The yeast strains used in this study are listed in Table 1. S. cerevisiae Ta, a haploid strain derived from commercial wine strain T73, was previously constructed in the laboratory of A. Querol. Standard complex media (0.5% peptone, 2% glucose, and 0.5% yeast extract) or SC-Ura medium [0.67% YNB, 2% glucose, 1.92 g/L Drop-out –Ura (Formedium, Norfolk, UK)] were used to grow stock cultures. DNA isolation, restriction, and gel electrophoresis were carried out by standard genetic methods. Strains were transformed by the lithium acetate procedure (Gietz and Woods, 2002).

The S. kudriavzevii ATF1 and ATF2 genes (SkATF1 and SkATF2) were amplified from the genomic DNA of S. kudriavzevii IFO1802. The S. uvarum ATF1 and ATF2 genes (SuATF1 and SuATF2) were amplified from the genomic DNA of S. uvarum CECT12600. The S. cerevisiae ATF1 and ATF2 genes (ScATF1 and ScATF2) were amplified by the PCR from the genomic DNA of S. cerevisiae T73. The primers that we used are listed in Table 2. PCR fragments were then independently cloned into the pGREG526 vector (Jansen et al., 2005), previously cut with NotI/SalI. Before the start codon of the individual ATF genes within the pGREG526 vectors, the TDH3 promoter (amplified by PCR from genomic DNA from the S288C strain) was inserted by homologous recombination. The resulting plasmids are listed in Table 3. The constructed plasmids that contained ATF genes behind the TDH3 promoter were then introduced into the BY4741atf1atf2iah1 strain (Table 1) used in the enzyme assays.

The strategy of replacing endogenous ATF1 and ATF2 in the Ta genome with the corresponding ATF genes from S. kudriavzevii or S. uvarum involved two steps (i) deletion of either ATF1 or ATF2 gene; and (ii) integration of the S. kudriavzevii or S. uvarum orthologs into the locus. The ATF genes deletions in the Ta genome were performed by integrating a nourseothricin resistance cassette by homologous recombination. Deletion cassettes were amplified using pAG25 (Goldstein and McCusker, 1999) as a template and specific primers (Table 2). The resulting strains were named JET02 (atf1Δ) and JET03 (atf2Δ). The integration of SkATF1 into JET02 is described to illustrate the second step. The integration of the other orthologs was performed in the same way. The integration cassette was amplified from plasmid pG-SkATF1-kX with primers pGSkATF1f and pG-ATF1-R. The resulting PCR fragment included the SkATF1 gene, followed by kanamycin resistance marker, which was used for selection in the subsequent JET02 strain transformation. The final Ta mutant that held SkATF1 gene was named JET02Sk. The same procedure was carried out with ScATF1 (and ScATF2), which resulted in the restoration of the endogenous ATF1 (and ATF2) gene by the undergone process. This strain, named JET02Sc, was used as a reference in the subsequent assays.

Cultivations were performed in triplicate with a synthetic medium that contained 0.17% YNB w/o AAs and (NH₄)₂SO₄ (BD DIFCO^TM, Madrid, Spain) and 2% glucose as the carbon source, as previously described (Stribny et al., 2015) with minimal modifications. Media were supplemented by individual amino acids leucine, phenylalanine, and valine as the nitrogen source. Concentrations were proportional to 5 g/L (NH₄)₂SO₄ to obtain the same nitrogen content, as follows: 10 g/L leucine, 12.5 g/L phenylalanine, 8.9 g/L valine (Bolat et al., 2013).

Starter cultures were prepared by pregrowing yeast in 15-mL tubes that contained 4 mL of standard complex media. Before inoculating the experimental culture, the grown precultures were washed with water and resuspended in the same synthetic medium (with a certain nitrogen source), as used in the assay. Cells were resuspended in a volume that allowed an OD₆₀₀ of 1.7 to be achieved. These precultures (100 μL) were used to inoculate 1.6 mL of synthetic media, when the initial OD₆₀₀ was 0.1. Cultivation was performed in 96-well plates with 2-mL-deep wells. Wells were covered by microplate sealer (Greiner bio-one, Frickenhausen, Germany) to allow yeast growth and avoid evaporation and loss of volatile flavor compounds. Cultures were incubated for 5 days at 25°C. The individual 1.7-mL cultures were later transferred to 2-mL tubes and were stored at -20°C for the analysis.

Yeast cells growth was followed using a 96-well plate. Synthetic media were supplemented with amino acids as described above. Then 100 μL of media were inoculated in a well with 2 μL of cell suspension with OD₆₀₀ = 1. Growth was monitored in a Spectrostar Nano absorbance reader (BMG Labtech, Ortenberg, Germany).

A synthetic wine must was prepared as described previously (Riou et al., 1997), but with 200 g/L of reducing sugars (100 g/L glucose + 100 g/L fructose) and with no anaerobic factors (Beltran et al., 2004). Nitrogen source 300 mg N/L was composed of amino acids (180 mg/L) and NH₄Cl (120 mg/L). The composition of the amino acids mixture was that described by Beltran et al. (2004). The following mineral salts were used: KH₂PO₄ 750 mg/L, K₂SO₄ 500 mg/L, MgSO₄ 250 mg/L, CaCl₂ 155 mg/L, NaCl 200 mg/L, MnSO₄ 4 mg/L, ZnSO₄ 4 mg/L, CuSO₄ 1 mg/L, KI 1 mg/L, CoCl₂ 0.4 mg/L, H₃BO₃ 1 mg/L, (NH₄)₆Mo₇O₂₄ 1 mg/L. The following organic acids were used: malic acid 5 g/L, citric acid 0.5 g/L, and tartaric acid 3 g/L. The following vitamins were used: myo-inositol 20 mg/L, calcium pantothenate 1.5 mg/L, nicotinic acid 2 mg/L, chlorohydrate thiamine 0.25 mg/L, chlorohydrate pyridoxine 0.25 mg/L, and biotin 0.003 mg/L. The final pH was adjusted to 3.3 with NaOH.

Fermentations were performed in 250-mL glass bottles that contained 200 mL of synthetic must. Fermentations were done in triplicate at 25°C with continuous orbital shaking (150 rpm). Flasks were closed with Müller valves and monitored by weight loss until a constant weight was obtained. Immediately after fermentation ended, yeast cells were removed by centrifugation, and the contents of higher alcohols and esters in the supernatants were analyzed by gas chromatography.

The samples stored in the 2-mL tubes were centrifuged (13,000 rpm, 2 min) and 1.5 mL of the supernatant was transferred to the 15-mL vials with 0.35 g of NaCl. The 20-μl volume of 2-heptanone (0.005%) was added as an internal standard. Higher alcohols and esters were analyzed by the headspace solid phase microextraction (HS-SPME) technique using a 100-μm poly-dimethylsiloxane (PDMS) fiber (Supelco, Sigma–Aldrich, Madrid, Spain). Solutions were maintained for 2 h at 25°C to establish the headspace-liquid equilibrium. The fiber was inserted through a vial septum into the headspace and was held for 7 min. The fiber was then inserted into the gas chromatograph inlet port for 4 min at 220°C with helium flow (1 mL/min) to desorb analytes. A Thermo Science TRACE GC Ultra gas chromatograph with a flame ionization detector (FID) was used, equipped with an HP INNOWax 30 m × 0.25 m capillary column coated with a 0.25-m layer of cross-linked polyethylene glycol (Agilent Technologies, Valencia, Spain). The oven temperature program was: 5 min at 35°C, 2°C/min to 150°C, 20°C/min to 250°C and 2 min at 250°C. The detector temperature was kept constant at 300°C. A chromatograph was recorded by the ChromQuest program. Volatile compounds were identified by the retention time for reference compounds. Quantification of the volatile compounds was determined using the calibration graphs of the corresponding standard volatile compounds.

Cell extracts were prepared as described previously (Rojas et al., 2002) with minor modifications. Yeasts were grown in SC-Ura at 25°C up to the late log phase and by taking optical density at 600 nm as a reference. Yeast cells were collected by centrifugation, washed twice with cold 0.85% NaCl solution and resuspended in ice-chilled disruption buffer [10 mM potassium phosphate, pH 7.5, which contained 0.8 mM MgCl₂, 10% (w/v) glycerol, 5 mM DTT]. Protease inhibitor cocktail tablet Complete Mini, EDTA-free (Roche Diagnostics, Mannheim, Germany) was also added to the disruption buffer (one tablet per 10 mL). Disruption was performed with glass beads in a MillMix 20 Bead Beater cell disrupter (Domel Tehtnica, Zelezniki, Slovenia) at a frequency of 30 s^-1 in 1-min intervals over a 10-min period. The resulting homogenate was centrifuged at 15,000 g for 30 min at 4°C and the resulting supernatant was used to assay alcohol acetyltransferase activity.

The protein concentrations in the cell extracts were estimated by the Bradford method (Bradford, 1976) using bovine serum albumin as a standard.

Enzyme assays were performed in a 25-ml glass syringe provided with a Luer lock cap, as described previously (Rojas et al., 2002) with minor modifications. Reaction mixtures consisted in glycerol buffer [50 mM potassium phosphate, pH 7.5, 10% (w/v) glycerol] that contained higher alcohol as a cosubstrate, glycerol buffer that contained acetyl-CoA as the other cosubstrate and a cell extract. The higher alcohol, acetyl-CoA and cell extract showed the following volume ratio 1:0.1:0.4. The final volume was determined by the number of samples (1.5 mL). Isoamyl alcohol (at a final concentration of 0.01–100 mM), isobutanol (60 mM) or 2-phenylethanol (30 mM) was used as the substrate together with acetyl-CoA (0.8 mM). Substrate concentrations were determined according to Rojas (2002). After adding all the components, entrapped air was removed by the plunger and the syringe was attached to an orbital shaker. After 30 min, the 1.5-mL samples were transferred to 15-mL vials with 0.35 g NaCl for ester quantification with gas chromatography. Enzyme activity was stopped by adding 60 μL of a saturated KSCN solution.

The presented values are averages of biological triplicates with standard errors. The differences between the measured volatile compounds were determined by a one-way ANOVA, followed by Tukey’s HSD test (statistical level of significance was set at P ≤ 0.05). The analysis was performed using the STATISTICA 7.0 software (StatSoft, Inc., Tulsa, OK, USA).

To verify the effect of the Atf1 and Atf2 enzymes from S. kudriavzevii and S. uvarum on the final content of higher alcohols and/or acetate esters, individual genes (SkATF1, SuATF1, SkATF2, and SuATF2) were cloned into the locus of ATF1 or ATF2 in a haploid strain of the wine S. cerevisiae T73 strain, maintaining in all strains the wild S. cerevisiae promoter. The constructed mutants were named JET02Sk, JET02Su, JET03Sk, and JET03Su, respectively (Table 1). The formation of the major aroma-active higher alcohols and acetate esters was compared to that of S. cerevisiae. To exclude any other mutations that may have occurred in the gene replacement step, the original ATF1 (or ATF2) gene was returned to its native position which resulted in the strain JET02Sc (or JET03Sc) that was then used as a reference in the aroma comparison.

Synthetic wine must fermentations by the strains with the different ATF genes were performed, and the production of higher alcohols and acetate esters was analyzed. Weight loss monitoring showed all the strains to have similar fermentation rates with no significant differences (Figure 1). Regarding aroma compounds production, the most significant differences were observed during 2-phenylethyl acetate production (Figure 2). The strains that harbored ATF1 or ATF2 from S. kudriavzevii and S. uvarum produced almost twofold larger amounts of 2-phenylethyl acetate compared to the corresponding reference strains with ScATF1 or ScATF2. Strain JET02Sk also revealed c. threefold and JET02Su c. twofold larger amounts of isobutyl acetate than JET02Sc. During synthetic must fermentation several ethyl esters (ethyl hexanoate, ethyl decanoate, and ethyl octanoate) were also detected. Nevertheless, the comparison of the concentrations of the ethyl esters produced by the strains with the different ATF genes did not reveal any significant differences (data not shown).

Since aroma active higher alcohols, and subsequently acetate esters, derive directly from their amino acidic precursors, our aim was to verify whether, and to what extent, the individual ATF genes played a role in the production of higher alcohols and acetate esters from their corresponding amino acidic precursors. Valine, leucine, or phenylalanine was therefore used individually as the sole nitrogen source in the cultivations, and subsequently production of the corresponding higher alcohols and their esters was analyzed. In the assay, all the strains presented a normal growth pattern under these conditions (Figure 3). This confirmed their capability to use these individual amino acids as the sole nitrogen source. The analysis of the higher alcohols and acetate esters produced by the JET02-derived strains (strains that harbored ATF1 genes from S. kudriavzevii or S. uvarum) showed significantly higher concentrations of 2-phenylethyl acetate in the phenylalanine-grown culture of the JET02Sk strain (Figure 4A). This strain, which harbored SkATF1, produced c. 2.7-fold higher concentrations than reference strain JET02Sc. With the other analyzed compounds only small and statistically insignificant differences were detected. In contrast, the JET03-derived strains (strains with introduced SkATF2 or SuATF2) grown with the individual amino acids exhibited more differences during higher alcohols and acetate esters production. When grown with valine as the sole nitrogen source, the strain that carried SkATF2 or SuATF2 genes produced both corresponding derivatives, isobutanol and isobutyl acetate, in larger amounts than the strain with ScATF2 (Figure 4B). Particularly, the isobutyl acetate concentrations were around threefold higher in both strains than that produced by JET03Sc. A similar upward trend in favor of JET03Sk and JET03Su was detected during isoamyl acetate production when their amino acidic precursor leucine was used as the nitrogen source. JET03Sk produced a twofold and JET03Su a 2.5-fold increase in the isoamyl acetate concentration. The phenylalanine-grown cultures of JET03Sk and JET03Su exhibited 2.2 and 1.9-fold larger amounts of 2-phenylethyl acetate than the reference JET03Sc.

To determine and compare the AATase activities of Atf1p and Atf2p from the three different Saccharomyces species, the individual ATF genes were cloned into the plasmid pGREG526 and expressed in a host S. cerevisiae strain with the deleted ATF1 and ATF2 genes. In order to obtain sufficient AATase activity, the ATF genes were expressed under the control of the constitutive TDH3 promoter. The host S. cerevisiae strain was also transformed with the empty pGREG526 vector as a negative control. In all assays with cell extracts of empty vector transformants, the concentrations of the measured acetate esters formed from the corresponding higher alcohols (isobutanol, isoamyl alcohol, and 2-phenylethanol) were below detection limit. This confirmed none or negligible alcohol acetyltransferase activity in these control strain cell extracts. The kinetic parameters (K_m and V_max) were assessed for isoamyl alcohol as a substrate since isoamyl alcohol acetylation is known to be catalyzed only by Atf1p and Atf2p. To avoid the isoamyl acetate breakdown, the measured catalytic reaction product, IAH1 gene that codifies ester hydrolase, was also deleted in the host strain (BY4741atf1atf2iah1). AATase activity was determined in cell extracts over a wide range of isoamyl alcohol concentrations (0.01–100 mM) and was measured by isoamyl acetate formation which was analyzed by head space gas chromatography. The activity measured in the cell extracts that expressed the ATF1 orthologs showed typical Michaelis–Menten saturation kinetics (see Supplementary File 1). In contrast, the activity of all the AATases encoded by ATF2 displayed an increasing tendency over the entire range of tested concentrations, without reaching saturation. Therefore, K_m and V_max were determined only for the enzymes encoded by the ATF1 orthologs. The K_m and V_max values calculated for isoamyl alcohol displayed considerable differences among the individual Atf1 enzymes (Table 4). SkAtf1p showed almost twofold higher and SuAtf1p c. threefold higher K_m than ScAtf1p. Regarding V_max, both SkAtf1p and SuAtf1p exhibited around twofold lower values than ScAtf1p.

To further characterize enzymatic properties, substrate specificities were compared for three different amino-acid-derived higher alcohols as follows: isobutanol, isoamyl alcohol, and 2-phenylethanol. Specific activities were analyzed at fixed substrate concentrations of 60 mM for isobutanol, 100 mM for isoamyl alcohol, and 30 mM for 2-phenylethanol. The cell extracts of all the strains that expressed the individual ATF1 and ATF2 orthologs exhibited activities for all the measured substrates. Interesting differences were observed between enzymes Atf1 and enzymes Atf2. When individual enzymatic activities were expressed as a percentage distributed among the total enzymatic activity of the measured substrates, enzymes Atf1 showed approximately equal percentage portions of the catalytic activities for isoamyl alcohol and 2-phenylethanol (Figure 5A). In contrast, enzymes Atf2 showed considerably stronger activity toward isoamyl alcohol than the other substrates. Moreover, while all three Atf1 enzymes displayed an almost identical percentage distribution of individual enzymatic activities, the proportion of the enzyme activity of SkAtf2p and SuAtf2p to 2-phenylethanol was almost twofold greater, and twofold lower to isobutanol than for ScAtf2p (Figure 5B).