BY Angela Lee Linderholm and
Linda F. Bisson
Department of Viticulture & Enology
University of California, Davis

Formation of hydrogen sulfide (H₂S) is a major problem in winemaking because, if untreated, it can leave a wine with an undesirable rotten-egg sensory characteristic. The fact that Saccharomyces cerevisiae is responsible for the sulfide defect in wine has been well-established.²,³,⁴,⁵ Production of hydrogen sulfide by Saccharomyces cerevisiae strains ranges from 0 mg/L to 290 mg/L, well above the human detection threshold of 11 ng/L.¹

Sources of hydrogen sulfide

Hydrogen sulfide can come from different sources (Figure I). If sulfur is used in the vineyard too close to harvest, the reductive conditions that are created during fermentation can chemically convert this sulfur to hydrogen sulfide. Volatile sulfur compounds can also be formed during degradation of the sulfur containing amino acids, methionine, and cysteine, which usually are in low concentration in most juices barring supplementation.

Some strains have been shown to reduce sulfite (used in wine as an antimicrobial or antioxidant) to hydrogen sulfide, but this trait appears to be rare. The majority of H₂S produced occurs during the process by which Saccharomyces makes sulfur-containing amino acids.

Saccharomyces cerevisiae is able to reduce sulfate from its environment to sulfide and then incorporate it into amino acids through several more enzymatic steps (Figure II). Sulfide that is not incorporated into these amino acids is converted to H₂S in a pH-driven reaction. Because winemaking conditions are typically at low pH, the abundance of hydrogen molecules favors the reaction to create the volatile gas, H₂S.

Hydrogen sulfide can be formed at different times during wine production. Volatile sulfide formation can occur during the active phase of fermentation or as the wine ages on yeast lees post-fermentation. In the first case, it is thought that the sulfide is derived from either of the sources listed above.

Post-fermentation, sulfide is believed to arise either from degradation of sulfur-containing compounds in the yeast lees or from the re-release of chemically-entrapped sulfide during ageing (Figure III). In the latter situation, the entire sulfide formed during fermentation was not released as hydrogen sulfide gas, but a fraction was bound to other compounds or trapped in a non-volatile form in the wine.

As ageing changes the reductive environment of a wine, this sulfur can be released from its bound or altered form, creating H₂S, a reaction favored by the low pH of the wine. Thus, a sulfide problem, thought to be cured, can return.

Factors affecting sulfide formation

Hydrogen sulfide production varies depending upon the growth conditions of the yeast. Environmental and nutritional factors such as levels of elemental sulfur,⁵ presence of sulfur dioxide,⁶ organic compounds containing sulfur,² nitrogen limitation,³,⁴ and vitamin deficiency⁷ have been associated with the production of volatile sulfur compounds.

However, the effect of these variables is dependent upon media composition and growth conditions. It has also been observed that production levels vary dramatically across strains in response to these conditions.²,⁸ This variation suggests that differences in internal enzyme regulation and activity impact H₂S production. The existence of such dramatic variation in sulfide production across yeast strains has made it difficult to devise generalized recommendations guaranteed to prevent the appearance of H₂S in wine.

The sulfate reduction pathway that generates hydrogen sulfide also generates many toxic intermediates and is biologically expensive in terms of consumption of cofactors and cellular energy stores. But, at the same time, the end products of this pathway are essential for ethanol tolerance and maintenance of the redox status of the cell, two critical issues challenging the yeast during wine fermentations.

Thus, strategies that simply knock out the activity of this pathway are not viable for the construction of strains with reduced sulfide production. Instead, it is important to understand the basis of the naturally occurring low sulfide production trait so that it can be exploited to generate strains with reduced tendency towards sulfide production.

Our goal is to determine the biological basis of variation in sulfide production in yeast strains under winemaking conditions, and to define the interaction between strain and environment so that sulfide formation can be more effectively controlled during wine production. Although it is true that copper treatments can eliminate sulfide from fermentation and wine via precipitation, the residual copper must then be considered. It is far more desirable to not have the sulfide formed in the first place.

Anecdotal information from wine producers experimenting with screwcap closures suggests that they are experiencing more problems from sulfide formation in the bottle than in the same wine with cork or synthetic closures. Although the exact explanation of this phenomenon will require some experimentation, it is possible that sulfide formed during or after fermentation may indeed be trapped in a non-volatile chemical form resistant to copper precipitation that reverts back to H₂S over time under the reductive conditions established with screwcap closures. Similar comments have been made about de-alcoholized wine and the re-appearance of a sulfide problem.

Although several studies have predicted the existence of such a sulfide reservoir in wine, the nature of the bound compound remains unknown in spite of significant effort to identify it. This is further impetus to discover and develop wine yeast strains that will not release sulfide.

Understanding role of strain
variation in sulfide production

The specific aim of our research program is to understand hydrogen sulfide production in Saccharomyces cerevisiae at both the genetic and physiological level, and to define the genetic changes that have arisen naturally in native wine strain populations that result in reduced sulfide formation by these strains. This is no small task, as the genetic variation seen indicates a complex inheritance, meaning multiple genes are likely involved.

As a first step, we have characterized sulfide production in a large group of strains of S. cerevisiae, including native and laboratory strain isolates and commercial wine and beer strains.⁸ We have found that some rare strains are consistently high or consistently low producers of sulfide regardless of the medium in which they are grown.

However, most strains show a strong influence of growth conditions and medium composition on sulfide formation, and produce varying levels of sulfide in response to those changing conditions. The levels of sulfide produced may vary over 2,000 fold for a given strain in response to changing environmental conditions. Other strains show very little difference in sulfide levels produced under the same conditions. The highest level of production observed in commercial or native isolates is roughly 300 mg/L of sulfide.

Regardless of the condition, high temperature, low temperature, high nitrogen, low nitrogen, vitamin deficiency, mineral excess or deficiency, strains can be found that will show varying sulfide production levels in response to the specific environmental condition evaluated. The magnitude of the response also varies, with some low level-producing strains suddenly showing a very high level of production if something in their environment dictates.

Predicting strain behavior

We have found it difficult, if not impossible, to predict strain behavior with respect to the levels of sulfide formed on the sole basis of physiological studies and medium composition.

Several different methods have been proposed to evaluate the H₂S-producing behavior of yeast strains. Since sulfide forms a dark-colored precipitate with heavy metal ions, colony color analyses have been developed that correlate the amount of dark pigment produced with the level of sulfide being formed under the given growth conditions.

One such medium is called “BiGGY” agar. BiGGY (bismuth glucose glycine yeast agar), is a differential medium that evaluates relative production of sulfide. Sulfide generated by the yeast colony forms a black precipitate and the intensity of color is dependent on the amount of sulfide produced. Some examples of the different colony colors that can be observed are shown in Figure IV.

Copper-based media can also be used in a similar way. Finally, the sulfide formed during growth in liquid media can be quantified in the gaseous phase released during fermentation. Since H₂S is volatile, it is driven off in the carbon dioxide stream produced during fermentation of sugar. The sulfide in this stream can be trapped and quantified. This can be done using a liquid trap-based method ⁴ or by allowing the gas stream to pass over a column containing lead that will bind to the sulfide. ³

Finally, sulfide from the headspace can be analyzed using gas chromato-graphy. We have used all of the above methods to characterize sulfide formation in wine strains. We find that the convenience of the lead trap method allows screening of large populations of organisms and that other more quantitative methods can then be used on selected strains.

Interestingly, there is not good agreement between the plate-based colored colony and gas stream data. Strains that yield dark colonies often do not give significant sulfide production in liquid media. Early reports in literature suggested that BiGGY-agar assesses sulfite reductase activity under conditions of the plate and that the color of colonies on BiGGY was correlated with levels of H₂S production.² Brown color colonies produced a detectable amount of H₂S by aroma while white colonies produced no detectable sulfide by aroma. In our research, this was not consistently accurate.

While white colonies did not produce H₂S in our trials, some lighter-staining colonies did make more than darker staining ones under fermentation conditions. This is true because grape juices and synthetic juices are different in composition. These compositional differences create differences in flux through the sulfate reduction pathway, therefore differing levels of sulfide production occur. Because of this, actual fermentations must be conducted in order to accurately gauge a particular yeast strain’s H₂S production.

To define H₂S production in yeast strains in our work, we perform small-scale 300mL fermentations in synthetic juice media (called Minimal Must Media.)³,⁸ This media is modeled after grape juice but allows us to control individual components such as sugar concentration and nitrogen levels and growth conditions such as temperature and degree of mixing or shaking. Fermentations are incubated at 25†C with slow shaking to insure mixing of nutrients.

Hydrogen sulfide production is monitored with lead acetate tubes attached to the flasks. Carbon dioxide produced during fermentation pushes the volatile gases through the tube and the sulfur reacts with the lead acetate-coated beads to produce a dark gray band. The height of the gray column reflects the amount of H₂S production produced.

Varying media composition and growth conditions indicated that there is indeed significant variation in strain behavior with respect to formation of hydrogen sulfide, more variation than previously thought from earlier studies that investigated fewer strains. Some strains formed more H₂S at lower nitrogen levels, some more at higher nitrogen levels, and some appeared insensitive to nitrogen content with respect to sulfide formation.

Effect of nutrient levels

A more detailed analysis revealed that, in most cases, strains do indeed respond to nitrogen similarly, but the trigger-levels at which their behavior changes will vary among strains. This means that as nitrogen is increased or decreased, varying levels of H₂S will be produced. But the nitrogen concentration and the nature of the nitrogen compound required for a given strain to decrease sulfide formation may be different than for other strains.

We have studied sulfide production in various strains between the ranges of 120mg/L to 430mg/L nitrogen equivalents. At the highest nitrogen concentration, the level of production of hydrogen sulfide varied from 17% to 70% of that seen at the lowest level for commercial strains. Some native isolates showed even less production at high nitrogen levels.

The magnitude of the amount of sulfide released by individual strains will vary as well as the change in amount with nitrogen addition. Sulfide production at the lower nitrogen level ranged from undetectable to 150 mg/L for the strains tested while, at the lower nitrogen level, production varied from undetectable for only a couple of strains to 300 mg/L.

Since most previous studies selected a single or a few nitrogen levels but did not examine yeast strains across a wide range of nitrogen concentrations, the consistency of response to increasing nitrogen that we observed has gone unnoticed. Our results indicate that for some strains, although nitrogen supplementation may decrease sulfide production, the residual level that is produced may still be well over the threshold of detection.

These findings suggest that winery protocols for nutrient supplementation to reduce sulfide formation must take strain variation into account. Similarly, vitamin limitation will impact strains differently with respect to the amount of sulfide released. Certain vitamins are needed in order for reduced sulfide to be incorporated into amino acids. Deficiency for these vitamins can mimic a nitrogen limitation and lead to elevated levels of sulfide.

Other researchers have found that the interaction between nitrogen and vitamin levels is also an important factor in sulfide formation.⁷ Limiting both nitrogen and vitamin levels can result in even higher levels of sulfide release than either one alone.

Other growth conditions also impact sulfide formation, but our data suggests that nitrogen and micronutrient levels play a major role in defined media studies. Interestingly, we have not seen a strong effect of sulfite addition in juice or wine on sulfide formation. However, we do know that in juice, factors other than nutrient levels are also important.

In one study, Pinot Noir juices were characterized for sulfide formation and amino acid analyses of the juices were performed.⁸ A synthetic juice was then created mimicking the amino acid levels of the Pinot Noir juice that had produced high levels of sulfide. That high level of sulfide formation was not found in the synthetic juice, suggesting that nutrition alone is not the sole factor dictating sulfide formation. Manipulation of nutrient levels may not lead to elimination of a sulfide problem.

Defining genetic basis
of sulfide production

An alternate approach toward eliminating sulfide formation in wine strains is to use genetic analyses to identify the genes most impacting sulfide production with the aim of altering those genes so that sulfide levels will be reduced.

One might think that eliminating the sulfate reduction pathway all together would be a good strategy to reduce sulfide formation. But this would then require supplementation of the media with sulfur-containing amino acids and result in possible formation of volatile sulfur degradation products from these amino acids by yeast or by bacteria present in wine. This solution is similar to curing one problem by creating another. We decided instead to focus our efforts on determining why some native yeast strain isolates identified from our studies do not produce sulfide.

The brewing literature suggests that the main driving force in sulfide release is a diminished ability to incorporate reduced sulfide into amino acids, caused by relatively lower levels of activity of the enzymes downstream of sulfite reductase (Figure II).

Published reports demonstrated that enzymes that consume sulfide, in the sulfate reduction pathway, are responsible for controlling H₂S production in brewing strains. They demonstrated that by modifying strains to over-express one of these enzymes, H₂S production was reduced in their strains.¹⁰,¹¹ However, only a few brewing strains were evaluated in these studies and it was not clear that over-expression of these enzymes would reduce sulfide production in wine strains.

Role of variation in genes
of the sulfate reduction pathway

Given this information, we therefore decided initially to evaluate the role of the level of expression of these enzymes in H₂S production in wine strains.¹² The enzymes we evaluated are the three immediately downstream of sulfite reductase encoded by MET17, MET6, and CYS4. We took multiple approaches to determining the role of these enzymes in sulfide formation in wine strains.

First, we obtained these genes that encode for the activity of these three enzymes from laboratory and brewing strains. We then used these genes to look at the effect of increased expression of their encoded enzyme levels on sulfide formation in wine yeast strains. Although we were able to show dramatically higher levels of activity of enzymes encoded by these genes, there was no universal reduction in H₂S production with CYS4, MET6 or MET17.

The second approach was to measure levels of the enzymes encoded by these genes in wine strains to assess variation in activity. We did see differences in the level of activity of these genes, but it was not correlated with sulfide production behavior, suggesting that biological entrapment of reduced sulfur into amino acids alone could not explain the variation observed in sulfide production.

Experiments described above examine levels of expression but not differences in possible regulation or activity of the enzymes themselves. To evaluate this factor, genes responsible for expression of MET17, MET6, and CYS4 enzymes were sequenced from 20 different wine strains.

The function or regulation of a protein can change without a measurable change in activity. To test this possibility, we sequenced three genes encoding proteins that function downstream of sulfite reductase in 20 strains that showed varying levels of sulfide production. Since all yeast strains alive today have been exposed to DNA alteration, how do we define the original undamaged DNA sequence?

Mutations or changes in the primary DNA sequence arise naturally in yeast populations for many reasons. Sometimes a mistake is made during DNA replication and an incorrect base is inserted into the sequence that fails to be recognized as an error by the DNA repair enzymes.

In other cases, the yeast strain has been exposed to environmental stress or DNA-damaging agents that harm the DNA sequence. The yeast cell will die if the damage is not repaired. In some circumstances, the damage can be repaired but not to the original sequence as that exact information has been irreplaceably lost. The result is a viable strain, but one that is genetically different from a strain that was not exposed to DNA damage.

Therefore, the original, undamaged DNA sequence must be defined by consensus, meaning that the sequence that most unrelated strains possess is thought to be the undamaged one with rarer changes in sequence believed to have arisen from either an error in replication of the sequence or from external damage of the DNA.

The nature of the mutation can also be important. Some mutations result in dramatic changes to the structure and composition of the protein or enzyme made from that gene. A mutation can destroy the sequence such that it is no longer recognized as encoding a protein or that the protein encoded is so defective that it is no longer functional.

These kinds of mutations lead to a loss of function of the protein or enzyme, which can be quite damaging to the yeast in its native environment. Other mutations are more subtle, that is, change one amino acid in the protein altering the activity or regulation of the protein but not eliminating its function altogether. These mutations can often only be found by sequencing the gene itself.

We decided to sequence the MET17, MET6, and CYS4 gene in 20 wine strains, both commercial and native isolates, to assess the level of naturally-occurring mutations found in members of this pathway. The 20 strains selected were unrelated to each other and vary from laboratory wild-type strains.

Interestingly, all strains analyzed had the identical sequence for the MET17 gene. This gene is highly conserved in yeast, meaning that mutations that alter its activity or regulation at the protein level are deleterious to the strains carrying that alteration. The CYS4 gene was identical in all strains but one, and the MET6 gene showed variation in four strains.

The sequence analysis confirmed our earlier observations that differences in the level of activity of the enzymes involved in incorporating reduced sulfur into amino acids were not well correlated with sulfide formation in wine strains. The task then is to determine what genes do dictate sulfide formation behavior.

Identifying all genes that
affect sulfide formation

Our next approach to identify what genes affect H₂S production was more global. Completion of the genome sequence of Saccharomyces led to identification of all of its genes. This information resulted in creation of a set of strains carrying mutations in each and every gene. Some mutations were lethal, but most were not.

We decided to screen the entire set of the yeast-deletion strains, comprised of 4,827 mutants, for effects on sulfide production. We did this initially using BiGGY agar to identify all strains that lead to a lighter or darker colony color than the wild-type parent strain.

The parent yeast strain of the deletion collection was tan on BiGGY and produced no H₂S during fermentation, and mutants ranged from white to brown. Wine strains on this medium also display colors ranging from white to brown.

The mutants were classified into six different color groups: white (4), light tan (257), tan (4,476), light brown (61), brown (28), and black (1) (Figure V). The light brown and brown colonies found by the screen were subjected to a second screen in synthetic juice media MMM and Pinot noir juice, to evaluate production of hydrogen sulfide. Fourteen of the 89 strains produced detectable amounts of H₂S consistently using the lead trap method.

Five of the 14 mutations are in the methionine synthesis or transport pathway, with others involved in various pathways such as amino acid, or nitrogen compound transport or utilization. While the reasons why some of the mutations affect H₂S production are obvious due to the pathways they are in, some are not. This inconsistency opens up new ideas as to what is causing some wine strains to produce H₂S.

In a related experiment, we have examined mRNA expression profiles of a collection of 30 wine strains of Saccharomyces that vary in sulfide production. The aim of this study is to identify genes that consistently change in expression in relation to production of H₂S.

Preliminary analysis of this data set suggests that it is the relative ratio of levels of expression of the upstream and downstream enzymes and relative level of activity of the entire pathway that influences the ability to produce sulfide. Interestingly, the two lowest-producing strains that we have identified appear to have levels of expression consistent with a high level of sulfide production, yet they produce virtually undetectable amounts during fermentation.

One of these strains carries point-mutations in five out of seven genes that were sequenced in the genome so has clearly been exposed to significant DNA damage. We are currently analyzing this strain to define the gene or genes responsible for the loss of sulfide formation. If we can identify specific genes that are responsible for H₂S production characteristic, we could cross those genes into more robust yeast strains to produce a commercial wine yeast strain with low H₂S production, and thereby eliminate H₂S formation by Saccharomyces during wine fermentations — a goal that once seemed unrealistic but that now is within sight.

Summary

Our research has shown that there is significant strain variation in sulfide production and that while nitrogen supplementation does tend to reduce sulfide levels, even with high nitrogen supplementation, some strains will still produce sulfide at an order of magnitude over the threshold of detection under these conditions.

It is also difficult to predict strain behavior in response to nutrient limitation. Some strains do not show a dramatic response while in others, sulfide levels may vary over 2,000-fold.

Analysis of known mutations has revealed a set of genes that, when lost, result in high sulfide production or in the absence of detectable sulfide. These mutants produced up to 700 mg/L of sulfide, higher than ever observed in commercial strains under any condition.

Now that we know which genes impact sulfide production, we can determine which ones are varying in native and commercial yeasts and define the best strain “genotype” for low sulfide production and attain our goal of eliminating the appearance of this off-character during wine production.

Acknowledgements:

This research was supported by a grant from the American Vineyard Foundation and the California Competitive Grant Program for Research in Viticulture and Enology, and by funds of the Maynard A. Amerine Endowment. Angela Lee Linderholm was recipient of the Paul Monk Scholarship.


References