Chemical characterization of phenols and taste-activesubstances in wine using liquid chromatography - massspectrometry

Chemical characterization of phenols and taste-active

substances in wine using liquid chromatography - mass

spectrometry

By

Erika Johansson

Candidate for Degree of Bachelor of Science School of Science and Technology

Örebro University May 2018

Abstract

Wine characterization of taste and odor of a wine is important to keep up with consumers interest and ensure quality of wine. In Sweden, wine is sent for destruction every year and one of the reasons is that some wines have a deviant taste and odor profiles than described on the label. The aim for this project was to perform a chemical analysis of wine to identify different taste-active compounds to connect to different taste sensations such as sweetness, acidity, bitterness and the mouthfeel of astringency. Wine profile patterns were also created to distinguish between the composition of these compounds in sixteen different wines of different colors. The targeted compounds were fifteen compounds which included phenolic compounds, amino acids and organic acids. The analysis were performed by liquid chromatography mass spectrometry. A literature survey was performed for selection of an appropriate method for the analysis of the targeted compounds. Fourteen out of fifteen compounds were detected in the wine samples. Fifteen out of sixteen wines contained tartaric acid, while proline was found in all wines and vanillin only in one sparkling wine. The red wines showed higher concentrations in compounds connected to bitterness and astringency and sparkling wine contained higher amounts of the compounds connected to sweet and sour taste. The wine profile patterns were compared to see both similarities and differences within the same color wines.

Table of content

1. Introduction

In Sweden, large quantities of wine are sent for destruction every year. One reason for this is that some wines have a deviant taste and odor profile compared to the description on the label. If the actual taste and smell of a wine is not the same as promoted on its label it could be less attractive to consumers with difficulties being sold. These wines could still be of good quality although the profiles deviate. With a new classification with regards to the taste and odor these wines could be re-labeled and sold on the market again. This would decrease the destruction of perfectly good quality wine and give the consumers what is anticipated favoring a more sustainable wine industry.

This project is one of two projects performing a chemical characterization of substances contributing to the aroma and taste of different wine profiles. Gas chromatography will be performed to screen for volatile substances connecting to smell of the wine profiles and in this project liquid chromatography will be used to identify taste-active compounds.

1.2 Wine characterization

There are different aspects of characterization of wine. Wine tasting is a traditional method to taste and smell the aroma and flavor of wine. Sensory techniques are used to minimize the bias that can occur during wine tasting and follows different protocols to analyze and statistically

interpret data of the characteristics of different wine 1. Sensory analysis is used to

characterize sensory properties such as aroma and flavor of wine, develop new styles and approaches of wine with consumers’ interest, to ensure quality and lack of spoilage and to characterize different origins of wine 2.

When characterizing for aroma and flavor compounds in wine, different chemical analysis can be performed. Gas mass spectrometry (GC-MS) and gas chromatography-olfactometry (GC-O) can be used to identify volatile compounds in wine corresponding to

aroma 3. Identifying taste-active compounds in wine, such as phenols, organic acids and

amino acids could be performed by liquid chromatography-mass spectrometry (LC-MS) or

1.3 Taste-active compounds

Receptors within the taste buds on the tongue generate different gustatory (taste) perceptions. In wine analysis, the three key tastes are sour, sweetness and bitterness. Astringency is also an important key factor in describing the mouthfeel of wines. Astringency is felt by free nerve ends that are placed around the oral cavity 5. The combination of taste, mouthfeel and smell becomes flavor. In this project, the focus were on these three sensations of taste and the mouthfeel of the wine and the connection to the targeted compounds.

1.3.1 Acids

Acids are contributing to both color and taste in wine and are responsible for most of the sour taste in wine. In wine and grapes, the two major acids present are tartaric acid and malic acid 6. They make up about 90% of the acidity content in grapes, but other acids such as citric acid, succinic acid and lactic acid have also been observed. The concentrations of the acids differ in wines dependent on the type of grape and could vary between different origins and wine producers. Fatty acids, such as decanoic acid and octanoic acid are compounds present in wine during fermentation and they could contribute to the taste profiles of wine 7. In a study on red wines, Cortés-Diéguez et al. 8 discusses that decanoic acid has a fatty odor, described as rancid fat, and octanoic acid is described as sweet and cheesy.

1.3.2 Amino acids

Amino acids can be used to characterize wine in different ways. Sufleros et al. 9 analyzed amino acids to help characterize different French wines by origin, aging and type. In a study by Cao et al. made on Chinese rice wine 10, amino acids were chemically analyzed by high performance liquid chromatography (HPLC) to study different flavors. Amino acids, such as leucine and phenylalanine, arginine and valine corresponded to a bitter taste, proline, serine and threonine to a sweet taste, and asparagine and glutamic acid to umami taste in the wine. Niu et al. 4 chemically analyzed amino acids as taste-active compounds in cherry wines. The study gave the result of amino acids methionine and proline to correspond with sour taste, while serine, phenylalanine, glycine and leucine corresponded to bitter taste.

1.3.3 Phenolic compounds and flavonoids

Phenolic compounds and flavonoids are mostly found to be corresponding to bitterness, astringency, color and flavor to wine 11. Bitterness is felt by receptors on the tongue and is a taste 12. Astringency on the other hand is not a taste but instead is described as mouth-feel, and a tactile sensation. A description of astringent is roughness and dryness. Dryness when friction is felt in the oral surface. Astringency is important for high quality wines, especially red wines, but there needs to be a balance. Lack of astringency gives the wine a uninteresting and flat character, but in excess it could overpower some of the other flavors and make the wine dry. Non-flavonoid phenolics, like vanillin and gallic acid contributes to flavor and other non-flavonoids to the color of white wine 12. Red wines contains a higher amount of catechins than white wine which gives more bitterness and astringency.

1.4 Liquid chromatography mass spectrometry

Liquid chromatography with UV spectrometry has been widely used to analyze phenolic compounds, but liquid chromatography-mass spectrometry (LC-MS) has a better selectivity 13. Derivatization and the use of stable isotope labeled internal standards help reduce matrix effects. Analysis on phenolic compounds in wine has been performed both with and without

derivatization 4, 5, 13, 14. Amino acids in wine has been analyzed with liquid

chromatography-tandem mass spectrometry (LC-MS/MS ) 13 and LC-UV 4, 10. Organic acids in wine were in the study by Niu et al. analyzed with LC-UV 4. In this project, analysis of the targeted compounds were performed without derivatization and analyzed using LC-MS. 1.5 Aim

The aim for this project is to perform a chemical analysis on wine to identify different phenolic compounds, amino acids and organic acids. These compounds are connected to taste and will help create the different wine profiles. Compounds will be identified and quantified using liquid chromatography connected to mass spectrometry (LC/MS). An appropriate method for analyzing these compounds will be selected and optimized. The aim includes characterization of wines by;

- Finding a method to analyze targeted compounds in wine using LC-MS - Identify and quantify targeted compounds in different wines.

- Connecting targeted compounds to taste

- Comparing wine profiles depending on color, production year and grape type

2. Experimental

2.1 Chemicals and reagents

Chemicals used for mobile phases were methanol (≥98%), formic acid (≥98%) and acetonitrile were bought from Fisher scientific (Loughborough, UK), all of Analytical reagent grade. Ethanol (99,7%) from Solveco, (Rosersberg, Sweden). Acetic acid, citric acid (≥98%), proline (≥99%), serine (≥99%), decanoic acid (≥98%), octanoic acid (≥99%), caffeic acid (≥98%), catechin (≥96%), leucine (≥98%), phenylalanine (98%), syringic acid (≥98%), vanillic acid (≥97%), vanillin and syringol were purchased from Sigma-Aldrich (St.Louis, USA). Gallic acid were purchased from MERCK (Darmstadt Germany) and tartaric acid bought form Kebo AB (Stockholm, Sweden).

2.2 Targeted compounds and wine samples

Targeted compounds

Fifteen different compounds were targeted, including organic acids, fatty acids, amino acids, phenolic compounds and flavonoids. The different compounds are presented in Table 1. The compounds were chosen due to their presence in other studies about taste-active compounds in wine 4, 5, 10, 13, 14. Sugars were not included although they contribute to the taste of wines. The chemical structure of all targeted compounds are shown in the Appendix (A1).

Table 1. Targeted compounds for chemical characterization of taste active compounds in wine by liquid chromatography-mass spectrometry (LC-MS ) analysis.

Amino acids: ● Leucine ● Phenylalanine ● Proline ● Serine Acids: ● Tartaric acid ● Citric acid ● Decanoic acid ● Octanoic acid

Phenolic compounds & flavonoids:

● Caffeic acid ● Catechin ● Gallic acid ● Syringic acid ● Vanillic acid ● Vanillin ● Syringol Wine samples

A total of sixteen wine samples were included in the project. Prior to analysis the wine bottles were opened and tasted by a student in training to become a sommelier. The wines were tested to eliminate any spoiled wines from the analysis. There were nine different types of wine and most of them had bottles from two different production years. After tasting, the amount of wine needed for analysis was taken out and the bottles were filled with nitrogen gas and sealed with plastic wine corks for cold storage.

Table 2. Wine samples used for characterization of taste-active compounds.

Type Producer & Name Country/Region Grape

Production year

White

Pellerin Chardonnay France/Bugey Chardonnay 2014 & 2016 Causse-Marines Greilles France/Gaillac

Mauzac, Loin de

L’oeil, Muscadelle 2014 & 2015

Orange

Causse-Marines Zacm’orange France/Gaillac Mauzac 2015 & 2016

Red

Pellerin Gamay France/Bugey Gamay 2014 & 2015 Causse-Marines Peyrouzelle France/Gaillac

Braucol, Syrah,

Duras 2014 & 2016 Karim Vionnet Beaujolais Nouveau France/Beaujolais Gamay 2016 & 2017 Domaine la Rabidote France/Saint Chinian

Grenache, Carignan,

Syrah 2015 & 2016

Sparkling

Balviet Cordon du Bugey France/Bugey Gamay, Poulsard 2013 Labarthe Ancestrale France/Gaillac Mauzac 2014

2.3 Preparation of standard solutions

A “wine blank” was prepared to mimic the matrix of a wine sample making a final content of 6.5% ethanol solution, pH adjusted to 3.5 with acetic acid. The ethanol percent was chosen to

match the excepted percent of the wines diluted 1:1. The wine blank contained MilliQ water, acetic acid and ethanol.

For all target compounds (except the amino acids), 0,1 g were weighted and diluted in 10 mL of either methanol or MilliQ water to a concentration of 10 mg/mL individually. These were then further diluted to a concentration of 1 mg/mL. The amino acids were already prepared in stock solutions of 1 mg/mL. A “fake wine matrix” was prepared to use as a standard for quantification with a final concentration of 10 mg/L of the target compounds in 6.5% ethanol. Since most of the compounds were prepared in methanol the methanol content of the fake wine matrix was about 23 %. The fake wine matrix was used as the highest point of the calibration curves. Five calibration solutions were made to concentrations of 10 mg/L, 5 mg/L, 2.5 mg/L, 1 mg/L and 0.5 mg/L. Dilution and concentration of standard solution and fake wine matrix and calibration curve are shown in the Appendix (A2).

An injection standards of flurotelomer sulfonamide alkylbetain (FTAB) was used to monitor instrumental repeatability. FTAB was added to the fake wine matrix, wine blank and the wine samples. The injection standard was kept at -14°C and all other solutions at 6°C.

2.4 Sample preparation

Prior to analysis the wine bottles were opened and tasted to test whether any wine had been spoiled. After sample preparation, they were filled with nitrogen gas and a synthetic cork were used to seal the bottles. The wine was stored at 8°C.

Sample preparation was done by filtering wine samples through a 0.2 µm, 13 mm Acrodisc

Syringe GHP filter. Then 500 µl wine was diluted 1:1 (v/v) with Milli Q water. Before analysis, the samples were spiked with 10 µl of the injection standard.

2.5 Optimization of LC-MS separation and detection

A literature survey was performed for selection of an appropriate method for the analysis of the targeted compounds. A number of studies has reported the analysis of phenolic compounds

and some amino acids 4, 5, 13, 14. All of these studies used C18 or C8 columns for

chromatographic separation applying gradient elution of water with formic acid, ammonium

formate with formic acidand methanol or acetonitrile. Several of the methods used a similar

analytical approach. For comparison, a list with different parameters from these studies was compiled (see Appendix, A4). Different settings from these methods were tested when setting up an analytical method for this study. The method in this study is based on the study by Saenz-Navajas et al. 5 with some changes to the gradient program and mobile phases. The study presented herein covers a broader range of analytes compared to Saenz-Navajas et al. Different solvents for mobile phase B were tested and optimized between acetonitrile, methanol, methanol with 5% formic acid, acetonitrile with 5% formic acid and methanol with 2 mM

NH4Ac and 5 mM 1-MP. A BEH C18 column (50 mm x 2.1 mm, 1.7 um, waters) as well as a

BEH C18 column (100 mm x 2.1 mm, 1.7 um, waters) were tested to optimize the method.

Targeted compounds were monitored in both electrospray positive and negative ion mode, monitoring the molecular ion (M) as well as M+1 _{and M}-1_{. Dwell times were optimized to have} about 10-15 points across each peak. Initial experiments were performed on the fake wine matrix in full scan mode, scan range 50-400 in both positive and negative mode but did not present good result and single ion recording (SIR) of expected m/z values was used instead to detect the compounds. The m/z values of all compounds are presented under instrumental

methods in Table 4.

In this project, organic acids are analyzed with the same instruments and settings. This is an approach which differ this project from the compared articles above.

2.6 LC-MS Analysis

2.6.1 Instrumental settings

Chemical analysis was performed by using a Waters Acquity Ultra Performance LC system

(Waters), with a BEH C18 column (100 mm x 2.1 mm, 1.7 µm, waters). The oven temperature

was kept at 40°C, the flow rate at 0.2 mL/min with an injection volume of 10 µl. Mobile phases used were (A) 5% formic acid in MilliQ water and (B) methanol.

Mass spectrometry analysis was performed using one quadrupole on a Waters Quattro Premier XE instrument in single ion recording (SIR) mode. Electrospray ionization (ESI) was used in both positive and negative mode with a capillary voltage of 3.00kV. The cone voltage was set to 25 V for all compounds except for the injection standard at 20 V. The source temperature was kept at 100°C and desolvation temperature at 350°C. The collision gas used was argon.

2.6.2 Instrumental Methods

Two different instrumental methods were used to quantify all targeted compounds. The first one, method 1 (M1) was used for 12 of the targeted compounds and the injection standard. Gradient elution for M1 was performed at the flow rate of 0.2 mL/min(if not otherwise stated) with mobile phases (A) 5% formic acid in MilliQ water and (B) methanol with the following program: 0-0.1 min 95% A, 0.1-12.0 95-70% A, 12.0-12.5 min 70-50&% A, 12.5-13.5 min 50-0% A, 13.5-17.0 min 50-0% A, 17.0-17.1 0-95%A, 17.1-17.5 95% A (flow 0.4 mL/min),

17.5-21.5 95% A. The method used ionization mode ESI+

with 3 different SIR functions, from 0.5 to 10 minutes, from 10 to 21.5 minutes and the last one scanning for the injection standard from 12.00 to 20.00 minutes.

The second method (M2) included three of the targeted compounds. Same mobile phases as M1 but with a different gradient program: 0-0.1 min 95% A, 0.1-2.0 95% A, 2.0-7.0 min 95-70% A, 7.0-7.5 min 70-50&% A, 7.5-8.5 min 50-0% A, 8.5-8.6 min 0-95% A, 8.6-9.0 0-95%A

from 0.3 to 12 minutes. In M2 injection standard were not used, since no suitable injection standard was available to use in electrospray ionization negative mode with the chosen parameters.

Table 4. Targeted compounds and MS parameters for analysis of phenolic compounds, organic acids and amino acids in wine.

Method Compound m/z Predicted Time

(min)

Dwell time (s) Cone voltage (V) Ionization mode Mobile phases M1 Leucine 132 3.40 0.2 25 + A / B Phenylalanine 166 4.70 0.2 25 + A / B Proline 116 1.24 0.15 25 + A / B Serine 104 1.65 0.2 25 + A / B Decanoic acid 173 15.70 0.1 25 + A / B Octanoic acid 145 15.50 0.15 25 + A / B Caffeic acid 181 5.81 0.15 25 + A / B Catechin 291 4.15 0.15 25 + A / B Syringic acid 199 7.75 0.15 25 + A / B Vanillic acid 169 6.37 0.15 25 + A / B Vanillin 152 8.19 0.15 25 + A / B Syringol 155 12.12 0.2 25 + A / B FTAB (IS) 571 145.20 0.5 20 + A / B M2 Citric acid 191 1.09 0.08 25 - A / B Tartaric acid 149 1.03 0.08 25 - A / B Gallic acid 169 2.97 0.08 25 - A / B 2.6.3 Sample analysis

The wine samples were run in triplicates on each method. A methanol blank, a fake wine matrix standard of 1.0 mg/L and 5.0 mg/L (for positive and negative mode respectively) and a wine blank were analysed with every nine injections, corresponding to three wine samples.

A calibration curve were run for both methods. For M1 a four point calibration curve were run with the concentrations of 0.5 mg/L, 1.0 mg/L, 5.0 mg/L and 10 mg/L. For M2 the calibration curve included the concentrations 0.5 mg/L, 1.0 mg/L, 2.5 mg/L and 5.0 mg/L.

3. Results

3.1 Quality control

The calibration curve for M1 had four points with standard solutions with concentrations of 0.5 mg/L, 1.0 mg/L, 5.0 mg/L and 10 mg/L. The calibration curve for M2 had concentrations of 0.5 mg/L, 1.0 mg/L, 2.5 mg/L and 5.0 mg/L. The calibration curves showed R2 _{values of >0.98}

for all compounds. Limit of quantification (LOQ) was extracted from the signal to noise ratio from the chromatogram of the 0.5 mg/L standard for all compounds and then multiplied by ten. Decanoic acid were not detected in the 0.5 mg/L or 1.0 mg/L standard and were first found in 5.0 mg/L giving it a much higher LOQ. Relative standard deviations (RSD) were calculated for each compound on the repeated injections of 1.0 mg/L standard in M1 and the 5.0 mg/L standard in M2.

The injection standard, FTAB, was used to normalize against for calculations of concentrations of wine samples. The relative standard deviation was calculated on the area of FTAB in all samples analyzed and was 16%.

Table 5. Results from calibration curve for compounds analyzed in M1 (positive mode). Includes R2_{, limit of}

detection and relative standard deviation.

Compound R2 LOQ (mg/L) RSD Serine 0.986 0.012 12% Leucine 0.969 0.022 13% Proline 0.981 0.00302 6% Vanillin 0.990 0.018 9% Vanillic acid 0.994 0.015 17% Phenylalanine 0.984 0.0043 13% Caffeic acid 0.992 0.18 20% Syringic acid 0.987 0.021 12% Catechin 0.989 0.047 23% Octanoic acid 0.987 0.26 7% Syringol 0.990 0.017 7% Decanoic acid* _{0.986 5 -}

*Decanoic acid was only detected in calibration standard 5.0 mg/L and 10 mg/L and RSD could not be calculated.

Table 6. Results from calibration curve for compounds analyzed in M2 (negative mode). Includes R2_{, limit of}

detection and relative standard deviation.

Compound R2 LOQ (mg/L) RSD

Tartaric acid 0.999 0.12 43%

Citric acid 0.987 0.12 35%

3.2 Compounds detected in wine samples 3.2.1 White wines

Four white wines were included in the analysis. The four wines are of two different types, Pellerin Chardonnay and Causse-Marines Grelles with different productions years. Each wine were analyzed in triplicates and the concentrations are displayed as a mean value of the three. RSD values of all compound in each wine are also calculated on the triplicates. Figure 1 displays the concentrations of the targeted compounds in each wine and Figure 2 displays a comparison pattern between the compounds found in the wines.

Figure 1. Concentrations of targeted compounds in white wine; Pellerin Chardonnay 2014 & 2016 and Causse-Marines Grellies 2014 & 2015.

Figure 2. Comparison of wine profiles between compounds found in white wine. 1. Pellerin Chardonnay 2014, 2. Pellerin Chardonnay 2016, 3.Causse-Marines Grellies 2014, 4. Causse-Marines Grellies 2015.

Vanillin, vanillic acid, citric acid and decanoic acid were not detected in any of the white wines. Syringol were not detected above 0.1 mg/L in any wine and catechin was present only in the Causse-Marines Grellies wines. Proline was present in all wines but in Causse-Marines Grellies 2014 detected outside of the calibration curve and the results are thus tentative. Tartaric acid in the Causse-Marines Grellies wines were also tentative. RDS values in all white wines for all compounds except for tartaric acids were <16%. Tartaric acid in Pellerin Chardonnay 2016 and Causse-Marines Grellies 2014 were 82% and 32% respectively.

3.2.2 Orange wines

Two orange wines were included in the analysis. The two wines are of the same type, Causse-Marines Zacm’orange but from different production years, 2015 and 2016. Each wine was analyzed in triplicates and the concentrations are displayed as a mean value of the three. RSD values of all compound in each wine is also calculated on the triplicates. Figure 3 displays the concentrations of the targeted compounds in each wine and Figure 4 displays a comparison pattern between the compounds found in the wines.

Figure 3. Concentrations of targeted compounds in orange wine; Causse-Marines Zacm’orange 2015 & 2016.

Figure 4. Comparison of wine profiles pattern between compounds found in orange wine. 1. Causse-Marines Zacm’orange, 2015, 2. Causse-Marines Zacm’orange, 2016.

Serine, vanillin, vanillic acid, citric acid and decanoic acid were not detected in the orange wines. Caffeic acid and catechin were detected in Causse-Marines Zacm’orange from 2016 but

not in the 2015 sample. The concentrations of proline, catechin and tartaric acid were detected outside of the calibration curve and are thus tentative results. RSD values were <14% for all compounds in both wines.

3.2.3 Red wines

Eight red wines were included in the analysis. The eight wines were of four different types, Pellerin Gamay, Causse-Marines Peyrouzelle, Karin Vionnet Beaujolais Nouveau and Domaine la Rabidote with different productions years. Each wine was analyzed in triplicates and the concentrations are displayed as a mean value of the three. RSD values of all compound in each wine was also calculated on the triplicates. Figure 5 displays the concentrations of the targeted compounds in each wine and Figure 6 displays a comparison pattern between the compounds found in the wines.

Figure 5. Concentrations of targeted compound in red wine; Pellerin Gamay 2014 & 2015, Causse-Marines Peyrouzelle 2014 & 2016, Karin Vionnet Beaujolais Nouveau 2016 & 2017 and Domaine la Rabidote 2015 & 2016.

Figure 6. Comparison of wine profiles pattern between compounds found in red wine. 1. Pellerin Gamay 2014, 2. Pellerin Gamay 2015, 3. Causse-Marines Peyrouzelle 2014, 4. Causse-Marines Peyrouzelle 2016, 5. Karin Vionnet Beaujolais Nouveau 2016, 6. Karin Vionnet Beaujolais Nouveau 2017, 7. Domaine la Rabidote 2015, 8. Domaine la Rabidote 2016

Serine, vanillin, decanoic acid and citric acid were not detected in any of the red wines. The concentration of proline was detected outside of the calibration curve and are thus a tentative result in all wines. In Pellerin Gamay 2014, catechin is a tentative result and for both the 2014 and 2015 caffeic acid and tartaric acid had tentative results. All because they were detected outside of the calibration curve. In the Causse-Marines Peyrouzelles 2014 wine, catechin and tartaric acid had tentative results. In Karin Vionnet Beaujolais Nouveau 2017, caffeic acid and catechin had tentative results, as well had catechin in Domaine la Rabidote for both the 2015 and 2016. Caffeic acid in the 2015 and tartaric acid in the 2016 were also detected outside of the calibration curve. RSD values were <16% in all red wines for all compounds except for syringol and tartaric acid. Syringol had RSD values <16% for all wines except for Karin Vionnet Beaujolais 2016 where it was 23%. Tartaric acid had RSD values between 2-87% in all red wines.

3.2.4 Sparkling wines

Two different sparkling wines were included in the analysis, Baviet Cordon du Bugey 2013 and Labarthe Ancestrale 2014. Each wine were analyzed in triplicates and the concentrations are displayed as a mean value of the three. RSD values of all compound in each wine is also calculated on the triplicates. Figure 7 displays the concentrations of the targeted compounds in each wine and figure 8 displays a comparison pattern between the compounds found in the wines.

Figure 7. Concentrations of targeted compounds in sparkling wine; Baviet Cordon du Bugey 2013 and Labarthe Ancestrale 2014.

Figure 8. Comparison of wine profiles pattern between compounds found in sparkling wine. 1. Baviet Cordon du Bugey 2013, 2. Labarthe Ancestrale 2014.

Serine and decanoic acid were not detected in either of the sparkling wines. Vanillin and octanoic acid were not detected in the Baviet Cordon du Bugey 2013, while vanillic acid, caffeic acid and syringol were not detected in the Labarthe Ancestrale 2014. Proline had tentative results in Baviet Cordon du Bugey 2013 and tartaric acid and citric acid were detected outside of the calibration curve and had tentative results in both wines. RSD values vere <9% for all compounds detected in the Labarthe Ancestrale 2014. In Baviet Cordon du Bugey 2013, RSD values ranges from 5-29%.

3.2.5 Gallic acid in all wines

Gallic acid were detected in all wines except for three of the white wines; Pellerin Chardonnay 2014, Pellerin Chardonnay 2016 and Causse-Marines Greilles 2014. Concentrations of gallic acids in the wines were detected to more than twice as high as any other of the targeted compounds. The RSD values were calculated on the triplicates analyzed for every wine sample and ranged from 0.7%-8.8%. All detected signals were above the calibration curve and the concentrations are thus tentative. Gallic acid compared to all other target compound were the only one with a decreasing retention time throughout the analysis decreasing by a whole minute from the first to last injection.

Table 7. Concentrations of gallic acid and relative standard deviation measured in the analyzed wine samples.

Wine Concentration (mg/L) RSD

Causse-Marines Greilles 2015 103 0.70%

Causse-Marines Zacm’orange 2015 /2016 166 / 155 0.40% / 1.9% Pellerin Gamay 2014 / 2015 111 / 149 3.7% / 2.8% Causse-Marines Peyrouzelle 2014 / 2016 140 / 93 1.7% / 8.8 % Karim Vionnet Beaujolais Nouveau 2016 /

2017 130 / 123 2.8% / 7.4%

Domaine la Rabidote 2015 / 2016 130 / 141 8.8% / 1.8%

Balviet Cordon du Bugey 2013 116 5.0%

Labarthe Ancestrale 2014 97 1.4%

3.2.6 Summary of targeted compounds in all wines

Gallic acid had the highest concentrations with a range from 93-166 mg/L and were detected in 13 out of 16 wines. Tartaric acid were found in all wines except for one white wine and had a concentration range between 5.2-80 mg/L. Proline was present in all wines with a concentration range between 19.4-35.4 mg/L. Decanoic acid was not detected in any of the wines and vanillin was only found in the sparkling wine Labarthe Ancestrale. Vanillic acid was only detected in red and sparkling wines. Citric acid was only found in sparkling wine and serine only in white wine. Table 8 displays the targeted compounds and their overall range in concentrations throughout all analyzed wine samples.

Table 8. Concentration range of targeted compounds in analyzed wines.

Compound Concentration range (mg/L) Compound Concentration range (mg/L)

Serine 0.6-4.2 Catechin 0.2-54.4

Leucine 0.3-12.6 Octanoic acid 0.9-6.3

Proline 19.4-35.4 Syringol 0.1-0.2

Vanillin <LOQ - 0.1 Decanoic acid <LOQ Vanillic acid 1.2-5.2 Tartaric acid 5.2-80.1 Phenylalanine 0.3-13.2 Citric acid 26.5-45.5 Caffeic acid 5.7-34.0 Gallic acid 93-166 Syringic acid 0.6-13.9

4. Discussion

4.1 Compounds detected in wine samples and correlation to taste

Catechin contributes to a bitterness and astringency to the wine and is said to be present in higher amounts in red wine [12]. This can clearly be seen in the results of this study where the four white wines, one of the orange and both sparkling wine showed levels of catechin below

1 mg/L. The orange wine Causse-Marines Zacm’orange from 2016 and all the red wines

showed higher amounts of catechin, in most of them higher than 20 mg/L. This would lead to these wines having a higher bitterness and astringency. In previous studies on amino acids by Cao et al. [10] and Niu et al. [4], phenylalanine and leucine were both correlated with bitter taste but Cao et al describes serine and proline to have a sweet taste while Niu et al. presents serine as a bitter taste and proline as a sour taste. In this study, serine was detected in very small amounts and only in the white wines. Serine whether it would give the wine a sweeter more bitter taste would probably not affect the flavor profile that much when other compounds were present in higher amounts. Proline on the other hand was detected in all wines with one of the highest concentration of all compounds in most samples. Proline would thereby probably contribute a lot more to the taste of the wines and whether or not it has a sweet or sour taste

cannot be determined without a proper sensory tasting of all the sixteen wines.The sparkling

wines were the only two which contained any citric acid. Citric acid contributes to citrus flavor and acidity [6]. Citric acid could be added in the process of wine making just for more acidity, but the wines used in this project were organic wines where additions of compounds to alter flavor are limited. The three most prominent compounds (except for gallic acid) in the sparkling wine were tartaric acid, citric acid and proline. Together these three compounds correspond to more than 80% of the wine profile patterns. With tartaric acid also contributing to acidity. Together with proline these wines would have a higher fresh sweet and sour taste compared to the rest of the wines. Syringic acid, vanillic acid and caffeic acid are in results presented by Saenz-Navajas [5] compounds that contribute to astringency. Amongst the results of the sixteen wines in this study syringic acid, vanillic acid and caffeic acid were most prominent in the red wines. This would make them more astringent compared to the other color wines. Phenolic

compound are corresponding to both astringency and bitter taste [11] and the highest amount of phenolic compounds were found in red wines.

4.2 Comparison between different wine profile patterns

The wine profile patterns were compared between the different grape types of the sixteen wines. When comparing the wines not by color or production year but the grape type, there seems not to be any clear similarities in the distribution of compounds found in the wines.

Figure 10. The wine profile pattern for all wines, sorted by the grape type.

Instead there seems to be more clear similarities and differences in the compound distribution between the wine profile patterns from the same color wines and the same wine but with different production years. For the red wines, the similarity in the wine profiles showed that for all eight wines the same compounds were detected. This is true with one exception of the Causse-Marines Peyrouzelles from 2016 where caffeic acid was not detected. All of the wine profile patterns are shown in Figure 10. The red wine profiles show that the first pair of wines, (1 and 2 in the diagram) Pellerin Gamay 20114 and 2015 were the most close in distribution of the compounds. The Causse-Marines Peyrouzelles pair (3 and 4) were the most different from each other of the red wines and this is mostly due to the caffeic acid.

Figure 10. The wine profile pattern for all the color wines.

For the white wines, the two Pellerin Chardonnay 2014 and 2016 (1 and 2 in the diagram) has little or no tartaric acid and would probably differ in acidity from the taste form the Causse-Marines Greilles 2014 and 2015 which has a high distribution of tartaric acid. But the wines seems to not differ that much depending on the age. The sparkling wines has the citric acid which as mentioned makes then stand out from the other fourteen wines, and though they are not the same brand, the Balviet Cordon du Bugey and Labarthe Ancestrale has similar profiles which distinguish them as sparkling wine in this project. The orange wines had very different profiles where the Causse-Marines Zacm’orange form 2016 contained both catechin and caffeic acid which were not detected in the same wine produced in 2015. The presence of catechin and caffeic acid would suggest the younger wine to be more bitter tasting and more astringent.

4.3 Quality control

The calibration curves used for both positive and negative mode did not go higher than 10 mg/L and many of the compounds detected in the wine samples did give a higher signal outside of the calibration curve. A more extended calibration curve would help to assure the linearity for the higher signals. Another solution to this problem with tentative results could be fixed by diluting the wine samples further before analysis. In this project to dilute the wine samples further and re-analyze them, or to make and analyze a calibration cure with higher concentrations would confirm the results better.

4.4 Method used for further analysis

The method used in this project display that it is possible to detect and quantify these taste-active compounds. Most compounds eluted fine but it was harder for the organic acids tartaric acid and citric acid. They eluted very quickly, around 1 minute and coeluted, so it was difficult to get a good separation of these compounds on the used column with the used settings. Tartaric acid also had RSDs varying very much from wine to wine with RSDs up to 82%. In an application note by RESTEK [15] a HPLC method for analyzing organic acids is presented. Here it states the difficulty with analyzing polar organic acids on conventional reversed phase columns. Here the compounds tartaric acid, quinic acid, malic acid, citric acid and fumaric acid were separated using a column designed for this application. The column was an Allure Organic Acids Column (300 mm x 4.6 mm, 5 µm). The method was performed with a 100 mM phosphate buffer, pH 2.5 as mobile phase at a flow rate of 0.5 mL/min. If further analysis on taste-active compounds would be performed, the method presented in this herein project would be applicable for almost all compounds. For organic compounds such as tartaric acid and citric acid another method with a different column, as presented above by RESTEK would be preferable to get a better separation.

5. Conclusion

The LC-MS method used to analyze wine were able to identify and also quantify fourteen different taste-active compound, such as phenolic compounds, organic acids and amino acids. The results show similarities and differences between wine profiles of the white, red, orange and sparkling wine. The different compound could also connect the wines with different taste sensations such as sweetness, acidity, bitterness and the mouthfeel of astringency. Tartaric acid and citric acid were harder to detect than the other compounds and for further studies a different method is suggested to optimize separation and detection of organic acids.

6. References

1. Lesschaeve, I. (2006) Sensory Evaluation of Wine and Commercial Realities: Review of Current Practices

and Perspectives. American Society for Enology and Viticulture.

2. Lesschaeve, I., Noble, A.C. (2010) Managing wine quality, Vol 1, Viticulture and wine quality, Woodhead Publishing Limited, Cambridge UK. ch 7, 189–217

3. Barata, A., Campo, E., Malfeito-Ferraira, M., Loureiro, V., 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 Chemistry., 2011, 59 (6), pp 2543–2553

4. Niu. Y., Zhang, X., Xiao, Z., Song, S., Jia, C., Yu, H., Fang, L., Xu, C. (2012) Characterization of

taste-active compounds of various cherry wines and their correlation with sensory attributes. J. Cromatogr. B 902,

55-60.

5. Saenz-Navajas, M.P., Ferreira, V., Dizy, M., Fernandez-Zurbano, P. (2010) Characterization of taste-active fractions in red wine combining HPLC fractionation, sensory analysis and ultra performance liquid

chromatography coupled with mass spectrometry detection. Analytica Chimica Acta 673, 151-159.

6. Beelman, R.B., Gallander, J.F. (1979) Advances in food research. Academic press inc, New york, VOL 25, 5-8.

7. Lafon-Lafourcade, S., Geneix, C., Ribereau-Gayon, P. (1984) Inhibition of Alcoholic Fermentation of Grape

Must by Fatty Acids Produced by Yeasts and Their Elimination by Yeast Ghosts. Applied and environmental

microbiology Vol. 47, No. 6, 1246-1249

8. Cortés-Diéguez, S., Rodriguez-Solana, R., Domínguez, J.M., Díaz, E. (2015) Impact odorants and sensory profile of young red wines from four Galician (NW of Spain) traditional cultivars. J. Inst. Brew. 2015; 121:

628–635

9. Soufleros, E, Barrios, M.L. and Bertrand, A. (1998) Correlation between the content of biogenic amines and

other wine compounds. Int J Food Microbiol 49, 266–277.

10. Cao, Y., Xie, G., Wu, C. and Lu, J. (2010) A Study on Characteristic Flavor Compounds in Traditional

Chinese Rice Wine - Guyue Longshan Rice Wine. Journal of the institute of brewing, VOL. 116, NO. 2,

11. Kennedy, J. A., Saucier, C., & Glories, Y. (2006) Grape and wine phenolics: History and perspective. American Journal of Enology and Viticulture, 57, 239–248

12. Baumes, R. M. (2009) V. Moreno-Arribas, M.C. Polo (eds.), Wine Chemistry and Biochemistry, DOI 10.1007/978-0-387-74118-5 11, C Springer Science+Business Media, LLC (2009) ch. 8C, 8D. 251-261

13.Malec, P.A., , Oteri, M., Inferrera, V., Cacciola, F., Mondello, L., Kennedya, R.T. (2017) Determination of

amines and phenolic acids in wine with benzoyl chloride derivatization and liquid chromatography–mass spectrometry. J. Chromatogr. A 1523 (2017) 248–256

14. Gruz, J., Novák, O., Strnad, M. (2008) Rapid analysis of phenolic acids in beverages by UPLC–MS/MS. Food Chemistry 111 (2008) 789–7

Appendix

A1. Targeted compounds and chemical structures.

Serine

Leucine

Proline

Phenylalanine

Tartaric acid

Citric acid

Decanoic acid

Octanoic acid

Caffeic acid

Catechin

Gallic acid

Syringic acid

Vanillic acid

Vanillin

Syringol

A2. Dilution and concentration of standard solution and fake wine matrix and calibration curve.

Solvent Concentration 1 Concentration 2 Fake Wine Matrix Calibration points mg/L Amino acids

Leucine Methanol 1 mg/mL 10 mg/L 0.5 , 1, 5, 10 Phenylalanine Methanol 1 mg/mL 10 mg/L 0.5 , 1, 5, 10 Proline 1:1 Methanol/H2O 1 mg/mL 10 mg/L 0.5 , 1, 5, 10

Serine Methanol 1 mg/mL 10 mg/L 0.5 , 1, 5, 10 Acids

Citric acid H2O 10 mg/mL 1 mg/mL 10 mg/L 0.5, 1, 2.5, 5 Decanoic acid Methanol 10 mg/mL 1 mg/mL 10 mg/L 0.5 , 1, 5, 10 Octanoic acid Methanol 10 mg/mL 1 mg/mL 10 mg/L 0.5 , 1, 5, 10 Tartaric acid H2O 10 mg/mL 1 mg/mL 10 mg/L 0.5, 1, 2.5, 5 Phenolic

compounds & Flavonoids

Caffeic acid Methanol 10 mg/mL 1 mg/mL 10 mg/L 0.5 , 1, 5, 10 Catechin Methanol 10 mg/mL 1 mg/mL 10 mg/L 0.5 , 1, 5, 10 Gallic acid Methanol 10 mg/mL 1 mg/mL 10 mg/L 0.5, 1, 2.5, 5 Syringic acid Methanol 10 mg/mL 1 mg/mL 10 mg/L 0.5 , 1, 5, 10 Vanillic acid Methanol 10 mg/mL 1 mg/mL 10 mg/L 0.5 , 1, 5, 10 Vanillin Methanol 10 mg/mL 1 mg/mL 10 mg/L 0.5 , 1, 5, 10 Syringol Methanol 10 mg/mL 1 mg/mL 10 mg/L 0.5 , 1, 5, 10

A3. Relative standard deviation (RSDs) of all wine and compounds in %. Wine Serin e Leucin e Prolin e Vanilli n Vanillic acid Phenylalan ine Caffeic acid Syringic acid Catech in Octanoic acid Syring ol Tartaric acid Citric acid W1 5 1 1 1 4 3 12 W2 1 0 1 1 8 11 3 82 W3 1 1 0 1 6 4 0 10 32 W4 1 1 1 1 0 8 4 2 15 4 W5 9 1 6 13 1 8 3 W6 0 1 1 1 2 2 5 9 2 W7 0 0 3 1 2 1 2 5 3 10 W8 1 0 2 1 1 1 1 2 3 2 W9 0 1 3 1 1 2 1 2 3 11 W10 1 1 4 1 2 1 1 6 52 W11 2 1 15 1 2 3 2 3 23 8 W12 5 2 1 2 2 2 1 6 5 28 W13 1 1 3 1 0 1 4 3 9 87 W14 1 0 6 1 1 1 2 4 10 6 W15 12 13 8 11 15 9 22 25 29 5 W16 1 2 8 1 6 2 1 5 3 3

A4. Table of LC-MS methods compared between different studies

Autho r

Instrument Column Colu mn temp

Mobile phase Flow rate

Gradient program Detector ESI details Gases Compounds Samp

le size

Yunw ei Niu UPLC system

Acquity (Waters, Massachuset ts, USA) BEH C18 (100 mm × 2.1 mm, 1.7 um,) 25°c A: H2O 5% formic acid B: methanol 5% formic acid 0,25 ml/mi n *0-20 min: 5-15%B *20-25min 15-10% B *25-30min 10-5% B.

DAD interfaced with ESI trap MS (Negative mode) source block temperature 100 ◦ C, desolvation temperature 350 ◦ C, capillary voltage 3.2 kV, cone voltage 40 V) Argon =collision Nitrogen = desolvation *Gallic acid *Chlorogenic acid *Vanillic acid *Caffeic acid Paige. A Waters (Milford, MA) nanoAc- quity UPLC Acquity HSS T3 C18 column (1 mm x 100 mm, 1.8 um, 100 Å pore size) 27°c A: 10 mM ammonium formate with 0.15% formic acid. B: acetonitril 100 ul/mi n *Iinitial 0% B *0,01min 15%B *0,5min 17%B *14min 55%B *14,5min 70% B *18min 100% B *19 min 100% B *19,1min 0% B *20min 0% B

Agilent (Santa Clara, CA) 6410 B triple quadrupole (dMRM) mode ESI positiv mode 4kV

ESI positiv mode 4kV

The gas temperature was 350 ◦ C, gas flow was 11L/min, and the nebulizer was at 15 psi. *Arginine *Asparagine *Caffeic Acid *Gallic acid *Glycine *Methionine *Phenylalanine *Proline *Serine *Threonine *Leucine *Vanillic acid *Vanillin 5ul J.Gru z ACQUITY Ultra Performance LCTM system (Waters, Milford, MA, USA) BEH C8, 1.7 lm, 2.1 um 150 mm, Waters, Milford, MA 30°C A: aqueous 7.5 mM HCOOH B: acetonitril 250 ul min *5% B 0,8min *5-10% B0,4min *isocratic 10% B ,0,7min *10-15% B 0,5min *isocratic 15% B 1,3min *15-21% B 0,3min *isocratic 21% B 1,2min *21-27% B 0,5min *27-50% B 2,3min *50-100% B 1min *100-5% B 0,5min End initial for 2,5min

PDA 2996 photo diode array detector (Waters, Milford, MA, USA) Micromass Quattro microTM API benchtop triple quadrupole mass spectrometer (Waters MS Technologies, Manchester, UK) Scanning range 210 - 600nm

ESI negative mode source block temperature 100 °C, desolvation temperature 350 °C, capillary voltage 2.5 kV, cone voltage 25 V) Argon= collision Nitrogen = desolvation *Gallic acid *Chlorogenic acid *Caffeic acid *Vanillic acid *Syringic acid Maria

-Pilar Waters Acquity Ultra Performance LC system (Waters). BEH C18 column (100 mm × 2.1 mm, i.d., 1.7 um particle size, Waters), 40°C A: water/formic acid 5% B:acetonitril o,2 ml/mi n *0-12min 90-70% A *12-12,5min 70-50% A *12,5-13,5min 50-0% A *13,5-13,6min 0-90% A microTOF II high-resolution mass spectrometer Apollo II ESI/APCI

ESI positive & negative mode for

catechin,

The capillary potential was set to 2.5 kV; the drying gas temper- ature was 200 ◦ C and its flow 5 L min−1 ; the nebulizer gas was set to 2 bar and 25 ◦ C *Syringic acid *Caffeic acid *Vanillic acid *Catechin 5 ul