Improving isobutanol productivity through adaptive laboratory evolution in Saccharomyces cerevisiae

Background: Isobutanol is an ideal second-generation biofuels due to its lower hygroscopicity, higher energy density and higher-octane value. However, isobutanol is toxic to production organisms. To improve isobutanol productivity, adaptive laboratory evolution method was carried out to improve the tolerance of Saccharomyces cerevisiae toward higher isobutanol and higher glucose concentration. Results: We evolved the laboratory strain of S. cerevisiae W303-1A by using EMS (ethyl methanesulfonate) mutagenesis followed by adaptive laboratory evolution. The evolved strain EMS39 with significant increase in growth rate and viability in media with higher isobutanol and higher glucose concentration was obtained. Then, metabolic engineering of the evolved strain EMS39 as a platform for isobutanol production were carried out. Delta integration method was used to over-express ILV3 gene and 2μ plasmids carrying ILV2 , ILV5 and ARO10 were used to over-express ILV2 , ILV5 and ARO10 genes in the evolved strain EMS39 and wild type W303-1A. And the resulting strains was designated as strain EMS39V2δV3V5A10 and strain W303-1AV2δV3V5A10, respectively. Our results shown that isobutanol titers of the evolved strain EMS39 increased by 30% compared to the control strain. And isobutanol productivity of strain EMS39V2δV3V5A10 increased by 32.4% compared to strain W303-1AV2δV3V5A10. Whole genome resequencing and analysis of site-directed mutagenesis of the evolved strain EMS39 have identified important mutations. In addition, RNA-Seq-based transcriptomic analysis revealed cellular transcription profile changes resulting from EMS39. Conclusions: With the aim of increase productivity of isobutanol in S. cerevisiae , sparing in the yeast proteome in response to sulfur demand.

improving tolerance toward higher isobutanol and higher glucose concentration via EMS mutagenesis followed by adaptive evolutionary engineering was conducted. An evolved strain EMS39 with significant increase in growth rate and viability had been obtained. And metabolic engineering of the evolved strain as a platform for isobutanol production was carried out. Furthermore, analysis of whole genome resequencing and transcriptome sequencing were also carried out.
Tolerance toward both higher glucose and isobutanol concentration does not seem to be a monogenic trait. As a complex phenotype, the evolution of such complex traits requires synergistic actions of many genes that are widely distributed throughout the genome. To circumvent this limitation, we enhance tolerance of S. cerevisiae to higher glucose and isobutanol concentration by adaptive laboratory evolution. Adaptive laboratory evolution was served as a genome-wide method for improving desirable phenotypes without the knowledge of genetic determinants for network information about those phenotypes [22][23]. And it was successfully employed to identify biological solutions to biofuel and alcohol toxicity in S. cerevisiae [24][25][26][27][28][29][30]. In this study, we used adaptive laboratory evolution to enhance yeast strain's tolerance toward higher glucose and isobutanol concentration in fermentation. It was based on a EMS (ethyl methanesulfonate) mutagenized W303-1A. After mutagenesis, strains were subjected to a 15 days stringent selection. By using this approach, we isolated one evolved strain EMS39 with significant increased growth rate and viability in fermentation.
Metabolic engineering of the evolved strain EMS39 as a platform for isobutanol production were carried out. 2 µ plasmids carrying ILV2, ILV5 and ARO10 were used to over-express ILV2, ILV5 and ARO10 genes in the evolved strain EMS39 and wild type strain W303-1A. It was reported that the enzyme dihydroxyacid dehydratase (encoded by ILV3) might be limiting the isobutanol pathway [11]. To break through this restriction point, we used δ-integration method to increase the integrated copy number of ILV3 in EMS39 and W303-1A. It was reported that δ-integration method could be used to over-express genes in yeast strains [31][32][33], Then, fermentation characters of strains EMS39V2δV3V5A10 and W303-1AV2δV3V5A10 were investigated. In addition, analysis of whole genome resequencing and transcriptome sequencing were also carried out to identify important mutations and significant changes in transcriptional levels that may caused higher isobutanol tolerance and higher isobtuanol titers.
In summary, to increase productivity of isobutanol in S. cerevisiae, improving isobutanol toleranceand higher glucose concentration via EMS mutagenesis followed by adaptive evolutionary engineering was conducted. Evolved strain with significant increase in growth rate and viability has been obtained. Metabolic engineering of the evolved strain as a platform for isobutanol production revealed advantages of the evolved strain for micro-aerobic production of isobutanol. Whole genome resequencing and analysis of site-directed mutagenesis of the evolved strain have identified important mutations that has caused improved in growth rate and viability in S. cerevisiae. RNA-based transcriptomic analysis were also carried out to identify transcriptome perturbations that may have caused higher isobutanol titers in strain EMS39V2δV3V5A10.

EMS and adaptive evolutionary engineering improved micro-aerobic growth and isobutanol tolerance of S.cerevisiae
Our strain development strategy is outlined in Fig.1. We used methods of adaptive laboratory evolution as generalized protocols, where ethyl methane sulfonate (EMS) was used on wild-type strain W303-1A to obtain a randomly mutagenized and genetically diverse initial population. The resulting population was used for 15 days Erlenmeyer flasks selections under higher glucose (100 g/L) and isobutanol (16 g/L) conditions throughout the cultivations. Individual mutant colonies are randomly selected from the final population and tested for their isobutanol tolerance using spot assay test. Meanwhile, relative viability rates were also determined. A strain designated as EMS39 with higher tolerance toward both glucose and isobutanol was identified. As shown in Fig.2a, both wild-type and the evolved strain EMS39 population could grow in control medium at all their dilutions (from 10 0 to 10 -4 ). 6 However, wild type could barely grow at its 10 0 dilution in YPD with 16 g/L isobutanol (stress containing medium). Additionally, the evolved strain EMS39 could grow nearly at all of its dilutions (10 0 to 10 -4 ) in YPD with 16 g/L isobutanol, which demonstrated its higher resistance properties against the stress factor. As shown in Fig.2b, the evolved strain EMS39 conferred a significantly improved cellular viability (over the course of 60 hours of culturing) above that of the control strain, even at concentrations as high as 20 g/L isobutanol. All these results indicated that the evolved strain EMS39 might be a predominant strain with higher tolerance toward isobutanol. So we used the evolved strain EMS39 as a platform to produce isobutanol.

Application of the evolved strain EMS39 as a platform for isobutanol production
Metabolic engineering of the evolved strain EMS39 as a platform for isobutanol production were carried out. To over-express ILV2, ILV5 and ARO10, 2μ plasmids YEplac195-PGK1p-ILV2, YEplac112-PGK1p-ILV5 and YEplac181-TDH3p-cox4-ARO10 were transformed into the evolved strain EMS39, wild type strain W303-1A and strain HZAL-7 using LiAc/ssDNA/PEG methods, the resulting strains were designated as strain EMS39V2V5A10, strain W303-1AV2V5A10 and strain HZAL-7V2V5A10, respectively. And in order to increase the integrated copy number of ILV3 (encoding enzyme dihydroxyacid dehydratase limiting the isobutanol pathway in S.cerevisiae), δ-integration method was used to over-express ILV3 in EMS39V2V5A10, W303-1AV2V5A10 and HZAL-7V2V5A10. And the resulting strains were donated as strain EMS39V2δV3V5A10, strain W303-1AV2δV3V5A10 and strain HZAL-7V2δV3V5A10, respectively. To determine whether the increased isobutanol tolerance of the evolved strain EMS39 could improve isobutanol yield, we examined performances of strain EMS39V2δV3V5A10, strain W303-1AV2δV3V5A10 and strain HZAL-7V2δV3V5A10 in micro-aerobic batch fermentation in YPD medium with 40 g L -1 glucose and 130 g L -1 glucose as carbon source in shaker flasks with OD 600 =0.5 and 3.0, as the initial inoculums size, respectively. And strain W303-1AδHis3 carrying plasmids YEplac181, YEplac195 and YEplac112 was used as the control strain.
As shown in Fig.3a and b, the control strain has the lowest growth rate and glucose consumption rate and glucose consumption was complete at 48h. Growth rate of strain EMS39V2δV3V5A10 was slightly higher than that of the control strain. And strain EMS39V2δV3V5A10 used up glucose at 32h. Meanwhile, growth rate of strain W303-1AV2δV3V5A10 and strain HZAL-7V2δV3V5A10 were higher than that of strain EMS39V2δV3V5A10. In addition, strain W303-1AV2δV3V5A10 and strain HZAL-7V2δV3V5A10 consumed glucose faster than strain EMS39V2δV3V5A10.
Isobutanol titers of strain EMS39V2δV3V5A10 increased by 35.1% and 5.2% compared with that of strain W303-1AV2δV3V5A10 and strain HZAL-7V2δV3V5A10, respectively. These data suggested that the increased isobutanol tolerance of the evolved strain EMS39 was useful for improving isobutanol titers.
To further gain insights into fermentation characteristics of the evolved strain EMS39 and strain EMS39V2δV3V5A10 in higher glucose concentration, we carried 8 out fermentations in YPD medium with 130 g L -1 glucose in shaker flasks with OD 600 =3, as the initial inoculums size. As shown in Fig.4a and Fig.4b, the growth rate and glucose consumption of strain EMS39YEplac181YEplac195YEplac112 were markedly higher and faster than that of the other four strains. While strain EMS39 V2δV3V5A10 resulted in lower growth rate and glucose consumption rate than strain EMS39YEplac181YEplac195YEplac112. In addition, there were no obviously differences between the control strain and strain W303-1AV2δV3V5A10 in the growth rate and glucose consumption. Finally, the growth rate and glucose consumption rate of strain HZAL-7 V2δV3V5A10 were slightly lower than that of the control strain.
As shown in Fig.4c, the control strain and strain EMS39YEplac181YEplac195YEplac112 produced 0.807 g L -1 isobutanol at 36h and 1.33 g L -1 isobutanol at 32h, respectively. The increased isobutanol titers in strain EMS39YEplac181YEplac195YEplac112 further indicated that increased isobutanol tolerance was useful for improving isobutanol titers. Meanwhile, strain EMS39 V2δV3V5A10 generated 2.79 g L -1 isobutanol at 24h. These results indicated that over-expression of ILV2, ILV5, ILV3 and ARO10 could increase isobutanol yield markedly. But after 24h, isobutanol titers of strain EMS39 V2δV3V5A10 decreased slightly. This perhaps due to the exhaustion of glucose. In addition, we found that strain W303-1AV2δV3V5A10 and strain HZAL-7V2δV3V5A10 gained 4.20 g L -1 and 3.45 g L -1 isobutanol at 48h, respectively. These results suggested that overexpression of ILV2, ILV5, ILV3 and ARO10 in strain W303-1A could markedly improve isobutanol titers. But over-expression of BAT2 and deletion of PDC6 is not useful for increasing isobutanol titers in strain W303-1A V2δV3V5A10.

9
Ethanol was one of the main byproducts in isobutanol fermentation in yeast. As shown in Fig.3d, the control strain and strain W303-1AV2δV3V5A10 generated 3.45g L -1 ethanol at 48h and 3.51 g L -1 ethanol at 36h, respectively. And the strain EMS39 V2δV3V5A10 produced 4.34 g L -1 ethanol at 32h. While strain HZAL-7 V2δV3V5A10 in the first 32 h fermentation. These data suggested that the evolved strain EMS39 also might be a predominant strain for producing ethanol.

Prelimiary investigation on the genetic basis of improved phenotype of evolved strain EMS39
To identified the genetic basis of improved phenotype in the evolved strain EMS39, whole genome resequence of the evolved strain EMS39 were carried out. More than 59 genes had mutations (including nucleotides insertions, nucleotides deletions and base changes) in their ORF or in upstream and downstream regulatory regions of genes. As shown in Table S1 regions. These results suggested that isobutanol tolerancerequired synergism of polygenic. But further investigations need to be carried out to explore mutations that can lead to improved isobutanol tolerance in S.cerevisiae.

EMS39V2δV3V5A10 resulted in transcription perturbations
In order to investigate cellular transcription profile changes in strain  Table S2). Glucose is phosphorylated by hexose-glucose kinase after uptake, and then enters the glycolytic pathway. The high-affinity glucose transporter genes HXT6 and HXT7 in strain EMS39V2δV3V5A10 were up-regulated by 8.1-fold and 11.5-fold compared with that in W303-1A V2δV3V5A10, respectively. The up-regulation of HXT6 and HXT7 in strain EMS39V2δV3V5A10 might promote its glucose assimilation ability.
Transcriptional levels of gene TPS2 and TSL1 in strain EMS39V2δV3V5A10 increased by 8.1-fold 5.7-fold, respectively. The increased transcriptional level of TSL1 and TPS1 might resulted in accumulation of trehalose, which could increase the stability of the cells and stimulates the secretion of heat shock proteins [34]. SOL4 and SOL1 were up-regulated in strain EMS39V2δV3V5A10. These two genes encode 6-phosphogluconolactonase in the pentose phosphate pathway. The enhanced pentose phosphate pathway activity perhaps might provide a large amount of nicotinamide adenine dinucleotide phosphate (NADPH) for isobutanol biosynthesis. FBP1, encoding fructose 1,6-bisphosphatase, was found up-regulated by 16.4-fold. ERR2, encoding a phosphopyruvate hydratase, was down-regulated by 17.2-fold. PCK1 , encoding phosphoenolpyruvate carboxykinase, was up-regulated by 17.9-fold. The perturbations of FBP1, ERR2 and PCK1 perhaps could promote regeneration of 2phosphoglycerate. It was reported that ADH1 knockdowns conferred increased tolerance toward both isobutanol and 1-butanol [35]. While our result indicated that alcohol dehydrogenase genes ADH1 and ADH5 were up-regulated by 2.8-fold and 3.0-fold in strain EMS39V2δV3V5A10, respectively. But alcohol dehydrogenase gene ADH4 was down-regulated by more than 42-fold. ADH1 is required for the reduction of acetaldehyde to ethanol, while ADH4 is involved in the degradation of ethanol and thereby contribute to ethanol detoxification to ensure cell survival. We found that FPKM values of ADH1 increased by 78.1-fold compared with that of ADH4.
Hence, the enhanced transcription of ADH1 in strain EMS39V2δV3V5A10 might confer it higher ethanol biogenesis. MPC3, encoding the highly conserved subunit of mitochondrial pyruvate carrier, was up-regulated by 33.2-fold in strain EMS39V2δV3V5A10. The up-regulation of MPC3 might promote pyruvate uptake into mitochondrial matrix. CIT1 was found up-regulated by 6.7-fold in strain EMS39V2δV3V5A10. Cit1p catalyzes the first reaction of the TCA cycle that is condensation of acetyl-CoA and oxaloacetate to form citrate. And Cit1p functions as a rate-limiting enzyme of the TCA cycle [36]. IDP2, encoding cytosolic NADP-specific isocitrate dehydrogenase Idp2 in the TCA cycle, was also up-regulated in strain EMS39V2δV3V5A10. The up-regulation of genes CIT1 and IDP2 might enhance the 13 activity of TCA cycle, Under anaerobic conditions, the TCA cycle can work as a reducing cycle, reducing the excess NADH [37]. Additionally, MLS1, encoding malate synthase in the glyoxylate cycle, was also up-regulated by 9.4-fold in strain It was speculated that the synthesis of sulfur-containing amino acids was increased in order to increase the sulfur reserve in advance to ensure the subsequent synthesis of glutathione (GSH) [39]. Other groups down-regulated DEGs in strain

Discussion
Isobutanol would be an ideal substitute for fossil fuels, because they have highenergy density and low hygroscopicity and can drop in directly in the current infrastructure and engines, preferred especially in aviation and diesel fuels. S. cerevisiae can produce low amount of isobutanol via valine synthesis pathway.
However, isobutanol is toxic to yeast cells, lowering the efficiency and costeffectiveness of these processes. So strain's isobutanol toleranceis one of important factors that restrict improvements of isobutanol fermentation performance in S.
cerevisiae. The evolution of such complex traits requires simultaneous modification in many genes' expression levels. Adaptive laboratory evolution could be used to screen desired phenotype that is generally dependent on multiple factors, which could not be identified unless genome-wide or large-scale approaches are used [25][26]. In this paper, we employed adaptive laboratory evolution strategies to obtain enhanced isobutanol and glucose stress-resistant S. cerevisiae mutants. And a strain designated as EMS39 was identified. Our results showed that the evolved strain EMS39 has higher isobutanol toleranceand improved cellular viability (Fig. 2).
These results suggested that adaptive laboratory evolution could be used to screen mutants with higher tolerance toward both isobutanol and glucose.
To investigate whether the enhanced tolerance toward isobutanol is useful to increase isobutanol yield, fermentation characteristics of strain EMS39V2δV3V5A10 in micro-aerobic batch fermentation with 40 g L − 1 initial glucose and 130 g L − 1 initial glucose were investigated. Our data suggested that strain EMS39YEplac181YEplac195YEplac112 and strain EMS39V2δV3V5A10 had higher growth rates compared to the control strain and strain W303-1AV2δV3V5A10 in fermentation with 130 g L − 1 initial glucose. In addition, glucose consumption rates of strain EMS39YEplac181YEplac195YEplac112 and strain EMS39V2δV3V5A10 were higher than the other three strains. This indicated that the evolved strain EMS39 had higher adaptability in fermentation with higher initial glucose. Meanwhile, strain EMS39V2δV3V5A10 produced higher isobutanol titers in micro-aerobic batch fermentation and in the first 24 h of fermentation with 130 g L − 1 initial glucose.
And strain EMS39V2δV3V5A10 gained the highest isobutanol productivity. Our data also shown that strain EMS39YEplac181YEplac195YEplac112 and strain EMS39V2δV3V5A10 produced higher ethanol titers. All these indicated that the evolved strain EMS39 might be a high-efficiency predominant strain for isobutanol fermentation and ethanol fermentation. In addition, our results suggested that improving yeast strain's isobutanol tolerancewas useful to increase its ability to produce isobutanol.
We have reported that deletion of PDC6 could improve isobutanol yield [9]. And we also reported that strain pILV2pARO10(W303-1AV2A10)produced 2.98 mg isobutanol per g glucose [13]. Here, we demonstrated that strain W303-1AV2δV3V5A10 produced 31.7 mg isobutanol per g glucose. Isobutanol yield of strain W303-1AV2δV3V5A10 increased 10.6-fold compared to that of strain pILV2pARO10 W303-1AV2A10). This indicated that over-expressing of ILV3 (encoding enzyme dihydroxyacid dehydratase limiting the isobutanol pathway) by using δ-integration method could markedly increase isobutanol yield.
To investigate the genetic basis of improved phenotype in the evolved strain EMS39, whole genome resequence of the evolved strain EMS39 were carried out. Furthermore, mutations in GPR1 perhaps conferred to the higher growth rate and higher glucose consumption rate of the evolved strain EMS39. Because GPR1 acts as G protein-coupled receptor that senses glucose and controls filamentous growth, and it is an important protein in Ras-cAMP pathway that regulate multi-aspects of cell growth. In addition, mutations in AAD4 and SOD2 (encoding putative arylalcohol dehydrogenase and mitochondrial manganese superoxide dismutase, respectively) perhaps conferred to higher isobutanol tolerancein the evolved strain EMS39. Furthermore, mutations in ASG1, TAF6 and MSS11 perhaps acted as transcriptional regulators or transcription factors to regulate multi-aspects of cell growth in the evolved strain EMS39. All these genes might work together corporately to confer the evolved strain EMS39 improved isobutanol tolerance.
To investigate the perturbations of transcriptome resulting from EMS39, we conducted differential expression analysis based on the RNA-Seq data. It was reported that knockdown of Hsp70p heat shock proteins improves isobutanol tolerance [35]. In the present study, we found that three heat shock proteins were up-regulated in strain EMS39V2δV3V5A10. One is Hsp10p that is involved in maintaining the stability of proteins in the mitochondria, ensuring their correct folding. The other one is Hsp26p that acts on the cytoplasm to prevent irreversible aggregation of proteins, which cannot properly folded. The third is Hsp30p that focuses on proteins on the plasma membrane and regulates the plasma membrane H + -ATPase to prevent rapid depletion of energy in the cell. The increased transcriptional levels of these heat shock proteins might be important adaptation mechanisms in S. cerevisiae, which could be induced by a series of intracellular adjustments. In addition,down regulation of biotin metabolism and regulatory subunit of the vacuolar transporter chaperone (VTC) complex coding gene VTC1 were only found in our study. Furthermore, Gene Ontology (GO) and KEGG 22 enrichment analysis suggested that down-regulation of genes involved in zinc ion transmembrane transport and vacuolar transporter chaperone complex were only found in this study. The down-regulation of zinc ion transmembrane transporter activity perhaps indicated the decreased synthesis of zinc ion-containing amino acids in strain EMS39V2δV3V5A10. The down-regulated gene VTC1 perhaps indicated decreased membrane trafficking and DNA replication stress in strain EMS39V2δV3V5A10. Finally, all these genes' transcriptional perturbations might work together corporately to confer strain EMS39V2δV3V5A10 improved isobutanol titers. It was reported that mutations in the eIF2 and eIF2B complexes greatly improve tolerance to these medium-chain alcohols [41]. While other studies had examined effects of overexpression of a single gene or individual gene knockout on the phenotype of improved alcohol tolerance[40, 46,47]. Their results suggested that each gene was insufficient to confer alcohol tolerance individually. Our results suggested that multi genes' transcriptional perturbations might work together corporately to confer strain EMS39 increased isobutanol tolerance and strain EMS39V2δV3V5A10 improved isobutanol titers.

Conclusions
In this study, we successfully obtained the evolved strain EMS39 that had enhanced tolerance toward higher isobutanol and glucose concentration by EMS (ethyl methanesulfonate) mutagenesis followed by adaptive laboratory evolution. In addition, metabolic engineering methods were used to improve isobutanol yield.
Further, we gained higher isobutanol titers (4.20 g L − 1 ) and higher isobutanol productivity (0.116 g/L/h). Finally, preliminary investigation of whole genome resequencing of the evolved strain EMS39 revealed the molecular mechanisms of 23 mutations affecting isobutanol tolerance in S.cereviase. And analysis of transcriptional perturbations in strain W303-1AV2δV3V5A10 and strain EMS39 V2δV3V5A10 revealed transcriptional perturbations of relative genes conferred yeast S.cereviase increased isobutanol tolerance and improved strain's ability to produce isobutanol.

Yeast strains and growth conditions
W303-1A was used as the parental strain and all strains used in this study and their genotypes were listed in Table 1. Plasmids and their descriptions were showed in Table 2. Escherichia coli DH5α was used as the cloning host and recombinant strains were cultured at 37°C in Luria-bertani medium (LB) (1% tryptone, 0.5% yeast extract and 1% NaCl) with 100 μg/mL ampicillin. Yeast extract peptone dextrose (YPD) medium (2% peptone, 1% yeast extract, 2% glucose) was used to routinely maintain and propagate yeast strains. Synthetic complete (SC) media (0.67% bacto yeast nitrogen base without amino acids supplemented with appropriate amino acid and 2% D-glucose) were used for selection of transformants.

DNA manipulation, plasmids and strains construction
Standard molecular genetic techniques were used for nucleic acid manipulations [43]. Primers used in this study were listed in Table 3. DNA templates used for PCR amplification of yeast genomic sequences were isolated from strain W303-1A.
Transformation of yeast cells was carried out with using LiAc/ssDNA/PEG method.

Mutagenesis by EMS treatment
1 mL fresh incubated cells (2 OD 600nm ) were collected by centrifugation, washed twice with 0.1 M potassium phosphate buffer (pH 7.0) and resuspended in 2 mL potassium phosphate buffer. Then four percent (v/v) EMS that gave rise to 85% lethality was added and mixed with the cell suspension by vortexing vigorously. The mixture was then incubated for 45 min at 30 °C with gentle agitation. And the mutagenesis was stopped by adding an equal volume of freshly made 5% (w/v) 25 sodium thiosulfate. Finally, the mutagenized cells were collected by centrifugation, washed twice with 5% (w/v) sodium thiosulfate, and resuspended in sterile ddH 2 O for selection or stored in −80 °C with 15% (v/v) glycerol.

Adaptive laboratory evolution
Adaptive laboratory evolution was performed by a serial batch transfer procedures and by batch fermentation in a 100 ml Erlenmeyer flasks in the presence of 16 g/L isobutanol as a selective pressure. Firstly, a single micro-aerobic culture of mutagenized W303-1A was initiated in 10 ml YPD supplemented with 16 g/L isobutanol in 50 ml Erlenmeyer flasks and grown at 30 °C and 150 rpm for 24 hours.
Then this culture was passaged daily in fresh YPD with 16 g/L isobutanol. After 15 daily passages, a diluted aliquot of the final culture was plated for single colonies on YPD plates with 16 g/L isobutanol. Finally, isobutanol tolerance of yeast strains derived from a number of single colonies were tested (see "Spot Assay Test" below) and compared with the parental strain. The best performing strain was designated as EMS39.

Spot Assay Test
Firstly, 1 mL fresh incubated cells (1 OD 600nm ) were collected by centrifugation, washed once with sterile ddH 2 O and resuspended in 1 mL sterile ddH 2 O. Then, for each culture, prepare 10 0 , 10 -1 , 10 -2 , 10 -3 and 10 -4 fold dilutions and spot 5μL of each dilution onto YPD plates (as the control plates) and YPD plates with 16 g/L isobutanol (as the stress containing plates), respectively. Finally, incubate the plates for 2-3 days at 30 °C and observe the growth of the cultures (Fig. 1).

Viability curve assays
Cells were incubated in 10 ml YPD medium with 20 g/L glucose in 100 ml Erlenmeyer 26 flasks for 16 hours at 30°C. Then, approximately 5 ml were centrifuged at 3000 rpm for 5 minutes. And cell pellets were resuspended in 30 ml YPD with 20 g/L of glucose and 16 g/L isobutanol to a final cell concentration of OD 600 =1, and incubated at 30°C with 200 rpm orbital shaking. 100 μL samples were collected (following vortexing to ensure homogeneity) every four hours (including the zero time point) and diluted and plated onto YPD plates. Finally, these YPD plates were then incubated for 2-3 days to allow for colony formation and colony forming unit counts.
Both control strains and the mutant were incubated in biological replicate.

Isobutanol fermentation and metabolite analysis
All micro-aerobic batch fermentation cultivations were carried out in 250 mL capcovered Erlenmeyer flasks with a working volume of 100 mL medium. Inoculums were cultured in SC medium (YNB without amino acids (Difco) supplemented with 20 g L -1 glucose and amino acids according to strains demands) until OD 600nm ≈5.0 in 100 mL Erlenmeyer flasks at 30°C. Then, cells were collected and used to inoculate 100 mL YPD medium with 40 g L -1 or 130 g L -1 glucose in Erlenmeyer flasks with OD 600 = 0.5 (or 3.0), as the initial inoculums size. During fermentations, the flasks were kept at 30°C with 100 rpm agitating to create micro-aerobic conditions. Samples were collected during fermentation.
Methods for measurements of cell growth, glucose, isobutanol and ethanol were illustrated in our previous report (Zhang A.L. et al. 2016).

Whole genome resequencing
The evolved strain EMS39 was used for whole genome resequenceing to identify important mutations occurred in the evolution. Strains were recovered with YPD medium from frozen stock for genomic DNA extraction with wild type strain W303-27 1A as a control. Genomic DNA were prepared via a standard phenol-chloroform method. Then genome DNA prepared was sent to Biomarker Technologies (Beijing, China) for library construction for next-generation sequencing according to the standard procedure of genome resequencing pipeline. Detailed procedures were carried out as described below. Firstly, qualified genomic DNA was then sheared on a Bioruptor Pico System Diagenode, Belgium to average length of 350bp. Then overhangs generated from the prepared fragmentation were repaired and an "A" base was added to the 3' end of the blunt phosphorylated DNA fragments; adapters were then ligated to the ends of the DNA fragments by using End Rrep Enzyme Mix (ExCellBio, Shanghai, China). The desired fragments were purified by using DNA clean beads (Vazyme Biotech Co.,Ltd, Nanjing, China) and selectively enriched by PCR amplification. Indexed tags were introduced into the adapters at the PCR stage.
The libraries were prepared with TrueLib DNA Library Rapid Prep Kit (Illumina, San Diego, CA, USA). And the qualified libraries were used for next-generation sequencing via the Illumina HiSeq X-ten platform (San Diego, CA, USA). Clean reads were purified from raw data by removing low-quality reads and then mapped to the S.cerevisiae S288C genome (R64-1-1.20110203.tgz) obtained from Saccharomyces Genome Database genome using bwa software (Version: 0.7.10-r789) [44]. The mapped reads were then sorted and duplicated reads were marked using Picard (Version: 1.94(1484)). Single Nucleotide variation (SNV)/InDel calling was done using the HaplotypeCaller algorithm of GATK, and the SnpEff output was used to decipher the potential effects of the mutations [45]. The distribution of variation results of different types is shown by using Circos diagram. Bioinformatics analysis was done using NR database, SwissProt database, GO database, COG database and KEGG database.