OUP user menu

Isolation of sulfite reductase variants of a commercial wine yeast with significantly reduced hydrogen sulfide production

Antonio G. Cordente, Anthony Heinrich, Isak S. Pretorius, Jan H. Swiegers
DOI: http://dx.doi.org/10.1111/j.1567-1364.2009.00489.x 446-459 First published online: 1 May 2009


The production of hydrogen sulfide (H2S) during fermentation is a common and significant problem in the global wine industry as it imparts undesirable off-flavors at low concentrations. The yeast Saccharomyces cerevisiae plays a crucial role in the production of volatile sulfur compounds in wine. In this respect, H2S is a necessary intermediate in the assimilation of sulfur by yeast through the sulfate reduction sequence with the key enzyme being sulfite reductase. In this study, we used a classical mutagenesis method to develop and isolate a series of strains, derived from a commercial diploid wine yeast (PDM), which showed a drastic reduction in H2S production in both synthetic and grape juice fermentations. Specific mutations in the MET10 and MET5 genes, which encode the catalytic α- and β-subunits of the sulfite reductase enzyme, respectively, were identified in six of the isolated strains. Fermentations with these strains indicated that, in comparison with the parent strain, H2S production was reduced by 50–99%, depending on the strain. Further analysis of the wines made with the selected strains indicated that basic chemical parameters were similar to the parent strain except for total sulfite production, which was much higher in some of the mutant strains.

  • hydrogen sulfide production
  • MET10
  • MET5
  • Saccharomyces cerevisiae
  • sulfite reductase
  • wine


The production of unpleasant volatile sulfur compounds in wine, such as hydrogen sulfide (H2S) and the related mercaptans, is a chronic problem in the global wine industry. H2S has a low sensory threshold, between 50 and 80 μg L−1 in wine, and it imparts an undesirable sulfurous, rotten egg-like off-flavor (Rauhut, 1993). The production of H2S during wine fermentation is a frequently encountered problem in winemaking, and, if it is not treated, the resulting wine will be tainted, leading to a loss in quality and the possibility of being rejected by consumers. Therefore, factors affecting H2S production have attracted the attention of numerous researchers over the last few years (Giudici & Kunkee, 1994; Jiranek, 1995a; Wang, 2003; Linderholm, 2008).

It is well established that Saccharomyces cerevisiae is responsible for H2S off-flavor in wine and that the production is strain dependent (Acree, 1972; Mendes-Ferreira, 2002; Nowak, 2004), even though not all wine strains produce H2S because it has been reported that about 1% of the naturally occurring wine strains are unable to produce this off-flavor (Zambonelli, 1984). Other factors affecting the production of H2S include the availability of sulfur compounds, fermentation conditions and the nutritional status of the environment (Rauhut, 1993; Spiropoulos, 2000). The majority of H2S produced during winemaking occurs as a result of the biosynthesis of the sulfur-containing amino acids methionine and cysteine, which occur in low concentrations in grape juice, through the sulfate reduction sequence (SRS). These amino acids are essential for the growth of S. cerevisiae, and if they are not present, or depleted, in the growth medium, then sulfur must be assimilated. The most common sulfur source in S. cerevisiae is extracellular sulfate, which naturally exists in high amounts in grape juice.

In the first step of the SRS pathway, sulfate is transported into the cell by two specific permeases before a two-step activation with the aid of two molecules of ATP. The first reduction step produces sulfite, which is, in turn, reduced by sulfite reductase to sulfide. At this point, the sulfide produced is combined with a nitrogenous precursor, O-acetyl serine or O-acetyl homoserine, to ultimately form cysteine and methionine. The yeast NADPH-dependent sulfite reductase is a heterotetramer protein, consisting of two α- and two β-subunits (α2β2). The α-subunit is encoded by the MET10 gene, whereas the β-subunit is encoded by the MET5 gene (alias ECM17). Each α-subunit binds one flavin adenine dinucleotide (FAD) as a prosthetic group and contains a NADPH-binding domain, while each of the β-subunits binds one flavin mononucleotide and one siroheme (Thomas & Surdin-Kerjan, 1997). Sulfite reductase uses a siroheme coupled to an iron–sulfur cluster, Fe4S4, to perform the reduction of sulfite to sulfide.

If there is a deficiency of assimilable nitrogen in the grape must, O-acetyl serine or O-acetyl homoserine becomes limiting, and sulfide builds up and is converted to the volatile gas H2S, which then diffuses from the yeast cell into the wine (Vos & Gray, 1979). The conversion of H2S from sulfide is a pH-driven reaction and it is favored at the low pH of the wine; therefore, a high rate of H2S production can be observed in response to yeast nitrogen deficiency during wine fermentation (Jiranek, 1995a). Subsequently, many winemakers routinely add a source of nitrogen in the form of diammonium phosphate to the grape juice. However, there are restrictions on the amount of diammonium phosphate that can be added to the wine (Jiranek & Henschke, 1991), and in addition, there is currently a market trend toward wine that is additive-free; therefore, finding alternatives to diammonium phosphate supplementation is desirable.

Several genetic engineering strategies have been used for limiting H2S production, which generally consisted of overexpression or inactivation of some of the genes involved in the SRS pathway. Constitutive expression of the MET25 gene (alias MET17), which encodes a bifunctional O-acetylserine/O-acetylhomoserine sulfhydrylase, lowered H2S by twofold in a brewing yeast (Omura, 1995). In a similar study, the overexpression of the same gene in a strain of S. cerevisiae greatly reduced H2S formation in a wine ferment, but this was not the case for another strain (Spiropoulos & Bisson, 2000). Overexpression of the CYS4 gene, encoding cystathionine β-synthetase, was also shown to reduce H2S production (Tezuka, 1992). Although the above strategies resulted in partial reduction of H2S levels, they may not overcome the excess H2S production during fermentation because overexpression of both MET25 and CYS4 would require an abundance of assimilable nitrogen precursors. Thus, altering sulfite reductase activity has been considered a better approach for limiting H2S formation in yeast, because reducing the production of sulfide is a better approach than trying to consume it in a later metabolic step. The inactivation of MET10 in a brewer's yeast using a gene replacement procedure resulted in increased sulfite accumulation during beer production and increased flavor stability (Hansen & Kielland-Brandt, 1996). Furthermore, strains with no functional MET10 showed no sign of H2S production. In another study, several amino acids identified to potentially play an important role in the function of MET10 were replaced in an attempt to modify sulfite reductase activity. The mutation of four of these amino acid residues resulted in an inhibition of the enzyme activity (Sutherland, 2003). The complete inactivation of MET10 results in a methionine auxotrophic strain that requires the supplementation of methionine in the media for its growth. Unlike in brewing yeast, this approach is not suitable for wine yeast because the methionine concentration in grape juice (1–5 mg L−1) (Henschke & Jiranek, 1993) can be more limiting than in beer wort (59 mg L−1) (Garza-Ulloa, 1986), making sulfate assimilation necessary. Therefore, a reduction in the activity of sulfite reductase, rather than its elimination, may be a more feasible approach for wine yeast. Indeed, it has been shown that there is a correlation between H2S production and sulfite reductase activity for different wine-producing yeasts (Nowak, 2004). Recently, a screen of the S. cerevisiae deletion library was used to identify genes affecting H2S production (Linderholm, 2008). Four of the mutant strains produced nondetectable levels of H2S in a qualitative plate assay, and all four strains were devoid of genes that are necessary for an active sulfite reductase enzyme (MET5, MET10, MET1 and MET8).

In this study, we have used a classical mutagenesis approach to isolate a number of strains from a diploid commercial wine yeast strain that produce significantly lower quantities of H2S during alcoholic fermentation in both synthetic and grape juice. We identified specific mutations in the MET5 and MET10 genes responsible for the low H2S phenotype. Fermentations with these strains indicated that H2S production was reduced by up to 99%. These strains offer a promising solution to H2S-related taints generated during wine production, thereby increasing the organoleptic quality of the product and potentially the general appeal to consumers.

Materials and methods

Microorganisms and culture conditions

The yeast strains used in this study are listed in Table 1. All strains were obtained from The Australian Wine Research Institute (AWRI) culture collection. Yeast cultures were maintained on solid YPD agar plates (2% glucose, 2% peptone, 1% yeast extract and 2% agar). Yeast strains were transformed with the plasmids using the lithium acetate procedure (Schiestl & Gietz, 1989). For selection of yeast transformants, minimal synthetic defined medium was used, containing 0.67% yeast nitrogen base without amino acids, 2% glucose and 2% agar. Amino acids were added to all media according to the specific requirements of each strain.

View this table:

Strains and plasmids used in this study

Strains or plasmidsGenotypes or constructsReferences or sources
Escherichia coli DH5αsup E44, ΔlacU169 (φ80 lac ZΔM15), hsd R17, rec A1, gyr A96, thi-1, rel A1Stratagene
Saccharomyces cerevisiae
Laboratory strains
BY4742Δmet5MATα his3 leu2 lys2 ura3 met5AWRI
BY4742Δmet10MATα his3 leu2 lys2 ura3 met10AWRI
Industrial strains
Maurivin PDMCommercial wine yeast strainMauri Yeast Australia
PDM-derived strainsLow H2S-producing strainsThis study
pGEM-T-easySubcloning vectorPromega
pGEM-T-easy-MET10wtThis study
pGEM-T-easy-MET5wtThis study
pRS416Yeast expression vector, CEN6/ARS4, URA3ATCC 87521
pRS416-MET10wtThis study
pRS416-MET10W59XThis study
pRS416-MET10L606FThis study
pRS416-MET10W841XThis study
pRS416-MET10E929KThis study
pRS416-MET10T990IThis study
pRS416-MET10T997IThis study
pRS416-MET5wtThis study
pRS416-MET5A979TThis study
pRS416-MET5G1115DThis study
pRS416-MET5E1356KThis study

Modified MS300 medium, simulating standard grape juice, was used for synthetic wine fermentations (Varela, 2004). The initial sugar concentration was 240 g L−1 (120 g L−1 of glucose and 120 g L−1 of fructose), and the yeast assimilable nitrogen (YAN) concentration was 100 mg L−1. Filtered-sterilized Riesling grape juice was used for wine fermentations. The initial sugar concentration was 221 g L−1, the YAN concentration was 145 mg L−1, the total sulfur dioxide (SO2) levels were 32 mg L−1 and the pH was 3.1. The grape juice was cold-settled at 4 °C for 48 h, and prefiltered using a 1.2-μm pore diameter filter. The synthetic medium and the grape juice were filter-sterilized using a VacuCap 60PF filter unit (0.8/0.2-μm pore diameter, Pall Life Sciences).

Mutagenesis and strain isolation

The mutants were generated using a classical mutagenesis technique applied to the commercial Maurivin PDM strain (Lawrence, 1991), hereafter referred to as PDM. The alkylating agent ethylmethane sulfonate (EMS) was used as it induces high frequencies of base-pair substitutions and minimal lethality. Cells were inoculated into liquid YPD medium and grown overnight at 30 °C with vigorous shaking. Cells were washed twice in 50 mM potassium phosphate buffer (pH 7.0), diluted and resuspended in the same buffer. To the suspension, different volumes of EMS were added, cells were incubated for 30 min at 30 °C, and the mutagenesis was stopped by adding an equal volume of freshly made 10% sodium thiosulfate. Cells were then washed twice with sterile water and spread on YPD agar plates and incubated at 30 °C for 2 days, and the numbers of cells were counted to determine survival rates. In order to isolate the low H2S-producing PDM mutants, a concentration of EMS that resulted in a survival rate of about 50% was used. After the mutagenesis, colonies grown on YPD plates were replica-plated both on BiGGY agar (a commercially available bismuth-containing indicator agar from Oxoid), and on sulfate–BiGGY agar plates, and then incubated at 30 °C until the color of the yeast colony was developed.

Plasmid construction

Standard procedures for the isolation and manipulation of DNA were used (Ausubel, 1994). Both MET10 and MET5 genes were amplified by the PCR method using the PfuTurbo® DNA polymerase (Stratagene), and genomic DNA from PDM as a template. Primers Met5-Xba.for (5′-GACTTCTAGACTAAGTAACCAACACACGCACAAC-3′) and Met5-Xho.rev (5′-GACTCTCGAGCATAGTTTTCGTTTCTCATCTGC-3′) were used to amplify the complete MET5 coding region (4329 bp), plus a fragment of 690 bp upstream and 190 bp downstream of the coding region. A fragment of about 5.2 kb was obtained, purified and subcloned into the pGEM-T-easy vector (Promega) yielding pGEM-T-easy-MET5wt. To enable cloning into the centromeric S. cerevisiae expression plasmid pRS416, an XbaI site (underlined in the forward primer), and an XhoI site (underlined in the reverse primer) were introduced in the same PCR mixture. Then the 5.2-kb DNA fragment containing MET5 promoter and its coding region was obtained from pGEM-T-easy-MET5wt after digestion with XbaI and XhoI. This fragment was then purified and cloned into the XbaI and XhoI sites of plasmid pRS416, resulting in construct pRS416-MET5wt.

Primers Met10-Sac.for (5′-GACTGAGCTCGAAGGACAATCCTCGTTGAAACC-3′), and Met10-Bam.rev (5′-GACTGGATCCGATTTCAAGATGGTACCTCCGTAG-3′) were used to amplify the MET10 coding region (3108 bp), plus a fragment of 530 bp of the upstream region that covers the known promoter region (Hansen, 1994), and 210 bp downstream of the coding region. A fragment of about 3.8 kb was obtained, purified and subcloned into the pGEM-T-easy vector yielding pGEM-T-easy-MET10wt. To enable cloning into the plasmid pRS416, a SacI site (underlined in the forward primer), and a BamHI site (underlined in the reverse primer) were introduced in the same PCR. The pGEM-T-easy-MET10wt plasmid was digested with SacI and BamHI, and the 3.8-kb DNA fragment containing MET10 promoter and coding region was purified and cloned into the SacI and BamHI sites of pRS416, resulting in the pRS416-MET10wt construct.

Sequence analysis

The sequence analysis of MET5 and MET10 was performed in the parental PDM strain and six of the low H2S-producing strains. Genomic DNA was isolated, and amplification of both genes was carried out using PfuTurbo® DNA polymerase. The primers used to amplify the MET10 coding region, and its promoter, were as follows: Met10(-500).f (5′-GTGGAAGACTTTGCATATGGCTTG-3′) and Met10(3300).r (5′-GATTTCAAGATGGTACCTCCGTAG-3′). The primers used to amplify the MET5 coding region and a fragment of 700 bp upstream were as follows: Met5(-700).f (5′-CTAAGTAACCAACACACGCACAAC-3′) and Met5(4500).r (5′-CATAGTTTTCGTTTCTCATCTGC-3′). The PCR products were cleaned using the QIAquick PCR Purification Kit (Qiagen), and used as templates for the sequencing reaction. Sequencing was carried out on an ABI PRISM 3730xl DNA Analyzer (Applied Biosystems) using the Australian Genome Research Facility Ltd sequencing service (Adelaide). The PDM MET5 and MET10 sequences were submitted to GenBank with the following accession numbers: EU871588 and EU871589 for MET5 alleles 1 and 2, respectively, and EU871590 and EU871591 for MET10 alleles 1 and 2, respectively.

RNA isolation and cDNA synthesis

RNA isolation was carried out using the RiboPureTM-Yeast kit (Ambion) as recommended by the manufacturer. cDNA was synthesized from the isolated RNA using the SuperScript III First-Strand Synthesis System for reverse transcriptase-PCR (Invitrogen) as recommended by the manufacturer using random hexamers. The synthesized cDNA was eluted in 20 μL of autoclaved milliQ water and stored at −20 °C. For the PCR reactions, 2 μL of the cDNA were used in a 30-μL reaction. Primers Met10(1500).f (5′-TATCATCGATGGTGAGACCAC-3′) and Met10(2350).r (5′-ACTCCACATCTTGAAATCTCTTC-3′) were used to amplify a fragment of 850 bp of MET10, and primers Met5(1900).f (5′-GACTTCTTAGACCGTGAGAATAAGC-3′) and Met5(3050).r (5′-TCCAGCCAAACTTCGTGGTAG-3′) were used to amplify a 1150-bp fragment of MET5.

Site-directed mutagenesis of MET10 and MET5

The Met5p mutants A979T, G1115D and E1356K were constructed using the Quick Change PCR-based mutagenesis procedure (Stratagene) with the pRS416-MET5wt plasmid as template. The Met10p mutants W59X, L606F, W841X, E929K, T990I and T997I were constructed using pRS416-MET10wt as template. The primers were designed so that both mutagenesis primers contained the desired mutation and annealed to the same sequence on opposite strands of the plasmid (Table 2). The appropriate substitutions, as well as the absence of unwanted mutations, were confirmed by sequencing the inserts.

View this table:

Oligonucleotides used to introduce the indicated amino acid substitutions into MET5 and MET10

Amino acid substitutionsOligonucleotide sequences (5′–3′)
  • * The exact complementary oligonucleotide was also used for each mutation. The mutated nucleotide is underlined in the oligonucleotide sequence.

Fermentation conditions

Laboratory-scale fermentations for the initial screening of the isolated mutants were run in triplicate in 250-mL conical flasks fitted with an air lock and side-arm port sealed with a rubber septum for sampling. Fermentations were incubated at 18 °C (standard temperature for white wine fermentation) with shaking (100 r.p.m.), and their progress was monitored by measuring the sugar concentration. Yeast starter cultures were made in YPD medium, and incubated aerobically at 29 °C with shaking for 24 h to stationary phase. Cells were then centrifuged and washed with sterilized water, and used to inoculate 200 mL of the synthetic or grape juice at a density of 1 × 106 cells mL−1. Samples of culture medium were obtained at different stages of the fermentation from each flask to determine OD600 nm, dry cell weight and for analysis of sugar concentrations. At the end of the fermentation, samples were centrifuged for 5 min at 5000 g and the cell-free supernatants were stored at −20 °C for analysis of the concentrations of residual sugars, ethanol, glycerol and acetic acid. Dry cell weight was estimated by filtering 10 mL of medium through a Supor®-200-membrane filter (0.2-μm pore diameter, Pall Life Sciences), washing twice with distilled water and drying to a constant weight at 85 °C.

Evaluation of H2S production

Three methods were used to detect H2S production by the different strains.

In the first method, H2S production was evaluated on BiGGY indicator agar. The qualitative amount of H2S production on this indicator medium was determined by the color of the colonies, which ranged from white through brown to near-black, depending on the extent of production (Giudici & Kunkee, 1994; Jiranek 1995b). A variation of the commercial BiGGY agar method was used, i.e. sulfate–BiGGY agar, in which the sulfur source was sodium sulfate (3 g L−1) instead of sodium sulfite.

In the second method, the qualitative amount of H2S formed during fermentation was detected by commercial lead acetate test strips (Fluka). For this purpose, fermentations were carried out in flasks equipped with these test strips attached to the top of the flask, away from the liquid. The degree of blackening of the strips correlates to the amount of H2S produced during fermentation (Giudici & Kunkee, 1994).

With the third method, the analytical quantification of H2S liberated from cultures during fermentation was achieved by a liquid trap-based method, using cadmium hydroxide as trapping agent, as described elsewhere (Jiranek, 1995a).

Analytical methods

The concentrations of sugars, ethanol, glycerol and acetic acid were measured by HPLC using a Bio-Rad HPX-87H column as described previously (Nissen, 1997). Sulfite measurements were carried out using the Flow Injection analysis using the Lachat Instrument QuickChem(r) FIA+8000 series. The sulfite in the sample is acidified with a dilute hydrochloric acid solution and the acidified sample passes through two channels where it is subsequently heated on one channel but not the other. The heat and acid hydrolyze the acetaldehyde-α-hydroxysulfonate bond and other bisulfite complexes to release SO2. The resulting SO2 gas passes through a Teflon membrane and reacts with formaldehyde and pararosaniline dye in an acid solution to form a red–violet colored complex. The absorbance of the colored complex is measured at a wavelength of 560 nm, and it is directly proportional to the SO32− in the sample (limit of detection is 4 mg L−1).


Random mutagenesis

To isolate low H2S-producing yeast derived from the Maurivin PDM commercial wine yeast, a classical mutagenesis technique was used, based on the incubation of yeast cells with the alkylating agent EMS. The rate of survival of the diploid PDM strain was determined following EMS mutagenesis, in order to obtain a survival rate of 50%. Several concentrations of EMS were tested, from 1% to 8%. A concentration of 6% EMS produced the desirable survival rate (51%), and this concentration was chosen for further experiments.

A total of 16 000 mutant PDM strains were isolated in YPD plates after incubation of the yeast cells with EMS. Cells were replica plated on commercial BiGGY agar, which allows rapid and reliable assessment of the production of sulfide. In addition, cells were replica-plated onto a variation of the standard BiGGY agar medium that contained sulfate instead of sulfite (sulfate–BiGGY). Whereas standard BiGGY agar medium only allows the assessment of the sulfite reductase reaction, the modified sulfate–BiGGY provides extra information about the upstream steps of the SRS pathway. H2S formation on these indicator media was determined by the color of the colonies, which ranged from white through brown to near-black (Fig. 1). PDM displayed a dark brown phenotype on both BiGGY and sulfate–BiGGY agar. Two white mutants were isolated from BiGGY agar plates (designated strains 1.1 and 45.2), and an additional white mutant (designated strain 85.1) was found on the modified sulfate–BiGGY agar plates. A total of five light tan colonies were found (designated strains 22.1, 23.2, 51.1, 75.1 and 92.1), and three high H2S producers (black mutants) were identified. In addition, 211 colonies that displayed a lighter phenotype than PDM, tan or brown, were also identified. Furthermore, the isolated colonies were spotted on solid minimal medium with and without methionine. While all 225 colonies grew in the minimal medium supplemented with methionine, two of the three white colonies (strains 45.2 and 85.1), two of the three black mutants, and one light tan colony (strain 92.1) failed to grow on agar media without methionine (data not shown).


BiGGY plate example of the types of colony color obtained due to sulfide production: white (W), light tan (LT), tan (T), brown (B), dark brown (DB) and black (BL). The PDM control strain displayed a dark brown coloration in this media. The darker the colony, the more H2S is being formed.

Small-scale fermentation experiments in synthetic grape juice medium

Screening of 13 of the low H2S-producing strains

A total of 13 low H2S-producing mutants (including the three white: strains 1.1, 45.2 and 85.1; the five light tan: strains 22.1, 23.2, 51.1, 75.1 and 92.1; and five of the tan strains: 4.2, 4.4, 6.3, 9.1 and 23.1) were selected and used for a series of fermentations in a synthetic grape juice medium with a YAN concentration of 100 mg L−1. The concentrations of the sulfur amino acids cysteine and methionine were 8.5 and 22 mg L−1, respectively.

The progress of the fermentations, in the form of sugar consumption, was followed (Fig. 2). The time required for the control PDM strain to reach dryness (<2 g L−1 sugar) was 36 days. Tan strain 4.2 showed a faster fermentation rate than that of the PDM parental strain as the ferment was dry after only 30 days (Table 3). A number of the other strains (1.1, 23.2, 51.1, 75.1, 85.1 and 4.4) fermented at a similar rate than the PDM strain, while strains 45.2, 22.1 and 23.1 fermented significantly slower because they were only dry after 42 days. Finally, strains 92.1, 6.3 and 9.1 displayed very slow fermentation rates, and ferments were not dry even after 45 days, when the fermentation was stopped.


Fermentation course of 13 of the low H2S-producing strains in synthetic grape juice. (a) Fermentation profile of PDM (▪), and strains 1.1 (◻), 45.2 (▲), 85.1 (◆), 22.1 (△), 23.2 (●), 51.1 (○) and 75.1 (◊). (b) Fermentation profile of PDM (▪), and strains 92.1 (◻), 4.2 (▲), 4.4 (△), 6.3 (●), 9.1 (○) and 23.1 (◆). Fermentations were stopped after 42 days.

View this table:

Properties at the end of the alcoholic fermentation in synthetic media

Residual sugar (g L−1)
Glycerol (g L−1)
Acetic acid (g L−1)0.830.680.810.840.830.810.841.050.990.920.900.650.650.99
Ethanol (%)14.414.314.214.414.214.414.514.314.114.314.
Days to dryness3637423942393439>423039>42>4242
  • * Fermentations were stopped after 42 days.

  • The SD did not exceed 10% for any of the values.

The lead acetate strips attached inside the flasks did not show any color change in any of the ferments, not even in the PDM control ferment. At the end of fermentation, a basic HPLC analysis of some of the fermentation metabolites was conducted, and none of the strains tested showed significantly different properties than those of the PDM control strain (Table 3). Furthermore, the strains were isolated from the ferments and plated on BiGGY agar, at the beginning and at the end of the fermentation. All strains displayed the same color (phenotype) at both time points; therefore, the phenotype of the mutagenized strains was stable throughout the fermentation. Of the 13 strains tested, two white strains (1.1 and 45.2), three light tan strains (22.1, 23.2, 51.1) and one of the tan strains (4.2) were selected for the next screening trial based on the fermentation rates and the color of the colonies on BiGGY agar. Although the fermentation rate of the strains 45.2 and 22.1 was slower than PDM, this is not necessarily a negative characteristic, as PDM is not used in red wines because the fermentation rate is too fast. Therefore, a PDM version that does not produce H2S and ferments slower could be useful for red wine fermentations.

Screening of six of the low H2S-producing strains

For the second small-scale fermentation screening, the total YAN concentration was also kept at 100 mg L−1, but unlike the medium used in the former screening, the concentrations of all the amino acids were reduced. Particularly, the sulfur amino acids cysteine and methionine were reduced to 1 mg L−1 in order to encourage the production of H2S by the mutant strains, because it is known that methionine negatively regulates the synthesis of some of the enzymes of the SRS pathway (Thomas & Surdin-Kerjan, 1997). A concentration of <0.05 mM (7.5 mg L−1) could be considered as nonrepressive.

The fermentation rates of the six selected strains and PDM are shown in Fig. 3. The ferment inoculated with the PDM strain was dry after 37 days, and again, the strain 4.2 showed a faster fermentation rate than the PDM strain. Strains 1.1, 23.2 and 51.1 fermented at a similar rate as the PDM strain. Finally, strains 45.2 and 22.1 fermented at a significantly slower rate than PDM. This time, the production of H2S could be detected using the lead acetate strips. The PDM strain produced a considerable amount of H2S. Strains 1.1 and 4.2 produced some H2S during fermentation, although to a lesser extent than the PDM parental strain, with the remaining strains producing very small amounts of H2S, with the strips remaining white or near white (Fig. 4).


Fermentation profile of PDM (▪), and strains 1.1(◻), 45.2 (▲), 22.1 (△), 23.2 (●), 51.1 (○) and 4.2 (◆) during fermentation of a synthetic grape juice. Fermentations were stopped after 40 days.


Lead acetate strips after fermentation of synthetic medium by PDM and strains 1.1, 45.2, 22.1, 23.2, 51.1 and 4.2.

No SO2 was detected in the PDM control ferment using the current analytical technique (limit of detection of 4 mg L−1), and strain 4.2 only accumulated 5 mg L−1 of total SO2. However, the remaining strains accumulated considerable amounts of total SO2, and both white strains were the highest SO2 producers (Table 4). Almost all the SO2 accumulated in the ferments was in the bound form, and the concentrations of free SO2 were very low for all the strains tested (<5 mg L−1).

View this table:

Properties at the end of the alcoholic fermentation in synthetic media

Residual sugar (g L−1)02.0572970.50
Glycerol (g L−1)
Acetic acid (g L−1)0.750.580.600.730.730.720.79
Ethanol (%)14.814.711.112.914.514.914.9
Days to dryness3740>40>40>404031
SO2 (free/total) (mg L−1)0/05/1982/2410/634/1620/1780/5
  • * Fermentations were stopped after 40 days.

  • The SD did not exceed 10% for any of the values.

The strains were isolated from the ferments and plated on BiGGY agar, at the beginning and at the end of the fermentation. Unlike the other strains, strains 1.1 and 45.2 produced white colonies at the start of the fermentation but showed a partial reversion at the end of fermentation because half of all the colonies were brown at that stage. We attempted to stabilize these two strains so that the H2S-negative phenotype could not revert during the fermentation. This was done by growing the strains in a liquid minimal media without amino acids for 72 h, then spreading them on YPD plates, and finally replica-plating them onto BiGGY agar. White colonies were selected each time, and the procedure was repeated four times. We were only able to stabilize strain 45.2, and only white colonies were observed after two cycles. We were not successful in stabilizing strain 1.1, and due to its unstable phenotype it was not used for further fermentation experiments.

Small-scale fermentation experiment in grape juice

Small-scale fermentations were carried out using filter-sterilized 2007 Riesling juice, in order to study the fermentation performance of the five selected strains in real grape juice, and to obtain a quantitative measurement of the H2S productivity of these strains in these conditions. Strain 51.1 fermented at a similar rate as the PDM control strain, and strain 4.2 fermented even faster (which was consistent with the previous experiment). Both strains 22.1 and 23.2 fermented slightly slower than PDM, and strain 45.2 showed a very slow fermentation rate (Table 5). In general terms, the strains fermented faster in the grape juice than in the synthetic grape juice.

View this table:

Properties at the end of the alcoholic fermentation in grape juice

Residual sugar (g L−1)
Glycerol (g L−1)
Acetic acid (g L−1)0.280.460.390.110.210.33
Ethanol (%)14.511.314.514.114.614.6
Days to dryness12>1715171210
SO2 (free/total) (mg L−1)0/320/1640/1430/1794/1730/39
H2S production (μg L−1)72 ± 120.4 ± 0.10.7 ± 0.0300.9 ± 0.0442 ± 3
  • * Fermentations were stopped after 17 days.

  • The SO2 content of the grape juice was 10 mg L−1 of free and 32 mg L−1 of total SO2.

  • The SD did not exceed 10% for any of the values.

PDM did not accumulate any detectable SO2, because its total levels were exactly the same as in the grape juice (32 mg L−1). The strain 4.2 only produced small amounts of total SO2 (7 mg L−1), while the remaining strains accumulated considerable amounts of SO2. PDM produced considerable amounts of H2S throughout fermentation, and maximum levels were observed after 3 days of fermentation; after that point, the rate of H2S production dramatically diminished with time (Fig. 5). Tan strain 4.2 produced a reasonable amount of H2S throughout the fermentation, and its total H2S production was about 60% of that of the control PDM (Table 5). Low rates of H2S liberation were observed for the remaining strains (45.2, 22.1, 23.2 and 51.1); in all the cases, the total H2S production during fermentation was about 1% of that of the control strain.


Production of H2S of PDM (▪) and strains 45.2 (▲), 22.1 (△), 23.2 (●), 51.1 (○) and 4.2 (◆) during fermentation of a Riesling grape juice.

Sequence analysis of the sulfite reductase genes

The use of BiGGY agar as a tool to screen for colonies with a lighter phenotype than parental strain PDM allowed us to target for strains that were likely to have an impaired sulfite reductase activity. Therefore, we sequenced the MET5 and MET10 genes, which encode for the catalytic subunits of the sulfite reductase enzyme, in some of the low H2S-producing strains to look for mutations in those genes. Firstly, we sequenced the diploid PDM strain, and several allelic variants were identified with respect to the haploid reference strain S288c. Six heterozygous allelic variants were found in MET5, five in the coding region and one in the upstream regulatory region (Table 6). One of the MET5 variants corresponds to an in-frame deletion of 3 bp (859–861delGTA); as a consequence, one of the two MET5 alleles in PDM encodes for a protein that will have one amino acid less than S288c Met5p. Two of the variants in MET5 result in a change of amino acid, whereas the other two variants are silent. As for MET10, a total of 10 allelic variants were found and all of them are single nucleotide substitutions. Six of the variants result in a change of amino acid, whereas the other four are silent.

View this table:

Sequence comparison between the haploid S288c and the diploid PDM strains

Base pair positionsS288c sequencesPDM sequences (1/2)Amino acid substitutions
  • * Both alleles of the PDM strain were arbitrarily assigned as 1 and 2.

The sequence analysis of MET5 and MET10 was also performed in six of the low H2S-producing strains. At least one mutation was found in either MET5 or MET10 in all the strains (Table 7). As expected from EMS mutagenesis, all mutations were transitions at G–C sites (G-to-A and C-to-T transitions). A total of seven single nucleotide substitutions were found in MET10; all of them are present in heterozygosity. Two of the mutations in Met10p, W841X and W59X, result in a stop codon in the ORF. Two other mutations, G911S and E929K, are in partially conserved amino acids within the NADPH-binding domain of the ferredoxin reductase family; in the case of the glutamic acid 929, an aspartic acid can be found in this position. The other mutations found in Met10p, L606F, T990I and T997I are in amino acids that are not conserved between sulfite reductases of different species. Three of the strains (22.1, 51.1 and 4.2) showed two mutations in MET10. To elucidate whether both mutations are in the same allele or in different alleles, each of the MET10 alleles were subcloned and sequenced. The analysis of the results showed that in strains 51.1 and 4.2, each mutation is in a different allele, whereas in strain 22.1, both mutations are in the same allele.

View this table:

Mutations found in both MET5 and MET10 genes for the PDM low H2S-producing strains

1.12935G>A (A979T)
3344G>A (G1115D)
45.24066G>A (E1356K)
22.12523G>A (W841X)
2731G>A (G911S)
23.22939G>A (G980D)176G>A (W59X)
51.12785G>A (E929K)
2969C>T (T990I)
4.21816C>T (L606F)
2990C>T (T997I)
  • The amino acid change is shown in parentheses.

Four mutations were found in MET5, and all of them are located in the C-terminal domain of the protein (Table 7). The Met5p G1115D mutation is present in an amino acid that is highly conserved within the iron–sulfur/siroheme binding site of the nitrite and sulfite reductases. This mutation was found in the 1.1 strain, which has an additional mutation (Met5p A979T), and each of them is located in a different allele. The glutamic acid 1356, which is substituted by a lysine (Met5p E1356K) in strain 45.2, is also highly conserved between sulfite reductases of different species. Finally, the fourth mutation, Met5p G980D, was found in strain 23.2, and this glycine is not conserved in sulfite reductases of different species.

Effect of MET5 and MET10 mutations in sulfite reductase activity

To examine whether the mutations observed in the MET5 and MET10 genes were responsible for the phenotype of the low H2S-producing strains, the mutations were introduced individually in the centromeric plasmid pRS416 containing either MET5 or MET10, which were then transformed into the BY4742 laboratory strain, in which the chromosomal copy of MET5 or MET10 was deleted. Both genes were expressed under the control of their respective promoters in order to obtain an insight into the possible effect of the mutations on sulfite reductase activity.

Growth in minimal medium in the absence and presence of methionine was tested. With regards to MET5, three mutations (Met5p A979T, G1115D and E1356K) were studied. On solid media, all three mutants failed to grow in the absence of methionine, as did the cells expressing the empty plasmid. In contrast, cells transformed with the pRS416-MET5wt plasmid grew normally (Fig. 6b). In liquid culture, cells expressing Met5p G1115D and E1356K were unable to grow in the absence of methionine, but those expressing Met5p A979T started to grow after 48 h of incubation, showing that this mutant has a leaky phenotype (data not shown). Cells were also spotted on BiGGY agar plates to check if they had the same phenotype as the corresponding mutagenized PDM strains. All three mutants showed a white phenotype on BiGGY agar plates, the same as cells expressing the empty plasmid, whereas cells expressing the wild-type Met5p showed a tan phenotype (Fig. 6d). In addition, to confirm that the mutated versions of the MET5 gene were being expressed under our experimental conditions, we conducted a reverse transcriptase-PCR experiment. The results showed that MET5 is being transcribed in all three MET5 mutants, as well as in cells expressing the wild type MET5, but not in cells expressing the empty plasmid (data not shown).


Expression of MET10 (left) and MET5 (right) genes in BY4742Δmet10 and BY4742Δmet5 cells, respectively. Saccharomyces cerevisiae BY4742Δmet5 was transformed with the pRS416 vector alone, and with pRS416 containing MET5wt, and the mutants A979T, G1115D or E1356K. Saccharomyces cerevisiae BY4742Δmet10 was transformed with the pRS416 vector alone, and with pRS416 containing MET10wt, and the mutants W59X, L606F, W841X, E929K, T990I or T997I. Growth of the transformants in the absence of methionine was compared (a, b), and the color of the transformants in BiGGY agar was also assessed (c, d).

As for MET10, six of the mutations (Met10p W59X, L606F, W841X, E929K, T990I and T997I) were studied. On solid media, mutants W59X, W841X and E929K failed to grow in the absence of methionine, whereas cells expressing the wild-type protein, and mutants L606F, T990I and T997I grew normally in these conditions (Fig. 6a). Interestingly, the same three mutants that failed to grow in the media without methionine showed no sulfite reductase activity because they displayed a white phenotype on BiGGY agar (Fig. 6c). We also confirmed that the MET10 gene is being transcribed in all six mutants, but not in the cells expressing the empty plasmid (data not shown).


In this work, we clearly demonstrate that it is possible to develop commercial wine yeast strains, in a nongenetically modified way, which exhibit significantly reduced H2S production capacity. The simplicity of the method used for sulfide detection made it possible to screen a large number of colonies for the desired phenotypic characteristic. These results confirm that classical mutagenesis techniques, such as UV irradiation or treatment with EMS, commonly used with haploid laboratory strains, can be efficiently used for mutational and molecular breeding of diploid industrial yeast strains (Hashimoto, 2005).

Among the 225 colonies isolated, five methionine auxotrophs were obtained, of which two were white colonies (strains 45.2, 85.1), two were black colonies and one a light tan colony. Strain 45.2 displayed a white phenotype on both BiGGY and sulfate–BiGGY plates; therefore, we suspected that this strain could have one or more mutations in the genes that encode the catalytic or the regulatory subunits of the sulfite reductase enzyme. Our sequence analysis confirmed this assumption as this strain bears a mutation in the MET5 gene and our experimental results demonstrated that this mutation was responsible for the methionine auxotrophy of the strain. Interestingly, strain 85.1 showed a brown phenotype on BiGGY but a white phenotype on sulfate–BiGGY agar. This indicates that this strain is able to metabolize sulfite but not sulfate. This strain most probably bears one or more mutations in the genes that encode proteins involved in the upstream steps of sulfite reductase in the SRS. However, given the large number of mutant strains generated, we did not conduct further studies on this strain. The fermentation experiments indicated that in spite of the relatively high dose of chemical mutagenesis, some of the selected strains still managed to retain their basic machinery for fermentation, as all four of the strains selected could ferment synthetic and natural grape juice to dryness. However, there are differences in fermentation rates, indicating that some minor changes did occur. In the case of strain 45.2, its fermentation rate was much slower than that of the PDM control strain and of the remaining mutant strains. This could be due to the fact that it is a methionine auxotroph, and therefore it struggles to grow in a medium with low sulfur-containing amino acid concentrations. In all the fermentations we conducted, the number of cells was lower for the strain 45.2 by 50%, while the remaining strains showed similar growth patterns as PDM.

The basic wine parameters remained the same for most strains at the end of fermentation. The only concern from a chemical point of view at this stage is the high SO2-producing capacity of the low H2S strains, because it is known that SO2 can inhibit malolactic fermentation (MLF) by lactic acid bacteria in wines. The chemistry of SO2 in wine is complex because upon its addition to wine, SO2 enters a pH-dependent equilibrium consisting of a bound SO2, molecular or free SO2 and bisulfite and sulfite ions. Together, these different forms represent the total SO2. It is considered that a total SO2 concentration of 100–150 mg L−1, or free SO2 of 1–10 mg L−1 are sufficient to disturb the growth of lactic acid bacteria (Rankine & Bridson, 1971; Somers & Wescombe, 1982). Although the strains we developed were able to produce high amounts of total SO2 (between 100 and 150 mg L−1), only a small quantity of the SO2 is in the highly toxic-free form (usually <5 mg L−1). Preliminary experiments we conducted indicated that total SO2 might not be as toxic as expected as MLF could be completed in 21 days in a red wine (pH 4.0) produced with one of the low H2S strains, which accumulated 90 mg L−1 of total SO2 and only 6 mg L−1 of free SO2 (data not shown). However, a more controlled and full-scale investigation, which is out of the scope of this paper, is needed to confirm this effect.

The analytical measurement of the H2S production by the PDM control strain and the low H2S strains during fermentation in a Riesling grape juice showed that there is an inverse correlation between H2S and SO2 production for these strains. These results are not surprising because it has been shown that strains with an inactive sulfite reductase cannot reduce sulfite to sulfide; therefore, sulfite accumulates and diffuses from the cell into the media (Zambonelli, 1984; Hansen & Kielland-Brandt, 1996). Furthermore, the degree of blackening of lead acetate strips also showed a good correlation with the amount of H2S produced by these strains. The use of this simple qualitative method also offers a reasonable estimate of the extent of H2S produced during fermentation.

Furthermore, a full scale, complete metabolic and phenotypic analysis is needed to determine exactly how these yeast strains have been changed and how they could possibly differ in their metabolic effect on wine quality. Indeed, in future work, we intend to investigate some of these strains in this way, particularly in connection with its ability to produce flavor compounds, as small variations in the concentration of these compounds could have a dramatic effect on wine quality.

It is not surprising that six of the seven mutations observed in Met10p are located in its C-terminal domain, which shares a higher degree of protein sequence similarity among species than its N-terminal domain. Both FAD- and NADPH-binding domains, which are crucial for the function of the enzyme, are located in its C-terminus, which explains why mutations of investigated strains are concentrated in this part of the protein. Our experiments suggest that W59X, W841X and E929K mutations in Met10p strongly affect enzyme activity, because expression of plasmid-borne genes containing each of these mutations was unable to complement the methionine auxotrophy of a laboratory strain lacking the MET10 gene. Furthermore, none of these strains produced a detectable amount of sulfide on BiGGY agar plates. In the case of the two nonsense mutations, W59X and W841X, this result was expected because both mutations result in a premature stop codon and a truncated protein.

The Met10p W841X mutation, which results in a stop codon, is located just before the NADPH-binding site; hence, it is tempting to speculate that the resulting protein will not be able to bind this important cofactor. This domain is critical for the enzyme's activity because the flow of electrons within sulfite reductase occurs from NADPH to the prosthetic group FAD, both bound to the α-subunit. Electrons are then transferred to the β-subunit, where sulfite is reduced at the metal center of the siroheme. Therefore, the presence of this mutation in the strain 22.1 is sufficient to explain its light tan phenotype. In this strain, the MET10 allele bearing the W841X mutation will produce a nonfunctional protein, but the other allele will produce a functional Met10p; hence, we can expect at least a reduction of 50% of the normal enzyme activity. But the decrease in activity is likely to be higher because yeast sulfite reductase has an α2β2 oligomeric structure, and two functional α-subunits are needed for its function. Therefore, if the truncated protein is still able to interact with the normal α-subunit, and/or the β-subunits, it will have a sequestering dominant negative effect, and the percentage of active sulfite reductase will be less than the expected 50%. Strain 22.1 also has a second mutation in Met10p, G911S. The possible effect of this mutation in this conserved glycine was not studied further, because the mutation was found to be in the same allele as W841X; therefore, the protein will prematurely terminate at the aspartic acid at amino acid position 840.

The nonsense W59X mutation occurs at the beginning of Met10p, and the resulting truncated protein will only have 58 of the 1035 amino acids of the full-length protein; thus, this mutation is very likely to produce a nonfunctional protein. The presence of the W59X mutation in strain 23.2 also explains its light tan phenotype. The Met10p E929K mutation is located in an amino acid that is partially conserved in the NADPH-binding domain, and our data showed that the substitution of the negatively charged glutamic acid at this position by a positively charged lysine totally impairs sulfite reductase activity. We can expect that this mutation will have a dominant negative effect on sulfite reductase activity, as discussed for the W841X mutation. The Met10p E929K mutation was found in the 51.1 light tan strain, which also has an additional mutation in Met10p (T990I), but in different alleles. The presence of the Met10p E929K mutation in the 51.1 strain is sufficient to explain the low H2S-producing phenotype of this particular strain.

As for Met5p, all the mutations are located in its C-terminus. Again, this is not surprising, because both flavin mononucleotide and iron–sulfur/siroheme binding sites, crucial for the activity of Met5p, are located in its C-terminus. Our results showed that one of the two mutations observed in the strain 1.1, Met5p G1115D, strongly impairs sulfite reductase activity, because the expression of a plasmid carrying this mutation was unable to complement the methionine auxotrophy of a strain lacking the MET5 gene, and showed a white phenotype in BiGGY agar. This mutation occurs in a residue that is highly conserved in the nitrite and sulfite reductase iron–sulfur/siroheme binding site, and leads to a nonconservative change of a neutral glycine by a negatively charged aspartic acid, which suggests that this mutation is likely to affect the enzyme activity.

The effect of the other mutation observed in strain 1.1, Met5p A979T, remains uncertain. It is likely that this strain still conserves some residual enzymatic activity, and this would explain why cells expressing this mutation eventually grow in liquid media without methionine. Although this residual activity might be enough to support growth in these conditions, no H2S could be detected on the plate assay. The combination of the individual effects of both Met5p A979T and G1115D mutations, present in different alleles in the 1.1 strain explain the particular phenotype of this strain, which produced white colonies on BiGGY agar and grew slowly in a liquid minimal media without methionine.

Our experimental results also showed that the Met5p E1356K mutation, the only mutation found in the 45.2 strain, leads to an impairment in the enzyme's activity. The expression of this mutation in BY4742Δmet5 cells showed exactly the same phenotype as the 45.2 strain, i.e. white colonies on BiGGY plates and methionine auxotrophy. This mutation occurs in a highly conserved residue, and leads to the nonconservative change of a negatively charged glutamic acid by a positively charged lysine. Glu1356 is located very close to the last of the four highly conserved cysteine residues, Cys1349, in nitrite and sulfite reductases. These cysteine residues are important for the enzyme's activity because they are involved in the covalent binding of the iron–sulfur center, and the last one also binds the siroheme group (Crane, 1995). We can hypothesize that a change in charge of the conserved Glu1356, in the proximity of the critical Cys1349, could lead to this impaired activity. In principle, the presence of the Met5p E1356K mutation in only one of the two alleles of MET5 will not be sufficient to justify the phenotype of the 45.2 strain, but after a closer examination of the sequences we found that this strain is not heterozygous for MET5. Therefore, the E1356K mutation is likely to have a dominant effect. It is possible that, as a result of the chemical mutagenesis procedure following the initial induction of a mutation in one allele, a loss of heterozygosity has occurred through mitotic gene conversion or mitotic crossover (Nakai & Mortimer, 1969; Hashimoto, 2005). Another option is that the part or even the whole X chromosome, which contains MET5 has been lost as a result of the mutagenesis procedure. Further analysis is required to confirm this hypothesis.

From these results it is clear that further work is needed to understand the exact mechanism behind the low-H2S phenotype. Indeed, our aim is to conduct enzymatic test on some of the mutated versions of Met5p and Met10p.

Finally, this work clearly demonstrates that it could be possible to solve the age-old H2S problem in wine production, and possibly also in other alcoholic beverage production such as saké and beer. We showed that these strains could ferment high concentrations of sugar without obvious adverse effects on taste and quality. Currently, some of these strains are being patented and commercialized, and wine industry trials are underway in order to test their effectiveness in an industrial environment. Although there are some naturally occurring S. cerevisiae wine strains that could be classified as non-H2S producers (about 1%) (Zambonelli, 1984), we wanted to isolate a stable and non-genetically modified organism variant of the already low-H2S-producing strain, PDM, because this strain has good fermentation capacity and robustness. Furthermore, the mutations generated in our study are potentially different from the naturally occurring strains; therefore, more fundamental knowledge could be generated through the understanding of the nature and consequences of these mutations. Preliminary data from industrial-scale trials indicated that these strains are likely to pose a commercially viable solution to the H2S problem. Furthermore, producing wines free of H2S could eliminate the use of copper sulfate and potentially eliminate, or at least reduce the amount of diammonium phosphate used in the wine industry. Finally, H2S-free wines could be better suited for screw cap closures as these are known to have a tendency to produce, under certain circumstances, ‘mercaptan’-like off-flavors during storage. Future industrial-scale testing would confirm if this is the case.


The Australian Wine Research Institute, a member of the Wine Innovation Cluster in Adelaide, is supported by Australia's grapegrowers and winemakers through their investment body, the Grape and Wine Research Development Corporation, with matching funds from the Australian Government. A.G.C. is financially supported by Mauri Yeast Australia. We are grateful to Dr Vince Higgins from the University of Western Sydney for the pRS416 plasmid. We are also grateful to Drs Paul Chambers and Peter Costello for the discussions about the experiments, and to Drs Cristian Varela and Darek Kutina for their help with HPLC analysis of wine samples.


  • Editor: Ian Dawes


View Abstract