Release of wine monoterpenes from natural precursors by glycosidases from Oenococcus oeni

Highlights

► Glycosidases of Oenococcus oeni are able to release terpenes from natural substrates. ► Bacterial glycosidases have a high capacity to release both primary and tertiary terpene alcohols. ► Both glucosidase and arabinosidase from O. oeni are more active in grape juice than glycosidases from Aspergillus niger. ► Riesling mash was treated with O. oeni glycosidases before alcoholic fermentation. ► Wines produced with the glycosidases from O. oeni received positive ratings by a professional panel.

Abstract

It is now well established that wine-related lactic acid bacteria (LAB), especially Oenococcus oeni, possess glycosidase activities that positively contribute to wine aroma through the hydrolysis of grape-derived aroma precursors. In our recent studies, we have identified and characterised several LAB glycosidases with potential in these terms. Here, we report that both a glucosidase and an arabinosidase from O. oeni can release high amounts of monoterpenes from natural substrates under optimal conditions, indicating that these intracellular enzymes might play a significant role in the hydrolysis of aroma precursors during malolactic fermentation. The enzymes from O. oeni exhibited broad substrate specificities (release of both primary/tertiary terpene alcohols) and were even active in grape juice. Further, a sensory panel clearly preferred enzyme-treated Riesling wines over the controls and affirmed that the glycosidases from O. oeni could improve the typical Riesling aroma.

1. Introduction

Wines are highly complex beverages, various combinations of flavour components, such as acids, sugars, phenols and volatile aroma compounds generate a multitude of sensorial variations (Jackson, 2008). Although over 800 wine aroma compounds have been identified, only a limited number thereof makes a significant contribution to the wine aroma (Rapp & Mandery, 1986). Volatile constituents of the primary grape aroma, especially monoterpenes that are formed in the grapes during ripening, are the key components of the varietal wine bouquet. As demonstrated by Rapp (1992, 1998), GC fingerprint analysis of only a selected number of wine terpenes can be used to distinguish between grape varieties and even to determine the region of origin.

In addition, the sensory properties of wines are influenced by the microorganisms (yeasts, malolactic bacteria, molds) involved in the winemaking process and by the choice of vinification techniques (types of fermentors, aging style, blending, bottle closure, etc.) (Jackson, 2008). Market and consumer preferences exhibit a considerable influence on the style of wines produced as well, which not only affects the choice of grape varieties planted but also the applied viticultural and enological practices (Bruwer, Saliba, & Miller, 2011).

The recent developments in winemaking and marketing practices show that wine aroma composition has gained increasing importance in recent years (Bruwer et al., 2011). An important aspect in wine aroma tailoring is the fact that a significant fraction of the aroma compounds present in grapes and wine occurs as non-volatile odourless glycosides (Gunata, Bayonove, Baumes, & Cordonnier, 1985); these are mainly found in the grape juice rather than in the skin and pulp (Strauss, Wilson, Gooley, & Williams, 1986). The precursors of important monoterpenes (e.g., linalool, geraniol, nerol, β-citronellol and α-terpineol), C₁₃-norisoprenoids, benzene derivatives and phenols are synthesised during the early development of the grape berry. These precursors have been identified as monoglucosides and diglycosides; in the latter group, glucose can further be conjugated to apiose, arabinose, rhamnose or xylose (Gunata, Bitteur, Brillouet, Bayonove, & Cordonnier, 1988; Williams, 1993). With the aim of improving the characteristic varietal wine aroma, many authors have investigated the possibilities of sequential enzymatic hydrolysis of these aroma precursors by glycosidases (glucosidase, arabinosidase, rhamnosidase, apiosidase) (Maicas & Mateo, 2005; Palmeri & Spagna, 2007). It has been shown that fungal glycosidases that are often present as side activities in pectolytic enzyme preparations are suited for such a purpose (Maicas & Mateo, 2005). On the other hand, detailed studies have been committed to the impact of wine microorganisms, especially yeasts, on wine aroma (Antonelli, Castellari, Zambonelli, & Carnacini, 1999; Kotseridis & Baumes, 2000). Other authors have focused on the aroma change caused by lactic acid bacteria (LAB) involved in malolactic fermentation (MLF) (Boido, Lloret, Medina, Carrau, & Dellacassa, 2002; D’Incecco et al., 2004; Ugliano, Genovese, & Moio, 2003). Grimaldi, Bartowsky, and Jiranek (2005a, 2005b) presented a comprehensive survey demonstrating that wine-related LAB (Oenococcus oeni, Lactobacillus spp. and Pediococcus spp.) possess the ability to hydrolyse various synthetic glycosides. Furthermore, it has been shown that high variations in glycosidase activities exist among isolates of O. oeni (Gagné et al., 2011; Ugliano & Moio, 2006). These studies indicated that wine LAB, in particular O. oeni, are indeed capable of releasing attractive aroma compounds during MLF and that LAB might be a promising source of novel glycosidases with oenological potential (Matthews et al., 2004). Therefore, it is of interest to identify the mechanisms that enable LAB, especially O. oeni, to release glycosylated aroma compounds. In our previous work, we were able to identify a glucosidase and an arabinosidase from O. oeni (Michlmayr, Schümann, Kulbe, & del Hierro, 2011; Michlmayr, Schümann, Barreira Braz Da Silva, Kulbe, & del Hierro, 2010). In the present study, we continued our research to determine if these glycosidases are capable of releasing monoterpenes from natural glycosidic precursors. Therefore, samples of Austrian wine and grape juice were prepared to perform assays with the aim of evaluating these enzymes’ performance on different natural substrates under varying conditions (pH, sugar content) and in comparison to fungal glycosidases. Additionally, the results of applying both O. oeni glycosidases at an early stage (cold maceration) in the production of a typical Austrian white wine variety (Rheinriesling) are presented.

2. Materials and methods

2.1. Enzymes used in this study

A list of all enzyme preparations used in this study is provided in Table 1. The physicochemical and kinetic properties of the bacterial glycosidases involved have been reported before (references in Table 1). The fungal enzyme preparations are commercial products. The abbreviations (letter codes) as displayed in Table 1 are used throughout the paper, especially in the results section.

Table 1.

Enzyme preparations included in this study.

Enzyme	GHa	Source organism	Code	Reference/supplier
β-d-glucosidase	3	Oenococcus oeni ATCC BAA-1163	GO	Michlmayr, Schümann, Wurbs et al. (2010)
β-d-glucosidase	3	Lactobacillus brevis SK3	GL	Michlmayr, Schümann et al. (2010), Michlmayr, Schümann, Wurbs et al. (2010)
β-d-glucosidase	3	Aspergillus niger	GA	Megazyme, Wicklow, Ireland
α-l-arabinofuranosidase	51	Oenococcus oeni ATCC BAA-1163	AO	Michlmayr, Schümann et al. (2011)
α-l-arabinofuranosidase	51	Aspergillus niger	AA	Megazyme, Wicklow, Ireland
α-l-rhamnopyranosidase	78	Pediococcus acidilactici DSM 20284	R	Michlmayr, Brandes et al. (2011)
Naringinaseb	–c	Penicillium decumbens	N	Sigma–Aldrich, Vienna, Austria
Maceration Cd	–	Aspergillus niger	MacC	Preziso (Lagerhaus), Austria
Trenolin Super DFe	–	(not specified)	–	Erbslöh, Geisenheim, Germany

^aGlycoside hydrolase family (Cantarel et al., 2009; Henrissat & Davies, 1997).

^bCommercial rhamnosidase preparation, side activities are shown in Fig. 1.

^cNot specified by the supplier.

^dMaceration preparation with side activities of β-d-glucosidase (3.5 U/g), β-d-xylosidase (1.3 U/g) and α-l-arabinofuranosidase (5.3 U/g).

^ePectinase preparation with side activities of β-d-xylosidase (0.2 U/mL) and α-l-arabinofuranosidase (3.9 U/mL).

All bacterial glycosidases (GO, GL, AO, R) were heterologously expressed and purified as previously described (Michlmayr, Brandes et al., 2011; Michlmayr, Schümann et al., 2011; Michlmayr, Schümann, Wurbs et al., 2010). The resulting enzyme fractions were further purified by ion exchange chromatography (Source Q for GL, AA and Source S for GO, R; both from GE Healthcare, Uppsala, Sweden) following the suppliers’ recommendations. The resulting enzyme fractions were dialysed over night against 20 mM citrate phosphate buffer, pH 7 (McIlvaine, 1921), at 4 °C and stored in this buffer. If required, the enzyme solutions were concentrated, using Amicon Ultra centrifugal filters (MWCO 10 kDa) (Millipore, Billerica, MA). All enzyme preparations were stored at 4 °C.

Glycosidase activities were determined with synthetic p-nitrophenyl (pNP) glycosides (all from Sigma–Aldrich, Vienna, Austria). The substrates used were pNP-β-d-glucopyranoside, pNP-β-d-galactopyranoside, pNP-β-d-xylopyranoside, pNP-α-l-arabinofuranoside, pNP-α-l-arabinopyranoside and pNP-α-l-rhamnopyranoside. The synthetic glycosides were dissolved in 10% (v/v) dimethyl sulfoxide. Unless mentioned otherwise, the conditions for all enzyme assays were: 10 mM substrate in 0.1 M McIlvaine buffer, pH 5.5, 37 °C, 10 min incubation time. The reactions were stopped with 0.5 M Na₂CO₃ (2-fold volumetric excess). The absorbance of p-nitrophenol was measured at 400 nm (ε₄₀₀ = 18.300 M⁻¹ cm⁻¹ at pH 10.2) in a Beckman DU 800 spectrometer (Palo Alto, CA).

One unit of glycosidase activity is expressed as 1 μmol of p-nitrophenol released per min at 37 °C.

2.2. Enzyme assays (terpene release) with commercial samples of wine and grape juice

Samples of wine and grape juice were prepared, to obtain controlled conditions for enzyme assays. White wine (Traminer 2010, Klöch, Austria) was evaporated (Heidolph Rotovapor; Heidolph, Schwabach, Germany) at 35 °C to remove ethanol and the background of volatile compounds; the pH was adjusted to 5.5 with KHCO₃. After readjustment to the original volume, the wine extract was sterilised by filtration (0.22 μm filter, Millipore). Two brands of commercial red grape juice (“St. Laurent”, Stift Klosterneuburg, Austria and “Happy Day”, Rauch, Rankweil, Austria) were sterilised by filtration as described above; if required, the pH was adjusted to 5.5 with KHCO₃ before filtration. The results of an analysis of the ingredients of wine extract and grape juices are shown in Table 2.

Table 2.

Composition of grape juice and wine extract used in enzyme assays. Analyses were performed using validated standard methods in an accredited test laboratory (ISO 17025).

	Grape juice “St Laurent”	Grape juice “Happy Day”	Wine extract (Traminer)
Fructosea (g/L)	85.2	75.3	1.7
Glucosea (g/L)	85.5	74.6	nd
Total sugarb (g/L)	189	162	–
Titratable acidc (g/L)	8.2	6.3	1.6
Tartrateb (g/L)	4.6	5.2	1.0
Malateb (g/L)	5.0	3.0	3.0
Lactateb (g/L)	nd	nd	nd
SO2d (mg/L)	nd	nd	nd
K+e (mg/L)	1286	1591	>2000
Na+e (mg/L)	31	23	22
Mg2+e (mg/L)	111	76	83
Ca2+e (mg/L)	138	121	68
pH	3.3	3.0	5.5

nd, Not detectable.

^aEnzymatic determination (test kits from Roche diagnostics).

^bFourier transform infrared spectroscopy (Foss WineScan FT 120).

^cExpressed as tartrate.

^dDetermined per titration (NaOH).

^eAtomic absorption spectroscopy.

All enzyme assays (terpene release) were conducted using 10 mL of sample (triplicate determinations). The samples were treated with the enzyme preparations in excess (2 U/mL as determined with pNP-glycosides (Section 2.1) in different combinations. The arabinosidases (AO, AA) and a rhamnosidase (R) were each applied in combination with the glucosidase of O. oeni (GO). Naringinase (N) was applied alone or in combination with GO. All assays were performed under sterile conditions, the enzyme preparations were sterilised (0.22-μm filter) before application. The samples were incubated for 7 days at 15 °C. After the incubation period, the samples were frozen (−30 °C) until terpene analysis (Section 2.4) of the volatile fraction was performed.

2.3. Vinification experiments with glycosidases of O. oeni

Five hundred kilograms of Rheinriesling grapes, an aromatic white wine variety widely cultivated in Austria, were harvested (2010 vintage) at the vineyards of the College for Oenology and Viticulture in Klosterneuburg, Austria. After cleaning, destemming and sorting, the grapes were crushed (roller crusher QU75, Benczak GmbH & Co. KG, Siegendorf, Austria). During crushing, 125 mg/kg of dimethyl dicarbonate (DMDC) (Velcorin®, Lanxess GmbH, Leverkusen, Germany) were added to inhibit wild yeasts and lactic acid bacteria. The free run juice of the resulting mash had a pH of 2.9, a total acidity of 13.1 g/L and 163 g/L of reducing sugars. SO₂ (50 mg/kg as potassium metabisulfite; PMS) was added to the mash and the pH was adjusted to pH 4.0 using 480 g CaCO₃ and 275 g of KHCO₃. The mash was thoroughly mixed and kept at 8 °C for 24 h to give time for the DMDC to react. Subsequently, the mash was divided into pre-cleaned 45 L tanks and treated with enzyme preparations as follows: GO: 300, 200, 60 U/L; AO: 35 U/L; GO + AO: 150 + 25 U/L; Maceration C (Preziso, Austria) 3 g/hL; two tanks were kept without enzyme as controls. After thorough mixing, a further 20 mg of SO₂ (PMS) were added to each tank on the top of the mash. The tanks were tightly sealed and kept at 12 °C for 4 days. Before pressing, the mash of the recombinant enzyme treatments and one of the controls (C2) were supplemented with 8 mL/hL Pectinase (Trenolin Super DF, Erbslöh, Geisenheim) to facilitate must extraction (following the producers’ recommendations 2 h before pressing). After pressing, the musts were chaptalised to a must weight of 18 °KMW (“Klosterneuburger Mostwaage”, corresponding to 214 g/L total sugar or 12.8% (v/v) potential alcohol; Ribéreau-Gayon, Glories, Maujean, & Dubourdieu, 2006) with sucrose. Demijohn carboys (35 L each) were washed with 2% NaOH, 2% citric acid and rinsed with tap water. The musts were brought to 20 °C room temperature for the start of fermentation; 5 g of fermentation nutrient (Fermaid E, Lallemand, Vienna, Austria) were added to each demijohn. SO₂ was adjusted to 50 ppm free in each balloon to prevent wild yeast fermentation. The musts were fermented using Oenoferm Freddo yeast (Erbslöh Geisenheim, Germany) at the recommended rate for low temperature fermentation (15 g/hL). After the fermentation started and more than 1% (v/v) alcohol built up, the demijohns were cooled down to 12 °C. When the wines reached 8% (v/v) alcohol, a further 2 g of fermentation nutrient (Fermaid E) were added per demijohn and the fermentations were completed at 20 °C. The wines were then cooled to 4 °C and cross-flow filtered using a Lab4-102 (Romfil GmbH, Wolfsheim, Germany) filtration module of 0.2 μm at 1 bar. After filtration, 50 ppm SO₂ as PMS was added. All wines presented between 9.6 and 10.0 g/L total acidity and 6.8–7.0 g/L malic acid. Deacidification of the wines to 7 g/L total acidity was carried out by double salt deacidification, following the method proposed by Steidl (2001). After deacidification, all wines were adjusted to 45 mg/L free SO₂, microfiltered over a Cuno 3 M Zeta Plus H cartridge 80H05 (0.5 μm diameter pore cut-off), bottled in 375-mL bottles and stored at 10 °C.

2.4. Analysis of terpene compounds

Volatile compounds were analysed by gas chromatography–mass spectrometry (GC/MS). The analytical procedure is based on the method described by Skinkis, Bordelon, and Wood (2008). A 7890A GC system (Agilent technologies, Paolo Alto, CA) with a DB-5 capillary column (60 m × 0.25 mm, 0.25 μm; stationary phase 5% dimethyl polysiloxane, 95% phenyl polysiloxane), a CombiPal autosampler (CTC analytics, Zwingen, Switzerland) and a 5975C MS detector (Agilent) were used. The samples were prepared by solid-phase micro-extraction (SPME). Five millilitres of sample and 50 μL of the internal standard (4-chlorobutyl acetate) were added to a vial containing 2 g NaCl. SPME fibres (100 μm polydimethylsiloxane) from Supelco (Bellefonte, PA) were used as absorbant. Extraction was performed for 30 min at 50 °C, followed by desorption for 5 min at 250 °C. The samples were injected in splitless mode (3 min), the carrier gas was helium (99.999%; Air Liquide, Vienna, Austria) with a flow of 1.2 mL/min. The program for the oven temperature was as follows: initial temperature 50 °C for 3 min, temperature increase to 92 °C (1 °C/min), holding time 10 min; further increase to 127 °C (5 °C/min), then increase to 260 °C (40 °C/min), holding time 5 min. The transfer line temperature was 260 °C. Ionisation was performed at 70 eV. Ions were quantified by selected ion monitoring (SIM) relative to the internal standard 4-chlorobutyl acetate (m/z = 54). The following standards were applied (the m/z ratios used for quantification are shown in parentheses): cis/trans-linalool oxide (59), linalool (71), hotrienol (71), cis/trans-rose oxide (139), cis-limonene oxide (67), trans-limonene oxide (94), α-terpineol (93), β-terpineol (71) γ-terpineol (121), nerol (69), β-citronellol (69), geraniol (69), nerol oxide (68) and lavandulol (69). Rose oxide was obtained from Moellhausen (Vimercate, Italy), hotrienol from Treatt (Lakeland, Florida), nerol oxide from Roth (Karlsruhe, Germany). All other standards were obtained from Fluka (Sigma–Aldrich, Vienna, Austria).

The limit of quantification (LOQ) was determined as 0.3 μg/L, the relative standard deviation between repeated samples (repeatability) was below 6%.

2.5. Sensory analysis

The wines were evaluated in triplicate by a panel of seven trained tasters. The participants are officially approved tasters for the quality assessment of Austrian wines. The tasters are trained according to the Austrian wine law and their performance is evaluated annually. Assessment took place in a standard sensory analysis chamber (EN ISO 8589) equipped with separate booths under yellow light forcing the tasters to focus only on the aroma and taste of the wines. The tasters were presented with the wines in groups of five wine glasses each. The wines were presented in randomised order in coded standard tasting glasses (ISO 3591). In each group, the tasters were first asked to rank the wines on their aroma intensity using an unstructured scale ranging from 1 to 5 with 1 the highest and 5 the lowest aroma intensity. The wines where then sorted according to the perceived aroma intensity, following the method of Cartier et al. (2006).

Second, the tasters had to assess each wine based on their olfactory and taste sensations on an unstructured scale from 0 to 10, with 10 reflecting the highest intensity of each attribute. The attributes involved were typical Riesling descriptors (stone fruit, citric, pomaceous fruit), attributes usually associated with white wines but not with Riesling (freshness, spice, tropical, candy). Further attributes were “floral” (corresponding to terpenoid aroma compounds) and “typicality” (Riesling).

2.6. Statistical analysis

The data from terpene analysis (Section 2.4) were statistically analysed with the software package SPSS 18. One-way analysis of variance (ANOVA) was applied to test for significant differences between the individual treatments. The results were analysed by Student’s t-test (Student–Newman–Keuls) at a significance level of 95% (α = 0.05).

The results from the sensory evaluation (Section 2.5) were analysed for significant differences (α = 0.05) by ANOVA (Statgraphics).

3. Results and discussion

3.1. Effect of pH and sugar on terpene release

The first enzyme assays were performed under optimal enzyme conditions (pH 5.5, ethanol removed) using an extract from a white wine (Traminer, Austria). According to Mateo and Jiménez (2000), Traminer is classified as a non-Muscat, but aromatic variety, that depends on monoterpenes as major flavour components. The resulting concentrations of monoterpenes after enzyme treatment are shown in Table 3. Because β-citronellol and nerol could not sufficiently be separated by the analytical method applied, the corresponding results are displayed as sum of β-citronellol plus nerol throughout the paper. Regarding the used enzyme codes, the reader is again referred to Table 1.

Table 3.

Terpene concentrations (μg/L) of the white wine extract (Traminer, Austria) after glycosidase treatment. C, control; the enzyme codes are the same as displayed in Table 1. All data represent the average of triplicate determination. The results were grouped by a t-test (Student–Newman–Keuls; α = 0.05). Values labelled with the same letters (upper case) are not significantly different.

	C	GL	GO	GA	GO/AO	GO/AA	GO/R	GO/AO/R	GO/N
α-Terpineol	0.327A	6.17B	7.83B	6.71B	26.6D	17.4C	9.77B	32.6E	23.7D
Linalool	<0.3	3.08A	4.74B	3.05A	5.10B	5.53B	2.95A	4.82B	8.01C
cis-Linalool oxide	4.10A	31.8BC	32.8BC	18.1B	70.2D	46.9C	30.5BC	97.0E	67.5D
trans-Linalool oxide	2.45A	34.9AB	37.7AB	14.6A	154C	41.8AB	23.2AB	213D	58.6B
β-Citronellol + nerol	2.55A	78.8B	130C	62.4B	218E	247EF	165D	250EF	269F
Geraniol	11.7A	249BC	299C	206B	515E	610F	407D	593F	809G
Hotrienol	<0.3	1.23B	1.44BC	1.31B	1.79D	1.27B	0.74A	2.23E	1.67CD
Sum of terpenes	21.7A	406BC	515CD	314B	993E	971E	640D	1190F	1240F

As shown in Table 3, all β-glucosidase preparations (GL, GO, GA) were able to release monoterpenes, the highest concentrations were detected with GO. According to the scheme of sequential precursor hydrolysis as proposed by Gunata et al. (1988), arabinosidase and rhamnosidase preparations were always applied in combination with the β-glucosidase from O. oeni (GO). The use of the same glucosidase (GO) in all assays with enzyme combinations was intended to obtain comparable results. Fungal (GO/AA) and bacterial (GO/AO) arabinosidase could release equal amounts of total terpenes. Addition of the Pediococcus acidilactici rhamnosidase R to GO caused only a small further increase in terpene concentrations, compared to treatment with GO alone. The highest terpene concentrations were released when applying the combinations GO/AO/R and GO/N. At this point, it is important to note that N, which was applied as a fungal rhamnosidase preparation, is in fact a complex mixture containing additional activities of α-l-rhamnosidase, β-d-glucosidase, β-d-galactosidase, β-d-xylosidase, and α-l-arabinosidase (see activity profile in Fig. 1).

Fig. 1.

Influence of pH and glucose concentration on the glycosidase activities of naringinase as determined with synthetic substrates. C, control (pH 5.5); Glu, p-nitrophenyl-β-d-glucopyranoside; Gal, p-nitrophenyl-β-d-galactopyranoside; Xyl, p-nitrophenyl-β-d-xylopyranoside; Rha, p-nitrophenyl-α-l-rhamnopyranoside; Araf, p-nitrophenyl-α-l-arabinofuranoside; Arap, p-nitrophenyl-α-l-arabinopyranoside. The data displayed represent the average of duplicate determinations.

Subsequently, two brands of red grape juice (“St. Laurent”, “Happy Day”), both commercially available at Austrian markets, were used as substrates for enzyme assays. “St. Laurent” is a highly aromatic grape variety that is often cultivated in Eastern Austria (Lower Austria, Burgenland), while the latter is a commercial bulk product which is probably an undefined blend of several grape varieties. The aim of these assays was to take the effects of the juice matrix, especially of sugar inhibition at still optimal pH (adjusted to pH 5.5) into account (see Table 2 for juice composition). At first, the total amounts of released terpenes differed significantly between the two varieties (Table 4), most likely due to different concentrations of aroma precursors. The overall release of terpenes from “St. Laurent” was low, while higher concentrations were detected in “Happy Day” after enzyme treatment. Nevertheless, the results from both juices followed similar trends. Remarkably, both glucosidase (GA) and arabinosidase (GO/AA) from Aspergillus niger were almost inactive under these conditions (Table 4). These results are in agreement with the finding that the fungal glucosidase GA was strongly inhibited by glucose in tests with pNP-β-d-glucopyranoside (3.6% residual activity at 500 mM glucose, corresponding to 90 g/L). In contrast, GO still exhibited 36% residual activity at 500 mM glucose. However, the results for the fungal arabinosidase AA are rather surprising: AA was not inhibited by glucose in laboratory assays with pNP-α-l-arabinofuranoside (Michlmayr, Schümann et al., 2011). An explanation for its inactivity in the grape juice could be the effect of the complete juice matrix (Table 2). Although the combination GO/AA could release low amounts of α-terpineol, β-citronellol + nerol and geraniol (compared to GO alone, Table 4), regarding the sum of terpenes, no further significant increase of free terpenes could be observed by adding AA to GO. The relatively high activity of N in grape juice compared to the enzyme preparations from A. niger might be caused by the comparably low effect of glucose on the glycosidase activities of N. As shown in Fig. 1, the rhamnosidase activity of N was clearly inhibited by glucose (13% residual activity at 500 mM glucose), but other glycosidase side activities of N were affected less or even increased in the presence of high glucose concentrations.

Table 4.

Total release of terpenes (μg/L) from two red grape juice varieties through glycosidase treatment. C, control; the enzyme codes are the same as displayed in Table 1. All data represent the average of triplicate determination. The results were grouped by a t-test (Student–Newman–Keuls; α = 0.05). Values labelled with the same letters (upper case) are not significantly different.

	C	GO	GA	GO/AO	GO/AA	GO/R	GO/AO/R	N	GO/N
“St Laurent”, pH 5.5
α-Terpineol	2.08B	2.15B	1.91A	2.80C	2.20B	2.17B	3.19D	3.42E	3.91F
Linalool	1.50B	1.44B	1.03A	1.69B	1.37B	1.46B	1.43B	2.10C	2.38D
cis-Linalool oxide	6.04AB	8.38C	5.35A	10.0D	7.05BC	7.93C	11.6D	11.0D	14.4E
trans-Linalool oxide	2.80A	5.79C	2.30A	9.14E	4.48B	7.17D	11.3F	5.13BC	13.9G
β-Citronellol + nerol	<0.3	<0.3	<0.3	1.41B	0.590A	0.670A	1.96C	0.507A	1.30B
Geraniol	<0.3	0.397A	<0.3	4.61D	2.07C	1.30B	8.04E	5.15D	9.71F
Hotrienol	1.05D	1.00CD	0.307A	0.917C	0.540B	1.08D	1.09D	1.65E	1.65E
Sum of terpenes	14.4B	20.0C	11.5A	31.5E	18.8C	22.9D	39.5F	29.9E	48.6G
“Happy Day”, pH 5.5
α-Terpineol	6.95A	8.58B	6.71A	15.9E	11.9C	11.0C	16.6E	14.3D	32.5F
Linalool	1.56B	1.62B	1.11A	1.66B	1.44B	1.57B	1.56B	2.88C	3.35D
cis-Linalool oxide	8.82A	26.4C	9.5A	36.9E	19.4B	32.7D	46.5F	25.0C	93.5G
trans-Linalool oxide	5.57A	15.0C	5.54A	25.4E	11.3B	17.7D	32.5F	15.6C	55.8G
β-Citronellol + nerol	<0.3	2.65B	0.429A	11.4F	5.92C	6.82D	11.2F	7.66E	7.44E
Geraniol	<0.3	5.05B	1.30A	18.3E	10.8C	12.0D	18.6E	23.0F	26.5G
Hotrienol	<0.3	<0.3	<0.3	<0.3	<0.3	<0.3	<0.3	<0.3	0.320
Sum of terpenes	23.8A	59.9B	24.9A	110E	61.2B	82.5C	127F	99.0D	221G
“Happy Day”, pH 3.0
α-Terpineol	9.22B	10.1C	8.10A	11.1D	8.49A	9.53B	11.7E	11.6E	12.6F
Linalool	2.13B	2.49BC	1.90AB	2.86CD	1.42A	2.18B	2.61BC	3.74E	3.36DE
cis-Linalool oxide	13.7A	15.0AB	15.7ABC	14.1A	15.0AB	17.1C	16.8BC	35.3D	35.0D
trans-Linalool oxide	7.67AB	8.26B	8.05AB	8.26B	7.34A	8.13AB	8.34B	21.1C	20.9C
β-Citronellol + nerol	<0.3	1.02B	1.16B	1.84C	<0.3	0.660A	1.82C	3.63D	3.55D
Geraniol	0.420A	1.68C	2.16D	2.56E	2.68E	1.16B	2.68E	9.64F	9.53F
Hotrienol	0.710F	0.693E	0.533A	0.657DE	0.610BCD	0.647CDE	0.587B	0.633BCD	0.600BC
Sum of terpenes	35.2A	39.7BC	38.0AB	41.8C	36.0A	39.7BC	45.0D	86.2E	86.1E

At natural juice pH (Table 4, assays only performed with “Happy Day”) the bacterial enzymes could still release statistically significant amounts of terpenes, although at a low magnitude. Only the fungal preparation N could release higher amounts of terpenes at pH 3.0, which is consistent with the results obtained with synthetic glycosides shown in Fig. 1, suggesting a high increase of glycosidase activities toward lower pH. The addition of GO to N caused no further increase of terpene concentrations.

3.2. Terpene profiles

In addition to the total amount of terpenes released under given conditions, it is important to consider the characteristic profile of free terpenes generated by an enzyme preparation in more detail. The corresponding observations are discussed in the present section. For this purpose, the results shown in Tables 3 and 4 are additionally presented in graphical form as Supplementary online content (Figs. S1 and S2).

The resulting terpene profiles in the Traminer wine extract (Table 3, Supplementary Fig. S1) suggest rather similar substrate specificities for the β-glucosidases GL, GO and GA. Although all these enzymes are classified into the same glycoside hydrolase family (GH 3, see also Table 1), both bacterial glucosidases possess additional side activities of xylosidase and arabinosidase (Michlmayr, Schümann, Barreira Braz Da Silva, Kulbe, del Hierro, 2010; Michlmayr, Schümann, Wurbs et al., 2010), while such side activities could not be detected in GA. Although it might be expected that these side activities of GL and GO would contribute to a distinct aroma profile compared to GA, such an effect was not observed.

A rather interesting observation was that (in combination with GO) the arabinosidase from O. oeni (AO) significantly produced higher amounts of the tertiary terpene alcohols α-terpineol, cis/trans-linalool oxide and hotrienol than the arabinosidase from A. niger (GO/AA; Table 3, Fig. S1). In contrast, AA released higher amounts of the primary terpenols geraniol and β-citronellol + nerol than AO. A similar effect was observed comparing the combinations GO/AO/R and GO/N. While the overall terpene concentrations produced were equal, GO/AO/R released higher amounts of α-terpineol, linalool oxides and hotrienol than GO/N. In the case of linalool, which is a tertiary terpene alcohol as well, no significant increase could be detected after addition of AO and R to GO. In contrast, the highest concentrations of linalool were released by the combination GO/N.

Regarding the complex composition of N (Fig. 1), it is interesting to observe that although the addition of N to GO could further increase the total concentrations of free terpenes, the resulting terpene profiles of GO and GO/N were rather similar in the wine extract (Supplementary Fig. S1). The same effect was observed in “Happy Day” grape juice at pH 5.5 (Supplementary Fig. S2). Fig. S2 also shows that the profiles generated by N and GO/N are clearly distinct, as the addition of GO to N caused a further significant increase of the tertiary terpenols α-terpineol and cis/trans-linalool oxides, implying synergistic effects between these preparations. Further, comparing the terpene profiles generated by N at pH 3.0 and pH 5.5, it is obvious that the resulting profiles were remarkably different (Fig. S2). This may indicate that the enzymes that contribute to aroma release by N respond differently to pH.

Fig. S2 also demonstrates that in the grape juice (“Happy Day”, pH 5.5), addition of AO and/or R to GO could further increase the concentrations of free α-terpineol, cis/trans-linalool oxide, β-citronellol + nerol, and geraniol, compared to samples treated with GO only.

The results presented above indicate that the glycosidases from O. oeni are capable of releasing terpenes from natural glycosylated precursors, suggesting that these intracellular enzymes might contribute to the release of glycosylated aroma compounds during malolactic fermentation. Further, the bacterial glycosidases demonstrated interesting characteristics in comparison to the fungal enzymes. Besides the lower inhibition of the O. oeni glycosidases in juice conditions, a general observation made here is that the bacterial enzymes, especially the arabinosidase from O. oeni, possess capacities to release both primary and tertiary terpene alcohols (terpenols), while the fungal enzymes preferentially released primary terpenols. These findings seem to contradict the results of Ugliano et al. (2003), and Ugliano and Moio (2006), who reported that O. oeni mainly released primary terpenols during MLF. However, it remains to be investigated to what extent such glycosidase genes are distributed in O. oeni genomes and further, whether such enzymes are actually expressed during MLF. Due the reported genetic heterogeneity of O. oeni (Bartowsky & Borneman, 2011; Borneman, Bartowsky, McCarthy, & Chambers, 2010), it can be expected that variations with regard to the presence of glycosidase genes and their regulation exist between individual O. oeni isolates. This is also indicated by the fact that glycosidase activity profiles have been shown to be highly strain-dependent (Gagné et al., 2011; Grimaldi et al., 2005b). Furthermore, O. oeni possesses several GH 1 phospho-β-glucosidase genes related to the cellobiose/β-glucoside specific phosphotransferase system (Capaldo, Walker, Ford, & Jiranek, 2011a, 2011b). It is not yet established, whether this enzyme class can be made responsible for the release of glycosylated aroma compounds during MLF.

As far as possible, the fungal enzymes (A. niger) used in this study were chosen due to their assignment to the same GH families as the bacterial glycosidases involved (glucosidases GH 3, arabinosidases GH 51, Table 1). However, it should be noted that the above discussed differences in substrate specificities are most likely not directly related to the bacterial or fungal origin of the involved glycosidases. It would be worthwhile to investigate whether the capability to release primary and/or tertiary terpenols is related to the empirical distinction between aryl/alkyl glycosidases on one hand and glycosidases specific for short chain oligosaccharides on the other hand, which is especially well documented in the case of β-glucosidases (Bhatia, Mishra, & Bisaria, 2002). Our previous results suggest that both glucosidase and arabinosidase from O. oeni can be classified as true aryl/alkyl glycosidases, while both A. niger glycosidases showed a high preference in hydrolysing disaccharides (Michlmayr, Schümann et al., 2011; Michlmayr, Schümann, Wurbs et al., 2010). Further, our recent work (Michlmayr, Brandes et al., 2011) on two bacterial rhamnosidases, both assigned to GH 78, revealed that Ram (“R” in the present study) can be classified as an aryl-glycosidase, while Ram2 (not involved in the present study) displayed its highest catalytic efficiency with the disaccharide rutinose. Interestingly, Ram (R) could release both primary and tertiary terpenols in a Muscat wine extract, while Ram2 could only release primary terpenols under the same conditions.

3.3. Vinification experiments with glucosidase and arabinosidase from O. oeni

Small-scale vinification experiments were conducted to perform an initial evaluation on whether the glycosidases from O. oeni are in principle suited for application in winemaking. Therefore, both glucosidase and arabinosidase from O. oeni were applied during the cold maceration stage of a Riesling wine. The total terpene contents of the musts extracted from the Riesling mash after enzyme treatment and that of the resulting wines are shown in Table 5. Additionally, graphical representations of these data can be found in the supplementary online content of this paper (Supplementary Figs. S3 and S4). Interpreting these data, it is not clear whether the bacterial enzymes could hydrolyse aroma precursors during the cold maceration period. The highest concentrations of terpenes were detected in samples treated with the commercial preparation Maceration C (MacC), followed by the two controls (In C1, no pectinase was added before pressing). The increase of free terpenes in both controls compared to the mash (determined in the free run juice of the mash) before the maceration period might be explained by the extraction of terpenes from the skin during maceration or the action of a grape glycosidase (Maicas & Mateo, 2005). The total terpene concentrations in the musts treated with GO were even lower than that of the controls, an explanation for this could be the absorption of volatile compounds on hydrophobic regions of the protein. In contrast to the experiments described in Section 3.1, only low enzyme concentrations were used in these vinification experiments, causing an expectedly low release of terpenes. Nevertheless, with increasing dose of GO the total terpene concentrations increased as well, suggesting that GO was not completely inactive, although the detected changes were at a low level (see Table 5 and Fig. S3). Interestingly, both samples treated with GO/AO and AO displayed terpene concentrations equal to those of the controls. In tests with the Traminer extract, it was shown that AO could release only low levels of terpenes compared to the control without the presence of the glucosidase (21 ppm total compared to the controls). Considering the fact that it was reported that grape glucosidases mainly hydrolyse primary terpenols (Maicas & Mateo, 2005), and further regarding the differences in the release of primary/tertiary terpenols by different enzyme preparations as discussed in Section 3.2, it could be expected that different enzyme treatments would result in recognisably distinct terpene profiles in the musts. However, such an observation cannot be confirmed.

Table 5.

Terpene concentrations (μg/L) in the Riesling musts after enzyme treatment during cold maceration and of the corresponding final wines after alcoholic fermentation. M, mash before cold maceration; C1, control without pectinase treatment before pressing; MacC, Maceration C (Preziso, Austria), 3 g/hL; GO, with 60, 200, 300 U/L. AO, 35 U/L; GO/AO: 150 U/L GO + 25 U/L AO. All data represent the average of triplicate determination. The results were grouped by a t-test (Student–Newman–Keuls; α = 0.05). Values labelled with the same letters (upper case) are not significantly different.

	M	C1	C2	MacC	GO60	GO200	GO300	AO	GO/AO
Musts after enzyme treatment
α-Terpineol	0.543A	14.6D	15.2D	20.4E	8.08B	11.6C	13.0C	16.3D	15.5D
Linalool	0.963A	18.4DE	18.3DE	20.5E	10.2B	15.4C	16.5CD	19.4E	18.6DE
cis-Linalool oxide	1.76A	7.07C	8.35D	12.2F	5.84B	6.61BC	7.06C	9.52E	9.10DE
trans-Linalool oxide	0.753A	2.57B	2.71B	4.13D	2.28B	2.32B	2.68B	3.39C	3.26C
β-Citronellol + Nerol	<0.3	5.07E	4.48D	2.30A	3.12B	4.30CD	3.89BCD	3.57BC	3.72BCD
Geraniol	0.353A	17.4F	15.1E	6.90B	10.8C	15.2E	13.8DE	12.8CD	13.2DE
Hotrienol	2.92A	28.9D	31.8D	39.7E	16.1B	22.6C	24.1C	30.2D	30.5D
Sum of terpenes	7.48A	94.9D	96.8D	107E	56.8B	78.6C	81.7C	95.8D	94.6D
Riesling wines
α-Terpineol		7.25A	7.87AB	10.1D	7.25A	7.60AB	8.11B	9.19C	9.02C
Linalool		26.4ABC	27.7BC	30.1D	26.1B	25.7A	27.8BC	30.1D	28.3C
cis-Linalool oxide		2.91A	3.26BC	3.76D	3.12AB	3.26BC	3.52CD	3.68D	3.58CD
trans-Linalool oxide		1.06A	1.11A	1.35B	1.07A	1.09A	1.16AB	1.22AB	1.23AB
β-Citronellol + Nerol		3.57A	3.73AB	3.50A	3.92B	3.67AB	3.45A	3.73AB	3.88B
Geraniol		5.36A	5.23A	5.76B	5.72B	5.48AB	5.73B	6.32C	6.75D
Hotrienol		13.7A	17.9CD	22.2E	15.7B	16.4B	17.6C	19.2D	18.6CD
Sum of terpenes		66.3A	73.3B	83.3D	69.3AB	69.0AB	72.6B	80.0CD	77.9C

The analysis of terpenes after the alcoholic fermentation (Table 5, see also Supplementary Figs. S3 and S4) shows that the differences in terpene concentrations between the treatments are less distinct than before the alcoholic fermentation. Further, the total terpene concentrations in the wines were lower than in the musts. However, it is interesting to observe that the overall trends observed in the musts are still recognisable after alcoholic fermentation, which is evident when comparing the results in the Supplementary Figs. S3 and S4. This indicates that dependent on the glycosidase activity profile of the yeast involved, enzyme treatment at an early stage of winemaking, as presented here, may indeed affect the sensory properties of the final product.

The results of the sensory evaluation are shown in Fig. 2. “Riesling HBLA” was included as a further control for the sensory tests. This wine was produced from the same harvest of grapes without cold maceration but otherwise the same fermentation conditions. Interestingly, in the aroma intensity ranking (Fig. 2a), the highest intensity was recognised in the wines treated with AO only, while wines treated with GO/AO received the lowest rating (except “Riesling HBLA”). This is remarkable as treatment with AO and GO/AO did not result in analytically distinct terpene concentrations compared to the controls. Further, wines treated with β-glucosidase (GO) alone were marked as significantly different to the controls. The analysis of variance of the aroma intensity ratings (Fig. 2a) showed that in the glucosidase-treated wines, aroma intensity significantly (significance level = 0.01) correlated with increasing enzyme dose. Additionally, the perceived intensity of the glucosidase-treated wines highly correlates to the stone fruit (0.01 level), citrus (0.05 level) descriptors; the intensity perceived for the arabinosidase (AO) and arabinosidase with glucosidase (GO/AO) treated wines highly correlates with pomaceous fruits (0.001 level), citrus (0.05 level), stone fruit (0.05) and freshness (0.01 level) (Fig. 2b). Therefore, it seems that wines with the treatment of AO and GO/AO were described with the typical Riesling descriptors (stone fruit, citrus, pomaceous). However, the tasters did not see an increase in floral, candied, tropical aromas. Interestingly, in the typicality rating, the external control wine “Riesling HBLA” was not recognised as a typical Riesling wine by the tasters (rating 37%); the controls (MacC, C1 and C2) received ratings between 57% and 64%. The wines treated with the bacterial enzymes were most often marked as typical (GO/AO and GO200 treatments 78%, GO60 81%, AO 90% and GO300 93%).

Fig. 2.

Analysis of the sensory evaluation (professional panel, n = 7) of the enzyme treated wines. (a) Intensity ranking, increasing aroma intensity correlates with less points. (b) Spider plot of the intensity of the attributes included. C1, control without pectinase treatment; MacC, Maceration C (Preziso, Austria), 3 g/hL; GO, with 60, 200, 300 U/L. AO, 35 U/L; GO/AO: 150 U/L GO + 25 U/L AO. “Riesling HBLA” is a wine produced with the same grapes, but without cold maceration.

The major drawback in the results presented above is that a clear correlation between analytical and sensory evaluation cannot not be made. It is conceivable that due to the low perception thresholds of volatile compounds (Mateo & Jiménez, 2000), significant differences in aroma composition may already be recognised on a subjective level where the corresponding chemical changes are not even detectable/distinguishable by analytical methods. Synergistic/additive effects between aroma compounds resulting in lowered perception thresholds have been described as well (Rapp & Mandery, 1986). Therefore, the question whether a given enzyme is a valuable tool for winemaking may be a matter of sensory and personal preferences rather than an analytical one. Accordingly, apart from a biochemical characterisation, it is most important to understand how an enzyme preparation influences the characteristic varietal aroma bouquet in sensory terms.

4. Conclusion

To the best of our knowledge, this is the first study reporting the properties of cell-free glycosidases from O. oeni to release aroma compounds from natural substrates like wine and fruit juice. From a biological point of view, this is an essential step towards understanding how O. oeni is capable of releasing grape-derived aroma compounds from wine. It will further be necessary to determine how such glycosidase genes are regulated during the MLF. Further, due to the intracellular nature of both glucosidase and arabinosidase of O. oeni, studies on the mechanisms involved in substrate import will be required as well to gain a complete understanding of the mechanisms that govern the aroma release by wine lactic acid bacteria.

The results of the present study further indicate that the enzymes of O. oeni could be interesting tools for the early stages of winemaking, especially since the wines produced from them were preferred by the tasting panel, and enzyme treatment could evidently contribute positively to the “typical” Riesling aroma. However, further detailed experiments using different wine varieties and fermentation conditions (e.g., yeasts) will be required in order to confirm this conclusion. Regarding a possible application of such glycosidases, it is necessary to mention that a direct application of bacterial enzymes in winemaking is at present not realistic. A major obstacle is the necessity of recombinant enzyme production. The use of recombinant techniques in the food industry has a rather negative image due to consumer and market preferences. An attractive alternative (although recombinant as well) could be the use of LAB as GRAS/food grade expressions systems, which is a developing field of intensive research (Peterbauer, Maischberger, & Haltrich, 2011).