Engineering a wild-type diploid Saccharomyces cerevisiae strain for second-generation bioethanol production

Bioethanol produced from lignocellulose, i.e., second-generation bioethanol, is benefit for a sustainable energy supply. Its cost-effective production process depends on the complete and rapid utilization of all the sugars in the hydrolysates derived from lignocellulosic raw materials (Ko et al. 2016; Moysés et al. 2016; Peng et al. 2012; Zhou et al. 2012). The budding yeast Saccharomyces cerevisiae is a prominent microorganism that has traditionally been used in industrial bioethanol production because of its numerous inherent advantages (Demeke et al. 2013). However, this natural ethanol producer must overcome at least two new challenges when the fermentation substrate is lignocellulosic hydrolysates, rather than starch-based materials. First, natural S. cerevisiae cannot effectively metabolize pentose sugars, particularly d-xylose, the second most abundant sugar in lignocellulosic materials because it lacks an effective initial metabolic pathway (Hahn-Hägerdal et al. 2007; van Maris et al. 2006). Second, the individual and synergistic negative interactions derived from the numerous inhibitory compounds that are formed during the pretreatment process and the hydrolytic release of sugars exert serious negative effects on the fermentation performance of S. cerevisiae (Ko et al. 2016; Palmqvist and Hahn-Hägerdal 2000). Therefore, for the economically viable and sustainable production of lignocellulosic bioethanol, it is necessary to confer the capacity to co-ferment glucose and xylose on an S. cerevisiae strain and to enhance its resistance to harsh production environments (Demeke et al. 2013; Li et al. 2015; Sharma et al. 2016).

The genetic background of the host strain significantly affects the performance of the recombinant strain. Normally, despite strain variation, polyploid, wild-type S. cerevisiae is a better ethanol producer than the haploid strain, which is generally used in the laboratory (Brandberg et al. 2004; Li et al. 2015; Sonderegger et al. 2004; Yamada et al. 2011). Although the genetic background issue is complex, basic selection principles are feasible. By evaluating strains for their glucose-fermenting power, stress tolerance, and the ability to metabolize pentose, we can select a strain suitable for use as the chassis cell in lignocellulosic ethanol production (Li et al. 2015).

The following two heterologous initial xylose metabolic pathways (Fig. 1a) were introduced into an S. cerevisiae strain: the XR-XDH pathway, which is composed of xylose reductase (XR) and xylitol dehydrogenase (XDH), and the XI pathway, which is composed only of xylose isomerase (XI). For the XR-XDH pathway, xylose is first converted to xylitol by XR, which has a higher affinity for NADPH than NADH. Then, xylitol is oxidized into xylulose by XDH, which depends exclusively on NAD⁺ (Ho et al. 1999; Peng et al. 2012; Wang et al. 2004). The imbalance in redox metabolism caused by the different coenzyme preferences between XR and XDH result in the accumulation of the byproduct xylitol and a lower ethanol yield; cofactor engineering resulted in limited improvement (Hou et al. 2009a; Zha et al. 2012; Zhang et al. 2010). For the XI pathway, since xylose is directly isomerized to xylulose with no coenzyme participation, incorporating this pathway is a direct and effective strategy for initiating xylose metabolism in S. cerevisiae (Demeke et al. 2013; Diao et al. 2013; Zhou et al. 2012). Highly efficient XI activity is a prerequisite for rapid and efficient xylose fermentation in S. cerevisiae (Brat et al. 2009; Kuyper et al. 2003; Madhavan et al. 2009; Walfridsson et al. 1996; Zhou et al. 2012). Recently, we screened a bovine rumen metagenomic library and discovered the novel XI gene Ru-xylA. Ru-XI exhibited higher activity (1.31 U mg⁻¹) in S. cerevisiae (Bao et al. 2013; Hou et al. 2016a) than Pi-XI (the XI from Piromyces sp.), which is a prototypically active XI (Kuyper et al. 2003, 2004).

To compensate for the shortfall in downstream metabolic flux in ethanol production, the xylulokinase Xks1p and the enzymes in the non-oxidative PPP are normally overexpressed in a xylose-utilizing strain (Bamba et al. 2016; Peng et al. 2012; Sharma et al. 2016). Additionally, the inactivation of aldose reductase Gre3p, which is encoded by GRE3, to decrease the formation of the byproduct xylitol (Bamba et al. 2016; Fujitomi et al. 2012; Hahn-Hägerdal et al. 2007; Kuyper et al. 2005a; Peng et al. 2012), and the inactivation of alkaline phosphatase Pho13p, encoded by PHO13, to reduce ATP waste (Bamba et al. 2016; Lee et al. 2014; Shen et al. 2012; Van Vleet et al. 2008), were also the considered strategies. Normally, xylose is absorbed non-specifically and insufficiently by hexose transporters and is competitively inhibited by glucose in S. cerevisiae (Subtil and Boles 2012). Therefore, to address this problem, both wild-type and mutated endogenous and heterologous transporters were screened (Diao et al. 2013; Farwick et al. 2014; Moon et al. 2013; Nijland et al. 2014; Runquist et al. 2009; Shin et al. 2015; Wang et al. 2015, 2016). Moreover, with higher efficiency and/or specificity for xylose, a transporter lacking glucose inhibition may relieve the glucose repression effect in a xylose absorption node, thereby increasing xylose metabolism. Recently, we found the mutant MGT05196 ^N360F, which is derived from the Meyerozyma guilliermondii transporter gene MGT05196 and exhibits xylose-specific transport and no glucose inhibition (Wang et al. 2015).

In theory, such rational genetic modifications should endow S. cerevisiae with the capacity to ferment xylose; nonetheless, the xylose-fermentation efficiency of the resulting engineered strain was less than expected, indicating the influence of unknown key factors. Additionally, it remains difficult to rationally enhance the tolerance of a xylose-fermenting strain to the inhibitors present in lignocellulosic hydrolysates. Therefore, non-rational, adaptive evolutions were performed in media that contained xylose as the sole carbon source or toxic lignocellulosic hydrolysates to further select for enhanced characteristics both in xylose metabolism efficiency and robustness (Heer and Sauer 2008; Kuyper et al. 2005b; Moysés et al. 2016; Sharma et al. 2016).

The co-fermentation of hexose and pentose is considered important for the effective industrial production of bioethanol. Normally, the concept of co-fermentation usually encompasses only a strain’s ability to ferment xylose and glucose in the presence of both sugars. Therefore, little attention is given to the synchronous fermentation of both sugars. In fact, the obvious hysteresis of xylose fermentation compared with glucose has been extensively observed, manifestations of which exhibit a delay not only with respect to initial xylose utilization and the xylose consumption rate but also the time to maximum ethanol production; in other words, the ethanol production value increases incrementally after glucose exhaustion (Ko et al. 2016; Peng et al. 2012; Reider Apel et al. 2016; Sharma et al. 2016). Therefore, a yeast strain that co-ferments xylose and glucose with the same consumption rate (i.e., synchronously) would be ideal. We believe that the synchronization of sugar fermentation will reduce the apparent fermentation time, improve equipment utilization and employee effectiveness, and play a vital role in the realization of the scaled-up production of second-generation bioethanol.

In the present work, the diploid S. cerevisiae strain BSIF, which we have previously shown to be robust and to possess a good background for xylose metabolism (Li et al. 2015), was used as the chassis cell. For rational strain design, two novel proprietary heterologous genes, MGT05196 ^N360F (ZL 2014105263569) (Wang et al. 2015) and the Ru-xylA (US 8,586,336 B2, EP 2679686, ZL 201110042170.2) (Bao et al. 2013; Hou et al. 2016a), were integrated into the genome of diploid strain BSIF. Together with other strategies, including the overexpression of the genes XKS1 (xylulokinase), TAL1 (transaldolase), TKL1 (transketolase), RKI1 (ribose 5-phosphate isomerase) and RPE1 (ribulose 5-phosphate epimerase) in the non-oxidative PPP and the inactivation of GRE3 (aldose reductase) and PHO13 (alkaline phosphatase), a functional xylose metabolic pathway was constructed in BSIF. For non-rational strain design, the recombinant cells were alternately cultured in media using xylose as the sole carbon source or SECS liquor (the leach liquor of steam-exploding corn stover) as stresses for adaptive evolution.

Recently, strains of S. cerevisiae engineered for second-generation bioethanol production have been continuously improved, inching closer to industrialization. In general, multiple modifications have been used, including the rational establishment of a xylose metabolic pathway and non-rational adaptive evolution to improve xylose metabolic capacity and inhibitor tolerance (Kim et al. 2013; Ko et al. 2016; Kuyper et al. 2005a; Peng et al. 2012; Van Vleet and Jeffries 2009). Moreover, using robust yeast strains as chassis cells has also proved to be an effective strategy (Demeke et al. 2013; Diao et al. 2013; Li et al. 2015). In the present work, based on our previous study, the S. cerevisiae wild-type diploid strain BSIF, which is highly robust and possesses a good genetic background for xylose metabolism, was used as the chassis cell (Li et al. 2015). The following two key novel elements were introduced into the chassis cell: MGT05196 ^N360F, which encodes a xylose-specific, glucose-insensitive transporter derived from the M. guilliermondii transporter gene MGT05196 (Wang et al. 2015), and Ru-xylA, which encodes a xylose isomerase (XI) with higher activity in S. cerevisiae and was derived from a screen of a bovine rumen metagenomic library (Bao et al. 2013). Another highlight in strain construction was the use of multiple adaptive evolutions under xylose and inhibitor stresses.

Overall, these modifications resulted in strain LF1, which has an excellent xylose fermentation capacity in media prepared from both sugars (Fig. 3) and lignocellulosic hydrolysates (Fig. 5). When xylose was the sole carbon source, LF1 consumed all of the 40 g L⁻¹ xylose in 12 h with initial inoculum size of 0.5 g DCW L⁻¹, and resulted in an ethanol yield of 0.446 g g⁻¹. In contrast, CIBTS0735, a similarly engineered strain harboring the xylose isomerase gene xylA from Piromyces and the transporter gene GXF1 from Candida intermedia (Diao et al. 2013), required 16 h to consume nearly the same amount of xylose with initial inoculum size of 0.63 g DCW L⁻¹, and resulted in a lower ethanol yield of 0.412 g g⁻¹. Moreover, GS1.11-26, another similarly engineered strain that harbors the xylose isomerase gene xylA from Clostridium phytofermentans and overexpresses the endogenous transporter gene HXT7 (Demeke et al. 2013), resulted in a slightly higher ethanol yield of 0.460 g g⁻¹ but also required 17 h to ferment less xylose (35 g L⁻¹) with a larger initial inoculum size (1.3 g DCW L⁻¹). In a glucose–xylose mixture, LF1 consumed 80 g L⁻¹ glucose and 40 g L⁻¹ xylose in 16 h with an ethanol yield of 0.475 g g⁻¹, which was greater than 93% of the theoretical yield. In contrast, the corresponding values for CIBTS0735 (Diao et al. 2013) were 20 h and an ethanol yield of 0.454 g g⁻¹ in a fermentation with the same amount of mixed sugars. Additionally, GS1.11-26 (Demeke et al. 2013) required only 13 h to produce an ethanol yield of 0.46 g g⁻¹ but fermented less sugar (36 g L⁻¹ glucose and 37 g L⁻¹ xylose) with a larger initial inoculum size (1.3 g DCW L⁻¹) compared with the 0.5 and 0.63 g DCW L⁻¹ of LF1 used in the present work (Fig. 5) and CIBTS0735 (Diao et al. 2013), respectively. Other similarly engineered strains did not show improvement in key fermentation parameters, as discussed above (Ko et al. 2016; Romaní et al. 2015; Smith et al. 2014). Additionally, LF1 also performed well in the fermentation of two types of lignocellulosic hydrolysates, even when supplied with the simple nitrogen source urea rather than yeast extract and peptone (Fig. 5). Moreover, LF1 prominently displayed highly synchronized glucose and xylose utilization not only in a mixture of the sugars (with a utilization of 77.6% xylose and glucose depletion at 12 h) but also in SPPR hydrolysate, and ethanol production in LF1 peaked quickly (Fig. 3d vs. f, 5). This characteristic is a significant advantage in second-generation bioethanol production because it will save fermentation time, reduce staff input, and so on.

During approximately 40 years of metabolic engineering in yeast to co-ferment xylose and glucose, researchers have gradually realized that it is easier to establish a xylose metabolic pathway than to improve the efficiency of xylose fermentation and glucose-xylose co-fermentation. Although much research toward this goal has been performed, key factors in this strategy remain unknown. Therefore, non-rational adaptive evolution is generally used for strain engineering (Almeida et al. 2007; Heer and Sauer 2008b; Kuyper et al. 2005b). In the present work, we also focused on the strain breeding, therefore, the contribution for ‘key factors’ exploration maybe not enough.

Increasing the activity of xylose isomerase (XI), which isomerizes xylose directly to xylulose, is considered a good basis for establishing a xylose metabolic pathway in S. cerevisiae (Brat et al. 2009; Kuyper et al. 2003; Madhavan et al. 2009; Walfridsson et al. 1996; Zhou et al. 2012). The specific activity of XI in LF1 (0.46 U mg⁻¹) was ~31.0% lower than in XH7 (which exhibited the highest activity of the strains in this study, 0.67 U mg⁻¹) but was higher than in BSN3 (0.30 U mg⁻¹) (Fig. 6a). However, the xylose fermentation capacity of LF1 was much improved compared with that of XH7 and BSN3 (Fig. 3; Table 3). This result implied that maintaining a moderate XI activity has a positive effect on the enhancement of xylose fermentation. The overexpression of Xks1p and four enzymes in the non-oxidative PPP is a common strategy to strengthen xylose metabolism (Bamba et al. 2016; Kuyper et al. 2005a; Peng et al. 2012; Sharma et al. 2016). Previous reports have shown that moderate, rather than extremely high, Xks1p activity is more beneficial for growth and ethanol production from xylose because this enzyme catalyzes an ATP-consuming reaction (Jin et al. 2003; Peng et al. 2011). In the present work (Table 4), the Xks1p (encoded by XKS1) expression level fluctuated noticeably. However, the expression level remained relatively high and constant in both improved xylose-fermenting strains, XH7 and LF1. Therefore, such a Xks1p expression level is apparently suitable for xylose metabolism. However, BSN3 also exhibited a similar Xks1p expression level (Table 4), even though this strain has not undergone any adaptive evolution and utilizes xylose poorly (Table 3, Line 1). Obviously, then, a suitable Xks1p activity is not the sole factor governing xylose metabolic capacity. Except for that of ribose 5-phosphate isomerase (encoded by RKI1), the expression levels of the enzymes in the non-oxidative PPP showed no significant change (Table 4). After the first round of adaptive evolution under xylose stress, the expression of RKI1 was upregulated from 14.4 ± 1.3 in BSN3 to 238.8 ± 15.3 in XH7 (in fold change relative to the original chassis, BSIF). Then, inhibitor stress dropped its expression to 146.78 ± 3.5 in XHR11 and remained at 144.5 ± 0.6 in LF1 (Table 4). As discussed above, XH7 and LF1 (but not BSN3 and XHR11) showed improved xylose fermentation, which indicates that ribose 5-phosphate isomerase, rather than the other three enzymes in the non-oxidative PPP, are important for xylose fermentation, as well as other key factors.

The transport issue is another limiting factor for xylose fermentation. Mgt05196p^N360F, a xylose-specific transporter with no glucose inhibition (Wang et al. 2015), was therefore introduced into the chassis cells. As expected, the transporter enhanced xylose utilization to some extent by simple expression. Moreover, after the Mgt05196p^N360F-containing strain was evolved in the presence of xylose (Fig. 1b), glucose and xylose utilization efficiency and synchronicity improved (Fig. 3; Table 3). Strain XHR11, which lacked this transporter, did not show any change in its xylose metabolic capacity during the same adaptive evolution, and the copy numbers and transcription levels of MGT05196 ^N360F remained constant in XHR11-360F and LF1. Therefore, we deduced that Mgt05196p^N360F absorbed significantly more intracellular xylose, which resulted in stronger xylose stress tolerance in these cells and drove evolution beneficially toward xylose metabolism. Therefore, Xks1p expression recovered in LF1 relative to its parent. However, ribose 5-phosphate isomerase expression did not recover in LF1 (Table 4), which indicates that non-rational adaptive evolution led to random changes in other genes, and the key factors contributing to xylose metabolism must be numerous. An exploration of these unknown factors and their function via inverse metabolic engineering would be a significant undertaking. S. cerevisiae maintains high levels of glycolysis and PPP enzymes when glucose is present, and it shifts to respiratory metabolism after glucose is exhausted (Salusjarvi et al. 2008). Because xylose is metabolized through the PPP and glycolysis, its utilization will obviously drop as well. Furthermore, in respiratory metabolism, pyruvate prefers to enter the tricarboxylic acid cycle instead of producing ethanol. Therefore, using a specific transporter to facilitate the synchronized co-fermentation of glucose and xylose is beneficial for their co-utilization.

It was observed that xylose consumption was inhibited by the presence of glucose; therefore, the specific xylose consumption rates in the sugar mixture were lower than when xylose was the sole carbon source (Table 3, line 9 vs. 12, 10 vs. 13, 11 vs. 14) (Demeke et al. 2013; Diao et al. 2013; Peng et al. 2012; Shen et al. 2012), However, the same parameter was still kept lower (Fig. 3g) in the xylose consumption phase (Hou et al. 2009b) after the depletion of glucose from the sugar mixture. We defined such a phenotype—the continued impact of glucose on xylose metabolism after glucose depletion—as the “post-glucose effect” in the present work, which is an underappreciated phenomenon frequently evident in data shown in previous publications (Demeke et al. 2013; Diao et al. 2013; Peng et al. 2012; Shen et al. 2012). Exploration of the mechanisms underlying the post-glucose effect and the discovery of elements to overcome this phenotype will be the focus of future research.

In the present work, two novel proprietary heterologous genes (MGT05196 ^N360F and Ru-xylA) were combined with other multiple genetic modifications and three rounds of adaptive evolution to engineer the diploid S. cerevisiae chassis strain BSIF. The final resulting strain of this effort, LF1, consumed all xylose (40 g L⁻¹) and glucose (80 g L⁻¹) in 16 h with an ethanol yield that was over 93% of the theoretical yield. LF1 also performed well in lignocellulosic hydrolysates, with ethanol yields of over 80.0% of the theoretical yields. Moreover, LF1 exhibited highly synchronous utilization of glucose and xylose. These results highlight the great potential for the practical use of LF1 in the production of second-generation bioethanol. Despite the directed genetic changes detailed in this work, some unknown factors derived from adaptive evolution were also responsible for the superior fermentation performance of LF1. We also outlined the “post-glucose effect” phenomenon. An exploration of its underlying mechanisms remains key.