Background

The use of renewable resource such as lignocellulosic biomass (or LC biomass) as feedstock for industrial activities will play an essential role for establishing a more sustainable society [1]. While hemicellulose degradation mainly releases xylose, the enzymatic hydrolysis of cellulose only generates hexose. Many of the commercially interest fungal cellulases display low β-glucosidase activity, rendering the hydrolysis of cellodextrins especially cellobiose a rate-limiting step during enzymatic hydrolysis of LC biomass [2]. To achieve efficient LC biomass conversion, different strategies have been employed in recent years. These include metabolic engineering of the production microorganisms [3, 4], engineering of the cellulases [5, 6], strain evolution [7], and the design and construction of artificial microbial consortia [8], etc.

The unconventional yeast Yarrowia lipolytica is an attractive workhorse for a variety of applications in detergent, food and pharmaceutical industries [9]. Advantageously, its extraordinary ability to accumulate high cellular content of lipids (more than 30% of its dry cell weight) [10] and its “Generally Recognized as Safe” (GRAS) status, have made this yeast an outstanding host for the production of commercially-useful lipids [11,12,13]. Nevertheless, despite these advantages, native strain of Y. lipolytica is unable to use xylose and cellodextrins as carbon sources. Therefore, engineering Y. lipolytica capable of assimilating xylose and cellodextrins is a vital step towards a simultaneous saccharification and fermentation (SSF) process of LC biomass, allowing improved substrate conversion rate and reduced production cost due to low demand of external glucosidase.

In this respect, examples of recent work performed on Y. lipolytica are noteworthy. A xylose-fermenting Y. lipolytica was constructed by introducing Scheffersomyces stipitis xylose reductase (SsXR) and xylitol dehydrogenase (SsXDH), plus overexpressing the endogenous xylulokinase (YlXK). The resulting strain was able to produce citric acid and lipid from xylose [14]. We also made the first attempt to develop an engineered strain of Y. lipolytica to co-ferment cellobiose and xylose [15]. However, the xylose-fermenting Y. lipolytica demonstrated a preferred glucose utilization over xylose. Also, the reported consumption rates of cellobiose and xylose, and the lipid production yield of the engineered strain were inferior to those obtained in glucose fermentation.

Herein, we present work to tackle for the first time the engineering of Y. lipolytica to produce lipids from cellodextrins and xylose by employing rational and combinatorial strategies. First of all, the construction of a cellodextrins-phosphorolytic Y. lipolytica was achieved by expressing the Neurospora crassa cellodextrin transporter (NcCdt1) [16], Clostridium thermocellum cellobiose/cellodextrin phosphorylase (CtCbp/CtCdp) [17], and the S. cerevisiae phosphoglucomutase (ScPgm2p) [15] (Fig. 1). This strain is capable of cellodextrin-phosphorolysis and glucose-1-phosphate (Glc-1P) production, and is expected to display an energetic advantage over the β-glucosidase-producing strains, since less ATP is consumed in glucose phosphorylation in glycolysis [18]. Then, xylose-fermenting ability was introduced into the above strain by expressing the three key genes in xylose assimilation pathway (SsXR, SsXDH and YlXK), and the gene CiGXF1 for xylose uptake. Finally, the obese phenotype was conferred to the engineered Y. lipolytica by overexpressing the genes GPD1 and DGA2. Gpd1p is involved in the production of precursor for TAG [19], while Dga2p catalyses the synthesis of TAG [20]. In addition, the genes involved in lipid degradation were interrupted. Both xylose assimilation and lipids production were enhanced by increasing the cytoplasmic NADPH supply via the overexpression of the genes ZWF1 and GND1, encoding glucose-6-phosphate dehydrogenase and 6-phosphogluconate dehydrogenase, respectively, in the pentose phosphate pathwapy (PPP). This work paved way for the development of the engineered strains of Y. lipolytica to produce valuable chemicals from lignocellulosic hydrolysates at low cost for the advanced generation biorefinery.

Fig. 1
figure 1

The contrustion of cellodextrin-phosphorololytic Y. lipolytica strains and their evluation. a the schematic diagram of the strategies used in strain engineering. Heterologous genes overexpressed included those encoding: CDT cellodextrin transporter, cbp cellobiose phosphorylase, cdp cellodextrin phosphorylase, PGM phosphoglucomutase, Comparison of the growth of (b) ylBPT on glucose, (c) ylBPT on cellobiose and cellotriose, and (d) yl4BPT on cellobiose and cellotriose. Shown are biomass, glucose, cellobiose, and cellotriose concentration versus time. Each data point represents the mean of three independent experiments and the error bar indicates the standard deviation

Full size image

Results

Expression of cellodextrin transporter and cellodextrin phosphorylase in Y. lipolytica

To confer cellodextrin-phosphorolytic capacity to Y. lipolytica, the genes Ctcdp and NcCDT1 were introduced into Y. lipolytica Po1f (Fig. 1a). The Ura+ transformants were selected for their ability to grow on cellobiose, cellotriose and cellotetrose. The results revealed that yeast ylPT co-expressing NcCDT1 and Ctcdp can grow on cellobiose, cellotriose, but not on cellotetrose. Therefore, cellobiose and cellotriose were chosen as carbon sources for the following studies. In addition, the mono-transformant ylP (containing Ctcdp only) cannot grow in these conditions, despite the fact that phosphorylase activity was detectable (0.56 ± 0.04 U/mg-total protein on cellobiose and 0.19 ± 0.02 U/mg-total protein on cellotriose). Most likely, Y. lipolytica is unable to transport cellodextrins into the cell, or at least the rate of transport (and thus the feed rate to the phosphorylases) was too low to support the growth on these substrates.

The recombinant strains were further characterized in shake flask cultures. The results showed that the growth of ylPT (harboring the cellodextrin phosphorolysis pathway) on cellobiose was extremely poor, and only 4.0 g/L cellobiose was consumed over 36 h at a specific consumption rate of 0.17 g/g-DCW/h (Fig. 1, Table 1). Although it grew better on cellotriose, a lag phase of 12 h was detected. As a result, 7.5 g/L of cellotriose was consumed over 36 h and thereafter, the substrate remained unchanged despite prolonged incubation. The μmax and biomass yield of ylPT on cellotriose was calculated as 0.11 h−1 and 0.42 g-DCW/g-cellotriose, respectively, which was 58% and 78% of those obtained on glucose (a μmax of 0.19 h−1 and a biomass yield of 0.54 g-DCW/g-glucose) (Fig. 1, Table 1).

Table 1 Comparison of the growth of recombinant strains of Y. lipolytica on glucose and cellodextrins
Full size table

A previous work demonstrated that the low conversation of Glc-1P to Glc-6P, which resulted in Glc-1P accumulation and thus a greater flux towards glycogen synthesis, impeded the cellobiose phosphorolysis of an engineered yeast [15]. To investigate the limitations that characterize the phosphorolytic assimilation of cellodextrin in present study, the cellular concentrations of Glc-1P and the reserve carbon source of ylDPT grown on cellotriose were measured. The results showed that ylPT contained 30% more cellular Glc-1P and 5 times more glycogen than when it was grown on glucose (Fig. 2). Strikingly, this analysis also revealed the accumulation of intracellular cellobiose up to 20.0 mg/g-DCW in ylPT. Therefore, poor growth of ylPT on cellobiose is likely due to the low cellobiose phosphorolysis efficiency of Ctcdp. Moreover, cellobiose is also the primary product of cellotriose phosphorolysis whose accumulation may inhibit the following phosphorolysis reaction. Thus, the limiting factors of cellodextrin assimilation are the low conversion rate from Glc-1P to Glc-6P, and the low efficiency in the use of the released cellobiose (Fig. 1).

Fig. 2
figure 2

Comparsion of cellular concentration of glycogen, Glc-1P and cellobiose in phosphorolytic Y. lipolytica ylPT (pTEF-cdp1, pTEF-CDT1) compared with Y. lipolytica ylBPT (pTEF-NcCDT1, pTEF-Ctcdp, pTEF-Ctcdp, pTEF-ScPGM2) in areobic growth on 20 g/L glucose, or cellobiose or cellobiose. Samples were taken from the cells grown at the exponential phase

Full size image

Optimization of the cellodextrin-phosphorolytic pathway in Y. lipolytica

To tackle the issue of low growth rate of ylPT on cellobiose and -triose due to the Glc-1P accumulation, the gene PGM2 encoding phosphoglucomutase (PGM) in S. cerevisiae, which showed the preference in catalyzing the reaction from Glc-1P to Glc-6P [15], was introduced into ylPT. In addition, the cellobiose phosphorylase of C. thermocellum was expressed in ylPT to enhance cellobiose assimilation. The performance of the newly engineered strain ylBPT was evaluated in cellobiose and cellotriose fermentations. As illustrated in Fig. 1, ylBPT demonstrated a 12 h lag phase and a μmax of 0.14 h−1 when it was grown on cellobiose. As a result, it consumed all of the cellobiose (20 g/L) over 30 h at a specific consumption rate of 0.29 g/g-DCW/h. Similar results were obtained when the same strain grown on cellotriose. This was accompanied by the decreased cellular content of glycogen and cellobiose resulted from the increased conversion rate of Glc-1P to Glc-6P, which would provide thermodynamic ‘pull’ for cellodextrin phosphorolysis (Fig. 2).

Our last effort was to combine the genes Ctcbp, Ctcdp and NcCDT1, and overexpress them under 4UASTef promoter, aiming to construct a recombinant strain that is able to ferment a mixture of cellobiose and cellotriose efficiently. The resulting strain yl4BPT demonstrated a shortened lag phase of 6 h, a μmax of 0.18 h−1 and a biomass yield of 0.54 g-DCW/ g-cellobiose or g-cellotriose, and thus a shorter fermentation time (24 h), all of which were similar to those values of the same strain grown on glucose (Fig. 1, Table 1).

Engineering Y. lipolytica capable of fermenting xylose efficiently in the presence of glucose

In this work, we first enhanced the expression of the three key genes SsXR, SsXDH and YlXK for xylose assimilation using 4UASTef promoter. In addition, CiGXF1 was installed to facilitate xylose uptake and to bypass the glucose repression [21]. Moreover, the genes ZWF1 and GND1 of the pentose phosphate pathway, which may lead to increased cytoplasmic NADPH supply and enhanced xylose fermentation by intermediates feeding, were also overexpressed (Fig. 3) [22, 23]. Unlike S. cerevisiae for which the growth on xylose was always less efficient than glucose despites great endeavors made on pathway optimization [24, 25], our engineered strain yl4XRHK exhibited similar growth rate (0.18 h−1) and biomass yield (0.51 g-DCW/g-xylose) on xylose to that of obtained on glucose (Fig. 3, Table 1). Compared to the parental strain ylXHK, yl4XRHK exhibited a 6 h shorter fermentation time and a 30% higher specific xylose consumption rate. We also applied a high xylose and glucose ratio of 3:1 to investigate the impact of high glucose concentration on xylose consumption. Expressing CiGXF1 greatly relieved the glucose repression on xylose fermentation, as simultaneous co-consumption of xylose and glucose was observed for the strain yl4XRHK [21]. For all the cultures, xylitol production was negligible even at high xylose/glucose ratio, for which 24 g/L of xylose remained to be consumed after glucose depletion (Fig. 4d).

Fig. 3
figure 3

A schematic illustration of the strategies used in the current study to construct recombinant Y. lipolytica strains able to ferment xylose efficiently. The heterologous genes introduced are indicated in cyan box and the endogenous genes overexpressed are shown in purple box. XR xylose reductase, XDH xylitol dehydrogenase, GXF xylose facilitator, XK xylulokinase, ZWF phosphoglucose dehydrogenase, GND1 6-phosphoglucose dehydrogenase

Full size image
Fig. 4
figure 4

Comparsion of the growth of recombinant Y. lipolytica ylHXK (pTEF-SsXR, pTEF-SsXDH, pTEF-XK) and yl4XRHK (p4UTef-SsXR, p4UTef-SsXDH, p4UTef-XK, pTEF-GiGXF1, pTEF-ZWF1, pTEF-GND1) in areobic growth on (a) xylose; (b), (c) and (d) mixed glucose and xylose. Shown are biomass, xylose and glucose concentration versus time. Each data point represents the mean of three independent experiments and the error bar indicates the standard deviation

Full size image

Development of recombinant strains of Y. lipolytica to co-ferment cellodextrins and xylose efficiently

Encouraged by the success of the above work, we then pursued a more ambitious goal to render Y. lipolytica with both xylose utilization and cellodextrins catabolism ability (Fig. 5). To achieve this, the genes for xylose utilization (SsXR, SsXDH, YlXKS1 and CiGXF1) were introduced into the strain yl4BPT harboring cellodextrin-phosphorolysis pathway (Ctcbp1, Ctcdp1, NcCDT1 and ScPGM2), resulting the strain yl4BX. Please note that the key genes SsXR, SsXDH, ylXKS1, Ctcbp1, Ctcdp1 and NcCDT1 were expressed under 4UASTef promoter, while the rest genes were controlled by the normal TEF promoter.

Fig. 5
figure 5

A schematic illustration of the strategies used in the current study to construct a cellodextrin- and xylose-fermenting Y. lipolytica for lipid overproduction. The pathway for cellodextrin phosphorolysis is marked with light yellow box, while xylose assimilation pathway is marked with pink box. Lipid pathway engineering is indicated in cyan box. PPP pathway overexpression to enhance NADPH supply is marked with gray box. The heterologous genes are written in red and the native genes are given in blue. The interrupted genes are marked with a red deletion line. GPD1 glycerol-3-phosphate dehydrogenase, DGA2 acyl-CoA:diacylglycerol acyl transferase, MFE1 multifunctional beta-oxidation enzyme hydratase-dehydrogenase-epimerase, PEX10 peroxisome biogenesis, TGL TAG lipase

Full size image

Thereafter, the strain yl4BX was characterized in YNB media containing different carbon source. The results showed that yl4BX required 26 h, 30 h, 32 h and 32 h to consume the single carbon source, glucose (20 g/L), xylose (20 g/L), cellobiose (19 g/L) and cellotriose (18.6 g/L), respectively (Table 2). Generally speaking, the lag phase for the cultures containing glucose was slightly shorter than those were deficient in glucose, and simultaneous co-consumption of these carbon sources was achieved. These results confirmed the positive effect of so-called ‘helper substrate’ to promote the use of less favourable substrate by nourishing the cells to achieve good fitness [14, 26]. As a result, the time needed for the complete depletion of carbon source including xylose (10 g/L)/cellobiose (9.5 g/L), xylose (10 g/L)/cellotriose (9.3 g/L) and xylose (10 g/L)/cellobiose (4.8 g/L)/cellotriose (4.7 g/L) was similar (32 h), but slightly longer than that of the fermentation on the mixture of xylose (10 g/L)/glucose (10 g/L), xylose (10 g/L)/glucose (5 g/L)/cellobiose (4.8 g/L), xylose (10 g/L)/glucose (5 g/L)/cellotriose (4.7 g/L). Strikingly, similar biomass yields (≈ 0.52 g/g-glucose equivalent in average) and growth rates (≈ 0.18 h−1 in average) were achieved for the above fermentations (Table 2). The production of xylitol was less than 1.0 g/L for all the xylose containing cultures.

Table 2 Comparison of the growth of recombinant strains of Y. lipolytica on xylose and glucose
Full size table

Engineering Y. lipolytica to accumulate high cellular content of lipids on LC biomass derived sugars

After the construction of a recombinant Y. lipolytica that is able to ferment xylose and cellodextrins in the presence of glucose efficiently, we then explored how his strain can be applied for lipids production on these carbon sources. The strategy we used to increase lipid accumulation is to overexpress GPD1 and DGA2, which are involved in TAG formation, and to interrupt MFE1, PEX10 and TGL4 to prevent ß-oxidation, peroxisome biogenesis [27] and the release of fatty acids from the lipid body [28] (Fig. 5). The contribution of each single overexpression and gene deletion was verified in lipid production on glucose (Additional file 1: Fig. S1). The strain yl4BXP which exhibited the highest lipid accumulation yield on glucose (45% of the biomass) was inoculated into the media containing a mixture of xylose, cellobiose and cellotriose for further investigation. Lipid production was conducted in a 3L-bioreactor in the presence of 25 g/L of each sugar. A second addition of 25 g/L of each substrate was carried out when the total amount of carbon source dropped below 20 g/L. A C/N ratio of 60 was used according to the previous literature, which was reported as the best C/N ratio for lipid accumulation from different carbon sources [29].

As illustrated in Fig. 6, co-consumption of the three sugars by yl4BXP was observed, with xylose being the fastest fermented sugar followed by cellobiose and cellotriose. In the end, yl4BXP consumed a total amount of 150 g/L mixed sugars in less than 5 days, and produced 59.6 g/L of biomass, a yield of 0.39 g-DCW/g-sugar. Lipid accumulation reached 23.8 g/L as estimated from the sum of the extracted fatty acids, a production of 0.16 g-lipid/g-sugar, corresponding to 40.0% of biomass (in DCW) and a productivity yield of 0.22 g/L/h (Fig. 6). Analysis of the fatty acid profile of yl4BXP illustrated a twofold increase in oleic acid (C18:1, 67%) and 73% decrease in linoleic acid (C18:2, 9.7%) in the engineered strain compared with the parental strain po1f-control. While linoleic acid (36%) and oleic acid (34%) was the first and second most abundant fatty acid in the parental strain, palmitoleic acid (C16:1), represents only 13% of total fatty acids, was the second most abundant fatty acid in yl4BXP (Fig. 7).

Fig. 6
figure 6

Lipid prodution of recombinant Y. lipolytica yl4BXP (p4UTef-SsXR, p4UTef-SsXDH, p4UTef- XK, pTEF-GiGXF1, pTEF-ZWF1, pTEF-GND1, pTEF-GPD1, pTEF-DGA2) from mixed carbon source including xylose, cellobiose and cellotriose in YNB meida. Shown are xylose, cellobiose, cellotriose, biomass, xylitol, citrate and lipids concentration versus time

Full size image
Fig. 7
figure 7

Comparison of fatty acid profile of recombinant Y. lipolytica yl4BXP (p4UTef-SsXR, p4UTef-SsXDH, p4UTef- XK, pTEF-GiGXF1, pTEF-ZWF1, pTEF-GND1, pTEF-GPD1, pTEF-DGA2) with the parental strain po1f-control in fed-batch SSF of steam-pretreated wheat straw in bioreactor

Full size image

Lipid production using recombinant Y. lipolytica in fed-batch SSF of wheat straw in bioreactors

To investigate whether the recombinant Y. lipolytica could be useful in the real scenario of LC biomass conversion, fed-batch SSF of steam-pretreated wheat straw was carried out in a 3L bioreactor using the strain yl4BXP. A sequential co-fermentation scheme was employed comprising an initial phase of batch culture on hydrolysate liquor followed by three additions of the cellulase and unwashed solids of the wheat straw slurry. It has been suggested that such a fermentation strategy is advantageous due to the detoxification of the inhibitory compounds by yeast during the batch phase and the improved tolerance of cells through progressive adaptation [30].

During the initial 48 h of batch cultivation, 8.6 g/L glucose was depleted and 26 g/L of xylose representing 79 wt% of the total available xylose in hydrolysate was consumed, and 2.76 g/L of lipids were produced (Fig. 8). To evaluate the feasibility of the process, Cellic CTec2 was added at a dosage to avoid excessive hydrolysis of glucan into glucose (10.0 FPU/g cellulose), and also at a higher dosage of 15.0 FPU/g cellulose for comparison. During the subsequent fed-batch SSF at lower enzyme dosage, xylose was co-consumed with the released sugars from enzymatic hydrolysis of glucan. No accumulation of glucose and cellodextrins was observed, and this observation was correlated with lipid production (detected in the form of FAMEs), reflecting a continuous biomass formation. After 144 h of cultivation, the fed-batch SSF resulted in a lipid concentration of 8.5 g/L, which corresponded to an overall production yield of 0.1 g-lipid/g-sugar (Fig. 8). Glucose was depleted, and 2.3 g/L of xylose sustained in the end of fermentation. Xylose utilization reached 96 wt% of the total xylose loading. The co-consumption of glucose, xylose and cellodextrins were sustained throughout the 144 h of co-fermentation. Prolonging the fermentation did not yield further gain in lipid titer.

Fig. 8
figure 8

Lipid prodution of recombinant Y. lipolytica yl4BXP (p4UTef-SsXR, p4UTef-SsXDH, p4UTef- XK, pTEF-GiGXF1, pTEF-ZWF1, pTEF-GND1, pTEF-GPD1, pTEF-DGA2) by fed-batch SSF of steam-pretreated wheat straw in bioreactor. Shown are glucose, xylose, biomass, xylitol, citrate and lipid concentration versus time

Full size image

Discussions

It is interesting to note that phosphorolytic capability was recently conferred to S. cerevisiae (expressing NcCdt1 and a Cbp from Saccharomyces degradans) [18]. In the previous work, the energetic benefits of phosphorolysis were evidenced because S. cerevisiae endowed with the cellobiose phosphorolysis pathway produced more biomass and ethanol than a cellobiose-hydrolyzing strain. Consistent with these observations, the newly developed strain in this work showed better performance, e.g., shorter lag phase, faster rates of growth and substrate consumption, than the previously engineered yeast relied on cellobio-hydrolytic activity [31]. However, our results imply that the presence of the phosphorolysis pathway in Y. lipolytica did not procure an obvious energetic benefit, as the biomass yield of yl4BPT was similar to the previous reported Y. lipolytica possessing cellobiose hydrolytic activity [31]. It is noteworthy that further overexpressing either Ctcbp, Ctcdp or CDT1 using 8UASTef promoter did not improve the growth rate and biomass yield further. Nevertheless, yl4BPT is still the most efficient cellodextrin-fermenting recombinant yeast reported so far.

Several recent studies have shared the same interest in develo** a xylose-fermenting Y. lipolytica by employing different strategies. As such, although Y. lipolytica PO1g overexpressing SsXR and SsXDH was unable to grow on xylose, the following adaptation has enabled the growth of the strain on xylose with a doubling time of 25 h [26]. In a second work, Y. lipolytica YlSR001 was directly adapted in xylose culture which allowed the isolation of a mutant strain capable of slightly growing on xylose [32]. The author disclosed that the insufficient XDH activity was the limiting factor for xylose assimilation. The overexpression of XDH enhanced the growth of the strain on xylose, but unfortunately, still at an extremely slow rate [32]. Although our previous strategy to construct a xylose-fermenting Y. lipolytica by overexpressing SsXR, SsXDH and YlXK was successful, xylose assimilation in the presence of glucose of the engineered strain was not optimal as sequential consumption of xylose after glucose was still the case due to glucose repression [14]. In addition, the accumulation of xylitol on xylose occurred after glucose depletion when a high initial concentration of xylose was applied, which indicated Y. lipolytica suffered from the cofactor imbalance issue since ssXR consumes NADPH and XDH generates NADH [26]. In this respect, our current strategy was obviously more successful. CiGxf1p is a non-glucose preference transporter which shows highest efficacy in xylose uptake in the presence of glucose [21]. Its Vmax value is one order of magnitude higher than the xylose-H+ symporter CiGxs1p [33], and the overexpression of CiGxf1p in Kluyveromyces marxianus has greatly improved xylitol production due to the enhanced xylose uptake [Calculations and statistics

Specific rate of substrate consumption (g-s/g-DCW/h) was estimated according to the following equation:\({\mathrm{q}}_{\text{sub}}=\frac{{\mu }_{\mathrm{max}}}{{Y}_{\text{X/s}}}\), where \({Y}_{\text{X/s}}=\frac{\text{dX}}{{\text{ds}}}\) represents the biomass yield coefficient. The biomass and product yields were calculated as the ratio of the amount of biomass or products formed divided by the amount of carbon source consumed. The maximum specific growth rate μmax (h−1) was calculated from the plot of biomass concentration versus time for exponentially growing cells. The duration of lag phase was estimated using an online tool as described previously [50]. All the experiments were performed at least in triplicate, and the mean values ± standard deviation were shown.