Metabolic engineering considerations for the heterologous expression of xylose-catabolic pathways in Saccharomyces cerevisiae

Xylose, the second most abundant sugar in lignocellulosic biomass hydrolysates, can be fermented by Saccharomyces cerevisiae expressing one of two heterologous xylose pathways: a xylose oxidoreductase pathway and a xylose isomerase pathway. Depending on the type of the pathway, its optimization strategies and the fermentation efficiencies vary significantly. In the present study, we constructed two isogenic strains expressing either the oxidoreductase pathway (XYL123) or the isomerase pathway (XI-XYL3), and delved into simple and reproducible ways to improve the resulting strains. First, the strains were subjected to the deletion of PHO13, overexpression of TAL1, and adaptive evolution, but those individual approaches were only effective in the XYL123 strain but not in the XI-XYL3 strain. Among other optimization strategies of the XI-XYL3 strain, we found that increasing the copy number of the xylose isomerase gene (xylA) is the most promising but yet preliminary strategy for the improvement. These results suggest that the oxidoreductase pathway might provide a simpler metabolic engineering strategy than the isomerase pathway for the development of efficient xylose-fermenting strains under the conditions tested in the present study.


Introduction
Global climate change has accelerated efforts to find eco-friendly alternatives for fossil fuels. One idea is to use wood wastes and agricultural residues called lignocellulosic biomass, which does not interfere with food or the environment [1]. Lignocellulosic biomass, mainly composed of cellulose and hemicellulose, is hydrolyzed into glucose, xylose, and other simple and minor sugars, which can be transformed into biofuels and chemicals by microbial fermentation [2].

PLOS ONE
PLOS ONE | https://doi.org/10.1371/journal.pone.0236294 July 27, 2020 1 / 18 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 The yeast Saccharomyces cerevisiae is an industrial microorganism with superior sugar fermentation capabilities and stress tolerance. However, this yeast cannot metabolize xylose, requiring the introduction of a heterologous xylose pathway [3,4] as summarized in Fig 1A. The first step is to introduce either the NAD(P)H-specific xylose reductase/NAD + -specific xylitol dehydrogenase (oxidoreductase, XR/XDH) pathway derived from Pichia stipitis or the xylose isomerase (XI) pathway derived from various anaerobic microorganisms, both of which convert xylose to xylulose. Next, xylulose is converted into xylulose-5-phosphate by xylulokinase either by endogenous but overexpressed S. cerevisiae XKS1 or P. stipitis XYL3. Finally, xylulose-5-phosphate is metabolized into ethanol through the native pentose phosphate (PP) pathway connected to glycolysis in S. cerevisiae. In engineered strains of S. cerevisiae expressing the xylose oxidoreductase pathway, the rate of xylose consumption and ethanol productivity are relatively high, but xylitol, glycerol, and acetate are accumulated as byproducts [3,5]. This byproduct accumulation is mainly due to an unbalanced cofactor preference of the xylose oxidoreductase pathway, leading to a shortage of NAD + [3]. On the other hand, the xylose isomerase pathway is cofactor-independent, the expression of which in S. cerevisiae can lead to a high ethanol yield with minimal byproduct accumulation even under anaerobic conditions [6]. However, slow growth and xylose consumption were commonly observed in the engineered S. cerevisiae strains expressing the xylose isomerase pathway compared to those expressing xylose oxidoreductase pathway [5,7,8].
Adaptive evolution have been the most commonly used and the most effective approach to improve both the strains expressing the oxidoreductase pathway [3,9,10] and the strains expressing the isomerase pathway [11][12][13]. Some of the evolved strains were subjected to genome sequencing to identify genetic changes responsible for the improved phenotypes. In prior studies, the loss of function mutation of the PHO13 gene encoding phosphatase with a broad substrate spectrum was identified as a key mutation of an evolved strain expressing the xylose oxidoreductase pathway [3,14,15]. Deletion of the PHO13 gene (pho13Δ) now provides simple, effective, and transferrable to different strain backgrounds expressing the xylose oxidoreductase pathway [3,[14][15][16]. Moreover, it was further confirmed that pho13Δ leads to transcriptional and metabolic shifts toward efficient xylose fermentation [17,18]. However, it has not been clearly understood how strains expressing the isomerase pathway can be simply improved, although there have been several attempts of genome sequencing of the evolved strains expressing the isomerase pathway [19][20][21][22][23][24].
In this study, we constructed two isogenic strains expressing either the xylose oxidoreductase pathway or the xylose isomerase pathway through a precise Cas9-based genome integration strategy [25,26]. Deletion of the PHO13 gene, adaptive evolution, the upregulation of the PP pathway, and some other strategies were performed to identify the most critical and simple factor to improve the strain expressing the xylose isomerase pathway.

Plasmid construction and strain engineering
The strains and plasmids used in this study are summarized in Table 1 and S1 Table, respectively. The detailed materials and methods for plasmid and strain construction are available in the online supplementary information (S1 Text).

Volumetric growth rate analysis at various xylose concentrations
To compare growth rate after PHO13 gene deletion, all strains were pre-cultured in 10 mL of YP medium containing 20 g/L of glucose, and the pre-cultured cells were harvested at midexponential phase and inoculated into 3 mL of YP medium containing various concentrations of xylose after washing twice with sterilized water. Growth rate analysis was performed in 14-mL Round-Bottom Tubes (SPL, Pocheon, Korea) at 30˚C and 250 rpm with a low initial cell density (0.5 g DCW/L). The control (XYL123 and XI-XYL3) and pho13Δ (XYL123 pho13Δ and XI-XYL3 pho13Δ) strains were compared at 1-200 g/L xylose. Volumetric growth rates (g/ L-h) were calculated based on the starting and ending points of the exponential phase. All experiments were performed in biological triplicate.

Transcriptional analysis by RT-qPCR
RT-qPCR was performed by extracting RNA from cells of the exponential phase as previously described [18]. All of the strains were grown in YP media containing 20 g/L glucose or 40 g/L xylose. The cDNA solution, prepared from 1 μg of RNA using the ReverTra Ace1 qPCR RT Master Mix (TOYOBO, Osaka, Japan), was used directly with primers and iQ TM SYBR Green Supermix (Bio-Rad, Hercules, CA, USA) for quantitative PCR (qPCR). qPCR was performed using a CFX Connect TM Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA). The primers used for RT-qPCR are described in S2 Table. All of the measurements were performed in three technical replicates for each biological triplicate.

Adaptive laboratory evolution
After pre-cultivation in YP medium containing 20 g/L of glucose for 24 hours at 250 rpm, yeast cells were harvested by centrifugation at 15,928 ×g, at 4˚C for 1 min. The pre-cultured cells were washed with distilled water, and the cell pellet was inoculated into 20 mL of YP medium containing 40 g/L or 100 g/L xylose under oxygen-limited conditions (80 rpm). The initial cell densities were adjusted to 0.5 g DCW/L. Growth adaption was performed at 30˚C in a 100-mL Erlenmeyer flask using a rotary shaker at 80 rpm. The cells were transferred to fresh medium when they reached exponential phase. The growth adaption was continued for about 90 days. To confirm that the strains had evolved, three independent colonies were isolated from the YPD agar plate and evaluated by fermentation performances under oxygen-limited conditions (80 rpm).

HPLC analysis
Quantitation of xylose, xylitol, glycerol, acetate, and ethanol in the culture was analyzed by a high-performance liquid chromatography (HPLC; Agilent Technologies, 1260 series, USA) equipped with a Rezex-ROA Organic Acid H+ (8%) (150 mm × 4.6 mm) column (Phenomenex Inc., Torrance, CA, USA). Columns were eluted with 0.005 N H 2 SO 4 at 50˚C, and the flow rate was set at 0.6 mL/min, as described previously [28]. Acetate was not detected in all fermentations, and the results were omitted from the figures and tables.

Intracellular metabolite extraction and derivatization
Metabolite extraction was carried out with some modification of the previously described method [29]. Briefly, 5 mL of cell cultures at mid-exponential growth phase were quenched by quick injection into 25 mL of 60% (v/v) cold methanol (HEPES, 10 mM; pH 7.1) at -40˚C. The cells were centrifuged at 3,134 ×g at -20˚C for 5 min, then discard supernatant thoroughly. Subsequently, 1 mL of 75% (v/v) boiling ethanol (HEPES, 10 mM; pH 7.1) was added to the quenched cell pellet, then make sure that cell pellet should be suspended well with boiling ethanol solution. The mixture was then vortexed for 30 s in a max force, incubated at 80˚C for 5 min. The cell residues were separated from the extract by centrifugation at 15,928 ×g at 4˚C for 1 min. The supernatant was then vacuum-dried for 5 h using a speed vacuum concentrator (Labconco, Kansas City, MO, USA). The vacuum-dried samples were derivatized by methoxyamination and trimethylsilylation as previously described with some modifications [29]. For methoxyamination, 40 μL of methoxyamine hydrochloride in pyridine (40 mg/mL; Sigma-Aldrich, St. Louis, MO, USA) was added to the samples and incubated at 30˚C for 90 min. For trimethylsilylation, 40 μL of Nmethyl-N-(trimethylsilyl)trifluoroacetamide (Sigma-Aldrich, St. Louis, MO, USA) was added to the samples and incubated at 37˚C for 30 min.

Intracellular metabolite analysis using GC/MS
GC/MS analysis was conducted using an Agilent 6890 GC equipped with an Agilent 5973 MSD as described previously with some modifications [17]. A 1 μL aliquot of derivatized samples was injected into the GC in a split mode (10:1) and separated on an RTX-5Sil MS column (30 m × 0.25 mm, 0.25-μm film thickness; Restek, Bellefonte, PA, USA). The initial oven temperature was set at 75˚C for 1 min, and then ramped at 15˚C/min to a final temperature of 300˚C, held for 2 min. Helium was used as a carrier gas at a constant flow rate of 0.7 mL/min. The temperatures of ion source and transfer line were set at 230˚C and 280˚C, respectively. An electron impact of 70 eV was used for ionization. The mass selective detector was operated in scan mode with a mass range of 50-550 m/z.

Construction and comparison of two isogenic strains expressing xylose oxidoreductase pathway or xylose isomerase pathway
Two isogenic strains expressing either a xylose oxidoreductase pathway (XYL1-XYL2) or the xylose isomerase pathway (xylA) were constructed as follows ( Fig 1B). For the origin of the genes, XYL1 and XYL2 from yeast P. stipitis [28] and xylA from anaerobic fungus Orpinomyces sp. (Gen-Bank No. MK335957) were used which are known to have the highest catalytic activities among the same group of enzymes tested [30,31]. Because acetaldehyde dehydrogenase encoded by the ALD6 gene plays a major role in acetate accumulation [32], and because acetate is detrimental to xylose metabolism of the oxidoreductase strains [3] as well as the isomerase strains [33,34], the ALD6 gene was often selected as knockout target for xylose strains [35,36]. In the present study, therefore, the xylose pathway genes, XYL1-XYL2 or xylA, were genome-integrated by replacing the ALD6 gene by a Cas9-based genome integration strategy, resulting in the XYL12 (ald6::XYL1-XYL2) and the XI (ald6::xylA) strains, respectively. Next, the XYL3 gene encoding xylulokinase, of which overexpression is required for both pathways, was genome-integrated at an intergenic region (int#1, Fig 1B), resulting in the XYL123 and XI-XYL3 strains, respectively.
When fermenting 40 g/L xylose under oxygen-limited conditions with a low initial cell density (0.5 g DCW/L), the resulting strains showed different phenotypes; while the XYL123 strain consumed over 90% xylose and produced ethanol within 72 h (Fig 2A), the XI-XYL3 strain consumed 10% xylose in the same time period and no ethanol was detected (Fig 2B and  Table 2). The difference in the rate of xylose metabolism is primarily due to the thermodynamic advantage of the oxidoreductase pathway compared to the isomerase pathway, as previously reported [37]. Ethanol production by the XI-XYL3 strain was only possible to detect under anaerobic conditions with a high initial cell density (25 g DCW/L) ( Table 3). The accumulation of significant amount of xylitol by the XI-XYL3 strain (5.0 g/L) compared to the XYL123 strain (0.6 g/L) was likely due to endogenous non-specific xylose reductase activities (Gre3), which is more significant when the rate of xylose metabolism is slow [38].

Effects of the PHO13 deletion on xylose fermentation by two xylosemetabolizing strains
To determine the effects of PHO13 deletion, the XYL3 gene was genome-integrated by replacing the PHO13 gene of the XYL12 and XI strains, resulting in the XYL123 pho13Δ and XI-XYL3 pho13Δ strains ( Fig 1B). When 40 g/L of xylose was provided, the XYL123 pho13Δ strain consumed xylose completely in 48 h, resulting in the fermentation time being reduced by 33%, and the growth rate being increased by 1.54-fold as compared to those of the XYL123 strain (Fig 2C and 2D). In addition, the XYL123 pho13Δ strain exhibited 1.76-fold higher specific ethanol productivity and a 6.33-fold increase in xylitol yield as a by-product, than that seen in the XYL123 strain (Fig 2E and  Table 2). These results confirmed that pho13Δ improved the xylose fermentation rate in a strain expressing a heterologous xylose oxidoreductase pathway as previously described [3,18,43,44]. However, pho13Δ did not affect xylose consumption or by-product yields in the XI-XYL3 strain expressing the xylose isomerase pathway (Fig 2F and Table 2). Under anaerobic conditions with a high initial cell density (25 g DCW/L), pho13Δ rather decreased ethanol production by 20% (Table 3).

PHO13 deletion-induced transcriptional and metabolic changes in two xylose-metabolizing strains
It has been reported that pho13Δ induces significant changes at both transcriptional and metabolic levels in the strains expressing the xylose oxidoreductase pathway. First, pho13Δ increases the expression levels of TAL1, which encodes sedoheptulose-7-phosphate:D-glyceraldehyde-3-phosphate transaldolase in the PP pathway, under both glucose [18,39] and xylose conditions [18], which was confirmed in the XYL123 strain of the present study as well (Fig 3A). Also, pho13Δ leads to the reduction in intracellular sedoheptulose-7-phosphate (S7P) and sedoheptulose levels during xylose metabolism [17], as confirmed in Fig 3B. However, in the XI-XYL3 strain, the TAL1 activation by pho13Δ was observed only under glucose conditions (7.3-fold increase, Fig 3A) but not under xylose conditions. Moreover, S7P was not accumulated in the XI-XYL3 strain during xylose metabolism (Fig 3B). Therefore, it was concluded that pho13Δ does not contribute to xylose metabolism of the XI-XYL3 strain neither at the transcription levels nor at the metabolic levels under the conditions we tested. It is hypothesized that some metabolic conditions are required for PHO13 deletion-induced transcriptional activation of TAL1, which is independent from the type of a metabolizing sugar. Because the XI-XYL3 strain metabolizes xylose very slowly, a lack of ATP and/or a low level of some metabolic intermediates could be associated with undesirable conditions for the TAL1 activation.

Adaptive evolution of two xylose-metabolizing strains
Concentration of xylose higher than 10 g/L inhibits the growth of S. cerevisiae strains expressing a xylose oxidoreductase pathway, which provides driving force for adaptive evolution as described previously [3]. The XYL123 strains constructed in the present study also showed decrease in the growth rates when the xylose concentration exceeded 10 g/L (Fig 4A). Also, when the XYL123 strain was subjected to serial sub-cultures on 40 g/L xylose, gradual increase in the growth rate of the culture was observed, suggesting the progress of adaptive evolution (Fig 4C). In fact, some selected mutants isolated from the evolved cultures showed improved xylose fermentation capabilities that were comparable to the XYL123 pho13Δ strain (S1 and S2 Figs and S3 Table). It is interesting to note that the growth of the XYL123 pho13Δ strain was (E, F) Volumetric growth rates (g/L-h) and product yields (g/g) of the xylose fermentations. Fermentations were performed in YP medium containing 40 g/L xylose under oxygen-limited conditions (80 rpm), with a low initial cell density (0.5 g DCW/L). Asterisks denote statistically significant differences (Student's ttest, p < 0.05). n. d.; Not detected.
https://doi.org/10.1371/journal.pone.0236294.g002 All strains were cultured in YP medium containing 40 g/L xylose under oxygen-limited conditions (80 rpm) with a low initial cell density (0.5 g DCW/L). All parameters were calculated when either more than 90% of xylose was consumed or fermented for up to 240 h. Acetate was not detected during the xylose fermentation. Parameters: Y Xylitol , Xylitol yield (g xylitol/g xylose); Y Glycerol , Glycerol yield (g glycerol/g xylose); Y Ethanol , Ethanol yield (g ethanol/g xylose); P Ethanol � , Specific ethanol productivity (g/g cell/h); n. d., not detected. In the XI-XYL3 strain, however, the growth rate gradually increased up to 50 g/L xylose, and there was no initial growth observed at 100 g/L xylose (Fig 4B). At 40 g/L xylose, therefore, serial sub-cultures of the XI-XYL3 strain would not provide high selection pressure for better growers. In fact, until 110 generations, the culture of the XI-XYL3 strain did not show improvement in the growth rates (Fig 4C). At 100 g/L, meanwhile, serial sub-cultures of the XI-XYL3 strain did show slight improvement in the growth rates ( Fig 4D); however, the isolated mutants did not have advantages in xylose fermentation (S1 and S4 Figs and S3 Table). Also, it was confirmed again that pho13Δ in the XI-XYL3 strain was not as critical as in the XYL123 strain regardless of xylose concentrations (Fig 4B), and during serial subcultures on 40 g/L and 100 g/L xylose (S3 Fig). Therefore, it was confirmed that either pho13Δ or evolutionary engineering could be an efficient strategy to improve strains expressing the xylose oxidoreductase pathway; however, the strategy of either pho13Δ or evolutionary engineering to improve strain expressing the xylose isomerase pathway did not have more dramatic results than the strain expressing the xylose oxidoreductase pathway.

Additional copies of xylA improves xylose consumption significantly
Several pathway-targeted approaches have been reported for the improvement of the xylose isomerase pathway (Table 3 and Fig 5A). First, the deletion of GRE3 encoding aldose reductase and/or the deletion of SOR1 encoding sorbitol (xylitol) dehydrogenase were proposed to reduce xylitol accumulation [45,46]. Also, extra copies of xylA and/or XYL3 [22,39,42] were often accompanied with the overexpression of the PP pathway genes such as TAL1 to improve xylose consumption rates. In Fig 5B, the necessity and contribution of each factors above were evaluated. Although some mutants showed statistically significant increases in growth rate and decreases in xylitol accumulation, none of the single factors contributed to ethanol production from xylose under oxygen-limited conditions ( Table 2 and S3 Table). In fact, the most signifi-  Table 2). In addition, with the improved level of xylose consumption, pho13Δ was shown to contribute to ethanol yield of the δ(XI)-XYL3 strain while its xylose consumption was not affected (Fig 5C, S6 Fig). However, the xylose consumption rate of the δ(XI)-XYL3 pho13Δ strain was still lower than that of the XYL123 pho13Δ strain (0.93 g/L-h) as well as those of the previously reported strains with 15-36 copies of the xylA gene (1.32-2.08 g/L-h, Table 3) [22,39]. The result suggested that the expression level of the xylA gene is one of the most critical factor for efficient xylose consumption, and the δ(XI)-XYL3 strain may have not reached to an optimal level of the xylA expression.

Discussion
There have been numerous attempts to develop S. cerevisiae strains fermenting xylose efficiently for decades [37,47,48]. Broadening the substrate range of S. cerevisiae is required not only to support cellulosic bioprocesses but also to extend product spectrum and efficiency with alternative substrates other than glucose [26,49]. In fact, metabolic engineering for xylose fermentation is a preliminary step toward strain development for desired products. However, the approaches to design efficient S. cerevisiae strains expressing the xylose isomerase pathway varied greatly, and adaptive evolution was essential in most prior studies (Table 3) [11,13,20,22,[50][51][52]. It is contradictory to the fact that the optimization of strains expressing the xylose oxidoreductase pathway can be reproducibly achieved by two factors: the constitutive expression of XYL1, XYL2, and XYL3 from P. stipitis and the deletion of the PHO13 gene (pho13Δ) [3]. Although a prior study presented a reduction in the lag phase by pho13Δ in the strain expressing the xylose isomerase pathway, the improvement was not as significant as those achieved by adaptive evolution in the same study [11]. The other study, which reported an 8% increase in the ethanol yield by pho13Δ in the xylose isomerase strain, used an extreme condition with an initial OD of 40 [39]. Consistent with previous findings, we also found that pho13Δ improved ethanol yield but it was only in the strain expressing multiple copies of xylA but not in the single copy xylA strain (Fig 5). Also, the xylose consumption rate remained constant in both strains, suggesting the conditional and limited effect of pho13Δ in the xylose isomerase strains. One of the most recent studies reported that nine different expression cassettes of Piromyces sp. xylA, the overexpression of both XKS1 and six non-oxidative PP genes (RPE1, RKI1, TAL1, NQM1, TKL1, TKL2), and deletion of the GRE3 gene are required to construct a xylose-assimilating S. cerevisiae strain [50]. The resulting strain was able to produce ethanol after adaptive evolution, in which the loss of function mutation in the PMR1 gene was critical [50]. In another recent study, two copies of a mutant version of Piromyces sp. xylA (E15D, E114G, E129D, T142S, A177T, and V433I), the overexpression of XKS1 and TAL1, pho13Δ and GRE3 as well as laboratory evolution were required for xylose fermentation [51,52]. The study concluded that the laboratory evolution was partially contributed by the loss of function mutations in the PMR1 and ASC1 genes [51]. Although the metabolic engineering approaches are complicated and different between the two studies, the studies shared the idea that the xylose isomerase step is the most limiting; therefore, 1) either multiple integration or protein engineering of xylose isomerase is required, and 2) the homeostasis of its inorganic cofactor has to be modified (pmr1Δ). The expression of approximately 36 copies of Orpinomyces sp. xylA [22] and the mutation in ASK10 for proper folding of isomerase [20] were also proposed to overcome the limitation in xylose isomerase. The above results from recent studies are all consistent with the findings of the present study that the copy number increase in the xylose isomerase gene is the most critical and primarily required (Table 3). However, the optimal level of the copy number of the xylA gene varies greatly among studies with the same xylA gene derived from Orpinomyces sp (Table 3).
It should be noted that the comparison of the two pathways in the present study was limited to the genes originated from P. stipitis and Orpinomyces sp. for the oxidoreductase and the isomerase pathways, respectively. Considering that the xylA gene was originated from strictly anaerobic fungus Orpinomyces sp., its functional expression in yeast could have been limited compared to other xylA genes originated from bacteria and other fungi [53]. Also, we only compared fermentation properties under oxygen-limited conditions with a low initial cell density. Indeed, under anaerobic conditions with a high initial cell density, where the limited growth of the XI-XYL3 strain can be compensated, the XI-XYL3 strain could produce ethanol at a higher yield (0.37 g/g xylose) than those achieved in the XYL123 strain (0.27 g/g xylose) ( Table 3). Nevertheless, engineering an efficient xylose-fermenting strain using the xylose isomerase pathway remains challenging because of the difficulties in reproducing adaptive evolution successfully and achieving optimal copy numbers of the xylA gene, as previously reported.
The present study aimed to develop a simple method to optimize S. cerevisiae expressing the xylose isomerase pathway: a genome-integrated heterologous xylose isomerase gene (xylA) under a strong promoter. We found that adaptive evolution as well as some of the pathway-targeted approaches (gre3Δ, XYL3, TAL1) did not work as efficiently as previously reported. One the other hand, significant improvement in xylose fermentation was achieved by sor1Δ as well as multiple integration of the xylA gene with or without pho13Δ. However, the improved strain was still inferior to an isogenic strain expressing xylose oxidoreductase pathway: xylose reductase (XYL1) and xylitol dehydrogenase (XYL2). Because the above mentioned approaches for the xylose isomerase pathway were successfully demonstrated in other studies, we think that other unknown factors are required such as different source of the xylA gene [53,54], different strain backgrounds [55,56], and/or other metabolic engineering designs. Although recent studies successfully discovered several knockout targets (ISU1, HOG1, GRE3, IRA2, SSK2) to improve the xylose isomerase pathway, they still required a strain background with the overexpression of the genes in the pentose phosphate pathway and/or the expression of multiple copies of the xylA gene [22,24]. With the current level of knowledge regarding xylose isomerase and its functional expression in S. cerevisiae, therefore, the xylose oxidoreductase pathway provides a more reproducible strategy to engineer xylose-fermenting strains.