INTRODUCTION
The rising popularity of craft beers and the growth of the craft brewing industry means that more beer is being produced in facilities lacking pasteurization (
1). Pasteurization stabilizes beer against contamination by spoilage organisms, including yeast and bacteria. The shift away from pasteurization is likely because of the high capital costs of pasteurization equipment and increased energy and water usage (
2,
3). In addition, the dominant beer styles produced by the craft brewing industry are negatively affected by the high temperature of the pasteurization process. Beer styles that have been aggressively hopped post-boil, such as India Pale Ales (IPAs), will suffer from excessive exposure to oxygen when pasteurized (
4). This treatment is perceived to increase the rate of staling and can lead to off-putting flavors described as papery, wet, cardboard-like, leathery, or even “catty.” This means that many breweries will avoid pasteurizing as it will cause the degradation of delicate hop aromas. This aversion to pasteurization can increase the risk of spoilage in craft breweries. Although the antimicrobial properties of hops can protect beers from bacterial spoilage, yeasts are more resistant and represent a more problematic spoilage organism without pasteurization.
One group of yeasts that cause spoilage in craft breweries are
Saccharomyces cerevisiae strains that express the
STA1 gene to produce an extracellular glucoamylase enzyme. These strains have been referred to as diastatic yeasts, a name that is derived from diastase, an alternative nomenclature for amylase. Diastatic yeasts are an evolutionarily related group of
S. cerevisiae strains used commercially to produce high-gravity Belgian-style beers (
5,
6). Diastatic yeasts are unique because the
STA1 gene allows the hydrolysis of long-chain polysaccharides such as dextrin and starch. In non-Belgian-style beer, these carbohydrates remain after the primary fermentation has consumed simple di- and monosaccharides created in the mashing process. Dextrins and starches are usually unavailable to most commercial brewing strains as they lack the appropriate hydrolytic enzymes to break the glycosidic linkages between carbohydrate monomers. The
STA1 gene evolved due to a fusion of
FLO11 and
SGA1 (
7,
8). The gene fusion resulted in a chimeric protein with the N-terminus of the
FLO11 gene joined to almost the complete open reading frame of the
SGA1 glucoamylase. The 5′ end of
FLO11 fused to
SGA1 enabled the transport of Sta1 into the extracellular milieu, where it can hydrolyze residual dextrin and starch to glucose monomers. The resulting glucose is used to prolong fermentation (usually after packaging), referred to as over-, super-, or hyperattenuation. Hyperattenuation results in the overproduction of CO
2 and alcohol, imparting off flavors and promoting “gushing” and the explosion of containers. The
STA1 gene is also part of a family undergoing gene duplications and translocations, creating the paralogs
STA2 and
STA3 (
9). One of the few publicized examples of a major diastatic yeast contamination resulted in the recall of $2 million of packaged beer by Left-hand Breweries and was the subject of a $6 million lawsuit (
10). Significantly, the overall occurrence of spoilage by diastatic yeasts in Europe increased between 2008 and 2017 and is an important problem for the brewing industry worldwide (
11,
12).
Good hygiene, strain husbandry, and monitoring practices can reduce the likelihood of contamination by diastatic yeasts. However, viable cell counts on agar-based media can be somewhat unreliable and take days after sampling to identify contaminants (
13). The gold-standard molecular methods for rapidly detecting the
STA1 gene by PCR require specialized equipment, reagents, and personnel to perform and interpret these technically demanding assays (
14). Even if diastatic yeast contamination is detected, the primary course of action in a brewery without a pasteurizer is the destruction of the contaminated product. Therefore, there is an urgent need to develop cost-effective technologies that actively prevent or remediate contamination by diastatic yeasts.
Killer yeasts can produce extracellular proteinaceous killer toxins that inhibit the growth of competing species of fungi (
15–18). Many studies have shown the effectiveness of killer yeasts in preventing spoilage of fruits, silage, and wine fermentation (
19–26). These successes have led to the application of certain fungal species and genetically engineered crops as biological controls in agricultural processes (
27–30).
Saccharomyces yeasts were some of the first species identified as producing killer toxins (
31), and surveys have estimated that many strains of
S. cerevisiae are killer yeasts (
32,
33). Killer toxin expression often depends on cytoplasmic double-stranded RNAs (dsRNAs) replicated and encapsidated by viruses of the family
Totiviridae (
34). To date, nine dsRNA-encoded killer toxins are produced by different strains of
S. cerevisiae (K1, K2, K28, and Klus) and
S. paradoxus (K62, K1L, K21, K74, and K21/K66) (
35–42). At least two functional genome-encoded killer toxins exist in
S. cerevisiae (KHR and KHS) (
43,
44). These killer toxins share little primary sequence homology and can target susceptible cells by different mechanisms, such as by disrupting cell membranes (K1, K1L, and K2) or by arresting the cell cycle (K28) (
45,
46). The antifungal activities of killer toxins are generally limited to closely related species, and there is evidence of widespread resistance across different yeast lineages, which has limited their potential as broad-spectrum antifungals (
47,
48). However, despite notable resistance to killer toxins, some studies have shown that the canonical killer toxins of
Saccharomyces yeasts can successfully inhibit specific human and agricultural pathogens (
49–53).
This study demonstrates that killer toxins from Saccharomyces yeasts have potent antifungal activity against different strains of diastatic yeasts. After screening an extensive collection of Saccharomyces killer yeast, diastatic strains resistant to canonical killer toxins were found to be susceptible to a non-canonical variant of the K2 toxin named K2v. As proof-of-principle for applying killer yeasts to control diastatic contamination, a K2 killer yeast strain was able to prevent hyperattenuation in an industrial-scale fermentation with no adverse effect on the final gravity. This work provides a framework for using killer yeasts in the craft brewing industry to prevent future losses and lawsuits.
RESULTS
To determine the susceptibility of diastatic (
STA1+) strains of
S. cerevisiae to killer toxins, 34 diastatic strains were challenged by
Saccharomyces yeasts expressing eight different canonical killer toxins (K1, K1L, K2, K21/K66, K28, K62, K74, and Klus) (
Fig. 1A). Zones of growth inhibition and halos of methylene blue surrounding the killer yeast indicated the susceptibility of diastatic yeasts to the antifungal activities of killer toxins. Whereas zones of growth inhibition showed no growth of diastatic yeasts, methylene blue halos likely resulted from the initial growth of diastatic yeasts, followed by cell death due to sustained killer toxin exposure. Specifically, loss of viability results in the oxidation of methylene blue present in diastatic yeast cells and the appearance of blue-stained cells. The extent of growth inhibition was first qualitatively scored according to the degree of growth inhibition and methylene blue staining using a high-throughput plating assay (
Fig. 1A;
Table S1). Of all the canonical killer toxins assayed, K1 was judged to be the most inhibitory to diastatic yeasts and could prevent the growth of 91.2% of the diastatic strains tested. K2 could inhibit the growth of 58.8% of diastatic strains and, after K1, produced the largest zones of growth inhibition with methylene blue halos. The potency of K1 and K2 against diastatic yeast was further confirmed by quantitatively comparing the area of killer toxin inhibition against all diastatic strains tested (
Fig. 1B). This analysis again found that K1 was the most effective at inhibiting diastatic yeasts. Overall, the quantitative analysis agreed with the K1 qualitative assay and had only two false positives (OYL055 and OYL026) across the K2 dataset. Overall, K1 was significantly more potent than K2, with an average area of growth inhibition of 175.5 mm
2 (SD; 84.5), whereas the average for K2 was 87.6 mm
2 (SD; 92.9) (Student’s two-tailed
t-test,
P < 0.05). This demonstrated that diastatic yeasts are particularly sensitive to the K1 killer toxin produced by
S. cerevisiae.
Killer toxin production by
Saccharomyces yeasts is accompanied by immunity to the mature toxin. To determine whether the killer toxin-resistant diastatic yeasts had gained immunity due to killer toxin production, three K1-resistant diastatic yeasts and an additional ten strains resistant to K2 were used to challenge three lawns of
S. cerevisiae known to be susceptible to K1 or K2. Only three diastatic strains were identified as killer yeasts (APP, AQH, and AFB) (
Fig. 2A;
Fig. S1). To determine whether killer toxin production was due to viruses and associated dsRNA satellites, each of the 14 killer toxin-resistant diastatic yeasts was subjected to analysis by cellulose chromatography to purify dsRNAs. This analysis revealed that five strains contained dsRNAs with sizes the same as totiviruses (~4.6 kb), and three strains harbored an additional satellite dsRNA (~1.5 kb) (
Fig. 2B;
Fig. S1). Using total nucleic acid samples, reverse transcriptase PCR (RT-PCR) was used to detect the K2 killer toxin gene in the strains AFA, AFB, and AFP (
Fig. 2B). K1 was not detected in any strains assayed by RT-PCR, and PCR alone could not amplify K1 or K2, indicating that the DNA genome does not encode these killer toxin genes (
Fig. 2B). Exposure to cycloheximide was used to cure the satellite dsRNAs from the killer yeasts AFA, AFB, and AFP, as determined by cellulose chromatography and RT-PCR (
Fig. 2C). This curing treatment resulted in the loss of killer toxin production and susceptibility to K2, with K1 susceptibility remaining unchanged (
Fig. 2D). These data show that while K1 resistance of diastatic yeasts was independent of dsRNAs, K2 resistance was due to the presence of M2 dsRNA satellites.
Diastatic yeast strains were resistant to K1 and K2 killer toxins (AFA, AQH, and OYL-112) and K1 and K74 (AFA and OYL-112). The diastatic strain OYL-112 was resistant to all canonical killer toxins. Therefore, 192 previously identified and uncharacterized
S. cerevisiae killer yeasts were screened to determine whether they could inhibit the growth of killer toxin-resistant diastatic yeasts (
33). In total, 32 killer yeasts were able to cause growth inhibition of K1 and K2-resistant diastatic yeast (
Table S2). Three strains of killer yeasts (CHD, BSG, and ACP) were judged the most effective at inhibiting the growth of killer toxin resistant diastatic yeasts (
Fig. 3A). These killer yeasts also inhibited the growth of all other diastatic yeast strains except the K2-resistant diastatic strain AFB (
Fig. 3B). These three novel killer yeasts were analyzed for dsRNAs using cellulose chromatography, which found that all three harbored totiviruses and satellite dsRNAs. RT-PCR confirmed that these strains were K2 killer yeast (
Fig. 3C). This result was surprising as this novel K2 variant could inhibit strains AFA, AQH, and OYL-112, which were all resistant to the canonical K2 toxin (
Fig. 1).
Purification and sequencing of the dsRNAs from
S. cerevisiae strains CHD, BSG, and ACP confirmed that all three strains contained satellite dsRNAs with K2 killer toxin genes. These K2 genes had four non-synonymous mutations compared to canonical K2 (
Fig. 3D). To distinguish this mutant toxin from canonical K2, it will be referred to as K2-variant (K2v) and the satellite dsRNA as M2v. K2v and K2 genes were introduced into a plasmid for expression in a non-killer laboratory strain of
S. cerevisiae to directly compare the effect of the observed non-synonymous mutations on the spectrum of killer toxin activity. Comparing the galactose-induced expression of K2 and K2v from a high copy plasmid, it was found that K2v had a broader spectrum of antifungal activity that could inhibit 78% of diastatic yeasts compared to K2, which inhibited only 50% (
Fig. 3A). Galactose-induced expression of K2 was almost identical to the wild-type K2 killer yeast, but plasmid-expressed K2v inhibited less diastatic yeasts than the K2v killer yeasts CHD, BSG, and ACP (
Fig. 3A). Plasmid-expressed K2v could not inhibit the diastatic K2 killer yeast AFB that harbored an M2 satellite dsRNA, suggesting that K2 immunity function could protect this strain from the K2v killer toxin. Surprisingly, K2v could inhibit the K2-resistant diastatic strains AFA and AFP that also harbored M2, indicating that K2 in these strains is insufficient for K2v immunity. Overall, K2v is characterized as a variant K2 killer toxin with a broad-spectrum activity against diastatic yeast compared to canonical K2. Mutations in K2v likely caused changes in the killer toxin spectrum of activity and immunity that could inform the future development and application of K2 against diastatic yeasts.
To determine whether it was possible to use killer yeasts to prevent hyperattenuation by diastatic yeasts, two 1,000 L brewing trials were conducted using the non-killer brewing strain WLP-001 (
Fig. 4A). Both fermentations proceeded normally in the first 6 days, with some variability in the gravity readings in the first 36-h period due to the rapid evolution of CO
2 (
Table S3). After approximately 100 h of stable readings, fermentations were judged to have reached terminal gravity [~1.6° Plato (P)]. In trial one, the diastatic phenolic off-flavor (POF)+ yeast strain Belle Saison (Lallemand Inc.) was added to a final concentration of 5 × 10
4 cells mL
−1. The addition of the diastatic yeast cells resulted in a rapid drop in gravity to 1.06° P (
Fig. 4B) as well as an increase in pH (
Fig. 4C) and temperature (
Fig. 4D) before the trial was halted. This indicated that diastatic yeasts could ferment saccharides derived from the hydrolysis of residual starches and dextrins in the finished beer. For trial two, as the diastatic yeast Belle Saison was sensitive to the K2 killer toxin, remediation of a simulated contamination event was trialed by adding the K2 killer yeast strain Viva (Renaissance Yeast) that was chosen because of its routine use in the brewing industry (
Fig. 1). Moreover, Viva is a POF− strain with suitable alcohol tolerance, desirable ester profile, and reduced production of hydrogen sulfide and 4-vinyl guaiacol. Many of the characteristics of Viva are shared with the primary brewing strain WLP-001. The diastatic and killer yeast strains were added simultaneously to a final concentration of 5 × 10
4 cells mL
−1. In contrast to trial one, the gravity in trial two dropped by only 0.08° P before recovering to 1.80° P at the end of the trial (
Fig. 4B). The pH (
Fig. 4C) and temperature (
Fig. 4D) remained stable.
To assess the effect of diastatic remediation on flavor profile, a sensory panel of trained cicerones performed a hedonic rating like/dislike and off-flavor evaluation on a 10-point scale. The yeast strain used in these trials (WLP-001) is commonly used in brewing and is characterized by a clean and fruity aroma (
Fig. 4E). While it was evident that adding a K2 killer yeast prevented hyperattenuation, there was still a noticeable and undesirable flavor to the final brew. Specifically, while the beer produced from both diastatic trials maintained several desirable flavor characteristics (fruity/sweet/malty), they were very expressive of 4-vinyl guaiacol (4-Vg), which presented as clove or allspice, with a sensory score of 5 out of 10 in both fermentation trials (
Fig. 4F). This off-flavor was present with or without adding the POF− K2 killer yeast strain (Viva) despite preventing hyperattenuation. In addition, the trial with the K2 killer yeast had notes of an autolysis/meaty flavor (Auto) that we attribute to the successful killing and lysis of the diastatic strain in this trial by the K2 killer toxin.
DISCUSSION
For decades, killer toxins have been proposed as an alternative to synthetic and inorganic fungicides to control pathogenic and spoilage fungi. However, their narrow spectrum of antifungal activity and general instability has likely limited the application of killer yeast. The evolution of the
STA1 gene is a unique genome innovation present in two clades of
S. cerevisiae, one clade includes brewing yeasts while the other includes yeasts isolated from humans (
5,
6). Given the low genetic diversity of diastatic yeasts, these clades have similar killer toxin susceptibilities, as killer toxin sensitivity can be related to phylogenetic distance for some species (
47,
48). This would suggest a unique opportunity for the application of killer toxins as an approach to prevent diastatic contamination in craft breweries.
Several strains of diastatic yeasts are resistant to canonical K1 and K2 toxins. For K2, this resistance was due to the acquisition of totiviruses and M2 satellite dsRNAs that provided preprotoxin-mediated immunity essential for the self-protection of killer yeasts from their toxins (
15,
54). Although K1 immunity can also be linked to preprotoxin immunity, K1 resistance in diastatic yeasts was independent of satellite dsRNAs and likely due to unknown genome-encoded immunity determinants. Prior large-scale screens of genome deletion libraries have demonstrated that many cellular pathways can contribute to killer toxin immunity, and
S. cerevisiae can rapidly evolve K1-resistance in cell culture (
42,
55–59). A recent genome-wide association study of K28 resistance in
S. cerevisiae identified polymorphic alleles of
KTD1 that dictated K28 susceptibility (
60). Similarly, truncated killer toxin genes analogous to the minimal preprotoxin immunity domain of K1 have been found in the genomes of several species of Saccharomycotina yeasts (
41). The acquisition of dsRNA satellites and the presence of anti-toxin defenses in yeasts suggest that the application of killer toxins in craft breweries could drive the evolution of killer toxin resistance in diastatic yeasts. However, the prevalence and diversity of killer toxins in
S. cerevisiae motivate the screening for killer toxins that would overcome evolved resistance in diastatic yeasts (
32,
33). The mutations identified in K2v broaden the spectrum of antifungal activity against diastatic yeasts compared to canonical K2. Previous studies have identified a variant K2 toxin named K3 based on differences in the spectrum of activity and dsRNA satellite size (
61,
62). Similarly, polymorphisms in K1 have also been shown to alter the potency and antifungal specificity of K1 (
33,
63). Therefore, a better understanding of how mutations improve the efficacy of killer toxins will benefit their future application against diastatic yeasts and other pathogens and spoilage fungi.
Proof-of-concept fermentation trials show that killer toxins effectively prevent diastatic hyperattenuation resulting from the growth of
STA1+ Saccharomyces cerevisiae. Similar protection has been observed in the winemaking industry, where killer yeasts are widely used and can prevent contamination by undesirable strains of
Saccharomyces but not non-
Saccharomyces species of yeasts (
64). In situations where a brewery actively monitors for the presence of diastatic yeasts during fermentation, the addition of killer yeasts or enriched killer toxins to a contaminated fermenter could be an approach to prevent product loss. However, the success of remediation would likely depend on the extent of diastatic contamination, and there is a need to define the number of killer yeasts or concentration of toxin required to prevent hyperattenuation. Future experiments to investigate the population dynamics after killer yeast or toxin remediation would yield valuable insights into the viability of diastatic yeast cells after treatment and the long-term stability of yeast toxins in beer.
In the fermentation trials to remediate diastatic contamination using K2 killer yeasts, the beer produced had a noticeable and undesirable phenolic flavor despite preventing hyperattenuation. The brewing strain WLP-001 was used in the primary fermentation for its clean aroma profile and POF− status; thus, the phenolic flavor after diastatic contamination was attributed to the addition of the diastatic POF+ Belle Saison yeast used in the trial. In these trials, a high final concentration of diastatic yeast was added (5 × 10
4 cells mL
−1), considerably higher than the threshold for contamination in the brewing process (
13). Therefore, the large bolus of diastatic yeasts was expected to be responsible for the undesirable flavor characteristics of the beer produced by these trials. Under more realistic scenarios with lower numbers of diastatic yeasts invading the brewing process, lower concentrations of the killer toxin in beer would likely be sufficient to prevent hyperattenuation and undesirable flavors. Indeed, killer toxins can trigger the cell death of susceptible yeasts at lower concentrations than those required for cell lysis (
65,
66).
As many craft breweries do not actively monitor for diastatic contamination, an alternative approach to safeguard against contamination could be engineering brewing strains to produce killer toxins during fermentation. Killer toxin genes could be introduced into the yeast genome by selective breeding or direct genome editing, as has been demonstrated for winemaking yeasts (
67–69). Alternatively, totiviruses and satellite dsRNAs that encode killer toxins could be introduced into existing brewing strains by cytoduction (
70). Engineered brewing yeasts have solved many fermentation-related problems for craft brewers (
71). Practical examples include lactic acid-producing yeast, diacetyl-free yeasts expressing alpha acetolactate decarboxylase, and yeasts expressing β-lyase to produce aromatic thiols. These yeasts allow for much faster fermentation times and save brewers money in labor and materials while enhancing the taste and flavor of the beer. As yeasts are pitched into wort at high densities, killer toxin concentrations are predicted to increase rapidly during fermentation. Thus, killer toxins in wort could prevent the invasion of diastatic yeasts into the brewing process at any downstream production stage.
S. cerevisiae acidifies wort during fermentation to a pH of ~4.2, which is optimal for killer toxin activity (
72,
73). The stability of killer toxins in the finished beer remains to be investigated, but it is conceivable that at low pH and ambient temperatures, killer toxins would remain active during the packaging process and protect against diastatic contamination. Alternatively, killer yeast could also be used for “conditioning,” whereby yeast is added during packaging for natural carbonation. This would allow for killer toxin production in the packaged beer, protecting the finished product from diastatic yeast invasion. Regardless of the method of killer yeast application in craft breweries, the most crucial consideration would be to ensure desirable fermentation profiles, flavor, and shelf life. Therefore, developing killer brewing strains will be a priority to realize the successful industrial application of killer toxins.
ACKNOWLEDGMENTS
We thank the team at Precision Fermentation for their Brew IQ instrument and all the support we received while performing the trials at the 1,000 L scale. Laura Burns and Lance Sharner at Omega Yeast provided invaluable help with their catalog of diastatic yeasts. We would also like to acknowledge Dr. Antonia Santos (Complutense Yeast Collection, Complutense University of Madrid) for S. cerevisiae CYC1172 and CYC1058, Prof. Manfred J. Schmitt (Saarland University, Saarbrücken, Germany) for S. cerevisiae MS300c, Dr. Gianni Liti (University of Nice) for S. paradoxus K74, K62, and K21 killer yeast strains. We also thank Mark D. Lee and Dr. Mike Rolfsmeier for constructing and confirming the sequences of the killer toxin expression plasmids. We are also grateful for the help, advice, and mentorship provided to V.Z. and X.G. by Mark D. Lee, Lance R. Fredericks, Angela M. Crabtree, and Josie M. Boyer.
The research was supported by funds provided to P.A.R. by the Institute for Modeling Collaboration and Innovation at the University of Idaho (NIH grant #P20GM104420), the Idaho INBRE Program, an Institutional Development Award (IDeA) from the National Institute of General Medical Sciences of the National Institutes of Health (NIH grant #P20GM103408), and the National Science Foundation Division of Molecular and Cellular Biosciences (grant numbers 2143405 and 1818368). This work was partly supported by NIH COBRE Phase III grant #P30GM103324. The funders had no role in study design, data collection and analysis, the decision to publish, or the preparation of the manuscript.