Background

In industrialized countries, the overconsumption of sucrose and other specialties including honey, maple syrup, or high fructose corn syrup causes overweight, obesity, diabetes, and metabolic syndrome [1, 2]. As an alternative sweetener, low or non-caloric molecules, such as sugar alcohol, are more and more considered. Among them, erythritol (1,2,3,4-butanetetrol) is a four-carbon polyol with several exciting features. It has very low digestibility and does not raise the blood insulin level. It reduces lipid peroxidation, thereby protecting the damage caused by oxidative stress involved in the pathogenesis of diabetes [3]. It is not metabolized by Streptococcus mutans, the main causative agent of tooth decay. Erythritol has been determined as safe for human consumption even at high doses. The maximal recommended dose ranges between 0.6 and 0.8 g/kg of body weight, while it is only around 0.3 g/kg of body weight for the sweetener xylitol [4]. It can also act as an antioxidant in vivo and displays endothelium-protective effects, which may help to protect against hyperglycemia-induced vascular damage [5]. Due to endothelium-protective effects during hyperglycemia, erythritol can reduce the risk of diabetic complications [6].

Erythritol is predominantly produced from glucose by osmophilic yeast such as Candida magnoliae, Y. lipolytica, Torula sp., Moniliella megachiliensis, or Trichosporonoides oedocephalis [7,8,9,10]. Among these, Y. lipolytica is the most widely used for industrial erythritol production. Using metabolic engineering, much progress has been gained to improve the erythritol production titer in Y. lipolytica. For instance, Janek et al. [11] found that overexpression of the endogenous erythrose reductase gene (i.e., ER, YALI0F18590g) resulted in an enhanced erythritol production with the titer of 44.44 g/L, a yield of 0.44 g/g, and productivity of 0.77 g/L/h, which was 20% higher than the parental strain. Carly et al. [12] constructed the Y. lipolytica mutant FCY218 by overexpressing GUT1 (encoding a glycerol kinase, YALI0F00484g) and TKL1 gene (encoding a transketolase, YALI0E06479g) and by disrupting the EYK1 gene (encoding erythrulose kinase, YALI0F01606g). The engineered strain exhibited a 75% higher erythritol production titer as compared to that of wild type (80.6 g/L). In Y. lipolytica strain CGMCC7326, overexpression of two erythrose reductase genes (ER10, YALI0D07634g, and ER25, YALI0C13508g) and engineering of the NADPH cofactor metabolism by overexpression of 6-phosphogluconate dehydrogenase genes (GND1, YALI0B15598g) and glucose-6-phosphate dehydrogenase genes (ZWF1, YALI0E22649g) led to a significant increase in erythritol titer and yield as compared to the wild-type strain (i.e., 190 g/L and 0.63 g/g, respectively) [13]. However, the erythritol production process still suffers from several drawbacks, such as the production of unwanted byproducts such as mannitol and d-arabitol, which renders the purification process more challenging. In these processes, the bioreactor cooling cost could not be neglected at a large scale. Therefore, the improvement of the thermotolerance of the producing strain must also be considered.

To optimize strain CGMCC7326 for erythritol production, it has been impaired for the ability to synthesize mannitol and d-arabitol and to catabolize erythritol. For that purpose, we first identify five genes, namely AraDH1 (g1595.t1, YALI0F02211g), AraDH2 (g3858.t1, YALI0E05643g), MDH1 (g5130.t1, YALI0B16192g), MDH2 (g2069.t1, YALI0D18964g), XDH1 (g4121.t1, YALI0E12463g) and characterized them with d-mannitol or d-arabitol dehydrogenase activities. Then, the d-mannitol dehydrogenase gene that is mainly responsible for byproducts synthesis was deleted, together with gene EYD1 encoding erythritol dehydrogenase (g1570.t1, YALI0F01650g) to disrupt the erythritol utilization pathway. Thirdly, the rsp5 gene from S. cerevisiae encoding ubiquitin ligase was overexpressed, allowing the cell to grow at 35 °C. As the synthesis of erythritol is not a redox-balance reaction, ZWF1 and GND1 genes were overexpressed in Y. lipolytica, to recycle cofactor NADP. We also disrupt gene Ku70 involved in non-homologous end-joining to facilitate gene disruption procedure.

Results and discussion

Construction of a chassis strain derived from Y. lipolytica CGMCC7326

We previously engineered the wild-type strain CGMCC7326 for erythritol production using hygromycin resistance for transformant selection [13]. However, no further genetic modifications could be operated in the resulting strain HCY108 as it is resistant to hygromycin and lacks auxotrophic markers. Therefore, we intended to construct a chassis strain allowing multiple genome editions. For that purpose, we first disrupted gene Ku70 (YALI0C08701g) involved in non-homologous end-joining (NHEJ) to facilitate subsequent gene disruption. We also rendered the resulting strain auxotrophic for uracil by disrupting the URA3 gene (U40564.1) encoding orotidine 5′-phosphate decarboxylase.

In Y. lipolytica, NHEJ, the dominant form of DNA repair, is mediated by the heterodimer protein complex Ku70/Ku80 [14, 15]. It has been reported that the disruption of gene encoding Ku70 resulted in the loss of NHEJ repair and an increased rate of homologous recombination (HR) [16, 17]. Therefore, Ku70 disruption significantly increases the rate of correct targeted gene replacement to generate knock-out and knock-in mutants, and hence improve the precision of genetic edition in Y. lipolytica. The disruption cassette contained upstream and downstream fragments of the Ku70 gene and the hygromycin resistance gene flanked by loxP and loxR sequences [18]. The Ku70 disrupted strain was then cured of the hygromycin resistance gene by transient expression of the Cre gene encoding loxR/loxP recombinase [18]. This yielded strain HCY109 (ery929△ku70) (Table 1).

Table 1 Strains, gene cassettes, and primers used in this study
Full size table

In Y. lipolytica and other yeasts, the URA3 gene encoding orotidine 5′-monophosphate decarboxylase (OMP) involved in uracil synthesis has been used as a sensitive and versatile auxotrophic marker [18]. Therefore, URA3 was disrupted in strain HCY109 by gene replacement using a defective allele and a selection medium containing 5-fluoroorotic acid (5-FOA). The resulting chassis strain was designated HCY109-2 (ery929△ku70△ura3) (Table 1). Figure 1 shows the schematic flow to obtain the URA3 disruption mutant.

Fig. 1
figure 1

The schematic flow to obtain authentic URA3 disruption mutant. The 1.2 kb upstream (lane 1, a) and downstream (lane 2, a) of the ura3 gene was amplified separately, then the 2.4-kb fragment (lane 3, a) was obtained by fusion PCR to merge 1.2 kb up and downstream. Then the 2.4-kb fragment was transformed into HCY109 (ery929△Ku70), plated on YPDF medium (10 g/L yeast extract, 5 g/L tryptone, 10 g/L glucose, 2 mg/ml 5-FOA, 15 g/L agar powder, pH 6.5) and cultivated at 30 °C for 5 days (b). Colonies were grown (c) and transferred to a new YPDF plate cultivated at 30 °C for 5 days (d). Then the colonies were transferred to synthetic complex medium without uracil (SC-U, 10 g/L yeast nitrogen base, 5 g/L ammonia sulfate, 10 g/L glucose, without uracil) (e). If the ura3 gene was authentically disrupted, the mutants cannot grow on the SC-U medium (1, 2, 3 in e), or they are not authentic ura3 gene disruption (4, 5, 6 in e). Then two putative ura3-deficient mutants were streaked on SC-U plate (f) and YPDF plate (g), they showed opposite growth. A ura3-deficient mutant was further to be identified by PCR to amplify the 480-bp fragment in ura3 gene, and no 480-bp fragment can be amplified (lane 4 in h), but the 480-bp fragment can be amplified for the control strain (HCY109, ery929△Ku70) (lane 5 in h). The No. 1 colony in g designated as HCY109-2 was used as the host to conduct metabolic engineering

Full size image

Identification of genes encoding enzymes with mannitol/d-arabitol dehydrogenase activities in Y. lipolytica CGMCC7326

Although Y. lipolytica is one of the best erythritol producers, it also presents some drawbacks such as the synthesis of byproducts including mannitol and d-arabitol. This negatively impacts the yield from glucose and complexifies the purification steps [9, 13, 19,20,21,22]. Therefore, we thought of constructing a strain unable to produce these byproducts.

In Ascomycetes fungi, the mannitol cycle consists of two pathways [23]. In the first one, fructose-6P from glycolysis is converted into mannitol-1P by a NAD(H)-dependent mannitol dehydrogenase (Mpd). It is then dephosphorylated into mannitol by a mannitol-1P phosphatase. In the second pathway, fructose is oxidized into mannitol by a reversible NADP(H)-dependent mannitol 2-dehydrogenase (Mdh). In Y. lipolytica, the mannitol dehydrogenase encoded by gene YALI0B16192g has been characterized as an Mdh [24]. Indeed, its disruption impairs mannitol production from fructose, but not from glycerol or glucose. By contrast, its overexpression slightly improves mannitol production from fructose [23]. These findings pointed out that other pathways may exist to produce mannitol from glucose through fructose-6P, such as an Mpd–Mpp like pathway [23]. However, genes involved in that pathway remain unknown in Y. lipolytica. d-Arabitol is produced by an arabitol dehydrogenase(s) from ribulose, which is obtained by dephosphorylation of ribulose-5P, an intermediate of the pentose phosphate pathway.

Before disrupting these biosynthetic pathways, the first step was to identify those genes involving mannitol and/or d-arabitol biosynthesis. Genome mining for mannitol dehydrogenase by Dulermo et al. [23] led to the identification of two additional genes, namely YALI0D18964g and YALI0B16192g that were not characterized so far. Further blast analysis, using gene XP_001387287 from Scheffersomyces stipitis CBS6054 encoding mannitol dehydrogenase (MDH)-like, classical (c) SDRs (short-chain dehydrogenases/reductases) did not lead to the identification of additional candidate genes. To identify arabitol dehydrogenase encoding gene, the d-arabinitol 2-dehydrogenase gene from Scheffersomyces stipitis CBS6054 (XP_001385035) and Wallemia ichthyophaga (XM_009272203) was used as a query to blast the genome of Y. lipolytica CGMCC7326. Gene g1595.t1 (YALI0F02211g) was found with 54.73% identity to gene XP_001385035 while gene g3858.t1 (YALI0E05643g) showed 54.9% identity to gene XM_009272203. The identity of g1595.t1 and g3858.t1 genes is 44.39%.

Rodriguez et al. [25] revealed the xylose utilization pathway in Y. lipolytica and found the gene YALI0E12463g (GenBank no. XM_503864) has xylitol dehydrogenase (XDH) with NAD as a cofactor in vitro, when expressed in E. coli. When overexpressed in Y. lipolytica, Polf along with XKS (xylulose kinase gene) resulted in improved growth on xylitol. However, no other polyol substrates were tested with purified ylXDH in this article, so it is necessary to determine whether the ylXDH has mannitol or d-arabitol dehydrogenase activity and is involved in mannitol or arabitol synthesis or not. The counterpart gene in CGMCC 7326 was g4121.t1, with a 100% identity to YALI0E12463g.

Therefore, five genes, which encode putative enzymes with mannitol and/or d-arabitol dehydrogenase were identified, namely YlAraDH1 (g1595.t1, YALI0F02211g), YlAraDH2 (g3858.t1, YALI0E05643g), YlMDH1 (g5130.t1, YALI0B16192g), YlMDH2 (g2069.t1, YALI0D18964g), and YlXDH genes (g4121.t1, YALI0E12463g) for further biochemical characterization.

Characterization of purified YlAraDH1, YlAraDH2, YlMDH1, YlMDH2, YlXDH enzymes

The YlAraDH1, YlAraDH2, YlMDH1, YlMDH2, YlXDH encoding genes were expressed in E. coli, and the corresponding proteins were purified by Ni–NTA chromatography with the purity of above 95% (Fig. 2). For each purified enzyme, the enzymatic activity, both oxidation, and reduction, was measured for different substrates (Table 2). All enzymes except xylitol dehydrogenase (YlXDH) were able to oxidize mannitol, d-arabitol, xylitol, d-sorbitol. Among these dehydrogenases, YlAraDH1 has the most robust activity towards mannitol, d-arabitol, xylitol, and d-sorbitol, whereas YlMDH1 exhibited the weakest activity toward those four polyols. YlXDH has the strongest xylitol dehydrogenase activity. We also tested other polyols such as ribitol, erythritol, and glycerol as well as primary alcohols (n-butanol, isobutanol, ethanol) as substrates. However, no activities were observed for all five enzymes (data not shown). The purified enzymes were also tested for their ability to reduce substrates such as fructose, l-sorbose, or d-xylulose. As shown in Table 2, all tested enzymes except YlXDH were able to reduce these substrates. For substrates such as ribose, mannose, xylose, glucose, and fructose-6P, no activity could be detected (data not shown). According to Table 2, all enzymes have mannitol dehydrogenase activity, and YlAraDH1, YlAraDH2, YlMDH1, YlMDH2 have d-arabitol dehydrogenase activity. Regarding cofactor specificity, YlAraDH1, YlAraDH2, and YlXDH were defined as NAD(H)-dependent, while YlMDH1, YlMDH2 are strictly NADP(H)-dependent. For YlMDH1, these results are in accordance with Dulermo et al. [23] and Napora et al. [24].

Fig. 2
figure 2

YlAraDHl, YlAraDH2, YlMDH1, YlMDH2, YlXDH gene expression in E. coli

Full size image
Table 2 Substrates/cofactors specificity of YlAraDH1, YlAraDH2, YlMDH1, YlMDH2, YlXDH
Full size table

In order to verify if these enzymes are involved in the synthesis of mannitol and d-arabitol from glucose, the corresponding encoding genes were disrupted in Y. lipolytica HCY109. The correctness of the mutant genotype was verified by analytical PCR with primers listed in Table 1. The lack of PCR amplification in mutant strains as compared to the parental strain confirmed the correct gene disruption, namely YlAraDH1, YlAraDH2, YlMDH1, YlMDH2, and YlXDH (Fig. 3a). The disrupted strains were then grown in a shake flask for 120 h in the YPD300 medium. As shown in Fig. 3, no significant differences could be observed in both cell growth and polyols synthesis for all the tested disrupted strains as compared to the parental strain (ery929Δku70Δura3), except strain HCY113 disrupted for gene YlMDH2 (YALI0D18964g). For this latter, production of both mannitol and d-arabitol was impaired while erythritol production was slightly improved by 7.9%, reached 177 ± 3 g/L from 164 ± 3 g/L (Fig. 3c–e). The above results demonstrate that mannitol is produced from the enzymatic activity of YlMDH2 as disruption of YlAraDH1, YlAraDH2, YlMDH1, YlXDH had no effects on mannitol synthesis. This was further evidenced by the synthesis of mannitol in the ∆YlMDH2 mutant complemented with the YlMDH2 gene (data not shown). Finally, it is worth mentioning that although YlMDH1 and YlMDH2 shared 74.8% similarity in amino acid, they present different catalytic activities.

Fig. 3
figure 3

Effects of gene disruption on cell growth, mannitol, and erythritol synthesis. a Molecular identification of 7 genes disruption (Ku70, YlAraDH1, YlAraDH2, YlMDH1, YlMDH2, YlXDH, YlEYD1 genes), no corresponding band can be amplified from the disruptants, but the band can be amplified for wild-type control strain (Y. lipolytica CGMCC7326, ery929), indicating Ku70, YlAraDH1, YlAraDH2, YlMDH1, YlMDH2, YlXDH, or YlEYD1 genes were disrupted individually. b Growth curves of different disruptants on liquid medium containing 300 g/L glucose. c Effects of various gene disruption on erythritol and mannitol biosynthesis, only mdh2 gene disruption can result in no mannitol synthesis. d, e The HPLC chromatography of wild-type strain and mdh2 gene disruption, showing no mannitol when the mdh2 gene was disrupted

Full size image

To explain the discrepancy between the enzymatic activity of the purified enzyme in vitro and the phenotype of disruption mutants, we speculated that genes YlAraDH1, YlAraDH2, YlMDH1, YlXDH were not or poorly expressed in those experimental conditions (i.e., glucose medium). Therefore, gene expressions were quantified during the growth of wild-type strain CGMCC7326 in the YPD300 medium. As shown in Fig. 4, the expression level of gene YlMDH2 was high as compared to that of genes YlAraDH1, YlAraDH2, YlMDH1, YlXDH.

Fig. 4
figure 4

Comparison of YlAraDH1, YlAraDH2, YlMDH1, YlMDH2, YlXDH expression levels relative to β-actin. YlAraDH1, YlAraDH2, YlMDH1, YlMDH2, YlXDH gene mRNA quantification of wild-type Y. lipolytica (CGMCC7326) grown on 300 g/L glucose, mRNA levels are represented as copy number relative to the copy number of β-actin

Full size image

Dulermo et al. [23] speculated that similarly to other Ascomycetes, two pathways may exist in Y. lipolytica for mannitol synthesis. Different genes with putative mannitol and/or arabitol dehydrogenase activity have been identified, and the corresponding enzymes were purified. Most of them were found able to reduce fructose into mannitol and xylulose into arabitol in vitro. However, mutant disrupted for gene YALI0D18964 (YlMDH2) was the only one unable to produce mannitol from glucose, suggesting that YlMDH2 is the only active mannitol dehydrogenase inside the cell. This was confirmed by quantification of the gene expression level, which was significantly higher for YALI0D18964 than for the other tested gene. From the substrate specificity, it may be deduced that YlMDH2 is instead a mannitol dehydrogenase (Mdh) than a mannitol-1P dehydrogenase (Mpd). Indeed, it could not reduce fructose-6P, suggesting thus that it is not involved in an Mpd–Mpp-like pathway.

Disruption of YlEYD1

Although Y. lipolytica can synthesize erythritol in response to osmotic stress, it also has the ability to reconsume it in isotonic conditions. In this catabolic pathway, erythritol is first converted into erythrulose by an erythritol dehydrogenase encoded by gene EYD1 [26]. Therefore, the gene was disrupted in strains HCY109-2 (△ku70△ura3) and HCY113 (△ku70△ura3△mdh2). Gene disruption in the chassis strain HCY109-2 yielded an 8% increase in erythritol titer (178 ± 3 g/L, Fig. 3C) as compared to the parental strain (HCY113). Glucose and erythritol consumption rates were determined during the culture of strains HCY113 (△ku70△ura3△mdh2) and HCY115 (△ku70△ura3△mdh2△eyd1) in a medium containing 20 g/L glucose and 20 g/L erythritol. As shown in Fig. 5, strain HCY115 was unable to catabolize erythritol while it showed a fivefold increased glucose uptake rate (0.83 g/L h) in the first 12 h of growth. At the industrial scale, the erythritol production process is operated at high glucose concentration (up to 300 g/L). In these conditions, no erythritol is reconsumed by the cell due to the high osmotic pressure. However, at the end of the production process, glucose must be depleted in the culture medium for ease of erythritol purification. It is known that at low glucose concentration (around 20 g/L and below), erythritol starts to be reconsumed by the cell together with the remaining glucose. To avoid this erythritol consumption, the erythritol catabolic pathway could be disrupted. Carly et al. [27] disrupted EYK1 encoding erythrulose kinase, the second enzyme of the pathway. Although this strategy led to mutant strain impaired in erythritol catabolism, the disrupted strain was still able to convert erythritol into erythrulose leading to a mixture of erythritol and erythrulose. To avoid this, we disrupted the first gene of the pathway, EYD1.

Fig. 5
figure 5

Effect of disruption of YlEYD1 gene on glucose and erythritol utilization. a HPLC profile of glucose and erythritol utilization by wild-type strain harboring the EYD1 gene. b Glucose and erythritol utilization rate by wild-type strain harboring the EYD1 gene. c Glucose and erythritol utilization rate by the EYD1 gene disruption strain. d HPLC profile of glucose and erythritol utilization by the EYD1 gene disruption strain

Full size image

Improving thermoresistance by expression of Sc.rsp5 gene in Y. lipolytica

The optimal growth temperature of Y. lipolytica ranges between 28 and 30 °C [13, 28, 29], and at higher temperature, cell growth ability decreases drastically. Yeast metabolism could be considered as an exothermic process for instance; glucose catabolism has a specific heat production of 1798 kJ/mol [30]. Since Y. lipolytica is known to be able to grow at high cell density and to sustain high glucose concentration, the energy, and thus the cost, requested for bioreactor cooling is to take into account, especially at industrial scale. Therefore, increasing the thermotolerance of strain HCY115 (Δku70Δura3Δmdh2Δeyd1) for several degrees is of interest to develop an efficient erythritol production process. Several strategies have been used for that purpose in yeast, such as overexpression of heat shock proteins (HSP) or transcription factors [31,32,33,48, 49] with some modifications. MDH activity was measured by monitoring the oxidation of NADH or NADPH (or reduction of NAD or NADP) at 340 nm. The reduction of fructose or fructose 6-P was assayed in 100 mM sodium acetate buffer (pH 6.0) containing 2 mM NADPH or NADH, purified enzyme, and 100 mM d-fructose or fructose 6-P in a total volume of 1.5 mL. The reactions were initiated by adding d-fructose or fructose 6-P in the mixtures. The oxidation of mannitol was assayed in 100 mM Tris–Cl (pH 8.0) containing 2 mM NADP or NAD, purified enzyme, and 100 mM d-mannitol in a total volume of 1.5 mL. One unit of activity was defined as the amount of enzyme which catalyzes the oxidation of 1 μmol of NADPH or NADH per min. For cofactor specificity, enzyme activity for mannitol (100 mM) was determined as described above using cofactor (NAD or NADP) at a concentration of 2 mM. To study the substrate specificity, other polyols such as d-sorbitol, d-arabitol, xylitol, ribitol, erythritol, glycerol were also tested. The reaction mixture contained (1.5 mL final volume) 50 mM substrate, 2 mM NADP or NAD, 50 μL of purified enzymes, and 10 mM Tris–Cl buffer (pH 8.0). Specific activity was expressed as U/mg of protein.

Transketolase (TKL) activities were assayed by measuring the decrease in NADH concentration at 340 nm [50]. Briefly, TKL activity was measured at 30 °C in a reaction mixture with the following composition: 84 mM triethanolamine buffer (pH 7.6), 0.9 mM xylulose 5-phosphate, 1.2 mM ribose 5-phosphate, 0.3 mM thiamine pyrophosphate, 0.3 mM NADH, 0.34 U of GDH, and 1 U of TIM. In this assay, xylulose 5-phosphate and ribose 5-phosphate are used as substrates for TKL, and the enzymatic formation of glyceraldehyde 3-phosphate in a coupled reaction with GDH and TIM is measured. One unit of enzyme activity was defined as the amount of enzyme that oxidizes 1 μmol NADH per min. These enzyme activities were determined using a UV/VIS-2450 spectrophotometer (Agilent).

Construction of the Ku70 deletion mutant and marker excision

The Ku70 gene of erythritol-producing strain Y. lipolytica CGMCC7326 was 100% similar to that of Y. lipolytica CLIB122 with gene tag YALI0C08701g. However, the identities of 1.2 kb upstream sequence of Ku70 in CGMCC7326 and CLIB122 was 90%, the identity of 1.2 kb downstream sequences of Ku70 in CGMCC7326 and CLIB122 was only 41%, these differences indicating the genome of the two strains are somewhat different. The Ku70 deletion cassette sequence is shown in Additional file 1 (sequence 1), containing the marker gene hygromycin resistance (hph) gene. The full-length Ku70 deletion cassette (4.3 kb) was synthesized in vitro and cloned into the pUC19 derivative vector to yield the Ku70 gene knockout vector pSWV-ΔKu70-hph, in which two loxP sites were located at the flanking sequence of hph sequence. Then the Ku70 deletion cassette was cut with EcoRI and NotI to release the 4.3 kb fragment, which was transformed into Y. lipolytica CGMCC7326 by lithium acetate transformation [13]. The cells were plated on YPD-hph solid medium containing 400 μg/mL hygromycin B. Transformants were transferred to a new YPD-hph medium, cultivated for 4 days, and repeated to obtain pure clones. Then total genomes of transformants were extracted according to Cheng [51]. Primers PKu70-knockout-F and PKu70-knockout-R listed in Table 1 were used to verify the Ku70 gene disruption. Hph gene marker was rescued from Ku70::hph by transformation with the replicative plasmid pUB4-XDH, in which the Gluconobacter oxydans 621H xylitol dehydrogenase gene (XDH) was used to replace the hph gene of pUB4-Cre [18]. The fragment EcoRI-hp4d-621XDH-TT-hp4d-Cre-SalI (sequence 2 in Additional file 1) was synthesized and used to replace the 4050 bp EcoRI-SalI fragment of pUB4-Cre yielding the rescuing plasmid pUB4-XDH. After the plasmid pUB4-XDH rescued the hph gene, the transformant was cultivated again in YPD liquid medium without hygromycin B and xylitol to lose the episomal plasmid pUB4-XDH, giving rise to the strain ery929△Ku70 (HCY109 in Table 3), which was used as chassis to delete or express other genes, for example, the URA3 and EYD1 genes.

Construction of the URA3 deletion mutant based on ery929△Ku70 (HCY109)

To disrupt the ylURA3 gene in ery929△Ku70 (HCY109), ylURA3 disruption cassette was constructed by overlap PCR. The 1.2 kb upstream of the ylURA3 gene (GenBank accession no. YLU40564, 861 bp) was amplified using primers PURA3-upstream-F and PURA3-upstream-R, the 1.2 kb downstream of the ylURA3 gene was amplified using primers PURA3-downstream-F and PURA3-downstream-R. Then the 1.2 kb upstream and 1.2 kb downstream of ylURA3 fragments were ligated by overlap** PCR with primers PURA3-upstream-F and PURA3-downstream-R, yielding the 2.4-kb ylURA3 gene disruption cassette and was verified by DNA sequencing. The 2.4-kb fragment was used to transform HCY109 and plated on a solid YPD plate containing 2 mg/mL fluoroorotic acid (5-FOA) to negatively select URA3 gene-deleted strains. The transformants grown on 5-FOA containing YPD were transferred to the SD medium without uracil (SD-U) to select stable URA3-deficient mutants. The transformants that were unable to grow on the SD-U plate were verified by PCR with primers PURA3-verify-F and PURA3-verify-R to amplify the fragment of 480 bp in the URA3 gene. The authentic URA3 disruption strain was designated as HCY109-2 (ery929△Ku70△URA3).

Deletion of genes encoding mannitol dehydrogenases and/or d-arabitol dehydrogenases

Genes encoding the mannitol and d-arabitol dehydrogenases were deleted to verify whether mannitol and d-arabitol can still be synthesized in erythritol-producing Y. lipolytica CGMCC7326 (hereinafter as ery929). Five genes were identified to have both mannitol and d-arabitol dehydrogenase activities in ery929 genome: (1) d-arabitol dehydrogenase gene 1 (AraDH1, gene locus was g1595.t1 in ery929 and its counterpart is YALI0F02211g in Y. lipolytica CLIB122 CLIB122), (2) d-arabitol dehydrogenase gene 2 (AraDH2, gene locus was g3858.t1 in ery929, and its counterpart is YALI0E05643g in CLIB122), (3) d-mannitol dehydrogenase gene 1 (MDH1, gene locus was g5130.t1 in ery929, its counterpart is YALI0B16192g in CLIB122), (4) d-mannitol dehydrogenase gene 2 (MDH2, gene locus was g2069.t1 in ery929, its counterpart is YALI0D18964g in CLIB122), and (5) xylitol dehydrogenase gene (XDH1, gene locus was g4121.t1 in ery929, its counterpart is YALI0E12463g in CLIB122).

The sequences of the five genes deletion cassettes are shown in the Additional file 1. All genes contained the marker gene hygromycin resistance (hph) gene, full-length synthesized in vitro, and cloned into the pUC19 derivative vector to yield the corresponding gene knockout vector, in which two loxP sites were located at the flanking sequence of hph sequence. For instance, AraDH1 gene knockout cassette (1.5 kb upstream of AraDH1-loxP-hph-loxP-1.5 kb downstream) (total 4.4 kb) was synthesized and cloned into pUC19-derived vector to yield AraDH1 gene knockout vector pSWV-ΔAraDH1-hph (sequence 3 in Additional file 1). Similarly, the other four knockout vectors were constructed in the same strategy. Then the knockout vectors were linearized with EcoRI and NotI to release the knockout cassettes and were transformed into HCY109 (Y. lipolytica ery929ΔKu70), respectively. The identification of deletion mutants was verified according to the above method using the primers listed in Table 1. The identified AraDH1, AraDH2, MDH1, MDH2, XDH1 gene-deleted mutants were designated as ery929ΔKu70Δura3ΔAraDH1 (HCY110), ery929ΔKu70Δura3ΔAraDH2 (HCY111), ery929ΔKu70Δura3ΔMDH1 (HCY112), ery929ΔKu70Δura3ΔMDH2 (HCY113), ery929ΔKu70Δura3ΔXDH1 (HCY114).

Functional identification of mutants deficient in ylAraDH1, ylAraDH2, ylMDH1, ylMDH2 or ylXDH1 genes

To verify whether ylAraDH1, ylAraDH1, ylMDH1, ylMDH2, and ylXDH1 gene deletion can block byproducts mannitol or d-arabitol synthesis during the erythritol fermentation, the wild-type Y. lipolytica CGMCC7326 and mutant strains (HCY110, HCY111, HCY112, HCY113, and HCY114) were inoculated in YPD300 medium for erythritol synthesis. Shake-flask cultures for erythritol production were performed in triplicate using 2-L baffled flasks containing 250 mL rich medium (YPD300), at 30 °C and 220 rpm. Cultures were performed until glucose was depleted.

Construction of EYD1 gene-deletion mutants and marker excision

The EYD1 gene knockout cassette was also synthesized and contained the marker hygromycin resistance gene (hph) gene, in which two loxP sites were located at the flanking sequence of hph sequence. The synthesized cassette was cloned into the pUC19 derivative vector to yield pWSV-EYD1-loxP-hph-loxP (sequence 8 in Additional file 1), which was cut with EcoRI and NotI to release the 4.3 kb fragment to transform the strain HCY113 (△Ku70△ura3△MDH2). Identification of EYD1 gene-deletion mutants was verified according to the above method using the primers PEYD1- knockout –F and PEYD1- knockout –R listed in Table 3. The transformation of plasmid pHB4-621XDH rescued the hph gene. The identified EYD1 gene-deleted mutants were designated as HCY115 (quadruple mutant, △Ku70△ura3△MDH2△EYD1). The fermentation verification of the HCY115 strain was performed as the method used above.

Overexpression of S. cerevisiae RSP5 gene to improve thermoresistance of HCY115 strain

The ubiquitin ligase gene RSP5 of S. cerevisiae [35] was overexpressed in HCY115 (△Ku70△ura3△MDH2△EYD1). The S. cerevisiae RSP5 gene was cloned into the pINA1313 vector to give plasmid pINA1313-ScRSP5 (containing ura3 marker), which was cut with NotI to linearize this plasmid, then transformed to HCY115 strain, and was plated on YPD and cultivated at 34 °C. PCR verified transformants with the pair of primers PScRSP5-F and PScRSP5-R. The corrected strain was designated as HCY117 (△Ku70△MDH2△EYD1:: ScRSP5, no longer ura3-deficient due to revertant by ura3 gene in pINA1313). The erythritol fermentation and growth of the HCY117 strain were evaluated in the medium YPD300 at 30 °C to 35 °C.

Overexpression of genes ZWF1 and GND1 in HCY117

ZWF1 (YALI0E22649g) and GND1 (YALI0B15598g) genes from Y. lipolytica were overexpressed in the HCY117 strain using NotI treated ZWF-GND expression cassette (5′-26S rDNA-hp4d-zwf1-TT-hph-hp4d-gnd1-TT-26S rDNA-3′, (Sequence 9 in Additional file 1), transformed into HCY117 and plated onto YPDH medium (YPD + 400 μg/mL hygromycin). Transformants grown on this medium were identified by PCR to verify the extra ZWF-GND gene and to verify their expression level by qPCR. The true strain was designated as HCY 118 (△Ku70△MDH2△EYD1::ScRSP5:: ZWF1-GND1). The erythritol fermentation and growth of the HCY118 strain were evaluated in the medium YPD300 at 30–35 °C compared with the wild-type strain CGMCC7326 (ery929).

RNA isolation and transcript-level quantification

Shake-flask cultures were grown in rich medium. Cells were collected at an OD600 of 2.0 and stored at − 80 °C in the Trizol solution. Total RNAs were extracted using liquid nitrogen and the TRIzol kit from Sangon Biotech (Shanghai, China). cDNA was obtained using HiScript® III-RT SuperMix for qPCR with gDNA wiper (Vazyme Biotech Co., Ltd). These cDNA samples were used as templates for real-time PCR analysis (qRT-PCR) with the specific primer sets listed in Table 1. The qRT-PCRs were performed using ChamQ™ Universal SYBR®qPCR Master Mix (Vazyme Biotech Co., Ltd) and ABI7500 Real-Time PCR system. The qPCRs proceeded as follows: initial denaturation at 94 °C for 1 min, followed by 40 cycles of denaturation at 94 °C for 10 s, annealing at 62 °C for 30 s, and elongation at 72 °C for 20 s. Specific amplification was confirmed by analysis of melting curves from 65 °C to 95 °C. Gene expressions were normalized to that of the β-actin gene (∆CT method). The fold differences in gene expression between the transformants and the control strains CGMCC7326 were calculated by the 2−ΔΔCT method. All samples were analyzed in triplicate.

Analytical methods

Sugars and polyols were quantified by HPLC using a refractive index detector (Shodex RI101) and a Shodex SP0810 ion exclusion column (300 × 8 mm). Elution was performed at 70 °C using pure water at a flow rate of 1 mL/min. The mass yield of erythritol (Yery) was expressed in g/g from glucose and was calculated from the equation Yery = P/S. The volumetric productivity (Qery) was expressed in g/L·h and was calculated from Qery = P/V · t, where P is the amount of erythritol in the culture liquid at the end of fermentation (g); S is the total amount of glucose consumed (g); V is the initial volume of culture liquid (L), and t is the culture time (h). Glucose consumption rate (Rglu) was calculated as the amount of glucose consumed per hour and per liter of culture medium.