Identification of a glucose-insensitive variant of Gal2 from Saccharomyces cerevisiae exhibiting a high pentose transport capacity

As abundant carbohydrates in renewable feedstocks, such as pectin-rich and lignocellulosic hydrolysates, the pentoses arabinose and xylose are regarded as important substrates for production of biofuels and chemicals by engineered microbial hosts. Their efficient transport across the cellular membrane is a prerequisite for economically viable fermentation processes. Thus, there is a need for transporter variants exhibiting a high transport rate of pentoses, especially in the presence of glucose, another major constituent of biomass-based feedstocks. Here, we describe a variant of the galactose permease Gal2 from Saccharomyces cerevisiae (Gal2N376Y/M435I), which is fully insensitive to competitive inhibition by glucose, but, at the same time, exhibits an improved transport capacity for xylose compared to the wildtype protein. Due to this unique property, it significantly reduces the fermentation time of a diploid industrial yeast strain engineered for efficient xylose consumption in mixed glucose/xylose media. When the N376Y/M435I mutations are introduced into a Gal2 variant resistant to glucose-induced degradation, the time necessary for the complete consumption of xylose is reduced by approximately 40%. Moreover, Gal2N376Y/M435I confers improved growth of engineered yeast on arabinose. Therefore, it is a valuable addition to the toolbox necessary for valorization of complex carbohydrate mixtures.

www.nature.com/scientificreports/ to take up pentoses in S. cerevisiae, all of them are naturally sensitive to competitive inhibition by glucose 14 . For instance, the endogenous galactose permease Gal2, which is particularly attractive due to its ability to transport both arabinose 3 and xylose 11 , is strongly inhibited in the presence of glucose 15 . To solve this problem, several groups have successfully employed different strategies. In our previous work, we designed a screening system, based on the hexose transporter deficient strain EBY.VW4000 16 , in which all genes encoding enzymes with hexokinase (Hxk) activity have additionally been deleted. Furthermore, in this strain background, XKS1, the noxPPP genes RKI1, RPE1, TAL1 and TKL1 and a heterologous XI have been overexpressed to enable the utilization of xylose 17 . The resulting strain (AFY10X) was suitable for screening of mutant transporter variants able to transport xylose in the presence of glucose. Thereby, a single mutation of a conserved asparagine residue located in the transmembrane domain 8 (TM8) of Hxt5, Hxt7 and Gal2 was found to be crucial to alleviate the binding of glucose as a competitor 17 . Similar observations regarding the role of the conserved asparagine have also been made in further studies [18][19][20][21][22] with different transporters and several mutant glucose-insensitive transporters are now available. As a rule, however, this beneficial property came at the expense of a reduced transport capacity (reflected by the kinetic parameter called maximal velocity, v max ). The corresponding values are systematically compared in a recent review article 10 . For instance, the v max for xylose transport of glucose-insensitive Gal2 variants N376F and N376V described in our previous study 17 was approximately threefold reduced compared to the wildtype protein. In fermentations, this causes a lower net xylose consumption rate and, consequently, prolonged fermentation times. Therefore, in this study, we used the AFY10X screening platform and performed additional rounds of random mutagenesis by error-prone PCR to select novel Gal2 variants that combine glucoseinsensitivity with an improved transport capacity.

Materials and methods
Construction and cultivation of yeast strains. The construction of the strains EBY.VW4000 16 and AFY10 17 was previously reported. The strain SRY027 is a derivative of the previously described HDY.GUF12 23,24 .
In brief, HDY.GUF12 contains genetic cassettes for overexpression of heterologous genes necessary for the utilization of xylose (codon-optimized gene for xylose isomerase XylA from Clostridium phytofermentans) and arabinose (codon-optimized genes for the arabinose transporter AraT from Scheffersomyces stipitis, arabinose isomerase AraA from Bacillus licheniformis, ribulokinase AraB and ribulose-5-P 4-epimerase AraD, both from Escherichia coli). In addition to these heterologous genes, the open reading frames (ORFs) encoding the endogenous xylulokinase Xks1, transaldolases Tal1 and Nqm1, transketolases Tkl1 and Tkl2, Ribose-5-phosphate ketol-isomerase Rki1, D-ribulose-5-phosphate 3-epimerase Rpe1 and the sugar transporters Gal2, Hxt2, Hxt7 and Hxt9 were placed under the control of strong constitutive promoters. To construct SRY027 as a fast pentose utilizing strain for testing engineered transporters without the interference with genomic copies of Gal2, all GAL2 alleles in HDY.GUF12 were deleted using the CRISPR technology 25 . The genotypes of all strains are listed in Table1. For the preparation of competent cells, plasmid-free cells were grown in standard YEP-media (1% (w/v) yeast extract, 2% (w/v) peptone) supplemented with 1% (w/v) maltose (M) for EBY.VW4000, 2% (v/v) ethanol (E) for AFY10 and 2% (w/v) of glucose (D) for SRY027. Frozen competent cells were prepared and transformed according to the established protocol 26 . The transformants of auxotrophic strains were initially plated on solid, selective synthetic complete (SC) medium with 1% (w/v) maltose (EBY.VW4000) or 2% (v/v) ethanol (AFY10), in which uracil (-Ura), leucine (-Leu), histidine (-His) and tryptophan (-Trp) were omitted as required depending on the plasmid selection markers. Amino acids were added as previously described 27 . SRY027 was plated on YEPD medium supplemented with 100 µg mL −1 ClonNAT (Nourseothricin) or 200 µg mL −1 G418 (Geneticin) for the selection of the natNT2 and kanMX markers, respectively. Table 1. Strains used in this study. The genotypes are annotated according to the standard nomenclature. Promoters and terminators are denoted by suffixes "p" and "t", respectively. Overexpressed ORFs resulting in functional proteins are underlined. The integrated genetic cassettes are shown in brackets. The occurrence of two identical alleles in a diploid genome is indicated by "2×". Gene deletions are indicated by a "Δ", followed by the range of deleted amino acids, where appropriate. Parental strains are indicated in bold under "relevant genotype". www.nature.com/scientificreports/ Plasmid construction and mutagenesis approaches. The DNA sequence of the GAL2 ORF that was used for plasmid construction and mutagenesis approaches corresponds to CEN.PK2-1C variant. For errorprone PCR (epPCR), the GeneMorph II Random Mutagenesis Kit (Agilent Technologies) was used as previously described 17 . Briefly, the mutagenesis reaction was performed with the primer pair Amp_GAL2_F/R using the p426H7_GAL2 plasmid as template. The epPCR fragments were purified by extraction from agarose gels (Nucle-oSpin Extract II; Macherey-Nagel) and subjected to a non-mutagenic second round of PCR using the Phusion polymerase (New England Biolabs GmbH) with the primer pair Clon_GAL2_F/R to add overhangs for homologous recombination with the p426H7 vector. AFY10X was transformed with the linearized p426H7 vector and the epPCR products and plated onto selective SCE plates. After colonies appeared, they were replica-plated to plates with various xylose-glucose concentrations (2 + 10, 2 + 20, 10 + 30, and 10 + 50 g L −1 ) and screened for the fastest growing clones. Site directed mutagenesis was performed using mutagenic primers listed in Supplementary Table 1 and the Phusion polymerase. Green fluorescent protein (GFP) was fused to the C-terminus of Gal2 variants as described previously 28 . Plasmids were isolated from yeast by the standard alkaline lysis protocol and electroporated into E. coli strain DH10B for amplification. Plasmid isolation from overnight E. coli cultures was carried out using a GeneJET Plasmid Miniprep Kit (Thermo Scientific) according to the manufacturer's instructions and sequenced. The plasmids used in this study are listed in Table 2.
Fermentation experiments and HPLC analysis. Cells were inoculated into selective media as indicated in the Results. Fermentations were performed in 300 mL shake flasks containing 50 mL of the culture medium. The cultures were cultivated at 30 °C with agitation at 180 rpm. For HPLC analysis, 450 µL of the culture supernatant were mixed with 50 µL 50% (w/v) 5-sulfosalicylic acid and centrifuged (16,000 g, 5 min, 4 °C). From this supernatant, 10 µL were subjected to high performance liquid chromatography (Dionex Ultimate 3000, Thermo Scientific) equipped with the ion exchange column HyperREZ XP Carbohydrate H + (7.7 × 300 mm; Thermo Scientific) and a refractive index detector. The liquid phase was 5 mM sulfuric acid, run at 0.6 ml min −1 and 65 °C. The Chromeleon 6.8 software was employed for quantifications and Prism 9 (GraphPad) for statistical analysis and presentation of the results. Table 2. Plasmids used in this study. Under "relevant properties", promoters and terminators are denoted by suffixes "p" and "t", respectively; ORF, open reading frame.

Plasmid name Relevant properties References
YEp181-kanR_optXI 2μ, LEU2, codon-optimized xylose isomerase gene xylA of C. phytofermentans under control of HXT7p -1-392  www.nature.com/scientificreports/ Radiolabeled xylose uptake assay. The uptake assay with radiolabeled xylose was performed according to the published protocol 29 . In brief, EBY.VW4000 yeast cells expressing the indicated Gal2 variants were grown overnight at 30 °C and 180 rpm in SCM-Ura medium. The next day, cells were harvested by centrifugation (4000 g, 5 min, 20 °C). The pellets were washed twice in ice-cold 0.1 M potassium phosphate buffer (pH 6.5), weighed, and resuspended in 0.1 M potassium phosphate buffer to a wet-weight concentration of 60 mg ml −1 .
Aliquots of 110 µl were prepared and cooled on ice. The residual cell suspension was centrifuged (4,000 g, 5 min, 4 °C) and the pellet was stored at − 80 °C before it was freeze-dried for cell dry weight determination. D-[1-14 C] xylose (55 mCi mmol −1 ) was obtained from American Radiolabeled Chemicals Inc. (St. Louis, MO) and diluted with unlabeled sugar to approximately 8 nCi mmol −1 and a total xylose concentration of 1500 mM (3 × concentration of that needed in the final uptake mixture). The cell suspension aliquots and the sugar solution were pre-warmed at 30 °C for 5 min. The uptake reaction was started by mixing 100 µl of the cell suspension and 50 µl of the 3 × sugar solution (yielding 500 mM total xylose in the uptake assay). The mixture was incubated for 20 s at 30 °C. The uptake was terminated by pipetting 100 µl of the assay mixture into 10 ml of ice-cold uptake buffer containing 500 mM xylose (quenching buffer). The suspension was filtered through a PVDF membrane filter, 0.22 µM pore size, 47 mm diameter (Durapore, Merck), which was subsequently washed two times with 10 ml of ice-cold quenching buffer. The amount of the radiolabeled sugar was measured in a 10 min window using the Wallac 1409 Liquid Scintillation Counter (Perkin Elmer). The proportion of the intracellular sugar was calculated as a ratio of the counts on the filter (corrected for background binding) to total counts in the uptake mixture. The transport rate (nmol min −1 ) was calculated and normalized to the dry cell weight.

Results
Identification of improved Gal2 mutants. To identify Gal2 variants with improved xylose transport properties in the presence of glucose, the GAL2 wildtype open reading frame was subjected to error-prone PCR (epPCR) as described in "Materials and methods". The obtained PCR products were co-transformed with the linearized p426H7 plasmids into the strain AFY10X to allow for in vivo plasmid assembly and the transformants were initially plated on the permissive, ethanol-containing (SCE-Ura) agar plates. The resulting colonies were subsequently replica-plated on media containing glucose/xylose mixtures and screened for fastest growth based on the colony size. One of fastest growing clones (designated ep3.1) was selected for further characterization. The GAL2-plasmid was isolated and sequenced, revealing five distinct mutations causing amino acid exchanges, namely M107K, V239L, N376Y, M435I and L558S. The substitution N376Y was already described in our previous publication to enable glucose-insensitive xylose transport, but its transport activity for the pentose was rather low 17 . We therefore concluded that at least one of the additional mutations is responsible for the improved phenotype of the ep3.1 mutant. To test this assumption, we combined each of the four mutations individually with N376Y and tested the growth conferred by the resulting plasmids on different carbon sources. For growth tests on hexoses (glucose and galactose) the strain EBY.VW4000, which is deficient for all transporters capable of hexose transport in S. cerevisiae 16 was used, whereas the growth on pentoses was assayed in the screening strain AFY10X. In accordance with our previous observations 17 , the constructs containing the N376Y mutations were not able to confer growth on glucose ( Supplementary Fig. S1). On xylose or xylose/glucose mixtures, the superior properties of the ep3.1 variant compared to the N376Y single mutant could be clearly attributed to the M435I mutation ( Supplementary Fig. S2). The N376Y/M435I variant was the only double mutant that conferred strong growth on pure xylose and glucose/xylose mixture. To see more subtle differences between the different variants than is possible on agar plates, we grew liquid cultures and measured the consumption of xylose in AFY10X cells transformed with the wildtype, ep3.1 and the N376Y/M435I variants (Fig. 1). The media contained 0.5% (w/v) xylose with or without glucose addition (2% w/v). Comparing the growth and xylose consumption rates, the N376Y/M435I variant performed even better than the ep3.1 construct, suggesting that at least one of the remaining ep3.1 mutations (M107K, V239L and L558S) has a negative impact on the xylose transport activity.
Interestingly, the M435I mutation alone appears to have a negative influence on the transport of both glucose and xylose ( Supplementary Fig. S3), which implies that there is a synergistic effect of the N376Y and M435I mutations in improving the uptake of xylose. We therefore also combined M435I with N376F, which has a lower v max for xylose compared to wildtype Gal2 17 Fig. S4).

Xylose uptake capacity of mutant Gal2 variants.
To better understand the results described above and to analyze the transport activity of the Gal2 variants independently of cell growth, which is influenced by multiple factors, we performed sugar uptake assays with radiolabeled ( 14 C) xylose (Fig. 2). To estimate the transport capacity (defined by the maximum velocity, v max ), the assays were performed at nearly saturating xylose concentrations (500 mM). Strikingly, the measured transport velocities mirror the growth behavior of the transformants described above. Whereas the N376Y mutation alone negatively affects xylose transport 17 (see also Supplementary Fig. S2), this defect is more than compensated by the additional M435I mutation. The lower transport rate by the N376F and N376F/M435I variants is consistent with the results shown in Supplementary  Fig. S4. Taken together, the results show that the M435I substitution increases the transport capacity in combination with tyrosine (but not phenylalanine) at position 376 in a synergistic manner.   www.nature.com/scientificreports/

Fermentation performance of engineered strains expressing mutant Gal2 variants. After hav-
ing demonstrated the superior properties of the N376Y/M435I mutant in the screening strain AFY10X, which consumes xylose very slowly (see Fig. 1), we reasoned that the "true" utility of the new transporter variant for mixed-sugar fermentations could be best challenged in a system highly optimized for pentose fermentation. We selected the strain SRY027 (see "Materials and methods" and Table 1 for details), derived from the robust, diploid industrial strain HDY.GUF12 (Ethanol Red background), which was genetically modified by integrating overexpression cassettes necessary for xylose and arabinose utilization, and further optimized to consume xylose through adaptive laboratory evolution 23,24 . In SRY027, the genomic GAL2 alleles were deleted to avoid any interferences with the engineered GAL2 constructs. In this background, we overexpressed different Gal2 variants from a multicopy (2µ) plasmid and performed fermentations in glucose/xylose mixtures (30 g L −1 and 10 g L −1 , respectively). All transformants consumed the sugars sequentially, with a fast glucose consumption phase (up to 12 h), followed by a slower, albeit considerably efficient xylose utilization, where the pentose sugar was consumed within approximately 36 h. Despite the presence of glucose-insensitive transporter variants (all mutants in which N376 was substituted by either Y or F), none of the transformants was able to significantly coconsume both sugars. Nevertheless, they did differ in the rate of xylose consumption (Fig. 3) in a manner that mirrors the xylose uptake capacities shown in Fig. 2, i.e. the N376Y/M435I exhibiting the highest and N376F/ M435I the lowest velocity. One possible explanation for the inability of the cells to co-consume glucose and xylose could be the glucoseinduced internalization and degradation of Gal2 in the vacuole 33 . This process is induced by phosphorylation and ubiquitination of the N-terminal cytoplasmic tail and can be abolished by mutating six serine residues at positions 32, 35, 39, 48, 53 and 55 to alanine. The resulting Gal2 6SA is stable at the plasma membrane even in the presence of glucose 28 . Therefore, we combined the 6SA N-terminal tail with the N376Y/M435I double mutation. To investigate if the 6SA tail is stabilizing the N376Y/M435I mutant, we also generated green florescent protein (GFP) fusion constructs as previously described 28 . As shown by fluorescence microscopy, the N376Y/M435I variant is internalized in glucose/xylose mixtures, whereas the 6SA/N376Y/M435I is nearly exclusively localized at the plasma membrane (Fig. 4A), as expected. Therefore, we transformed the stabilized constructs (without fused GFP) into the SRY027 strain, and performed a fermentation experiment in mixed-sugar medium. Again, the N376Y/M435I variant enabled faster xylose consumption compared to the control (Fig. 4B,C), reducing the overall duration of total xylose consumption by approximately 20 h (corresponding to 40% of the total fermentation time). Still, the consumption of glucose and xylose was largely sequential. This suggests that the reason for the inability of the strain for the simultaneous fermentation of both sugars does not primarily lie at the level of transport but might be rather due to metabolic constraints (see "Discussion").
The N376Y/M435I variant confers improved growth on arabinose as well. Besides xylose, arabinose is a relevant constituent of plant biomass. Since Gal2 is well-known to transport arabinose, we wanted to test if the N376Y/M435I double mutation is beneficial for the utilization of this pentose as well. AFY10 was enabled to metabolize arabinose by transforming it with plasmids encoding L-arabinose isomerase (araA), L-ribulokinase (araB), L-ribulose-5-P 4-epimerase (araD) and different Gal2 variants (wildtype, N376F, N376F/ www.nature.com/scientificreports/ M435I and N376Y/M435I). Growth tests were performed on solid media containing arabinose or arabinose/ glucose mixtures with these transformants. Expectedly, all N376 mutants conferred glucose-resistant growth on arabinose. Moreover, on all media, the N376Y/M435I variant conferred the strongest growth (Fig. 5), similar to what was observed with xylose (see Supplementary Fig. S4). This demonstrates that the Gal2 N376Y/M435I could also be used to improve the fermentation of hydrolysates that are rich in arabinose and glucose, such as pectin-rich biomass hydrolysates.

Discussion
Here, we isolated an improved variant of Gal2 (N376Y/M435I), which combines two properties highly desirable for fermentations of lignocellulosic or pectin-rich hydrolysates -the ability to transport pentoses in the presence of glucose (Figs., 1B, 5) and an improved capacity for xylose (Figs. 1A, 2, 3, 4) and arabinose (Fig. 5) transport. Interestingly, both N376Y 17 (Fig. 2) and M435I ( Supplementary Fig. 3) mutations individually have a negative impact on xylose transport, which is more than compensated when they are combined (Fig. 2). Strikingly, another substitution at position 435 was identified among multiple mutations in a Gal2 variant that had been generated by error prone PCR and selected for growth on low xylose concentration 34 . The authors did not analyze the individual contributions of the four mutations (L301R, K310R, N314D, M435T), but considering our results it  www.nature.com/scientificreports/ is likely that M435T is at least partly responsible for the observed phenotype, possibly in a synergistic manner, as described here. In a screen aiming to identify hexose transporter variants with increased specificity towards xylose, Zhao and coworkers 35 isolated a double mutant of Hxt2 (T339P/M420I, where M420 corresponds to M435 of Gal2). Although this variant was not glucose-insensitive, its specificity (calculated as a ratio of the K M values for the respective sugar) was shifted from glucose to xylose compared to the wildtype and the mutations acted in a synergistic manner. These reports demonstrate the importance of the conserved methionine residue corresponding to M435 of Gal2 in a different experimental context. Since no crystal structure of Gal2 is available, it is only possible to speculate about the underlying mechanism on the basis of homology models. The residue 435 is located at the periphery of the transmembrane helix 10 (TM10), at the interface of the lipid bilayer and the extracellular space. Hence, it is distant from the residue 376 that is located close to the sugar binding pocket and decisive for discrimination between pentoses and hexoses (Fig. 6A). The introduction of tyrosine, a large hydrophobic residue at the position 376 instead of asparagine in TM8 could cause clashes and affect the positioning of the proximal TM10. In turn, substituting methionine at the position 435 by the more hydrophobic isoleucine could positively influence the mobility of TM10 within the lipid bilayer at the extracellular membrane interface and thereby alleviate the possible steric constraints introduced by N376Y (Fig. 6B).
It was previously postulated that the sequential consumption of glucose and xylose is mainly due to the competition of both sugars at the level of sugar transporters, which have a much higher affinity for the hexose 15 . Our results show that the expression of glucose-insensitive transporters (Fig. 3), even in an internalization-resistant form (Fig. 4) is not sufficient to enable simultaneous consumption of both sugars. This is consistent with more recent work showing that the interferences of glucose and xylose are more complex and also involve their metabolism as well as glucose repression mechanisms. For instance, laboratory evolution for forced glucose-xylose co-consumption revealed mutations affecting the activity of the hexokinase Hxk2, the Gal83 subunit of the Snf1 kinase complex and of the E3-ubiquitin ligase Rsp5 36 . In another study, it was demonstrated that reducing the hexokinase activity is crucial for achieving simultaneous consumption of glucose and xylose 37 . Interestingly, a reduction in the xylose consumption rate was also observed during the co-consumption of maltose 15 , a disaccharide that does not interfere with xylose transport, but is intracellularly cleaved into two glucose moieties, which are subsequently phosphorylated by hexokinases and thereby enter glycolysis. These observations collectively suggest that a fast glucose metabolism (and related signaling) impede the simultaneous consumption of pentoses, but the details of the underlying mechanism are still not well understood. It is conceivable that a buildup of the noxPPP intermediates derived from the fast glucose metabolism prevent the entrance of xylose (via Xu5P) into the equilibrium-driven "scrambling" reactions of this pathway. However, many other metabolic and signaling interferences are possible and further research is necessary to disentangle the individual contributing components.
In conclusion, although Gal2 N376Y/M435I alone does not facilitate co-fermentation of glucose and pentoses, it does significantly reduce the overall fermentation time and therefore should prove valuable for valorization of substrates, which contain mixtures of glucose, xylose and/or arabinose, such as lignocellulosic or pectin-rich biomass.