Brettanomyces bruxellensis Aroma-Active Compounds Determined by

Brettanomyces bruxellensis Aroma-Active Compounds Determined by SPME GC-MS Olfactory Analysis C.M. Lucy Joseph,1 Elizabeth A. Albino,1 Susan E. Ebeler,1 and Linda F. Bisson1* Abstract: A survey of 95 Brettanomyces strains was undertaken to identify strains that consistently give positive aroma characteristics (e.g., spicy, fruity, and floral). These strains were grown in a defined medium, and five human evaluators rated each strain according to aroma quality, and their ratings were coupled with a solid phase micro-extraction with gas chromatography (SPME GC-MS) analysis. None of the strains yielded universally positive aromas for the evaluators. A further characterization of nine of these strains grown with both aromatic amino acids (phenylalanine, tryptophan, and tyrosine) and hydroxycinnamic acids (caffeic, p-coumaric, and ferulic acids) indicated that low levels of compounds that were most important in differentiating the strains may contribute to a positive sensorial perception of Brettanomyces strain aromas under these conditions. To define the components associated with positive Brettanomyces character, the volatile aroma compounds produced by five Brettanomyces bruxellensis strains were analyzed. Compound detecting in SPME GC-MS was coupled with olfactory analysis (SPME GC-MS-O), using nine individual evaluators to identify compounds most associated with Brettanomyces character and to assess the breadth of descriptive terms used by different individuals for the same compound. Twenty-two compounds were identified as having an impact on aroma, including the well-known ethylphenols and vinylphenols, as well as several fatty acids, alcohols, esters, terpenes, and an aldehyde. Key words: Brettanomyces, SPME, olfactory, volatile phenol

Brettanomyces in wine, and results from a recent evaluation of Brettanomyces character in wine have highlighted the importance of chemical interactions in perceived aroma characteristics (Romano et al. 2009). Many descriptors that are not overtly negative are used when Brettanomyces is present in wine, particularly those defined as spicy, smoky, savory, generic fruity, and mild floral (Arvik and Henick-Kling 2002). A descriptive analysis of wines inoculated with different Brettanomyces strains showed that although most panelists used terms commonly associated with Brettanomyces offcharacters (barnyard, Band-Aid, soy, and leather), others did not (Wirz 2005). Two potential reasons may account for this observation: either the panelists were unable to detect these compounds or they did detect them but used unrelated descriptive terms to describe their aroma. To distinguish between these two possibilities, panelists were asked to evaluate specific aroma compounds that have been identified as being products of Brettanomyces in wine (Heresztyn 1986a, 1986b, Chatonnet et al. 1995, Licker et al. 1999, Fugelsang and Zoecklein 2003) and were used to spike synthetic wines. Some panelists were unable to detect specific compounds, and others used nontraditional terms to describe these aromas (Joseph, unpublished observations). Thus, both possibilities may contribute to the perception of these chemicals in wine. This finding is not surprising because individuals are known to perceive aromas differently (Hasin-Brumshtein et al. 2009). The importance of semantics in wine descriptions has been discussed previously (Parr et al. 2004). Furthermore, how individuals describe odors and whether an aroma is perceived as positive or negative is often based on experience (Pangborn 1988, Stevenson and Boakes 2003). A descriptive analysis of Brettanomyces strains in wine and subsequent

Brettanomyces bruxellensis is arguably the most important yeast contaminant in finished wines (Loureiro and MalfeitoFerreira 2003, Oelofse et al. 2008). This yeast can grow and produce aroma compounds in the low-pH, low-sugar, lownutrient, and high-alcohol environment of dry commercial wines. Although often considered a spoilage organism, many wine enthusiasts appreciate the complexity that Brettanomyces brings to a finished wine. An attempt was undertaken to determine what types of aroma compounds are produced by Brettanomyces under controlled conditions and to seek strains that are uniquely able to produce only those aroma compounds generally perceived as pleasing, such as spicy, fruity, or floral. Brettanomyces has a highly diverse genetic background and a diverse physiology (Conterno et al. 2006, Hellborg and Piskur 2009, Curtin et al. 2012, Piskur et al. 2012). The University of California, Davis, Viticulture and Enology culture collection curates Brettanomyces strains from around the globe, enabling the study of the metabolic diversity of this organism. Certain aroma characteristics are associated with Department of Viticulture and Enology, University of California, Davis, 595 Hilgard Lane, Davis, CA 95616. *Corresponding author ([email protected]) Acknowledgments: Parts of this research were supported by a grant from the American Vineyard Foundation. Supplemental data is freely available with the online version of this article at www.ajevonline.org. Manuscript submitted Jul 2014, revised Jan 2015, Apr 2015, accepted Apr 2015 Copyright © 2015 by the American Society for Enology and Viticulture. All rights reserved. doi: 10.5344/ajev.2015.14073 1

379 Am. J. Enol. Vitic. 66:3 (2015)

380 – Joseph et al.

PCA analysis also showed that some Brettanomyces-contaminated wines scored lower in negative aroma characters than uninfected wine (Wirz 2005). This observation suggests that some strains of Brettanomyces could have a positive impact on wine aroma. In a subsequent study, the aroma impact of different specific precursor compounds in synthetic wines infected with different strains of Brettanomyces was assessed, again using a group of individuals with experience in sensory analysis. Synthetic wines were inoculated with Brettanomyces strains and supplemented with the hydroxycinnamic acids (caffeic, coumaric, ferulic acid) and/or the aromatic amino acids (phenylalanine, tryptophan, and tyrosine) to determine the major compounds produced by Brettanomyces in the presence of these precursors. The major metabolites produced by Brettanomyces when these specific precursors were present were identified and the aromas associated with those compounds were identified from the literature (Joseph et al. 2013). In the present study, the screening of a much larger group of strains of Brettanomyces was undertaken, allowing the same panelists as in the previous study to use their own descriptors. The main goal of this work was to identify strains producing aromas such as floral, fruity, and spicy that are often considered to impart positive qualities to wines. Cultures from those strains that most panelists characterized as having pleasant aromas were analyzed by headspace solid phase micro-extraction with gas chromatography and mass spectrometer detection (HS SPME GC-MS) to determine the chemical identity of the volatile components. The role of specific precursors in the formation of specific aromas by five Brettanomyces strains was characterized by chemical analysis. Synthetic wines with defined precursors were inoculated with Brettanomyces strains and evaluated for aroma characteristics by headspace solid phase micro-extraction with gas chromatography and mass spectrometer detection coupled with olfactory detection (HS SPME GCMS-O). The results reported here confirmed the chemical identity of the major aroma compounds produced by Brettanomyces and also identified a spectrum of non-identical descriptors used by panelists to describe the aroma of these individual compounds.

Materials and Methods Strains and culture conditions. Brettanomyces strains used in this study are listed in Supplemental Table 1. The nine strains indicated with a superscript “b” were used in the analysis of aroma by SPME GCMS and the five strains indicated with a superscript “c” were used for the detailed analysis of aroma using SPME GCMS-O. The screening of the Brettanomyces strains for the positive aroma characteristics, typically floral, fruity, and spicy, was performed as follows: strains were grown in 25 mL of defined medium (Conterno et al. 2006) or of defined medium supplemented with all of the three aromatic amino acids phenylalanine, tryptophan, and tyrosine at 100 mg/L each for two weeks at 30°C without agitation in quadruplicate. For an HS SPME GC-MS analysis of cultures with strains selected as having some positive aroma characteristics (as judged by a sensory

panel of human odor testers described below), each strain was grown in the defined medium, supplemented with either phenylalanine, tryptophan, and tyrosine at 100 mg/L each or with these three amino acids and p-coumaric, ferulic, and caffeic acids each at 50 mg/L, for two weeks at 30°C. The cultures were centrifuged to remove the yeast cells, and 10 mL of the supernatant was transferred to brown 20-mL SPME vials. To the vials was added 10 μL of a 1 mg/mL 2-nonanone internal standard in 50% ethanol to a final concentration of 1 μg/mL 2-nonanone. For analysis of specific aroma compounds by HS SPME GC-MS-O described below, Brettanomyces strains were grown for two weeks in the defined medium at 30°C (Conterno et al. 2006) with 10% ethanol. Cultures were supplemented with caffeic (18 mg/L), p-coumaric (16 mg/L), or ferulic (19 mg/L) acid (Amerine and Ough 1974). Two aromatic amino acids, phenylalanine or tyrosine at 100 mg/L, were also used to assess their effects on aroma production. Five genetically distinct strains of Brettanomyces (UCD VEN 615, 2058, 2077, 2082, and 2091) were used for this analysis to determine the impact of strain as well as of supplementation on aroma profiles in the media. The cultures were centrifuged to remove the yeast cells, and 10 mL of culture supernatant was transferred to brown 20-mL SPME vials and stored in the refrigerator overnight prior to analysis. Screening for Brettanomyces with positive aroma characteristics. To determine whether Brettanomyces strains exist that produce primarily positive aroma characteristics, 95 strains from the UC Davis Department of Viticulture and Enology Wine Yeast and Bacteria Collection were screened for aroma intensity, positive and negative aroma characteristics, and aroma descriptors. Negative aroma characteristics would include those described as rancid, rotten, chemical, or sweaty. A panel of five experienced individuals from the same group of aroma evaluators used in the HS SPME GC-MS-O analysis described below were asked to rate the aroma characteristics of the Brettanomyces strains grown in the defined medium supplemented with the three aromatic amino acids. Samples were assigned random numbers from a random number generator and presented to the panel in randomized sets of three to four strains, with each strain replicated four times. The panelists were asked to rate the aroma as positive, negative, or as a mixture of positive and negative and to describe the perceived aromas. They were also asked to rate the intensity of the aroma on a scale from one to nine. This screen is by its nature subjective, and the evaluators evaluated the samples separately to prevent the panelists from influencing each other. Mixed positive and negative aromas were considered as negative to bias the results more conservatively; only those evaluations that the individuals considered to be completely positive were counted as a positive response. The strains were evaluated on the basis of the percentage of times their aroma profiles were ranked as positive or negative and were given a final score according to the percentage of positive scores they received. For example, a strain that never received a positive aroma rating would have a 0% rating, and a strain that always received a positive aroma score

Am. J. Enol. Vitic. 66:3 (2015)

Brettanomyces Aroma – 381

would have a 100% rating. A strain rated as having a positive aroma in only 10% of the ratings and was otherwise given a negative or mixed score in the other ratings would have a final score of 10%. Volatile compound production in defined media. Nine of the 95 strains were selected for a study of volatile compound production. Volatile compounds were measured by HS SPME GC-MS, using an established protocol for compounds in model wine (Joseph et al. 2013). Headspace analysis was performed as described below with the following modifications. The entire sample was sent directly to the mass spectrometer detector without splitting to an olfactory port. The scan parameters for the MS were set between m/z 48 and 400. Compounds were identified with the NIST spectral library and via data mining for compounds previously identified as important Brettanomyces aroma compounds. Peak areas were normalized by dividing them by the peak area for the 2-nonanone internal standard (Joseph et al. 2013). The identification of these compounds was tentative and based on the retention times and mass spectra and was not confirmed with authentic standards. Assessment of volatile aroma composition. HS SPME GC-MS-O was run on each sample in triplicate for each of nine assessors. The analysis was performed by GC using an Agilent 6890K gas chromatograph (Avondale, PA) paired with an Agilent 5973 mass selective detector with an electron impact (70 eV) source and an MPS2 autosampler (Gerstel, Mülheim an der Ruhr, Germany). The GC injector was equipped with an SPME inlet liner, with an internal diameter (i.d.) of 0.75 mm (Supelco Inc., St. Louis, MO). The oven parameters were as follows: initial temperature 40°C and held for 5 min, then increased to 280°C at a rate of 8°C/min (total cycle time of 35 min). The injector was held at 260°C, and the temperature of the transfer line to the MS detector was 280°C. Helium was the carrier gas with a total flow of 26.3 mL/min, and constant pressure was maintained at 0.17 MPa throughout the run. An HP-5 MS column (30 m × 0.25 mm i.d., 0.25 μm film thickness) (J&W Scientific, Folsom, CA) was used for all analyses. A PDMS-DVB SPME fiber (65 μm thickness; Supelco, St. Louis, MO) was conditioned for 30 min in an injection port at 260°C before first use. The MS detector was operated in scan mode with a mass range of m/z 50 to 550. Samples were heated to 40°C for 30 min with the SPME fiber in the headspace to sorb volatiles. After the extraction, the fiber was inserted into the injection port for 10 min to desorb any volatiles from the fiber. The injection was splitless for the first 5 min, after which the split flow (20 mL/min) was turned on for the remainder of the GC run. As the sample came off the column, it was split into two smaller uncoated, deactivated fused silica capillary restrictor columns (MS column i.d. 0.1 mm ID, ODP 0.15 ID) with the sample being split between the MS detector and the olfactory port. Humidified air was added to the olfactory port to enhance odors for the evaluators. Enhanced MSD ChemStation software (G1701DA, version D.00.01.27; Hewlett-Packard, Santa Clara, CA) was used for instrument control and data analysis.

A nine-member panel used the olfactory port to evaluate the odors from a total of 75 samples of media in which the Brettanomyces strains had been grown. The evaluators were experienced in sensory analysis and also trained in the use of the equipment, but were not trained in the use of specific descriptors. This was a purposeful decision to prevent them from being influenced by other evaluators. They were instructed to hold down an indicator button whenever they detected an aroma and to continue to hold it as long as they sensed the aroma. Using electronic recording equipment, they were also asked to describe the aroma verbally and to also note the time of detection. The chromatograms were correlated with the audio-recorded descriptions and compared with spectra from the National Institute of Standards and Technology library (NIST). For the SPME GC-MS study described above, chromatographic peaks were identified by comparison with the NIST library spectra, and the identities of the peaks were confirmed by comparison of retention times and mass spectra with those of authentic standards and with the judges’ descriptions of aromas associated with the known standards by HS SPME GC-MS-O. To be considered a valid aroma compound, a specific peak had to be identified by at least two of the nine judges in each of the individual replicates. Chemical standards were purchased from Fisher Scientific (Waltham, MA) and from Sigma-Aldrich (St. Louis, MO). Data analysis. The SPME GC-MS data were analyzed using MetaboAnalyst online software (Xia et al. 2009). The data were filtered using standard deviations and scaled with Pareto scaling.

Results Evaluation of Brettanomyces strains for positive aroma character. In total, 95 Brettanomyces strains were surveyed to identify those that gave consistently positive aroma characters. A panel of five sensory evaluators was used to determine both the aroma descriptors and the positive and negative perception of aromas from the Brettanomyces strains grown in defined medium supplemented with the three aromatic amino acids, phenylalanine, tryptophan, and tyrosine (Figure 1). Eleven strains produced little or no aroma and were excluded from the analysis. None of the strains were rated as producing only positively perceived aromas by any of the five evaluators. The majority, i.e., approximately two-thirds, of the strains gave a negative or mixed aroma perception. Approximately one-third of the strains tested (i.e., 31) received all three possible aroma scores: positive, mixed, and negative. The most positively perceived strains were rated as positive in up to 40% of the evaluations. A single judge may have scored the strain mixed or negative or positive depending upon the day and time of the session. Different panelists may have also scored the same sample differently from one another based upon personal preferences or sensitivities. We therefore note that this sensory measure is very subjective, and to achieve a true preference value, many more panelists would be needed to evaluate each aroma sample. In this case, the evaluation attempted to screen only these Brettanomyces

Am. J. Enol. Vitic. 66:3 (2015)


strains for a basis for further HS SPME GC-MS characterization and later testing. Metabolite analysis of positively scoring Brettanomyces strains. After the aroma screening of all B. bruxellensis strains in the culture collection, nine strains were selected for chemical characterization (Table 1 and Supplementary Table 1, indicated by superscript “b” in the latter table). Five of the strains were chosen because they had higher positive aroma scores. Two strains had a lower percentage of positive scores but a similar level of negative and mixed scores, and all tended to have low aroma-intensity scores. One strain was selected because it had previously been identified as having positive aroma characteristics (strain 2091; Conterno, personal communication), although it had a positive score of zero in this study. Finally, one strain (2092) was used as a negative control because it had a high negative aroma score and the highest aroma-intensity score, 6.08, of the selected strains. The selected strains were used to compare production of volatile compounds when the defined medium was supplemented with the aromatic amino acids and hydroxycinnamic acids together or with the aromatic amino acids alone. The HS SPME GC-MS results were analyzed as described by Joseph et al. (2013) using MetaboAnalyst (Xia et al. 2009). An

ANOVA indicated a significant difference among 16 of 18 compounds produced by the different treatments at the 99% confidence level (data not shown). The partial least squares discriminant analysis (PLS-DA) of these strains grown with only the aromatic amino acids added (Figure 2) indicated that one strain, 2051, was particularly distinct from the rest of the strains when grown with aromatic amino acids. Strain 2051 had a low positive attribute score in the screen (8%). These other strains clustered closer together, and the replicates of these strains formed very tight groupings. Strain 2504, which had a positive score of 19%, overlapped significantly with several other strains that had lower scores. PLS-DA separated the control strain (2092), which was almost universally scored as negative under these culture conditions, from the other strains but not as distinctly as strain 2051 (Figure 2). The loadings into the PLS-DA indicated that many of the compounds identified clustered with the major clustering of the strains. The patterns were different for strain 2051, the strain that formed a separate cluster from the other strains. The differentiation of strain 2051 appeared to be influenced by production of lower levels of many of the compounds and of slightly higher levels of phenethyl alcohol and ethyl octanoate by this strain than by the others. A separation of strains 2402 and 605 was driven by higher levels of decanoic Figure 1 Evaluation of Brettanomyces strains for positive and negative characters. Each strain that was in the initial screen was evaluated as having negative, positive, or mixed negative and positive aroma characteristics in defined medium supplemented with the volatile amino acids phenylalanine, tryptophan, and tyrosine. Mixed negative and positive perceptions are shown in gray, negative evaluations are shown in black, and positive evaluations are shown in white.

Table 1 Brettanomyces strains used to analyze volatile compounds by SPME GC-MS. UCD strain number 2402 2408 2504 605 2404 2051 2509 2091 2092d

Aroma intensitya 4.50 4.19 3.00 2.89 3.10 4.75 4.44 4.93 6.08

Perceived aroma attributesb (%) Positive Negative Mixed 25 19 19 16 10 8 8 0 0

13 44 31 58 40 42 50 42 88

63 38 50 26 50 50 42 58 13

Strain designation in other collection

Source

Port. 1679 nac SA-3 FST 71-12 Port. 1700 na 35/02C CBS 2797 CBS 2796

Wine, Portugal Wine, 2000 Rhone, France Wine, South Africa Beer Wine, Portugal Wine, 2002, Pinot noir, California Wine, Uruguay Wine, 1957, Bordeaux, France Wine, 1957, sparkling Moselle, Germany

a

A panel of five individuals scored aroma intensity on a scale of 1 to 9, and the data is presented as the mean of that ranking. The percentage of times the five-member panel rated the strain as positive, negative, or mixed in defined medium with aromatic amino acids added. c na: not applicable. d Negatively perceived strain included as control. b

Am. J. Enol. Vitic. 66:3 (2015)


acid, decanoic acid ethyl ester, and octanoic acid. Strain 2091 appeared to have somewhat higher levels of phenethyl alcohol and ethyl hexanoate than the other strains. The negative control strain, 2092, was distinguished primarily for coming closest to the zero point for all measured compounds; i.e., it had few distinguishing compounds and represented an “average” composition among all the strains evaluated. A different picture emerged when the strains were grown with the phenolic acids as well as the amino acids. The PLSDA analysis of the strains illustrated these differences, and a heatmap of that data mapped with the chemical compounds is shown in Figure 3 with the strains grouping on the left or right sides of the map. Two distinct groupings were identified in the cluster analysis, with the negative control strain 2092 at the bottom of the group. This lower group also included strains 2051, 2504, and 2509, which had positive aroma scores of 8, 19, and 8%, respectively. The group at the top of the map included the strains 2402, 2091, 2404, 2408, and 605, which had positive scores of 25, 0, 10, 19, and 16%, respectively. Figure 3 also indicates the relative levels of the volatile compounds produced by each strain and the importance of those compounds in the groupings determined by the cluster analysis. The chemicals are listed at the bottom of the map, and their groupings are on the top. Those that were the most

important in determining the separation of the strains are shown on the far right and far left and include 4-EG and 4-EP as well as ethyl decanoate, phenethyl alcohol, and ethyl octanoate. The groupings reflected the levels of these compounds, with the strains in the bottom group having higher levels of the volatile compounds overall and the group at the top having lower levels of these compounds. Identification of aroma-active compounds in Brettanomyces. The aforementioned analyses identified strains that produced aroma compounds perceived as positive and also indicated the possible chemical components associated with these aromas. To define product–precursor relationships, we undertook HS SPME GC-MS-O of five of the Brettanomyces strains (indicated by superscript “c” in Supplementary Table 1). Each substrate tested was analyzed in triplicate for each of the five Brettanomyces strains. In total, 22 metabolites were identified as valid aroma-active compounds produced in the defined medium with the supplemented substrates (Table 2). These aroma-active compounds fell into several broad chemical classes, including alcohols, esters, fatty acids, aldehyde, terpenes, ethyl phenols, and vinyl phenols. Some of the compounds appeared to be primarily associated with certain substrates and strains. We could not quantify these compounds because many of them accumulated only to very low levels.

Figure 2 PLS-DA of the SPME GC-MS data from the nine Brettanomyces strains chosen for further analysis after the initial screen of the culture collection; the strains were grown in defined medium supplemented with aromatic amino acids. Strains 605 and 2051 are both shown in red, with strain 2051 showing the greatest separation from the other strains. The relative loadings of the compounds identified are shown by the pink squares as follows: 1: decanoic acid ethyl ester; 2: decanoic acid; 3: octanoic acid; 4: ethyl octanoate; 5: phenethyl alcohol; 6: ethyl hexanoate; 7: phenethyl acetate; 8: isoamyl alcohol; 9: ethyl acetate; and 10: isovaleric acid, phenylacetaldehyde, hexanoic acid, ethyl butyrate, ethyl isobutyrate, octanol, and decanol. Numbers after strain designations given in the plot indicate data from specific replications; the first number denotes the biological replicate and the second the analytical replicate.

Am. J. Enol. Vitic. 66:3 (2015)


Trace amounts, with a signal-to-noise ratio between 3 and 8, were detected and tentatively identified by their mass spectra. Authentic standards were used to determine the retention time and mass spectrum and to confirm the aroma descriptors from the panel. The Brettanomyces strains produced some compounds in the absence of the supplemented substrates. These compounds included isoamyl alcohol, phenethyl alcohol, and undecanoic acid. However, not all of the strains produced detectable amounts of these compounds without added aromatic amino acids or hydroxycinnamic acids (data not shown). Seven of

these compounds (Table 2) were previously identified by SPME analysis as major compounds produced by Brettanomyces (Joseph et al. 2013). The ethyl and vinyl phenols were detected only in medium supplemented with hydroxycinnamic acids. The ethyl phenols, 4-EP and 4-EG, were produced by all strains; however, 2-methoxy-4-vinylphenol was identified only in strain 2058 grown with ferulic acid. This compound is a precursor for 4-EG and may have been detected in cultures of this strain because it converted vinyl phenol to the ethyl phenol less efficiently than the other strains during the limited growth period.

Figure 3 Heatmap representation of the results of the PLS-DA of the nine strains chosen for SPME GC-MS analysis grown in defined medium supplemented with both aromatic amino acids and hydroxycinnamic acids. The analysis yielded two main clusters with four strains in the bottom group and five strains in the top group (shown on the right). The volatile compounds divided into three groups (shown at the bottom), with the ethylphenols composing one group; decanoic acid, ethyl octanoate, and 2-phenyl ethanol making up the second group; and all other compounds composing the third group. Numbers after strain designations indicate data from specific replications; the first number denotes the biological replicate and the second the analytical replicate. Am. J. Enol. Vitic. 66:3 (2015)


The fatty acids butyric and heptanoic acid, whose aromas are described as rotten and rancid, were primarily found in cultures supplemented with the amino acids phenylalanine or tyrosine. The esters, which typically have fruity and floral aroma descriptors, were associated with both the aromatic amino acids and the hydroxycinnamic acids, as were the alcohols. Two terpenes, ocimene and bisabolene, were associated with samples from cultures supplemented with phenylalanine or caffeic acid, respectively. Each of the five strains tested produced only one of the terpenes. The only aldehyde identified, nonanal, was found only in strain 2091 when grown in the presence of tyrosine.

Discussion Previous studies have identified the metabolites primarily responsible for what is termed the Brettanomyces character in wine (Chatonnet et al. 1995, Licker et al. 1999). These compounds are the fatty acids and ethyl phenols, whose aroma profiles are usually described as sweaty, rancid, barnyard, or medicinal and which are therefore typically negatively perceived by many consumers. However, some aromatic complexity attributed to wines contaminated with Brettanomyces is often perceived as an enhancement to the character of the wines. A recent study also noted the importance of other compounds in producing the Brettanomyces character (Romano et al. 2009), and other work has highlighted the importance of minor odor components in the overall perception of odors

(Ryan et al. 2008). Our study has identified several minor products that are produced by Brettanomyces in a defined medium supplemented with hydroxycinnamic and amino acids that may also contribute to the complexity of wine aroma. These compounds included the ethyl and vinyl phenols, the fatty acids and their esters, alcohols, terpenes, and an aldehyde (Table 2). Esters, aldehydes, and alcohols often result in fruity and floral characteristics that may enhance the aroma of wine, but they may also impart a chemical or artificial character. Terpenes are usually associated with plant-derived aromatic characteristics and are common constituents of aromatic oils and incense. They often have strong floral and resinous aroma characteristics. Seven compounds that have been identified previously as being major products of Brettanomyces in synthetic medium (Joseph et al. 2013) were also detected in this study. These compounds included the ethyl phenols (4-EG and 4-EP), alcohols (isoamyl alcohol, 2-methyl-1-butanol, 2-ethylhexanol, and phenethyl alcohol), and an ester (ethyl 2-methyl butyrate). The remainder of the compounds were observed as minor products but consistently contributed to an odor that was discernible to the panel of aroma evaluators when assessed by HS SPME GC-MS-O. Many of these minor metabolites were esters of compounds present at higher levels in cultures, such as ethyl valerate and ethyl isovalerate. Some were related in other ways, such as the C11 undecanoic acid, which is similar to the C10 decanoic acid and also to the

Table 2 Aroma-active compounds identified by SPME GC-MS olfactory analysis in synthetic media; underlined aroma descriptors are generally considered negative characters. Compounda Isoamyl alcohol 2-Methyl-1-butanola Ethyl butyrate Pentyl formate Butyric acid Ethyl isovalerate Ethyl valerate Ethyl-2-methyla butyrate Nonanal 2-Ethyl-1-hexanola Ocimene Phenethyl alcohol 4-Ethylphenola Phenethyl formate Heptanoic acid 4-Ethylguiacola 2-Methoxy-4-vinylphenol 2-Methoxyphenethyl alcohol Octyl butyrate Amyl-octanoate Undecanoic acid Bisabolene

CAS # 125-51-3 137-32-6 105-54-4 638-49-3 107-92-6 108-64-5 539-82-2 7452-79-1 124-19-6 104-76-7 502-99-8 60-12-8 123-07-9 104-62-1 11-14-8 2785-89-9 7786-61-0 7417-18-7 110-39-4 638-25-5 112-37-8 495-62-5

Chemical type

Substrateb

Strain

Panel aroma descriptors

Alcohol None, caf, fer, tyr All Chemical, sweet, fruit, banana Alcohol Fer, phe, tyr 2058, 2091 Oxidized/canned fruit, plastic, sulfur Ester Fer 2058, 2091 Artificial fruit, bubblegum, pineapple Ester Phe 2077, 2058 Artificial fruit, candy, chemical Fatty acid Phe, tyr 615, 2082 Rotten, rancid, sewage, sulfur, sweaty Ester Phe, tyr All Fruity, pineapple, citrus, artificial grape Ester Tyr 2058, 2091 Artificial fruit, bubblegum, chemical Ester Caf, fer, tyr, phe 615, 2058, 2082, 2091 Minty, menthol, citrus, green apple Aldehyde Tyr 2091 Lemon furniture polish, air freshener, oily Alcohol Caf, fer, phe, tyr All Fake floral, chemical, fusel oil Terpene Phe 615, 2077, 2082 Sweet floral, hyacinth, jasmine Alcohol None, all All Rose, strong floral, pollen Ethyl phenol Cou All Band-Aid, medicinal, plastic, solvent Ester Phe 2058, 2091 Artificial floral, perfume, wild flower, solvent Fatty acid Phe, tyr 615, 2082 Sweaty, solvent, rotten, barnyard Ethyl phenol Caf, fer All Clove, spicy, smoky, burned rubber Vinyl phenol Fer 2082 Leather, burned, woody, hot electric Alcohol Fer 2058 White glue, plastic, mimeograph sheet Ester Phe 2091 Spicy, eucalyptus, floral, plastic Ester Caf, tyr 2058, 2091 Spicy, clove, chemical, plastic Fatty acid None, tyr 615, 2091 Coconut, balsam, oily, vanilla Terpene Caf 2058, 2091 Spicy, tropical, toasty, wood resin, minty

a

Compounds identified as major volatile products by Joseph et al. (2013); non-IUPAC (International Union of Pure and Applied Chemistry) chemical names and their IUPAC equivalents are presented in Supplemental Table 2. b caf: caffeic acid; cou: p-coumaric acid; fer: ferulic acid; phe: phenylalanine; tyr: tyrosine; none: no substrate added; all: all five substrates when added individually. Am. J. Enol. Vitic. 66:3 (2015)


methoxy-vinyl phenol that is a precursor to 4-EG. The terpenes identified here, ocimine and bisabolene, are related to b-farnesene, which was a major compound found previously (Joseph et al. 2013). The types of compounds identified here indicate that Brettanomyces strains can produce a wide variety of odoractive compounds under different conditions and suggest that many factors influence the formation of a “good” or “bad” odor from this yeast. Our study was undertaken to determine whether any of the strains yielded consistently positive aromas under controlled culture conditions. The initial screen indicated that some of the strains were perceived as having positive characteristics, but none of the strains universally yielded only positive aromas as judged by the sensory panel. When the defined medium was supplemented with the three aromatic amino acids phenylalanine, tryptophan, and tyrosine, one of the most important factors that determined the separation of the strains was the absence in some of the strains of some of the most common volatile compounds produced. Thus, absence of a strongly negative aroma character rather than the presence of positive characters was associated with a positive score for a given Brettanomyces strain. Because the initial screening scores were done without hydroxycinnamic acids in the medium, it was expected that the levels of the compounds directly related to the presence of the hydroxycinnamic acids would not correlate with the rating. The vinyl and ethyl phenols should have no correlation to the percent positive rating of the strain. However, ethyl phenols accumulated to the highest levels in the negative control, 2092, and in strain 2051 (8% positive). The vinyl phenols were highest in strains that were rated more positively: 4-vinylphenol in 2402 (25% positive) and 2-methoxy4-vinylphenol in 605 (16% positive). This may indicate that the strains with the highest positive scores were strains that metabolized slowly and did not complete the conversion of many of the precursor compounds into the final end products. This supports the idea that the absence of certain aromatic compounds is as important as the presence of others in overall perception of Brettanomyces-associated odor. Strain 2402 also showed relatively high levels of octanol that could lead to a more positive perception of the wine aroma. Upon a detailed analysis of the chemical products produced by the strains when both the aromatic amino acids and the hydroxycinnamic acids caffeic, coumaric, and ferulic were added to the medium, it appears that the strains perceived as positive or with a mixed aroma profile most often had low levels of the volatile metabolites and grouped at the top of Figure 3. This was true for strains 605, 2402, 2404, and 2408, but also for 2091. Strain 2091 was not perceived to have positive aroma characteristics in the initial screening of strains with only aromatic amino acids added but was included because it has been suggested in previous studies to be a strain with positive characteristics (Conterno personal communication). Only strain 2504 (19% positive) did not fit the same pattern as the other positively ranked strains and grouped instead with the strains at the bottom of Figure 3, along with the negative control strain 2092, despite having a

relatively high positive aroma score. Strain 2504 had a score similar to that of strain 2509 (8% positive), and these strains clustered next to each other in the PLS-DA analysis. In some cases, aroma compounds perceived as pleasant may be responsible for the positive aroma profile of a strain (Tominaga et al. 1998). Another possibility is that low levels of the negatively perceived compounds, below the level of perception by some individuals, may also contribute to the positive perception of some Brettanomyces strains. For example, when the levels of vinyl phenols are higher, the levels of ethyl phenols will be correspondingly lower and may result in increased positive odor scores. This could mean that the ethyl phenols are below the level of detection or that at low levels they are not perceived as being negative, or that the vinyl phenols are positively affecting the aroma, or some combination of these factors. Moreover, very high levels of aroma compounds may be overwhelming and may therefore be perceived as negative, even if these aromas are generally considered positive (Goode 2005).

Conclusion

Addition of specific aroma precursors to a defined medium helped characterize the aroma profiles of a set of Brettanomyces strains. SPME GC-MS-O identified compounds with aroma affects. These compounds are associated with the substrates available during the growth of the yeast, and while many of them are not the major compounds produced during metabolism, they are often related to those major end products. The variety of aroma-associated compounds produced from just the five substrates with five strains indicated the type of metabolic complexity Brettanomyces is capable of producing in a diverse precursor matrix. Nine strains were screened for positive characters, defined as being pleasant in aroma to the panelists. The negative control strain had the highest aroma intensity of the strains evaluated, and it received the lowest perceived positive character score. However, aroma intensity did not closely correlate with the character rating among the other strains. The most positively rated strain had an aroma-intensity rating of 4.50, while some strains with lower character scores had lower intensity ratings. However, another strain with a lower positive score had the next-highest intensity rating at 4.93. This study suggests that the absence of negative characters rather than the aroma intensity of positive characters is most closely associated with positive sensory perception of Brettanomyces strains. No single strain produced only positively perceived aroma compounds, and aroma compound production varied with added substrate.

Literature Cited Amerine, M., and C.S. Ough. 1974. Wine and Must Analysis. John Wiley & Sons, New York, London, Sydney, Toronto. Arvik, T., and T. Henick-Kling. 2002. Overview: Brettanomyces bruxellensis occurrence, growth, and effect on wine flavor. 31st Annual New York Wine Industry Workshop. Pract. Winery Vineyard. May/June:117-123. Chatonnet, P., D. Dubourdieu, and J.N. Boidron. 1995. The inf luence of Brettanomyces/Dekkera sp. yeast and lactic acid bacteria on the ethylphenol content of red wines. Am. J. Enol. Vitic. 46:463-468.

Am. J. Enol. Vitic. 66:3 (2015)

Brettanomyces Aroma – 387 Conterno, L., C.M.L. Joseph, T.J. Arvik, T. Henick-Kling, and L.F. Bisson. 2006. Genetic and physiological characterization of Brettanomyces bruxellensis strains isolated from wines. Am. J. Enol. Vitic. 57:139-147. Curtin, C.D., A.R. Borneman, P.J. Chambers, and I.S. Pretorius. 2012. De-novo assembly and analysis of the heterozygous triploid genome of the wine spoilage yeast Dekkera bruxellensis AWRI1499. PLoS One 7:1-10. Fugelsang, K.C., and B.W. Zoecklein. 2003. Population dynamics and effects of Brettanomyces bruxellensis strains on Pinot noir (Vitis vinifera L.) wines. Am. J. Enol. Vitic. 54:294-300. Goode, J. 2005. The Science of Wine: From Vine to Glass. University of California Press, Oakland. Hasin-Br umshtein, Y., D. Lancet, and T. Olender. 2009. Human olfaction: From genomic variation to phenotypic diversity. Trends Genet. 25:178-184. Hellborg, L., and J. Piskur. 2009. Complex nature of the genome in a wine-spoilage yeast, Dekkera bruxellensis. Eukaryot. Cell. 8:1739-1749. Heresztyn, T. 1986a. Formation of substituted tetrahydropyridines by species of Brettanomyces and Lactobacillus isolated from mousy wines. Am. J. Enol. Vitic. 37:127-132. Heresztyn, T. 1986b. Metabolism of volatile phenolic compounds from hydroxycinnamic acids by Brettanomyces yeasts. Arch. Microbiol. 146:96-98. Joseph, C.M.L., L.W. Gorton, S.E. Ebeler, and L.F. Bisson. 2013. Production of volatile compounds by wine strains of Brettanomyces bruxellensis grown in the presence of different precursor substrates. Am. J. Enol. Vitic. 64:231-240. Licker, J.L., T.E. Acree, and T. Henick-Kling. 1999. What is “Brett” (Brettanomyces) f lavor? A preliminary investigation. In Chemistry of Wine Flavor. ACS Symposium Series 714. A.L. Waterhouse and S.E. Ebeler (eds.), pp. 96-115. American Chemical Society, Washington, DC.

Loureiro, V., and M. Malfeito-Ferreira. 2003. Spoilage yeasts in the wine industry. Int. J. Food Microbiol. 86:23-50. Oelofse, A., I.S. Pretorius, and M. du Toit. 2008. Significance of Brettanomyces and Dekkera during winemaking: A synoptic review. S. Afr. J. Enol. Vitic. 29:128-144. Pangborne, R.M., J.X. Guinard, and R.G. Davis. 1988. Regional aroma preferences. Food Qual. Prefer. 1:11-19. Parr, W.V., K.G. White, and D.A. Heatherbell. 2004. Exploring the nature of wine expertise: What underlies wine experts’ olfactory recognition memory advantage? Food Qual. Prefer. 15:411-420. Piskur, J., et al. 2012. The genome of the wine yeast Dekkera bruxellensis provides a tool to explore its food-related properties. Int. J. Food Microbiol. 157:202-209. Romano, A., M.C. Perello, A. Lonvaud-Funel, G. Sicard, and G. de Revel. 2009. Sensory and analytical re-evaluation of “Brett Character.” Food Chem. 114:15-19. Ryan, D., P.D. Prenzler, A.J. Saliba, and G.R. Scollary. 2008. The significance of low impact odorants in global odour perception. Trends Food Sci. Tech. 19:383-389. Stevenson, R.J., and R.A. Boakes. 2003. A mnemonic theory of odor perception. Psychol. Rev. 110:340-364. Tominaga, T., C. Peyrot des Gachons, and D. Dubourdieu. 1998. A new type of f lavour precursor in Vitis vinifera L. cv. Sauvignon blanc: S-Cysteine conjugates. J. Agric. Food Chem. 46:5215-5219. Wirz, D.O. 2005. Descriptive Analysis of Brettanomyces-Infected Cabernet Sauvignon Wines. Thesis, University of California, Davis, Davis, CA. Xia, J., N. Psychogios, N. Young, and D.S. Wishart. 2009. MetaboAnalyst: a web server for metabolomic data analysis and interpretation. Nucleic Acids Res. 37:W652-W660 (Web Server issue).

Am. J. Enol. Vitic. 66:3 (2015)