Yeast extract peptone dextrose (YPD) medium (10 g L^-1 yeast extract, 20 g L^-1 tryptone, 20 g L^-1 glucose) was used for activation, culture of host strains, and spot dilution growth (with 2% agar).

Synthetic complete medium (SD) or Sc-URA medium (1.7 g L^-1 yeast nitrogen base, Sangon, China, 5 g L^-1 ammonium sulfate, Sangon, China, CSM or CSM-URA, MP Biomedicals, Solon, OH, USA) was used for activation and batch fermentation of recombinant strains, supplying 20 g L^-1 glucose. Vanillin was used as an inhibitor and was added in the medium as indicated. All of the cultures were grown at 30°C unless otherwise indicated.

The S. cerevisiae industry strain NAN-27 (Zhang et al., 2010) and the vanillin-tolerant strain EMV-8, which was derived from NAN-27 (Shen et al., 2014), were the strains whose genomes were sequenced. Genetic manipulation including deletion and overexpression were mainly conducted in the lab strain BY4741 (MATa, his3Δ1 leu2Δ met5Δ ura3Δ, EUROSCARF, Germany). Deletion of YRR1 was also conducted in CEN.PK2-1C (MATa; ura3-52; trp1-289; leu2-3,112; his3Δ1; MAL2-8C; SUC2, EUROSCARF, Germany). The genes for overexpression were amplified from the mutant strain EMV-8. DBP2 was amplified from cDNA of BY4741. The gene deletion was performed using homologous recombination. The primers of this work are listed in Supplementary Table S1. The destruction cassette contained the KanMX expression cassette amplified from plasmid pUG6 (Escherichia coli plasmid with segment LoxP–KanMX4–LoxP) (Güldener et al., 1996) and the sequences that were homologous with the genes for knockout. The DNA fragments of the destruction cassettes were then transformed into the lab strain BY4741, and the mutants were screened in YPD medium with 800 mg L^-1 G418. pJFE3 (Shen et al., 2012) is a 2 μ plasmid with the TEF1 promoter, the PGK1 terminator, and URA3 as the selection marker. pJFE1 (developed in the laboratory) is a centromere plasmid with the TEF1 promoter, the PGK1 terminator, and the URA3 expression cassette as the selection marker.

A single colony was inoculated into 3 mL YPD medium and cultured for 12 h. The cultured cells were transferred into a fresh 10 mL of YPD medium with an initial OD₆₀₀ of 0.2. After overnight culture, the cells were harvested and washed three times with ddH₂O, followed by suspension in 1 mL of ddH₂O for 9 h to consume endogenous substances. The suspended cells were normalized to an OD₆₀₀ of 1.0. Then, 4 μl of 10-fold serial dilutions were spotted onto YPD or YPD with different concentrations of vanillin and incubated at 30°C for 1 days (for YPD) or 2–3 days (with vanillin in YPD).

A single colony was cultured into 3 mL of SD or SC-URA for 24 h. Then, the cultures were transferred into 20 mL of fresh medium with OD₆₀₀ at 0.2 for overnight culture. Then, the overnight-cultured cells were inoculated into 100-mL flasks with 40 mL of fermentation medium with an initial OD₆₀₀ of 0.2. Then, the fermentation was performed at 30°C and 200 rpm.

The biomass concentration was calculated with reference to the measured OD₆₀₀-dry weight correlation. For BY4741, one unit of OD₆₀₀ (detected by a BioPhotometer plus, Eppendorf, Germany) equals 0.18 g L^-1 of biomass. The maximum growth rates were estimated as the linear regression coefficients of the ln (OD₆₀₀) versus time during the exponential growth phase. Specific consumption rates of vanillin were calculated using the following equation:

where r is the specific consumption rate during the phase from sampling point m to sampling point n; A, B, and t are the metabolite concentration, biomass concentration, and time, respectively, at sampling points n, i, and m, as previously described (Peng et al., 2012).

The concentrations of glucose and ethanol were measured using a Prominence LC-20A HPLC system (Shimadzu, Japan) furnished with an Aminex HPX-87H ion exchange column (Bio-Rad, USA) and refractive index detector RID-10A (Shimadzu, Japan), using 5 mmol L^-1 H₂SO₄ as the mobile phase with a flow rate of 0.6 mL/min at 45°C (Eliasson et al., 2000). A BioSil-C18 column (Bio-Rad, USA) with a mobile phase of 40% aqueous methanol at a flow rate of 0.6 mL/min at room temperature was used to test the concentrations of vanillin and vanillyl alcohol by HPLC. The peaks were detected via ultraviolet detection (SPD-M20A) at 210 nm (Ji et al., 2011).

Genomic DNA was extracted from EMV-8 and NAN-27 using a yeast genome DNA extraction kit (Sangon Biotech, China) after yeast cells were harvested during the mid-log phase. Three paired-end genomic DNA libraries (200, 500, and 6000-bp) were constructed from NAN-27, and a 500-bp genomic DNA library was constructed from EMV-8. The library quality was assessed using the Qubit@ 2.0 Fluorometer (Thermo Scientific) and Agilent Bioanalyzer 2100 system. Finally, the library was sequenced on an Illumina HiSeq platform, and 125-bp paired-end reads were generated (an Illumina Hiseq2500 system).

The pre-cultured cells of BY4741 and BY4741(yrr1Δ) in SD medium were transferred into fresh SD or SD containing 5 mM vanillin with an initial OD₆₀₀ of 0.2. The cells were harvested during the log phase (OD₆₀₀≈1.6-2.0, vanillin content in the medium approximately 0.73∼0.78 g/L) and were quick frozen using liquid nitrogen. The UNIQ-10 Trizol RNA Purification Kit (Sangon Biotech, China) was used to extract total RNA. mRNA was isolated, fragmented, and then used as templates to synthesize cDNA. The short fragments were connected with adapters to obtain suitable fragments for PCR amplification. Ultimately, the libraries were sequenced by the Illumina HiSeq^TM 4000 (BGI Shenzhen, China). The differentially expressed genes were screened out according to the following criteria: Fold change ≥ 2 and FDR ≤ 0.001. All of the analyses were performed in biological duplicate. The transcriptome data were deposited in the NCBI Gene Expression Omnibus database (GEO accession number: GSE89854). Gene functions were annotated using the tool on Saccharomyces genome database (SGD)¹.

The strains were cultured in SD medium. The cells were collected when the OD₆₀₀ reached 4.0. Then, the cells were disrupted using a Fast Prep cell homogenizer (Thermo Savant, Germany) with 33 mM Na₃PO₄ buffer (pH = 7.0) containing 0.5-mm glass beads and 1 mM PMSF, which was used to inhibit any protease activity. The supernatant was collected by centrifugation at 4°C for the enzyme activity assay (Larroy et al., 2002).

The vanillin dehydrogenase activity was measured using the assay described by Larroy (Larroy et al., 2002). The 0.6-mL reaction mixtures used to measure enzyme activities toward vanillin contained 33 mM sodium phosphate buffer (pH 7.0), 0.5 mM NADPH, and 1 mM vanillin (using cuvettes with 0.2-cm path length). The molar absorption coefficient (ε₃₆₅) was 7.71 mM^-1 cm^-1 for vanillin plus NADPH. One unit (U) of enzyme activity is defined as the amount of enzyme that can reduce 1 μmol of NADPH plus vanillin in 1 min. The BCA protein assay reagent kit (Beyotime, China) was used to measure protein concentrations of crude enzyme solution. The specific enzyme activity (U mg^-1 protein) refers to the enzyme activity per milligram of protein.

To identify the key genetic targets related to vanillin resistance, comparative genome analysis was performed between the vanillin-tolerant strain EMV-8 and its parent strain, NAN-27. The genome sequencing results identified a total of 450 non-synonymous SNPs in CDS and 44 CDS with InDels in strain EMV-8 compared to its parent strain, NAN-27. In the present work, we focused on genes with InDels. Using PCR amplification and sequence analysis, only seven genes with InDels were confirmed in the evolved strain (Table 1).

To determine whether these InDel-containing genes influence vanillin tolerance in the mutant strain, these seven genes were separately knocked out in the lab strain BY4741. The growth profiles of the knock-out strains for these genes were tested using a spot dilution growth assay on media with or without vanillin (Figure 1).

Deleting YRR1, as by the addition of a stop codon at 409 bp that shortened the final protein product from 811 to 141 aa in EMV-8, notably enhanced the strain growth on the plates containing vanillin (Figure 1). Meanwhile, deleting other genes did not improve the vanillin tolerance of the strain. The maximum specific growth rates of BY4741(yrr1Δ) were, respectively, 23 and 142% faster than those of BY4741 with 5 and 6 mM vanillin in SD medium (Figure 2 and Table 3).

To confirm that deletion of YRR1 was the decisive factor in vanillin detoxification, compensation of YRR1 was performed. YRR1 was inserted into the centromere plasmid pJFE1 under the control of the TEF1 promoter and the PGK1 terminator, producing the plasmid pJFE1-YRR1. Then, pJFE1-YRR1 was transformed into the BY4741(yrr1Δ), resulting in the YRR1-compensated strain BY4741(yrr1Δ, pJFE1-YRR1). Strains BY4741(yrr1Δ, pJFE1) and BY4741(pJFE1) were also constructed and used as controls. The growth test result showed that the lag phase was much longer in BY4741(yrr1Δ, pJFE1-YRR1) and BY4741(pJFE1), which expressed YRR1, than in BY4741 (yrr1Δ, pJFE1). The maximum growth rates of BY4741(pJFE1), BY4741(yrr1Δ, pJFE1-YRR1), and BY4741(yrr1Δ, pJFE1) were 0.0867 ± 0.0014, 0.0887 ± 0.0031, and 0.1214 ± 0.0012, respectively. The compensation of YRR1 prolonged the lag phase and the strain growth rate decreased by 27% in liquid Sc-ura with 6 mM vanillin (Figure 3).

To investigate whether the deletion of YRR1 produced a general effect on vanillin tolerance, the strain CEN.PK2-1C (yrr1Δ), which has a genetic background different from that of BY4741, was constructed. This strain also exhibited obviously higher vanillin tolerance compared to the parent strain CEN.PK2-1C (Figure 4A). This result suggested that YRR1 deletion may be an effective method for enhancing the vanillin tolerance of S. cerevisiae.

Yrr1p is a transcription factor containing a Zn₂-Cys₆ zinc-finger motif. YRR1 is a paralog of YRM1, which also contains a Zn₂-Cys₆ zinc-finger motif. To investigate whether Yrm1p has a function in vanillin tolerance similar to that of Yrr1p, the effect of YRM1 deletion was also studied. The deletion of YRM1 did not improve vanillin resistance (Figure 4B), demonstrating that this protein has a different function from Yrr1p in this context.

To investigate the mechanism by which Yrr1p influences vanillin tolerance, the transcriptomic differences between BY4741 and BY4741(yrr1Δ) were analyzed in SD with 5 mM vanillin or without vanillin. The genes showing significant differential regulation under different conditions are listed in Table 2. Genes that are underlined are down-regulated; the others are up-regulated. The genes for which 2 ≥ fold change ≥ 1.5 are marked with “^∗”. Other genes listed in the table are those for which fold change ≥ 2.

When the cells were cultured in SD medium with no vanillin stress, the transcriptome differences between BY4741 and BY4741(yrr1Δ) were very small. Only eight genes showed obviously different expression (Figure 5), and these genes do not share any functions. Next, we analyzed how the strains responded to vanillin stress. BY4741 and BY4741(yrr1Δ) gave similar response, for each strain, over 200 genes showed differential expression when vanillin was present compared to vanillin free condition. When vanillin was present, dehydrogenases and ATP-binding cassette (ABC) transporters were up-regulated in both strains. This may be a general mechanism for resistance to vanillin stress in S. cerevisiae. Interestingly, genes involved in ribosome biogenesis and rRNA processing were significantly down-regulated in BY4741, while these genes and other related genes were significantly up-regulated in BY4741(yrr1Δ). This result suggested that the vanillin may decrease cell growth by repressing ribosome biogenesis and rRNA processing, thereby decreasing protein synthesis, and deleting YRR1 may restore these processes. The marked up-regulation of ribosome biogenesis and rRNA processing in BY4741(yrr1Δ) compared with BY4741 in the presence of 5 mM vanillin in the culture medium also supports this model. Based on these findings, more experiments were performed to further investigate how Yrr1p affects the vanillin tolerance of S. cerevisiae.

Adh7p, a NADPH-dependent medium-chain alcohol dehydrogenase, has been reported to have reductase activity and to convert vanillin to vanillyl alcohol (Nguyen et al., 2015). In both BY4741 and BY4741(yrr1Δ), the expression of ADH7 was up-regulated in the presence of vanillin, which was consistent with the report that ADH7 was inducible under severe vanillin stress (Nguyen et al., 2015). However, in BY4741(yrr1Δ), the increase in ADH7 was eight times that in BY4741. The deletion of YRR1 significantly increased the specific consumption rate of vanillin by 39 and 51% in SD with 5 and 6 mM vanillin, respectively (Figure 2). In BY4741(yrr1Δ), 6 mM vanillin was completely reduced within 26 h, while at that time point, only ∼10% of the vanillin was reduced by BY4741. Furthermore, in the enzyme activity assay, the cell-free extract of the BY4741(yrr1Δ) exhibited 95% higher NADPH-dependent reductase activity toward vanillin than did BY4741, while the cell-free extract of the BY4741(yrr1Δ) exhibited no NADH-dependent reductase activity compared to the control (Figure 5). Together, these results suggested that the up-regulation of ADH7 might be one reason why deleting YRR1 enhanced the strain tolerance to vanillin. To verify whether this is the only reason, we constructed the BY4741(adh7Δyrr1Δ). Interestingly, the adh7Δyrr1Δ strain still grew better than BY4741 in the presence of vanillin (Figure 6). Therefore, there are other mechanisms related to vanillin tolerance in the BY4741(yrr1Δ).

The multidrug resistance proteins Pdr5p, Yor1p, and Snq2p occur in the plasma membrane and belong to the ABC transporter superfamily, members of which export a diverse range of intracellular compounds using the energy supplied by ATP hydrolysis (Ananthaswamy et al., 2010; Guo et al., 2012). The transcriptional levels of PDR5, YOR1, and SNQ2 in both BY4741 and BY4741(yrr1Δ) were significantly up-regulated under vanillin stress, and the degree of up-regulation of PDR5 in BY4741(yrr1Δ) was notably greater than that in BY4741. Moreover, in our previous transcriptome analysis comparing EMV-8 with NAN-27, several multidrug transporters were expressed at markedly higher levels in EMV-8 than in NAN-27, including Pdr10p (6.2 times) and Yor1p (3.6 times). Next, the abovementioned genes related to multidrug resistance were overexpressed in the lab strain BY4741. In Sc-ura medium, the strain overexpressing PDR5 and PDR10 grew much more slowly than the control BY4741(pJFE3) (Figure 7C). Nonetheless, when vanillin was present, the strain overexpressing PDR5, YOR1, and SNQ2 grew markedly faster than control (Figures 7A,B) and had a shorter lag phase which was defined as the time before OD₆₀₀ reached 1.0 (Table 4).

Ribosomes are responsible for protein synthesis in the cell (Pelava et al., 2016). Reduced or abnormal ribosome production occurs in response to cellular stress and leads to decreased growth and metabolic capacity of the strain (Nomura, 1999). Our transcriptional analysis result indicated that ribosome biogenesis and rRNA processing were repressed by vanillin in wildtype S. cerevisiae. Deleting YRR1 increased the expression of genes related to ribosome biogenesis and rRNA processing when vanillin was present. Nine RNA helicases, including DBP2, DBP3, DBP9, DBP10, PRP43, MTR4, HAS1, DHH1, and DRS1. It was reported that the proteins encoding by these genes are participated in ribosome biogenesis, rRNA processing, and translation initiation (Rodriguez-Galan et al., 2013), were identified from dozens of markedly up-regulated genes in the comparison between BY4741 and BY4741(yrr1Δ). To further confirm whether enhancing ribosome biogenesis and rRNA processing can increase the vanillin tolerance of a strain, the gene DBP2, which encodes an ATP-dependent RNA helicase that functions in mRNP assembly (Ma et al., 2013) and is the gene most significantly down-regulated in BY4741 under vanillin stress but up-regulated in yrr1Δ, was overexpressed in BY4741. Although the maximum specific growth rate of BY4741(DBP2) was decreased by 15% compared to the control under vanillin-free conditions, in the presence of vanillin, the growth and metabolic characteristics of BY4741(DBP2) were better than those of the control. In SD medium containing 6 mM vanillin, the maximum specific growth rate of strain overexpressing DBP2 was 7% higher than that of the control. The OD₆₀₀ of the DBP2-overexpressing strain reached 1.0 and 2.0 after only 24 and 30 h, respectively, while reach the same biomass in the control required 30 and 40 h, respectively. The specific vanillin consumption rate of the strain overexpressing DBP2 was 0.062 ± 0.002 g g^-1 h^-1, which was 59% higher than that of the control (0.039 ± 0.001 g g^-1 h^-1, Figure 8).