726 S. Saberi et al. /
726 S. Saberi et al. / Food Research International 48 (2012) 725–735Single ong>strainsong> ong>ofong> commercial active dry wine yeasts have been usedfor many years to control alcoholic fermentation; however, this hasresulted in the production ong>ofong> wines with a similar character throughoutthe world. Nevertheless, winemakers have the ability to influence thenature and complexity ong>ofong> their wine by utilizing new indigenous yeastong>strainsong> (Swiegers & Pretorius, 2005) or using mixtures ong>ofong> yeast ong>strainsong>to develop complexity in their wines. Howell, Cozzolino, Bartowsky,Fleet, and Henschke (2006) investigated the effects ong>ofong> ong>mixedong> knownS. ong>cerevisiaeong> ong>strainsong> on the chemical prong>ofong>ile and aromatic properties ong>ofong>Chardonnay wines. They determined that the chemical prong>ofong>iles ong>ofong> thewines fermented with individual and ong>mixedong> S. ong>cerevisiaeong> ong>strainsong> weredifferent and that it was not possible to blend wines produced by thesingle ong>strainsong> to create the same chemical prong>ofong>ile as a wine fermentedby the ong>mixedong> yeast cultures.To characterize the chemical prong>ofong>ile ong>ofong> a wine, the compoundsmust be first extracted and/or concentrated prior to gas chromatography(GC) or GC–mass spectroscopy (MS). This can be accomplishedusing a number ong>ofong> techniques including: static headspace, purgeand trap, solid-phase microextraction (SPME), as well as solvent-,supercritical-, microwave- and stir bar sorption–extraction methods(Malherbe, Watts, Nieuwoudt, Bauer, & du Toit, 2009). Since thesetechniques influence the presence and concentration ong>ofong> the metabolites,they in part explain the differences in compounds reported inthe literature.Patel and Shibamoto (2003) used solvent extraction and GC–FIDto quantify 53 volatile compounds from 20 yeasts ong>ofong> S. ong>cerevisiaeong> inSymphony wine. While 18 ong>ofong> the 20 yeast ong>strainsong> produced thesame compounds (alcohols, esters, acids), it was the difference inconcentration ong>ofong> these compounds which influenced the flavor prong>ofong>ileong>ofong> the wine. In contrast, Li, Tao, Wang, and Zhang (2008) utilizedSPME GC–MS to quantify 41 compounds in Chardonnay, ong>ofong> which13 were odor active. Like Komthong, Hayakawa, Katoh, Igura, andShimoda (2006), they used odor active values (OAVs) to evaluatethe sensory impact ong>ofong> the volatile compounds. Malherbe et al.(2009), used headspace SPME GC–MS to identify 68 volatile compoundsin red and white wine and were able discriminate betweencontrol and problematic fermentations.As such, headspace analysis was applied in this research to morethoroughly understand the volatiles among Chardonnay wines fermentedwith individual and ong>mixedong> yeast cultures, compared to sixcommercial yeast ong>strainsong>. This research was undertaken to: i) first documentthe uniqueness ong>ofong> two novel Burgundian yeast isolates, ii) quantifythevolatileprong>ofong>ilesong>ofong> wines from these novel Burgundian ong>strainsong>when fermented as individual and ong>mixedong> cultures and iii) estimate thesensory prong>ofong>ile ong>ofong> the resultant wines using OAVs and radar diagrams.2. Materials and methods2.1. JuiceChardonnay must was obtained from White Salmon Vineyard inCalifornia (2008). It had soluble solids (SS), pH, titratable acidity(TA) and yeast available nitrogen (YAN) ong>ofong> 24 °Brix, 3.46, 5.76 g/Land 131 mg nitrogen/L, respectively. The juice was stored at −20 °Cprior to use.2.2. Yeast ong>strainsong> and yeast culturingThe novel S. ong>cerevisiaeong> yeast ong>strainsong> were isolated from a vineyardin Burgundy region, France and named C2 and C6. To evaluate the impactong>ofong> ong>mixedong> fermentations on the complexity ong>ofong> the wine, theseong>strainsong> (C2:C6) were ong>mixedong> in four ratios 1:1, 1:2, 1:3 and 2:3, referredto as M1, M2, M3 and M4, respectively. The individual andong>mixedong> ong>strainsong> were compared to six widely used commercially availableyeast ong>strainsong> ong>ofong> S. ong>cerevisiaeong> (Blanc, Elegance, Fusion, CY3079,ICV-D254, X16). Blanc, Elegance, Fusion were produced by MauriYeast Australia (Sydney, NSW, Australia), CY3079 and ICV-D254 byLallemand (Montreal, QC, Canada) and one (X16) by Laffort (Petaluma,CA, USA); all were purchased from Scott Laboratories (Pickering,ON, Canada) as active dry yeasts. These ong>strainsong> were recommendedfor white wines especially Chardonnay to increase fruity aroma andcomplexity (AB Mauri, 2012; Laffort, 2009; Lallemand, 2012). Yeastswere cultured in Difco yeast peptone dextrose (YPD) broth (Becton,Dickinson and Co., Sparks, MD, USA) based on the standard methods(Ausubel et al., 1995). S. ong>cerevisiaeong> ong>strainsong> were stored at −80 °C inYPD broth with 15% glycerol.Individual yeast ong>strainsong> were genetically fingerprinted by a polymerasechain reaction (PCR) method. The PCR method discriminated yeastong>strainsong> based on the amplification ong>ofong> repetitive δ sequences ong>ofong> S. ong>cerevisiaeong>genome (Saberi, 2011; Schuller, Valero, Dequin, & Casal, 2004), which areong>ofong>ten associated with Ty1 transposons (Schuller et al., 2004). ThePCR was performed on a MJ Research Peltier Thermal Cycler 200(Walthman, USA) using the δ 2(5′-GTGGATTTTTATTCCAAC-3′) andδ 12 (5′-TCAACAATGGAATCCCAAC-3′) primers.Freezer stocks ong>ofong> S. ong>cerevisiaeong> ong>strainsong> were used to inoculate 5 mLliquid cultures ong>ofong> YPD; S. ong>cerevisiaeong> cells were grown overnight in a rotarywheel to stationary phase at 30 °C. Flasks containing 50 mL YPDcultures were subsequently inoculated for each strain at a rate ong>ofong>5×10 5 cells/mL and grown aerobically in a shaker bath (180 rpm) at30 °C for 24 h. Cells were then harvested by centrifugation (5000 gfor 5 min). Harvested cells were washed with sterile MilliQ waterand re-suspended in the fermentation medium (Saberi, 2011).2.3. FermentationsChardonnay fermentations were performed in triplicate at 16 °C and20 °C; these temperatures reflectcommercialwinemakingconditionsthat optimize retention ong>ofong> volatiles and reflect typical cellar conditions,respectively. The 250 mL fermentation bottles containing 200 mL Chardonnaymust were inoculated at the rate ong>ofong> 2×10 6 cells/mL. Mixedstrain fermentations were inoculated using the ratios as indicatedabove (M1, M2, M3, M4). Yeast ong>strainsong> were not ong>mixedong> before inoculation.All fermentation bottles were topped with disinfected (70% ethanol)rubber bungs and water-filled capped gas locks to provideanaerobic conditions. When fermentations were complete, 100 mg/Long>ofong> potassium metabisulflite was added to prevent oxidation. Anaerobicsampling was aseptically performed by removing approximately 1 mLsample through the rubber bung with a 12.5 cm hypodermic needle(Air-Tite, Virginia Beach, VA, USA) attached to a 3 mL syringe (BectonDickinson, Franklin Lakes, NJ, USA). Fifty mL wine samples were placedin 50 mL glass vials with screw cap closures, and stored at 4 °C for3–4 weeks until GC analysis.2.4. Headspace analysis by gas chromatography–mass spectroscopyHeadspace analysis ong>ofong> the volatile compounds in the Chardonnaywines was conducted by gas chromatography–mass spectrometry(GC–MS) analysis, according to the method utilized by Danzer,Garcia, Thiel, and Reichenbacher (1999). Other technologies are availablefor volatile analysis, such as solid phase microextraction (SPME);this methodology adsorbs the volatiles onto a fiber prior to GC analysis.While this concentrates the volatiles and enhances sensitivity, italso shifts the pattern ong>ofong> volatiles to those that are preferentiallyadsorbed by the fiber. In contrast headspace analysis samples the volatilesdirectly from the gas phase above the wine, which more closelyresembles the collection or pattern ong>ofong> volatiles that would be evaluatedby a human assessor.Wine samples (10 mL) were sterile filtered (0.22 μm) and placedin 20 mL glass GC headspace vials with 3 g ong>ofong> NaCl. Vials were sealedwith rubber septa and metal crimp tops. Vials were agitated, thenequilibrated at 85 °C for 10 min, prior to injecting 1 ml ong>ofong> headspacesample into the GC–MS (Agilent Technologies, Palo Alto, USA).
S. Saberi et al. / Food Research International 48 (2012) 725–735727The GC was equipped with a 60 m×0.25 mm ID, 0.25 μm thicknessDBWAX fused silica open tubular column (J&W Scientific,Folstom, CA, USA) and 5973N Mass Selective Detector (MSD) (AgilentTechnologies, Wilmington, DE, USA) for separation, detection andquantification ong>ofong> volatile compounds. Ultra high purity helium wasused at a flow rate ong>ofong> 1.3 mL/min. The headspace samples (1 mL)were injected through a valve that was maintained at 100 °C, whilethe temperature ong>ofong> the transfer line was kept at 110 °C. The initialtemperature ong>ofong> the GC oven was held at 40 °C for 5 min, raised to100 °C at a rate ong>ofong> 5 °C/min, then increased to 200 °C at a rate ong>ofong>20 °C/min. The MSD was set in scan mode with a mass range ong>ofong> 35–400 amu. Each sample was quantified in triplicate; 3-octanol wasused as an internal standard.2.5. Volatile selection and quantificationVolatile compounds were identified by GC–MS using the enhancedChemstation song>ofong>tware (Chemstation Build 75, Agilent Technologies,Palo Alto, CA, USA). Aroma compounds were identified by comparingthe peak retention times against those ong>ofong> authentic standards andmatching the mass spectra against the Wiley7Nist05 mass library(Wiley & Sons, Hoboken, NJ, USA). Peaks were quantified when thesignal-to-noise ratio was greater than 10. The single point internalstandard method (Alltech Associates, 1998) was used to quantify theheadspace concentration ong>ofong> the volatiles, by comparing their responsesto that ong>ofong> the internal standard (IS). Standards were preparedin a synthetic wine (~12% ethanol+tartaric acid, pH~2.3) with 3-octanol as an IS. The formula used for quantifying the specific volatiles(SV) was: concentration ong>ofong> SV=(concentration IS ×area SV ×IRF SV )/area IS , where IRF was the internal response factor.2.6. Statistical analysisA two-factor analysis ong>ofong> variance (ANOVA) with replication wasused to evaluate the effects ong>ofong> yeast ong>strainsong>, temperature as well astheir interaction (temperature×yeast) on the production ong>ofong> volatilecompounds. Since the effects ong>ofong> temperature and temperature×yeastwere none significant (p>0.05) for all compounds, only the effect ong>ofong>yeast strain was reported, i.e. mean values were averaged across bothfermentation temperatures. Differences among yeast ong>strainsong> were evaluatedusing Fisher's least significant difference (LSD) test (p≤0.05).Principal component analysis (PCA) using the correlation matrixwas conducted on mean volatile concentrations for the six individualindustrial, two individual Burgundian, and four ong>mixedong> Burgundianong>strainsong>. PCA assessed the volatile prong>ofong>ile among the yeast ong>strainsong>.PCA analyses were performed on all volatiles, as well as the collectionong>ofong> higher alcohols, ethyl esters and acetate esters. Volatile compoundswere represented as vectors. Principal component (PC) I, PC II and PCIII were calculated and bivariate plots prepared for PC I versus PC II,PC II versus PC III and PC I versus PC III. However, in the interests ong>ofong>brevity, only PC I versus PC II were reported in this manuscript. However,the other plots were examined to verify that the interrelationshipsamong the ong>strainsong> were retained in these higher dimensions(PC I versus PC III, PC II versus PC III). For clarity ong>ofong> presentation, vectorcoordinates were scaled by a factor ong>ofong> three times compared to thesample coordinates.Radar diagrams (MS Excel, Seattle, WA, USA) were used to representthe odor prong>ofong>iles, as well as the estimated sensory prong>ofong>iles for the 18volatile compounds. These prong>ofong>iles were created using mean volatileconcentrations (mg/L) and odor active values (OAVs), respectively.Group means for the individual Burgundian ong>strainsong> (n=12, 2 yeasts×3replications×2 temperatures) and ong>mixedong> Burgundian ong>strainsong> (n=24, 4yeast mixtures×3 replications×2 temperature) were compared to theindustrial yeast ong>strainsong> (n=6, 1 yeast×3 replications×2 temperatures),using a collection ong>ofong> three plots. Each diagram compared the individualBurgundian and ong>mixedong> Burgundian ong>strainsong> with two industrialong>strainsong>. A group mean for the industrial ong>strainsong> was not calculated, sincethe commercial ong>strainsong> were not similar to one another.Since the radar option did not have provision for different scales, itwas necessary to multiply the concentrations ong>ofong> the compounds by aconstant (0.05–200). This meant radar plots ong>ofong> the volatiles representedrelative values; exact concentrations for the higher alcohol,ethyl esters, and acetate esters can be obtained from Tables 2, 3 and4 respectively. The relative sensory impact ong>ofong> the volatile compoundswas represented using odor active values (OAVs), for the 18 volatilecompounds. OAVs were calculated by dividing the volatile concentrationby the absolute aroma threshold (mg/L). Since wine thresholdswere not available for most compounds, published water thresholds(Campo, Ferreira, Escudero, Marques, & Cacho, 2005; Culleré, Escudero,Cacho, & Ferreira, 2004; Francis & Newton, 2005; Schieberle &Hong>ofong>mann, 1997) wereutilized,asiscustomaryintheliterature(Li, Tao,Wang, & Zhang, 2008).Radar plots ong>ofong> the OAVs were constructed using a log scale; thisallowed widely different concentrations (1×10 − 4 –1×10 3 mg/L) tobe represented on the same figure. These estimated sensory prong>ofong>ileswere labeled with the sensory attribute typically associated withthe volatile as well as the volatile abbreviation (Table 1).In order to compare the individual and ong>mixedong> Burgundian ong>strainsong>with all the commercial yeasts, it was necessary to generate threeplots, each with two commercial yeast ong>strainsong>. In addition, the odorand estimated sensory prong>ofong>iles were placed on the same page, sothat the diagrams could be readily compared.ANOVA and radar plots were calculated using MS Excel (Seattle,WA, USA); whereas, PCA and cluster analyses were performed usingMinitab 15 (State College, PA, USA).3. Results3.1. Genetic characterization ong>ofong> yeast ong>strainsong>Genetic fingerprinting successfully differentiated two Burgundianong>strainsong> (C2, C6) and six industrial ong>strainsong> based on their differences inthe chromosomal regions between δ sequences. The PCR method discriminatedyeast ong>strainsong> based on the amplification ong>ofong> the δ fragmentsin the S. ong>cerevisiaeong> genome. The individual Burgundian ong>strainsong> C2 andC6 shared four common bands around 350, 450, 750 and 1000 bp; C6had an additional band around 250 bp, which distinguished it from C2(data not shown) (Saberi, 2011).Table 1Quantifiable volatile compounds in Chardonnay wine fermented by six individual industrial,two individual Burgundian and four ong>mixedong> Burgundian S. ong>cerevisiaeong> strain at16 °C and 20 °C. The quantifiable compounds were listed by their class.Volatile compound2,3-Butanediol2-Methyl-1-butanol3-Methyl-1-butanoln-Butanol1-HexanolIsobutanolPhenylethanolPropanolEthyl butanoateEthyl hexanoateEthyl octanoateEthyl decanoateEthyl laurateEthyl acetateIsoamyl acetateHexyl acetateAcetaldehydeAcetic acidClass/AbbreviationHigher alcohol (HA-1)Higher alcohol (HA-2)Higher alcohol (HA-3)Higher alcohol (HA-4)Higher alcohol (HA-5)Higher alcohol (HA-6)Higher alcohol (HA-7)Higher alcohol (HA-8)Ethyl ester (EE-1)Ethyl ester (EE-2)Ethyl ester (EE-3)Ethyl ester (EE-4)Ethyl ester (EE-5)Acetate ester (AE-1)Acetate ester (AE-2)Acetate ester (AE-3)Aldehyde (ACET)Acid (AA)