2 min at 4°C to collect the cells and determine the NADH/NAD⁺ ratio.

Fermentation.

Fermentations were performed under orbital shaking (BioShaker BR-42FL; TAITEC, Japan) at different levels of oxygen based on the ratio of liquid medium (in milliliters) in the flask and the volume of the Erlenmeyer flask (in milliliters). The different oxygen conditions were as follows: microaerobic (20:100), semiaerobic (20:200), and aerobic (20:300). The cell pellets were obtained from approximately 20 ml cell culture (OD₆₀₀ = 4), which was harvested from appropriate volumes of preculture via centrifugation at 1,600 × g for 5 min and then washed with 20 ml Milli-Q water. The conversion of glycerol into ethanol was conducted at 30°C with an agitation speed of 180 or 200 rpm. The YP and YNB media were supplemented with different initial concentrations of glycerol or glucose, which are listed in Fig. 2, Table 3, and footnote a of Table 4, in accordance with those obtained after the glycerolysis of biomass (16). For fermentation under strict anaerobic conditions, 50-ml Mighty vials (Maruemu, Japan) were sterilized with a precision seal septum cap before addition of the YNB medium and inoculation of the cell pellets. Then nitrogen gas was flushed into the medium through Terumo needles, which were also used for sampling.

Fermentation analysis.

The samples (100 μl) were picked from fermentative flasks under sterile conditions, diluted with 900 μl Milli-Q water in Eppendorf tubes, and mixed before centrifugation at 22,300 × g for 5 min. Subsequently, the supernatants were decanted using a 1-ml syringe and filtered through a 0.45-μm hydrophilic filter (PTFE) directly into new 2-ml high-performance liquid chromatography (HPLC) glass vials. Analyses were performed using HPLC (Shimadzu, Japan) on an Aminex HPX-87H column (Bio-Rad Laboratories, Hercules, CA) connected to refractive index (RID-10A; Shimadzu) and prominence diode array (SPD-M20A; Shimadzu) detectors using 5 mM H₂SO₄ as the mobile phase at 50°C with a flow rate of 0.6 ml/min. Glucose, glycerol, ethanol, acetic acid, pyruvate, succinic acid, and acetaldehyde were detected using the refractive index detector (RID). DHAP and DHA levels were quantified using the prominence diode array detector (SPD-M20A).

Ethanol efficiency calculation.

Ethanol was calculated according to following equations:

$${\text{C}}_6{\text{H}}_{12}{\text{O}}_6\to {\text{\hspace{0.17em}2C}}_2{\text{H}}_5\text{OH\hspace{0.17em}}\hspace{0.17em}+\hspace{0.17em}{\text{\hspace{0.17em}2CO}}_2$$ (1)

$$\text{1\hspace{0.17em}g\hspace{0.17em}}\to \hspace{0.17em}0.51\text{\hspace{0.17em}g\hspace{0.17em}}\hspace{0.17em}+\hspace{0.17em}\hspace{0.17em}0.\text{49\hspace{0.17em}g}$$

Ethanol (grams per liter) = (initial concentration of glucose − residual concentration of glucose) × 0.51

$${\text{C}}_3{\text{H}}_8{\text{O}}_3\hspace{0.17em}+\hspace{0.17em}½{\text{O}}_2\to {\text{\hspace{0.17em}C}}_2{\text{H}}_5\text{OH\hspace{0.17em}}\hspace{0.17em}+\hspace{0.17em}{\text{\hspace{0.17em}CO}}_2\hspace{0.17em}+\hspace{0.17em}{\text{\hspace{0.17em}H}}_2\text{O}$$ (2)

$$\text{1\hspace{0.17em}g\hspace{0.17em}}\to \hspace{0.17em}0.5\text{\hspace{0.17em}g\hspace{0.17em}}\hspace{0.17em}+\hspace{0.17em}\hspace{0.17em}0.\text{478\hspace{0.17em}g}$$

Ethanol (grams per liter) = (initial concentration of glycerol − residual concentration of glycerol) × 0.50

Data availability.

All necessary data required to assess our findings are available in this article or the supplemental material. Further details related to this article may be requested from the authors.