Gas Chromatography–Mass Spectrometry Method Optimized Using

Feb 21, 2015 - 2-isobutyl-3-methoxypyrazine, 3-octanone, fenchone, 1-octen-3-one, ... headspace of grapes or wine using solid-phase microextraction...
15 downloads 0 Views 3MB Size
Article pubs.acs.org/JAFC

Gas Chromatography−Mass Spectrometry Method Optimized Using Response Surface Modeling for the Quantitation of Fungal Off-Flavors in Grapes and Wine Navideh Sadoughi,* Leigh M. Schmidtke, Guillaume Antalick, John W. Blackman, and Christopher C. Steel National Wine and Grape Industry Centre, School of Agricultural and Wine Sciences, Charles Sturt University, Locked Bag 588, Wagga Wagga, New South Wales 2678, Australia S Supporting Information *

ABSTRACT: An optimized method for the quantitation of volatile compounds responsible for off-aromas, such as earthy odors, found in wine and grapes was developed. The method involved a fast and simple headspace solid-phase microextraction−gas chromatography−mass spectrometry (HS-SPME−GC−MS) for simultaneous determination of 2-isopropyl-3-methoxypyrazine, 2-isobutyl-3-methoxypyrazine, 3-octanone, fenchone, 1-octen-3-one, trans-2-octen-1-ol, fenchol, 1-octen-3-ol, 2-methylisoborneol, 2,4,6-trichloroanisole, geosmin, 2,4,6-tribromoanisole, and pentachloroanisole. The extraction of the temperature and time were optimized using response surface methodology in both wine base (WB) and grape base (GB). Low limits of detection (0.1−5 ng/L in WB and 0.05−1.6 in GB) and quantitation (0.3−17 in WB and 0.2−6.2 in GB) with good recoveries (83−131%) and repeatability [4.3−9.8% coefficient of variation (CV) in WB and 5.1−11.1% CV in GB] and reproducibility (3.6−10.2 in WB and 1.9−10.9 in GB) indicate that the method has excellent sensitivity and is suitable for the analysis of these off-flavor compounds in wine and grape juice samples. KEYWORDS: fungal off-flavor, design of experiment, optimization, Colletotrichum acutatum, Botrytis cinerea



INTRODUCTION Consumer demand for high-quality products free from defect or off-flavor dictates that rigorous quality control procedures are used to maintain product integrity. The development of off-flavors arising from fungal contamination of fruits, particularly grapes and grape products, such as wine, is increasingly an important quality control issue.1 The analysis of aroma compounds responsible for the main olfactory defects of grapes and wines is an important step in the evaluation of product quality. A rapid analytical method that enables the simultaneous quantitation of the principal molecules identified as responsible for the off-flavors is therefore particularly valuable. A range of fungal off-flavors in grapes and wine have been described, with the most widely recognized being a dusty, moldy, and musty odor. Pentachloroanisole,2 2,4,6-tribromoanisole,3 and 2,4,6-trichoroanisole4 are responsible for cork taint. These compounds have very low olfactory thresholds in wine, and therefore, powerful analytical methods are required for their detection and quantitation. Besides cork taint, geosmin, an aromatic volatile fungal metabolite with an earthy smell, has been detected in wines made from rotten grapes.5,6 In addition, methylisoborneol and 1-octen-3-ol produced by Botrytis cinerea cause a mushroom taint.6 Other compounds of fungal origin with earthy, muddy, or mushroom odor are 2-isobutyl-3-methoxypyrazine,7 2-isopropyl-3-methoxypyrazine, 3-octanone, 1-octen-3-one, fenchone, fenchol, and trans-2-octen1-ol, which have also been reported in musts or crushed grapes.8 While the concentrations of these compounds are typically very low (nanograms per liter), they can influence the wine aroma because of their low sensory thresholds.1 © 2015 American Chemical Society

The fungal off-flavor components listed in Table 1 comprise several classes of organic compounds that typically differ by several orders of magnitude in affected fruits and fruit products. Because the target molecules are volatile, previously reported procedures for their determination are based on sampling the headspace of grapes or wine using solid-phase microextraction (HS-SPME),9,10 purge and trap,11 or headspace gas analysis12 coupled with gas chromatography−mass spectrometry (GC−MS). A challenge in determining the concentrations of compounds in ranges that typically vary by orders of magnitude is determining optimum combinations of conditions for sample handling. The design of experiments and response surface modeling of analyte responses have been employed to determine the optimum time and temperature for SPME processes.13 However, when multiple analyte responses require optimization, the most favorable handling conditions for individual analytes becomes increasingly convoluted. To date, no single method has been developed that enables quantitation of a wide range of fungal off-flavors across several orders of magnitude of the MS detector. In this context, we describe the application of response surface modeling combined with a modified desirability method for optimization of multiple responses,14 to optimize HS-SPME extraction time and temperature conditions for the simultaneous quantitation of a range of fungal off-flavors. Our approach models multiple responses and maximizes analytical sensitivity, with Received: Revised: Accepted: Published: 2877

November 12, 2014 February 17, 2015 February 21, 2015 February 21, 2015 DOI: 10.1021/jf505444r J. Agric. Food Chem. 2015, 63, 2877−2885

Article

Journal of Agricultural and Food Chemistry Table 1. Optimized Extraction Times and Temperatures for Target Compounds in Grape Samplesa compound 2-isobutyl-3-methoxypyrazine 2-isoproyl-3-methoxypyrazine 3-octanone fenchone 1-octen-3-one fenchol 1-octen-3-ol trans-2-octen-1-ol 2-methylisoborneol 2,4,6-trichoroanisole geosmin 2,4,6-tribromoanisole pentachloroanisole sum weighted mean extraction temperature and time a

optimized time

optimized temperature

predicted response

relative peak size (RPS)

inverse RPSb

inverse RPS × OPT time

inverse RPS × OPT temperature

70.0 69.0 70.0 65.5 64.5 64 64.5 50.5 69.5 70.0 66.0 56.6 63.0

70.0 67.0 70.0 45.5 69.5 51.5 59.5 49.5 52.5 70.0 70.0 70.0 70.0

23479 9706 79560 1168575 57264 622317 1249123 41463 3965 4360 5283 6995 1701

0.018 0.008 0.063 0.935 0.045 0.498 1.000 0.033 0.003 0.003 0.004 0.005 0.001

53 129 16 1.1 22 2 1 30 315 286 236 178 734 2004 65.3 min

3724 8880 1099 70 1406 128 64 1521 21895 20054 15605 10107 46263 130820 66.7 °C

3724 8622 1099 48 1516 103 59 1491 16539 20054 16550 12500 51404 133714

Determined from the midpoint of the calibration curve. bRounding errors account for minor differences of expressed values. 210 °C at a rate of 10 °C/min, and held for 12 min. The total run time was 51.4 min. The flow rate of ultra-high-purity helium gas was constant at 3 mL/min. The MS source, quadrupole, and transfer line temperatures were set to 230, 150, and 275 °C, respectively. Compound Identification and Elution Profiles. Solutions of all listed compounds were prepared at the approximate midpoint of the calibration range, in a model wine consisting of tartaric acid (0.008 M), potassium hydrogen tartrate (0.011 M), and ethanol (3.0%, v/v). These solutions were used to confirm compound elution times and ion profiles. Samples (10 mL) were transferred to the heater block set at 50 °C with an agitation rate of 250 revolutions per minute (rpm) and equilibrated for 1 min. A divinylbenzene/carboxen/polydimethylsiloxane (DVB/ CAR/PDMS) SPME fiber, which is reported to possess a high level of sensitivity for the compounds of interest,15 had been preconditioned according to the instructions of the manufacturers and was then inserted into the headspace for 20 min at 50 °C, before the insertion into the GC injector. Mass spectral data was collected in both selective ion monitoring (SIM) at ionization voltage of 70 eV and scan mode (m/z 35−350). Final elution profiles were confirmed by matching mass spectral data with the National Institute of Standards and Technology (NIST) mass spectral search program (version 2.0) MS library (version 8.0) and the determination of Kovat’s retention indices (RIs). RIs were checked for each compound using a commercial mixture of n-alkanes (Sigma, Steinheim, Germany), an identical oven ramp profile, and gas flow rates as used for the final analyses. Preparation of Wine Base (WB) and Grape Base (GB) Solutions. A WB solution was prepared by diluting commercial Chardonnay wine (Charles Sturt University Winery 2009) with ultrapure water 1:4 to achieve an ethanol concentration of 3% (v/v) and adjusting the pH to 7.0 by the addition of sodium hydroxide (1 M). A GB solution was prepared by homogenizing Thompson seedless grapes (approximately 2 kg) with an ultra-Turax T25 (Jane & Kunkel Ika-labtechnik) for 2 min. This was diluted 1:5 with ultrapure water to achieve a carbohydrate concentration of 2.5% (w/v), and pH was adjusted to 7.0. Prior to the use of WB and GB for further method development, the absence of the compounds of interest was confirmed by HS-SPME−GC−MS using SIM mode, with all conditions as previously mentioned. Optimization of Sample Extraction Conditions. Optimization of extraction time and temperature was determined using a Box− Behnken experiment designed with a three-level factorial design as previously described.13,16 Compounds were tested in WB or GB with salt at approximately the one-tenth point of the calibration curve. SPME extraction was performed using a DVB/CAR/PDMS fiber. Temperature conditions of 30, 50, and 70 °C with extraction times of 30, 50, and 70 min were examined in a randomized order, with the central condition

limits of quantitation well below the reported sensory threshold for these compounds. However, when multiple analyte responses require optimization, the most favorable handling conditions for individual analytes becomes increasingly convoluted. This method has also allowed for the first reported investigation of compounds that contribute to the off-flavors found in wines made from ripe rot (Colletotrichum acutatum)-affected grapes.



MATERIALS AND METHODS

Chemicals. Analytical-grade chemicals and reagents for the preparation of standard solutions for GC−MS analysis were used without further purification. 2-Isobutyl-3-methoxypyrazine, 2-isopropyl3-methoxypyrazine, 3-octanone, 1-octen-3-one, trans-2-octen-1-ol, 1-octen-3-ol, 2-methylisoborneol, 2,4,6-trichoroanisole, geosmin, 2,4,6-tribromoanisole, methyl isobutyl ketone, 2-octanol, and geosmin-d3 were obtained from Sigma-Aldrich (St. Louis, MO). Fenchone and fenchol were obtained from Fulka (Buchs, Switzerland). 2-Methyld3-isoborneol, 2,4,6-trichloroanisole-d5, 2,4,6-tribromoanisole-d5, and pentachloroanisole-d3 were obtained from CDN Isotopes (PointeClaire, Quebec, Canada). Ethanol (VWR Prolabo, Fontenay Sous Bois, France), L-(+) tartaric acid (Sigma, Steinheim, Germany), and potassium hydrogen tartrate (BHD Chemicals, Ltd., Poole, U.K.) were used for preparation of model wine base solutions. Deionized water was obtained from a Milli-Q mixed bed resin system (18 MΩ/cm, 25 °C). Instrumentation. An Agilent 7890A gas chromatograph, equipped with a Gerstel multipurpose sampler with automated SPME capability and Peltier cooled sample tray, interfaced to an Agilent 5975C triple axis mass detector, was used for confirmation of compound identity, method development, and final sample analysis. MSD Chemstation E.02.00.493 (Agilent Technologies, Ltd.) and NIST MS Search 2.0, version 2008, were used to control the instrument and for mass spectra assessment. Prepared samples were placed into the Peltier cooler tray set at 8 °C until analysis, whereupon vials were transferred to a heater block with a 2 min pre-incubation time before insertion of the SPME fiber. A fused silica capillary column (DB-Waxetr, 60 m × 0.25 mm inner diameter, 0.25 μm film thickness, Agilent Technologies, Santa Clara, CA) was used for compound separation by GC. The injector block was fitted with a 2 mm internal diameter borosilicate liner (SGE), and the injector temperature set to 260 °C in splitless mode. Fibers were inserted into the injector for 1 min, withdrawn, and injected into a second injector set at 270 °C with a 50:1 split for 10 min with a 15 mL/min purge flow to clean the fiber, prior to the next sample analysis. The oven temperature program commenced at 40 °C for 3 min, increased to 115 °C at a rate of 5 °C/min, increased to 166 °C at a rate of 3 °C/min, a final increase to 2878

DOI: 10.1021/jf505444r J. Agric. Food Chem. 2015, 63, 2877−2885

Article

Journal of Agricultural and Food Chemistry

Figure 1. Response surface modeling of quantification ion counts to determine the optimized experimental conditions was undertaken in accordance with eq 1 for each analyte of interest. An asterisk above the coefficient indicates a significant term in the regression model (p < 0.05). Compound numbers: (1) 2-isopropyl-3-methoxypyrazine, (2) 2-isobutyl-3-methoxypyrazine, (3) 3-octanone, (4) fenchone, (5) 1-octen-3-one, (6) trans-2-octen1-ol, (7) fenchol, (8) 1-octen-3-ol, (9) methylisoborneol, (10) 2,4,6-trichloroanisole, (11) geosmin, (12) 2,4,6-tribromoanisole, and (13) pentachloroanisole. where Y is the predicted response for each compound, X1 is the extraction temperature, X2 is the extraction time, b0 is the constant term, b1 and b2 are the coefficients for the linear terms, b12 is the coefficient for the interactions, b11 and b22 are the coefficients for the squared terms or variables (time and temperature) for each compound model, and ε is the residual. The coefficients were calculated using regression analysis, and the significance of each variable was determined using a two-tailed t test. The coefficients bn and indication of their significance are shown in Figure 1. The response surface curve of geosmin for the extraction conditions is shown in Figure 2. A modified desirability method for the optimization of multiple response variables was used, in which the importance of

repeated 3 times to determine the extraction repeatability. Thus, the experimental design was constructed with independent variables at levels (−1, 0, and 1) for temperature (30, 50, and 70 °C) and time (30, 50, and 70 min). Compound quantitation ion results for each compound were exported to MATLAB R2007a (The Mathworks, Natick, MA), and extraction conditions were modeled using response surface methodology. The predicted maximum quantitation ion response at optimized conditions for each compound was determined from the response surface model using a full quadratic model to account for constant, linear, and squared terms17

Y = b0 + b1X1 + b2X 2 + b12X1X 2 + b11X12 + b22X 2 2 + ε

(1) 2879

DOI: 10.1021/jf505444r J. Agric. Food Chem. 2015, 63, 2877−2885

Article

Journal of Agricultural and Food Chemistry

Calibration Curves. Calibration solutions were prepared in either WB or GB by diluting a mix of the pure compounds to seven concentrations over the analytical range (Table 3) and internal standard (IS), as designated in Table 4, and saturated with NaCl (30%). Triplicate analysis was conducted for each calibration level in both WB and GB, and SPME extraction was conducted as previously mentioned. Limit of Detection (LOD) and Limit of Quantitation (LOQ). WB and GB spiked with low levels of target compounds were used to establish the LOD and LOQ of the overall method. The LOD (concentration for signal/noise = 3) and the LOQ (concentration for signal/noise = 10) were manually calculated from the ratio of the peak heights to the average noise before and after each peak. These data are listed in Table 3. Recovery Experiments in Red and White Wine and Grapes. Recovery experiments were conducted in triplicate in wines (diluted 1:4) and grape homogenates (diluted 1:5) adjusted to a pH of 7.0 using 1 M NaOH. Compounds were spiked in these solutions at approximately one-fifth and one-tenth of the calibration curve. NaCl (3.0 g) was placed into each 20 mL sample vial prior to introducing 10 mL of sample. Compound recoveries are reported as a percentage of the spiked quantity determined from the calibration curves for each analyte.18 Repeatability and Reproducibility. To evaluate repeatability, five white wines (Chardonnay) and five homogenate Thompson seedless grape samples were spiked with a concentrated mix of target compounds, as described above. To evaluate reproducibility spiked, samples were prepared and analyzed over a period of 5 days with results expressed as the coefficient of variation (% CV).18 Analysis of Healthy and Infected Grape and Wine Samples. Healthy and Infected Grape Samples. Bunches of Shiraz grapes were collected from a vineyard in the Riverina, New South Wales, Australia, during the 2012 vintage. Fruit was handpicked and sorted into two separate batches according to a visual assessment for the presence of obvious fungal bunch rots. On the basis of a visual examination, gray mold or B. cinerea19 was the predominant bunch rot disease in these grapes. Cabernet Sauvignon bunches were similarly collected from a commercial vineyard located in the Hunter Valley, New South Wales, Australia, during the 2013 vintage and sorted into two batches according to the presence or absence of obvious mold growth. On the basis of a visual examination, ripe rot or C. acutatum20 was the predominant bunch rot disease in these grapes. Triplicate samples from each batch of grapes were prepared the same as GB samples. Wine Made from Grapes Affected with Bunch Rot Molds. Grapes were obtained from a single block of Semillon vines located in the Riverina, New South Wales, Australia, during the 2011 vintage.

Figure 2. Response surface curve in WB for geosmin of varying sample incubation temperatures and fiber exposure time. specific responses were down-weighted.14 Briefly, relative peak size (RPS) coefficients for each compound were calculated from the inverse of the largest predicted maximum quantitation ion peak divided by each target compound peak area in the compound list (Tables 1 and 2). Weighted mean extraction times and temperatures were then calculated from the product of inverse RPS and optimized time or optimized temperature for the compounds. This enabled a comparison of the relative extraction and detection efficiencies for all compounds and assisted in minimizing the loss of analytical sensitivity for specific compounds that give low peak areas compared to compounds with large peak areas. Effect of Salting on Extraction Efficiency. Duplicate samples were prepared in either a WB or GB with or without salt. Samples were spiked with compounds to produce a concentration of one-tenth of the maximum calibration point. The effects of salting on the ionic strength were determined by placing 3.0 g of NaCl into each 20 mL sample vial prior to introducing 10 mL of sample. SPME extraction was conducted at final SPME optimized extraction conditions [for GB, 65 °C for 67 min (Table 1) and, for WB, 44 °C for 65 min (Table 2)] using a DVB/CAR/ PDMS fiber with quantitation ions counted in SIM mode. Quantitation ion counts for target compounds in WB or GB solution with or without the salt are presented in Figure 3.

Table 2. Optimized Extraction Times and Temperatures for Target Compounds in Winea compound 2-isobutyl-3-methoxypyrazine 2-isoproyl-3-methoxypyrazine 3-octanone fenchone 1-octen-3-one fenchol 1-octen-3-ol trans-2-octen-1-ol 2-methylisoborneol 2,4,6-trichoroanisole geosmin 2,4,6-tribromoanisole pentachloroanisole sum weighted mean extraction temperature and time a

optimized time

optimized temperature

predicted response

relative peak size (RPS)

inverse RPSb

inverse RPS × OPT time

inverse RPS × OPT temperature

70 70 70 30 70 70 37 70 70 30 70 31 30

70 70 70 30 70 70 30 70 70 70 70 46 70

13035 18217 92474 1335750 66676 247942 1301970 208108 15421 5791 20984 8835 8954

0.188 0.054 0.275 1.000 0.199 0.738 0.899 0.619 0.046 0.017 0.062 0.026 0.0267

3 18 4 1.0 5 1.3 1.1 2 22 58 16 38 37 205 43.9 min

180 1290 254 30 352 95 41 112 1524 1739 1120 1178 1125 9040 65.2 °C

180 1290 254 30 352 95 33 112 1524 4058 1120 1748 2624 13419

Determined from the midpoint of the calibration curve. bRounding errors account for minor differences of expressed values. 2880

DOI: 10.1021/jf505444r J. Agric. Food Chem. 2015, 63, 2877−2885

Article

Journal of Agricultural and Food Chemistry

Figure 3. Effects of salting in WB or GB solution on the quantitation ion counts for all target compounds. The WB or GB solution was spiked at a concentration at one-tenth of the maximum calibration point. GB, grape base; WB, wine base.

Table 3. Details of Target Compound Identification, Calibration Curves, and Method Evaluations number

compound

IS

quantitation iona m/z

qualifier ion m/z (%)

1

2-isobutyl-3-methoxypyrazine

IS1

124

2

2-isoproyl-3-methoxypyrazine

IS1

137

3

3-octanone

IS2

57

4

fenchone

IS2

81

5

1-octen-3-one

IS2

55

6

trans-2-octen-1-ol

IS3

57

7

fenchol

IS3

81

8

1-octen-3-ol

IS3

57

9

2-methylisoborneol

IS4

107

10

2,4,6-trichoroanisole

IS5

195

11

geosmin

IS6

112

12

2,4,6-tribromoanisole

IS7

346

13

pentachloroanisole

IS8

280

151 (20) 94 (21) 152 (31) 124 (41) 72 (73) 71 (63) 99 (75) 69 (44) 152 (17) 70 (95) 83 (16) 97 (25) 68 (31) 55 (46) 80 (61) 121 (14) 72 (18) 85 (11) 99 (6) 135(12) 95 (5) 210 (67) 212 (65) 111 (24) 125 (17) 344 (98) 329 (73) 265 (95) 237 (92) 267 (81)

a

calibration range (ng/L) 4.4−282 2.8−179 20−4010

33−26760 2−980

3−2480 34−6905 35−7000

3.1- 200 0.58−37.3 1.25- 80 0.82−52.5 3.47−55.5

sample typeb

r2

LOD (ng/L)

LOQ (ng/L)

WB GB WB GB WB GB

0.996 0.994 0.986 0.995 0.998 0.988

0.2 0.1 0.1 0.05 10 7

WB GB WB GB

0.993 0.981 0.999 0.981

GB WB WB GB WB GB WB GB WB GB WB GB WB GB WB GB

repeatability (% CV)

reproducibility (% CV)

0.6 0.3 0.3 0.2 20 20

7.4 9.7 6.8 5.1 9.6 11

8.7 8.9 8.1 7.2 10.2 9.8

1 1 0.3 0.3

4 4 1 0.8

9.8 9.6 7.1 8.7

4.9 5.7 8.8 8.6

0.993 0.991 0.993 0.997 0.997 0.996

1 1 1 1 1.5 1.5

3 3 3 3 3.5 3.5

9.8 9.6 6.4 8.1 5 9.4

4.9 5.7 6.4 6.7 8.0 6.1

0.992 0.996 0.998 0.994 0.955 0.997 0.997 0.997 0.919 0.991

3.1 3.1 0. 10 0.10 1.00 0.42 0. 2 0.1 0.12 0.5

6.2 6.2 0. 30 0. 30 2.5 1.5 0.8 0.3 0.4 1.8

4.3 8.6 8.5 11.1 5.2 7.4 9.7 8.7 7.9 10.3

3.6 6.7 7.14 5.5 5.17 8.7 8.9 1.9 8.04 10.9

As a proportion of the quantitation ion. bWB, wine base; GB, grape base.

Whole bunches were randomly selected along vine rows from both sides of the vine and sorted into two groups based on the visible level of infection. The control group had no perceived infection (i.e., clean fruit or 0% bunch infected), while the second treatment had high level infection (up to 45−50% of the bunch infected). These treatments were juiced and fermented separately to finished wine.21 A further

experimental wine made from Cabernet Sauvignon grapes that investigated the effect of ripe rot in an earlier study22 was also used in this investigation. Three commercially made wines were selected for analysis. These wines were a 2003 Semillon wine from the Hunter Valley, New South Wales, Australia, a 1978 Cabernet Sauvignon wine, and a 1987 vintage 2881

DOI: 10.1021/jf505444r J. Agric. Food Chem. 2015, 63, 2877−2885

Article

Journal of Agricultural and Food Chemistry Table 4. Details of IS Identification and Concentration

a

number

compound

retention time (min)

quantitation iona m/z

qualifier ion m/z (%)

concentration (ng/L)

IS1 IS2

2-isobutyl-3-methoxy-d3-pyrazine methyl isobutyl ketone

24.0 9.5

127 58

25 97

IS3

2-octanol

21.3

45

IS4

2-methyl-d3-isoborneol

26.0

110

IS5 IS6

2,4,6-trichloroanisole-d5 geosmin-d3

33.9 34.4

215 115

IS7 IS8

2,4,6-tribromoanisole-d5 pentachloroanisole-d3

44.6 46.2

349 283

154 (15) 85 (96) 100 (88) 55 (36) 56 (12) 70 (10) 138 (24) 55 (10) 217 (94) 129 (12) 128 (18) 347 (36) 285 (57)

60.2

100 20 30 20 20

As a proportion of the quantitation ion.

Table 5. Recovery of Two Different Addition Concentrations of Compounds of Interest in Wine and Grape Juice recovery (%)a

a

recovery (%)b

compound

red wine

white wine

grape

red wine

white wine

grape

2-isobutyl-3-methoxypyrazine 2-isoproyl-3-methoxypyrazine 3-octanone fenchone 1-octen-3-one trans-2-octen-1-ol fenchol 1-octen-3-ol 2-methylisoborneol 2,4,6-trichoroanisole geosmin 2,4,6-tribromoanisole pentachloroanisole

104 87 103 83 100 97 103 100 99 118 99 99 97

125 108 108 92 83 97 95 95 103 95 102 100 91

116 96 120 100 101 92 104 95 104 107 100 113 115

125 111 108 105 94 123 131 118 100 94 116 100 115

116 99 113 92 90 107 117 108 104 106 106 98 104

89 111 121 115 98 95 104 96 102 105 104 101 108

Recoveries of one-tenth addition of the maximum calibrate. bRecoveries of one-fifth addition of the maximum calibrate.

wine made from Botrytis Muscat of Alexandria grapes from the Riverina, New South Wales, Australia. Triplicate samples from each batch of wine were prepared the same as WB samples for the quantitation of the compounds of interest.

interaction terms are modeled for a two-factor (time and temperature) model. The magnitude of each b coefficient is dependent upon the analyte signal, and the significance for each coefficient can be determined using a t test. Inspection of the t test values for the b coefficients for each compound in WB infers that an optimized sample incubation temperature is a more important factor than fiber exposure time to achieve maximum analytical sensitivity (Figure 1). In GB, fiber exposure time and sample incubation temperature are both important factors to optimize for the analysis of some compounds. The diminished importance of fiber exposure time for optimizing WB analysis may reflect a modified polarity of the solvent base because of the presence of ethanol, which has increased the volatility of the analytes and, hence, headspace concentration. Consequently, equilibrium between the WB headspace concentration and binding of these compounds onto the fiber, once exposed to the sample, occurs more rapidly than in GB. The analysis of alcoholic beverages can be challenging for HS-SPME analysis because the competitive adsorption between ethanol and other interesting volatile compounds onto the SPME fiber may cause loss of analytical sensitivity. This is due to the relative concentration of ethanol far exceeding that of other compounds leading to fiber saturation.26 Dilution of samples to a lower ethanol concentration limits competitive binding of ethanol to the fiber, allowing for a lower detection limit for interesting compounds to be achieved.27



RESULTS AND DISCUSSION Optimization. Optimization of sample treatment parameters is arguably the most important aspect of method development for sensitive, accurate, and precise quantitation of volatile compounds.23,24 Optimized sample handling parameters were determined using response surface modeling and a desirability function, in which the importance of sample conditions for compounds with a high signal response relative to those with a low signal, is down-weighted by the calculation of a relative response factor.14 Weighted mean extraction times and temperatures were calculated from the product of inverse RPS and optimized time or optimized temperature for the compounds of interest (Tables 1 and 2 for WB and GB, respectively). This approach for HS-SPME optimization enabled lower limits of quantitation to be achieved than previously reported for the compounds of interest.25 Response surface modeling to determine the optimized experimental conditions was undertaken in accordance with eq 1 for each analyte of interest. Consequently, for each analyte and solvent system (WB or GB), there exists six b coefficients that correspond to the fitted response when linear, squared, and 2882

DOI: 10.1021/jf505444r J. Agric. Food Chem. 2015, 63, 2877−2885

Article

Journal of Agricultural and Food Chemistry Table 6. Sensory Threshold and LOD of Target Compounds in the Wine Matrix target compound 2-isobutyl-3-methoxypyrazine 2-isoproyl-3-methoxypyrazine 3-octanone fenchone 1-octen-3-one trans-2-octen-1-ol fenchol 1-octen-3-ol 2-methylisoborneol 2,4,6-trichoroanisole geosmin 2,4,6-tribromoanisole pentachloroanisole a

sensory threshold in wine (ng/L) 1−1631 231

7025 and 15 NR 4000025 5525 325 5025 2−525 1000025

odor

LOD (ng/L) present study

LOD (ng/L) previously reported

vegetal, green pepper muddy

0.1 0.2 10 1 0.3 1 1 1.5 3.1 0.1 1 0.2 0.12

3.125 3.125 NRa 1.825 0.7532 NR 1.625 125 5.1 0.05,33 0.1,34 2,35 0.8,36 and 0.325 0.535 and 1.525 0.09,33 0.1,34 and 0.225 0.525

muddy fungus muddy fungus muddy mold muddy mold dust

NR = not reported.

multiple response factors by weighting specific variables according to a global desirability function. Analysis of Healthy and Infected Grape and Wine Samples. The method has been applied to the analysis of infected grapes and wines. The growth of C. acutatum and B. cinerea on grapes is clearly associated with elevated levels of 2-methylisoborneol, geosmin, and 1-octen-3-one (Table 7). These compounds are also present in wine made from grapes infected with these fungal pathogens and are reported above in sensory thresholds. Previously reported levels of geosmin in grapes infected with B. cinerea support our findings for the production of these compounds in grapes infected with this mold.5−8 As shown in Table 7, geosmin was measured at levels higher than the reported sensory perception threshold in Semillon wine made from grapes infected with B. cinerea. Similarly, in Cabernet Sauvignon wines made from grapes infected with C. acutatum, 2-methylisoborneol levels were higher than the sensory perception threshold. It appears that 2-methylisoborneol contributes to the off-flavors found in wines made from ripe-rot-affected grapes, while geosmin has a greater role in contributing to the off-flavors found in wines made from gray-mold-affected grapes. While previous studies have examined off-flavors found in wine made from grape infected with gray mold (B. cinerea) and Penicillium rots8 and sour rot,11 as far as we are aware, this is the first study that has attempted to identify the chemical nature of the offflavors found in wine made from grapes affected with ripe rot (C. acutatum). Dilution of wines to an approximate ethanol concentration of 3% (v/v), increasing ionic strength in SPME fiber exposure, and sample extraction conditions have enabled a sensitive HS-SPME method to be devised. This method is also suitable for grape samples and has excellent analytical reproducibility and repeatability for volatile fungal metabolites. It has been validated across a large range of analytical levels and has a higher sensitivity than previously reported. The methods described enabled the determination of off-flavors in wine and grapes at nanograms per liter levels and achieve a linear, specific, accurate, and repeatable method in a single run. The levels of detection are below the reported olfactory thresholds of perception.25 The method had good recoveries (83−131%), with acceptable repeatability and reproducibility. The developed method requires little or no sample preparation, is easily automated and reliable, and has no requirement for solvent extraction. Because of its simplicity, the method is suitable for the analysis of fungal off-flavor

The incorporation of high ionic strength solutes into the sample matrix is frequently employed for the analysis of volatile compounds with HS-SPME methods,28,23 with sodium chloride or sodium sulfate most commonly used. An increased ionic strength sample matrix often results in higher analyte signal responses with HS-SPME methods. This is because an increased volatility of specific compounds arises from the altered solubility of polar compounds.29 Partition coefficients for the gaseous phase are therefore improved, and consequently, greater analyte quantities are absorbed by the SPME fiber. This phenomenon was observed for all compounds of interest in the present investigation (Figure 3). Calibration Curves and Method Evaluation. Calibration curves prepared in a diluted wine or grape homogenate matrix were constructed from the mean quantitation ion count of three replicate analyses, using a mix of naturally occurring and deuterated analogues as ISs. Good correlation coefficients (r2 > 0.981) for all compounds (Table 3) and recovery experiments (Table 5) indicate that matrix interferences were relatively minor. Recovery values expressed as the coefficient of variation ranged from 83 to 131 in WB and from 89 to 121 in GB and indicate that the wine matrix had a greater impact upon the analysis than the grape base. This is not unexpected because, even with dilution, the ethanol concentration in wine was still relatively high compared to the analyte concentration. Additionally, the wine matrix is comprised of a significantly greater number of volatile compounds arising from yeast metabolism, which may lead to competitive interactions on the fiber surface and contribute to the matrix effect. Recovery values for spiked red wine also indicate possible interactions of the matrix with analytes to a greater extent than in the white wine (Table 5). It is possible that the greater concentration of phenolic compounds in red wine may have caused interference with recovery. However, values from two spiked additions were acceptable for analytical use (Table 5). The limits of detection and quantitation for all analytes (Table 3) are lower than levels previously reported for wine and grape samples.30 These values for are remarkably lower than those found in the literature (Table 6) and, in any case, are well below the reported sensory threshold of these components in wine. To the authors knowledge, no prior single GC−MS method has been developed that enable rapid and sensitive analysis of the diverse compounds elicited from mold growth that impart earthy taint aromas in one single GC−MS run. Our approach demonstrates the value of using optimization of 2883

DOI: 10.1021/jf505444r J. Agric. Food Chem. 2015, 63, 2877−2885

Article

Samples: Sh, Shiraz; Semi, Semillon; and Cab, Cabernet Sauvignon. Treatment: Con, control; Gray, gray mold (B. cinerea affected); and Ripe, ripe rot (C. acutatum affected). ND, not detected; BLQ, below limit of quantitation. Mean ± standard deviation of triplicate determinations are reported. Mean values within grape and experimental wine sample columns (healthy and infected) followed by different letters (a and b) are significantly different at p < 0.05.

ND 15 ± 2 17 ± 3 ND BLQ BLQ 1657 ± 132 ND ND ND BLQ ND ND 1±0 4±1 Muscat Semi Cab

ND BLQ BLQ

1±1a ND 112 ± 8 b ND ND 140 ± 1 a 38 ± 13 138 ± 4 a ND ND ND ND ND 121 ± 31 ND ND ND ND ND ND ND ND 1±0a 1±0a Semi/Con Semi/Gray Cab/Con Cab/Ripe

compounds in wine and grapes and may have applicability for the analysis of other fruit and beverage products impacted by fungal growth.



ASSOCIATED CONTENT

S Supporting Information *

Response surface curves for all of the target compounds in WB (Figure S1) and GB (Figure S2) for varying sample incubation temperatures and fiber exposure times. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*Telephone: +61-2-6933-4047. E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS Kyra Boissery is thanked for collection of berry samples and production of Semillon wine affected with B. cinerea. Navideh Sadoughi was a recipient of a Charles Sturt University International Postgraduate Research Scholarship (IPRS).



ABBREVIATIONS USED WB, wine base; GB, grape base; IS, internal standard; LOD, limit of detection; LOQ, limit of quantification; CV, coefficient of variation



REFERENCES

(1) Steel, C. C.; Blackman, J. W.; Schmidtke, L. M. Grapevine bunch rots: Impacts on wine composition, quality, and potential procedures for the removal of wine faults. J. Agric. Food Chem. 2013, 61, 5189−5206. (2) Peña-Neira, A.; Fernández de Simón, B.; García-Vallejo, M. C.; Hernández, T.; Cadahía, E.; Suarez, J. A. Presence of cork-taint responsible compounds in wines and their cork stoppers. Eur. Food Res. Technol. 2000, 211, 257−261. (3) Chatonnet, P.; Bonnet, S.; Boutou, S.; Labadie, M.-D. Identification and responsibility of 2,4,6-tribromoanisole in musty, corked odors in wine. J. Agric. Food Chem. 2004, 52, 1255−1262. (4) Buser, H. R.; Zanier, C.; Tanner, H. Identification of 2,4,6trichloroanisole as a potent compound causing cork taint in wine. J. Agric. Food Chem. 1982, 30, 359−362. (5) La Guerche, S.; Chamont, S.; Blancard, D.; Dubourdieu, D.; Darriet, P. Origin of (−)-geosmin on grapes: On the complementary action of two fungi, Botrytis cinerea and Penicillium expansum. Antonie van Leeuwenhoek 2005, 88, 131−139. (6) Morales-Valle, H.; Silva, L. C.; Paterson, R. R.; Oliveira, J. M.; Venancio, A.; Lima, N. Microextraction and gas chromatography/mass spectrometry for improved analysis of geosmin and other fungal “off” volatiles in grape juice. J. Microbiol. Methods 2010, 83, 48−52. (7) Bañeras, L.; Trias, R.; Godayol, A.; Cerdán, L.; Nawrath, T.; Schulz, S.; Anticó, E. Mass spectrometry identification of alkyl-substituted pyrazines produced by Pseudomonas spp. isolates obtained from wine corks. Food Chem. 2013, 138, 2382−2389. (8) La Guerche, S.; Dauphin, B.; Pons, M.; Blancard, D.; Darriet, P. Characterization of some mushroom and earthy off-odors microbially induced by the development of rot on grapes. J. Agric. Food Chem. 2006, 54, 9193−9200. (9) Martins, J.; Esteves, C.; Limpo-Faria, A.; Barros, P.; Ribeiro, N.; Simões, T.; Correia, M.; Delerue-Matos, C. Analysis of six fungicides and one acaricide in still and fortified wines using solid-phase microextraction−gas chromatography/tandem mass spectrometry. Food Chem. 2012, 132, 630−636. (10) Antalick, G.; Perello, M.-C.; de Revel, G. Development, validation and application of a specific method for the quantitative determination

a

ND ND 21 ± 2 ND BLQ ND 75 ± 21 ND ND ND BLQ ND ND 17 ± 1 87 ± 18

ND ND ND ND ND ND ND 53 ± 8

BLQ BLQ ND ND

a b a b 1 5 1 3 ± ± ± ± 9 51 5 27

ND ND ND ND

ND ND ND ND ND ND ND ND 0a 10 b 0a 1b ± ± ± ± 12 118 7 24 ND ND ND ND ND 127 ± 13 ND 156 ± 27

Grape BLQ 935 ± 10 a BLQ 5647 ± 74 b BLQ 1087 ± 2 a BLQ 5735 ± 52 b Experimental Wine ND 1186 ± 24 a ND 1416 ± 14 b ND ND ND 166 ± 12 Commercial Wine 2542 ± 235 ND ND ND ND 13 ± 4 a a a a 10 21 10 13 ± ± ± ± 400 425 185 195 ND 2567 ± 48 ND 1190 ± 46 BLQ BLQ BLQ BLQ BLQ BLQ BLQ BLQ ND ND 1±0a 1±0a ND ND 2±0a 2±0a Sh/Con Sh/Gray Cab/Con Cab/Ripe

2-methylisoborneol 2,4,6-trichoroanisole 1-octen-3-ol fenchol trans-2octen-1-ol 1-octen3-one fenchone 2-isobutyl-32-isoproyl-3methoxypyrazine methoxypyrazine 3-octanone sample/ treatments

Table 7. Off-Flavor Compounds in Bunch-Rot-Affected Grape and Wine Samples and Commercial Winesa

geosmin

2,4,6-tribromoanisole pentachloroanisole

Journal of Agricultural and Food Chemistry

2884

DOI: 10.1021/jf505444r J. Agric. Food Chem. 2015, 63, 2877−2885

Article

Journal of Agricultural and Food Chemistry of wine esters by headspace-solid-phase microextraction−gas chromatography−mass spectrometry. Food Chem. 2010, 121, 1236−1245. (11) Barata, A.; Campo, E.; Malfeito-Ferreira, M.; Loureiro, V. l.; Cacho, J.; Ferreira, V. Analytical and sensorial characterization of the aroma of wines produced with sour rotten grapes using GC−O and GC−MS: Identification of key aroma compounds. J. Agric. Food Chem. 2011, 59, 2543−2553. (12) Aberl, A.; Coelhan, M. Determination of sulfur dioxide in wine using headspace gas chromatography and electron capture detection. Food Addit. Contam. 2012, 30, 226−233. (13) Ryan, D.; Watkins, P.; Smith, J.; Allen, M.; Marriott, P. Analysis of methoxypyrazines in wine using headspace solid phase microextraction with isotope dilution and comprehensive two-dimensional gas chromatography. J. Sep. Sci. 2005, 28, 1075−1082. (14) Derringer, G.; Suich, R. Simultaneous optimization of several response variables. J. Qual. Technol. 1980, 12, 214−219. (15) Ezquerro, O.; Tena, M. T. Determination of odour-causing volatile organic compounds in cork stoppers by multiple headspace solid-phase microextraction. J. Chromatogr. A 2005, 1068, 201−208. (16) Rebiere, L.; Clark, A. C.; Schmidtke, L. M.; Prenzler, P. D.; Scollary, G. R. A robust method for quantification of volatile compounds within and between vintages using headspace-solid-phase microextraction coupled with GC−MSApplication on Semillon wines. Anal. Chim. Acta 2010, 660, 149−157. (17) Leardi, R. Experimental design in chemistry: A tutorial. Anal. Chim. Acta 2009, 652, 161−172. (18) Lizarraga, E.; Irigoyen, A.; Belsue, V.; Gonzalez-Penas, E. Determination of chloroanisole compounds in red wine by headspace solid-phase microextraction and gas chromatography−mass spectrometry. J. Chromatogr. A 2004, 1052, 145−149. (19) Elmer, P. G.; Michailides, T. Epidemiology of Botrytis cinerea in orchard and vine crops. In Botrytis: Biology, Pathology and Control; Elad, Y., Williamson, B., Tudzynski, P., Delen, N., Eds.; Springer: Dordrecht, Netherlands, 2007; pp 243−272. (20) Whitelaw-Weckert, M. A.; Curtin, S. J.; Huang, R.; Steel, C. C.; Blanchard, C. L.; Roffey, P. E. Phylogenetic relationships and pathogenicity of Colletotrichum acutatum isolates from grape in subtropical Australia. Plant Pathol. 2007, 56, 448−463. (21) Boissery, K. Effect of bunch rot on the berry microflora and composition of semillon wine. Honors Thesis, Charles Sturt University, Wagga Wagga, Australia, 2011. (22) Meunier, M.; Steel, C. C. Effect of Colletotrichum acutatum ripe rot on the composition and sensory attributes of cabernet sauvignon grapes and wine. Aust. J. Grape Wine Res. 2009, 15, 223−227. (23) Rocha, S.; Ramalheira, V.; Barros, A.; Delgadillo, I.; Coimbra, M. A. Headspace solid phase microextraction (SPME) analysis of flavor compounds in wines. Effect of the matrix volatile composition in the relative response factors in a wine model. J. Agric. Food Chem. 2001, 49, 5142−5151. (24) Alpendurada, D. F. M. Solid-phase microextraction: A promising technique for sample preparation in environmental analysis. J. Chromatogr. A 2000, 889, 3−14. (25) Boutou, S.; Chatonnet, P. Rapid headspace solid-phase microextraction/gas chromatographic/mass spectrometric assay for the quantitative determination of some of the main odorants causing offflavours in wine. J. Chromatogr. A 2007, 1141, 1−9. (26) Khio, S. W.; Cheong, M. W.; Zhou, W.; Curran, P.; Yu, B. Characterization of the volatility of flavor compounds in alcoholic beverages through headspace solid-phase microextraction (HS-SPME) and mathematical modeling. J. Food Sci. 2012, 77, C61−C70. (27) Stackler, B.; Christensen, E. N. Quantitative determination of ethanol in wine by gas chromatography. Am. J. Enol. Vitic. 1974, 25, 202−207. (28) Howard, K. L.; Mike, J. H.; Riesen, R. Validation of a solid-phase microextraction method for headspace analysis of wine aroma components. Am. J. Enol. Vitic. 2005, 56, 37−45. (29) Fries, E. Determination of benzothiazole in untreated wastewater using polar-phase stir bar sorptive extraction and gas chromatography− mass spectrometry. Anal. Chim. Acta 2011, 689, 65−68.

(30) Cortada, C.; Vidal, L.; Canals, A. Determination of geosmin and 2methylisoborneol in water and wine samples by ultrasound-assisted dispersive liquid−liquid microextraction coupled to gas chromatography−mass spectrometry. J. Chromatogr. A 2011, 1218, 17−22. (31) Allen, M. S.; Lacey, M. J.; Harris, R. L. N.; Brown, W. V. Contribution of methoxypyrazines to sauvignon blanc wine aroma. Am. J. Enol. Vitic. 1991, 42, 109−112. (32) Culleré, L.; Cacho, J.; Ferreira, V. Validation of an analytical method for the solid phase extraction, in cartridge derivatization and subsequent gas chromatographic−ion trap tandem mass spectrometric determination of 1-octen-3-one in wines at ng L−1 level. Anal. Chim. Acta 2006, 563, 51−57. (33) Schmarr, H.-G.; Koschinski, S.; Sang, W.; Slabizki, P. Trace level analysis of corky off-flavor compounds: Development of a new analytical method based on solid phase extraction and analysis by multidimensional gas chromatography with mass spectrometric detection. J. Chromatogr. A 2012, 1226, 96−102. (34) Slabizki, P.; Schmarr, H.-G. Analysis of corky off-flavour compounds at ultra trace level with multidimensional gas chromatography−electron capture detection. J. Chromatogr. A 2013, 1271, 181− 184. (35) Weingart, G.; Schwartz, H.; Eder, R.; Sontag, G. Determination of geosmin and 2,4,6-trichloroanisole in white and red Austrian wines by headspace SPME−GC/MS and comparison with sensory analysis. Eur. Food Res. Technol. 2010, 231, 771−779. (36) Ö zhan, D.; Anli, R. E.; Vural, N.; Bayram, M. Determination of chloroanisoles and chlorophenols in cork and wine by using HS-SPME and GC−ECD detection. J. Inst. Brew. 2009, 115, 71−77.

2885

DOI: 10.1021/jf505444r J. Agric. Food Chem. 2015, 63, 2877−2885