Aroma and taste impact components in grapefruit juice


Material Information

Aroma and taste impact components in grapefruit juice
Physical Description:
xvi, 144 leaves : ill. ; 29 cm.
Jella, Prashanthi
Publication Date:


Subjects / Keywords:
Grapefruit juice -- Odor   ( lcsh )
Grapefruit juice -- Composition   ( lcsh )
Food Science and Human Nutrition thesis, Ph.D   ( lcsh )
Dissertations, Academic -- Food Science and Human Nutrition -- UF   ( lcsh )
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph.D.)--University of Florida, 1998.
Includes bibliographical references (leaves 136-143).
Statement of Responsibility:
by Prashanthi Jella.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 029544287
oclc - 40139348
System ID:

This item is only available as the following downloads:

Full Text






This thesis is dedicated to Shirdi Sai.


I would like to express my gratitude to my advisor Dr. Russell Rouseff for his

assistance and support during the course of my research. His constant encouragement

promoted independent thought and his words "You Can Do It", "Go Get Them"

challenged me at every turn and refused to let me settle for superficial solutions for critical

problems. He is a good teacher, an exceptional person and I am glad that I got a chance

to work with him.

I would also like to thank my committee members Dr. Gregory, Dr. O'Keefe, Dr.

Powell, Dr. Sims and Dr. Teixeira for their guidance in this project. Dr. Gregory is one of

the teachers I admire for promoting critical thinking in his students. His questions during

seminars were always topics to ruminate" for my friends and me.

I had an opportunity to sit through some of Dr. Teixeira's classes and they were

one of the most cherished experiences for me. I will never forget the definition of "a

thixotropic fluid" and the way he demonstrated it in the class. He is one of the best

teachers I had, who never took no for an answer but helped the students to work through

the problem.

There are no words to express my thanks to friends and room mates from Texas

A&M, who are like a second family to me. The bonding we have is a special one and I

will always cherish it. They are the one of the reasons for making my stay in US worth


Thanks to my friends at UF who made going to school an enjoyable experience. I

miss the time I shared with Mitwe, Pimpen, Cynthia, Alex, Rena and Jamie. They are the

people whom I admire for their qualities, and I am glad that I am friends with them.

I appreciate the learning experiences and help from Rusty, Kevin and Harold in our

lab. Words fail to express the gratitude for panel members at USDA and especially Uli.

Uli is the person I admire for his liberal outlook and broad knowledge about other cultures

of the world.

My deepest and most sincere gratitude is to my family. My parents were a

constant support and their guidance and encouragement is a yard stick for my

advancement. The importance they placed on good education and their philosophy of

"always strive for better but be happy with what you have" made my sister, brother and

me the kind of persons we are today. I owe it to them. My father's dynamism and my

mother's liberal thinking are source of inspiration to me to try anything.

If there is one person who was more proud of me and my achievements, it was my

grandfather. He was a great teacher, exceptional human being and philanthropist who

touched many lives other than his family. His memories are permanently etched in my

heart. My paternal and maternal grandmothers are the women I admire most. Their

strength and intelligence are a source of inspiration in my life. Special thanks to my

uncles, aunts and cousins for their emotional support.

My father- and mother-in-law are a great support to me. They treat me like their

own daughter and stand by me in all situations. They invited me in to their family against

all traditional Indian norms. I am ever thankful to them for it.

Of all the friends I have, the best of them is my life partner Rohini. His love,

support, patience, encouragement are sources for my strength. He is the happiness in my

life. Without him this would not be possible. All I can say to him is THANK YOU



ACKNOWLEDGMENTS ............. ............................ iii

LIST OF TABLES ............................ ... .... .. ix

LIST OF FIGURES .. ........................... ...... .............. xi

A B STR A C T ................................... .. .......... xiv


1. INTRODUCTION ........................... ........ 1

2. LITERATURE REVIEW ........ ........... ........... .. ...... .... 4
F lav o r .. ... .... . .. 4
Statistical Correlations ...................... .. .. 5
Bitterness .................. ............... 7
GC-Olfactometry .................... ....................... 9
Charm@ Analysis ................ ... ......... 9
AEDA ........ ...... ............. .......... 10
OSME ................ ......... ...... ....... 10
Sample Preparation .............. ..... 12
SPM E ..................................... .......... 13
Sulfur Compounds ........ ...... ..... ................ 15

3. MATERIALS AND METHODS ..................... ................. 17
Grapefruit Juice Sample Collection .. ........ ............. 17
Survey Sam ples ................. ..... .. ..... 17
Methylene chloride extracts .................... 17
Pentane-diethyl ether extracts ...................... 18
GC-Olfactometry Samples ............................... 18
Sam ple Preparation ................................. ......... 19
Liquid-Liquid Extraction with Methylene Chloride ................ 19
Liquid-Liquid Extraction with Pentane-Diethyl ether (1:1): ......... .20

Dynamic Head Space Purge and Trap Solvent Elution ............. 20
Extraction Procedure for Sulfur Compounds .................. 21
Extraction Procedure for GC-Olfactometry Analysis ............. 21
Instrumental Techniques ................... ........ .. ... 21
GC-Flame Ionization Detector ............................ 21
GC-Sulfur Chemiluminescence Detector ........................ 22
GC-M ass Spectrometry .................................... 24
Limonin and Naringin Analysis Using HPLC ..................... 24
Sample preparation ................................. 24
HPLC instrumentation ............................. 25
Peak Identification and Quantification ....................... ... 25
Sensory Analysis ...................................... .... .. 27
DOC Preference Panel ................................... 27
USDA Descriptive Panel ................................... 31
Training of Panelists ..................... ... ............ ....... 33
GC-Olfactometry Panel .................................... 33
Descriptive Panel ........................................ 34
Statistical Analysis ................ ................... ........ 34

4. RESULTS AND DISCUSSION ..................................... 36
Correlations Between Preference and Analytical Measurements ............ 36
Sensory Analysis ................... ........... .. ... .. 40
Statistical Analysis ........................... .. .. ... 40
Univariate analysis ... ... ................ 40
Multivariate analysis ............................... 43
Identification of the Peak at RI-1126 .......................... 54
Grapefruit Juice Aroma Extraction Methods .......................... 56
Chromatographic Separation and Analysis .................... 56
Extraction M ethods ................... .... ........ ... 58
Liquid-liquid extractions ............................. 58
Dynamic head space extraction ......................... 62
Static head space extraction using SPME ................. 64
GC-Olfactometry Studies ........................................ 65
Instrumental Detectors vs. Human Response ............ 65
Maturity and Processing Changes .................. .......... 69
Standard Descriptors Vs. Panelist's Descriptors .................. 76
Grapefruit Aroma ................... .......... ..... 80
D ilution Analysis ......................................... 81
Sulfur Compounds in Grapefruit ................................... 82
Detection ................... .......................... 82
Processing and Maturity Effects ............................ 84
p-menthene-8-thiol .............................. ......... 88
Correlation Between Aroma Components and Sensory Measurements ....... 89

Juice Classification .. .......... .......... 89
Sensory Analysis ............ ......... .. ......... .. 90
Univariate Analysis .................... ............. 93
Taste components ............................ .. 93
Aroma components ..................... ..... 100
Multivariate Statistical Analysis .. ...... .......... .... 104
Flavor models using taste components .............. .. ..105
Flavor models using aroma components ........ .. .. 107
Flavor models using aroma and taste components .......... 111

5. CONCLUSIONS ..................................... .. ..... 119
Correlation Between Preference and Analytical Measurements ............ 119
Aroma Extraction M ethods ...................................... 120
GC-Olfactometry ................. ........ ................. 121
Sulfur Compounds in Grapefruit Juice .............. ...... ... 122
Correlations Between Aroma Components and Sensory Measurements ..... 123



B. MASS SPECTRUM OF VANILLIN .......... ..... ........... 127


JUICE ................ ............................... 132

LIST OF REFERENCES ........ .......................... 136

BIOGRAPHICAL SKETCH .. .......... .............. ... .. 144


Iabk page

1. Calibration equations used for calculating the concentrations of components
detected in GC-FID ................... ................ ........ 29

2. Maximum, minimum and average area percent for components extracted with
methylene chloride. ..................................... ..... 38

3. Univariate correlations of selected volatile and non-volatile data with preference
category. ............ ...................... .. ......... 41

4. Forward stepwise discriminant analysis (methylene chloride extractions). ..... 50

5. Discriminant analysis classification results (methylene chloride extracts). ..... 51

6. Percent relative standard deviation for different aroma extraction methods in
grapefruit juice. ................................................ 60

7. Top note peak areas for different aroma extraction methods. .............. 61

8. Formation and loss of aroma attributes due to pasteurization in early season red
grapefruit juices. ...................................... ....... 72

9. Concentration levels (ppm) of components in early, mid and late season red
grapefruit juices. .............................................. 75

10. Aroma descriptors used by panelists from GC-O experiments of citrus standards 78

11. Comparison of standard (Arctander lexicon) with panelist descriptors ....... 79

12. List of components present in 16x concentrated juice extract and their intensities
and aroma attributes. .................. ...................... 83

13. Minimum and maximum descriptive sensory panel scores for grapefruit juices. 92

14. Univariate correlations between sensory and taste components (Brix, acid, ratio,
limonin and naringin). ............................ .. .. 94

15. Univariate correlations between 26 aroma active volatiles and sensory scores. 101

16. Squared mahalanobis distances for groups separated by taste components (Brix,
acid, ratio, limonin, and naringin). ........ .......... 108

17. Squared mahalanobis distances for 26 aroma and 5 taste components (Standard
Discriminant Analysis). .............. ..... ................ 112

18. Forward step wise discriminant analysis-volatiles and taste components (Number
of steps and corresponding component). ................ .... ... .. 116

19. Comparison of sensory and statistical classification of grapefruit juices. (Model
has been tested using 17 aroma components and 5 taste components) ....... 117



1 Standard curve used for calculation of Kovat's retention indices for volatile
com ponents........ .... ..... ........... ......... .......... 27

2 Calibration curves used for quantifying the volatiles. (A) propyl benzene, (B)
myrcene, (C) linalool, (D) nootkatone. ............................... 28

3 Calibration curve for s-methyl-thiobutanoate (sulfur compounds). .......... 30

4 Sample ballot for the grapefruit juice descriptive sensory panel. ........... 32

5 Chromatogram of methylene chloride extract of pasteurized (NFC) grapefruit
juice on a DB-5 column ....................... .. ... ............ 37

6a Eigenvector values of PC 1 vs PC 2 from principal component analysis of all 57
volatile and non-volatile components, where = high preference category, O =
medium preference category and A= low preference category. ............. 44

6b Eigenvector values of PC 1 vs PC 3 from principal component analysis of all 57
volatile and non-volatile components, where 0 = high preference category, o =
medium preference category and A= low preference category ............. 45

7a Peak Areas of linalool and caryophyllene from 29 grapefruit juice extracts analyzed
in triplicate, where = high preference category, o = medium preference category
and A= low preference category. ............................. 47

7b Peak Areas of myrcene and caryophyllene from 29 grapefruit juice extracts
analyzed in triplicate, where 0 = high preference category, o = medium preference
category and A= low preference category. ................... .. 48

8a Canonical Discriminant Analysis of using myrcene, linalool, Brix, and the peaks at
RI 1677 and 1126, where = high preference category, 0 = medium preference
category and A= low preference category. ........................... 52


8b Canonical discriminant analysis using thirteen variables (Brix/Acid ratio, RI-935,
cis linalool oxide, Nonanal, allo-ocimene, a-terpineol, Decanal, RI-1299, a-
copaene, 1-gurjunene, RI-1762, RI-1796) where = high preference category, o
= medium preference category and A= low preference category. ........... 53

9 Chromatogram classification of pasteurized grapefruit juice (pentane-diethyl ether
extraction). ................ ............ ........ ........... 57

10 Aroma extraction methods in grapefruit juice. A) liquid-liquid extraction (pentane-
diethyl ether 1:1), B) static head space extraction (solid phase microextraction -
SPME), C) dynamic head space purge and trap solvent elution (Tenax/charcoal
trap). ............... .. .... .... ........ ............ 59

11 Comparison of aromagram from OSME and chromatograms from FID and SCD.66

12 Formation of vanillin from ferulic acid. .................. .. .. 68

13a Number of aroma active components at different maturities in unpasteurized
grapefruit juice. A) early season, B) mid season, C) late season. ............ 70

13b Number of aroma active components at different maturities in pasteurized
grapefruit juice. A) early season, B) mid season, C) late season. ........... 71

14 Concentrations of components in grapefruit juice. A) unpasteurized juices, B)
pasteurized juices: (N) early season, (0) mid season, (n) late season. ........ 74

15 Acid catalyzed hydration of limonene. ................. ....... 77

16 Sulfur chemiluminescence reactions. .............. ....... 85

17a Total number of sulfur peaks at different maturities in pasteurized grapefruit juice.
A) early season, B) mid season, C) late season. ........... ......... 86

17b Effect of pasteurization on sulfur compounds in early season grapefruit juice. A)
unpasteurized, B) pasteurized. ...... ........... .. 87

18 Correlation between limonin concentration with bitterness score. ........... 95

19 Correlation between overall flavor score and sweet/tart balance. .......... 97

20 Correlation between aroma quality score and overall flavor score. ........... 99

21 Correlation between aroma quality and nootkatone peak area. ........... .103


22 Standard discriminant analysis using 5 taste components: (4) worst category, (0)
fair category, (A) good category, (U) best category juices. ............. 106

23 Forward stepwise discriminant analysis using 17 aroma and 4 taste components:
(4) worst category, (@) fair category, (A) good category, (U) best category
juices. .................................................. 114

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



Prashanthi Jella

August, 1998

Chairperson: Russell L. Rouseff
Major Department: Food Science and Human Nutrition

This study represents the most comprehensive determination of aroma active

volatiles and sulfur compounds in grapefruit juice reported to date.

Initial studies correlated GC peak areas of 52 volatile components (in methylene

chloride juice extracts) plus 5 taste components with sensory preference. Highly preferred

juices were associated with low myrcene, low linalool and intermediate levels of p-

caryophyllene. Since concentrated methylene chloride extracts contained few highly

volatile components, a search for a more complete aroma extraction procedure revealed

the superiority of pentane-diethyl ether extraction. This extraction gave 73% more top

note peak area than methylene chloride liquid-liquid extraction.

Approximately 80 peaks were separated using GC-FID, of which 37-49

components were aroma active. Twenty-five of these aroma active components had

intensities high enough to be considered key aroma components. Vanillin was one of the

aroma active peaks detected in grapefruit juice for the first time using GC-olfactometry

and GC-MS.

A routine method for grapefruit juice sulfur compounds using sulfur

chemiluminescence detection was developed for the first time. Twenty-two sulfur

compounds were detected. Total peak area increased with pasteurization but decreased

with maturity. p-menthene-8-thiol, a key aroma impact component, increased as much as

143% after pasteurization.

FID peak areas of 26 aroma active volatile components extracted with pentane-

diethyl ether and 5 taste components were correlated with sensory descriptive panel

results. Myrcene and a-terpineol correlated negatively with aroma intensity and quality.

Grapefruit aroma quality correlated significantly with overall flavor score (r=0.54 at

p<0.05). This is an important conclusion as current industry standards are based only on

taste components. The worst juices were effectively separated using taste components

whereas aroma active components separated best juices. One hundred percent separation

was obtained in the training set when 17 aroma active volatiles and 4 taste components

were used to classify the juices based on their quality. This model was tested by

evaluating 18 samples not used in the training set. Sixteen of 18 samples were correctly

classified within one flavor category. The main application of this flavor model is in

grapefruit juice processing industry where the processors can use it to predict juice



The citrus industry is one of the largest fruit crop industries in the United States.

Florida ranks first among the citrus producing regions in North America. Other citrus

producing areas include California, Texas, Arizona, and Mexico. The juice produced from

citrus fruits also constitutes the majority of fruit juices consumed in the United States and

around the world (Kimball, 1991).

Grapefruit (Citrusparadisi Mcfadyen) has a highly distinctive flavor with slight

bitterness and tanginess. Florida is the world's leading producer of this fruit with a record

production of 2.4 million tons in the year 1996-97. However, the value of the crop for

this season was $ 68,436,000 which is the lowest since 1969-70 (Citrus Summary, 1997).

In Florida, approximately half of the grapefruit grown is processed (Kimball, 1991). A

variety of products ranging from pasteurized not-from-concentrate to thermally

concentrated frozen grapefruit juice are processed .

Color, taste, and aroma quality of citrus juices can have a pronounced influence on

consumer preferences and purchase decisions. According to A.C. Nielsen numbers for

supermarkets, the demand for grapefruit juices has decreased (50 million gallons from

1991 to 42.2 million gallons in 1996), while production (61 million boxes of fruits for

1996) of grapefruit increased (Stinson and Barros, 1997). Decreasing popularity of this

fruit is reflected in the economic abandonment of 3 million boxes each of white and

colored seedless varieties of grapefruit (Citrus Summary, 1997). Considerable effort has

been expended towards the isolation, identification and quantitation of compounds

influencing the taste of grapefruit juice (Attaway, 1977; Rouseffet al., 1980; Fellers et al.,

1986). The taste factors influencing the flavor of grapefruit juice are sweetness, tartness,

balance of sweet/tart and bitterness. The current industry standards also depend on these

four factors. Since aroma is a key contributor for any perceived flavor, the purpose of this

study was to determine the relative contribution of aroma, and to determine which of these

components are important to overall acceptance of grapefruit juice. Specifically, the

objectives of this study were to

I. determine the flavor impact components in grapefruit juice, and

II. develop a model that can predict juice acceptance.

These goals can be achieved by

1. Identifying the volatiles in 40 processed NFC Florida grapefruit juices using high
resolution capillary gas chromatography and mass spectrometry;

2. Determining the concentrations of bitter compounds in the above juices using high
pressure liquid chromatography (HPLC);

3. Determining which volatile components have aroma activity using a gas-
chromatography-olfactometry technique (OSME);

4. Evaluating and testing different extraction techniques to determine the technique
that produces the most representative volatile profile;

5. Developing an analytical method to determine potent low level sulfur compounds
such as 1-p-menthene-8-thiol;

6. Training and conducting sensory panels (sniff and descriptive taste panel); and


7. Determining the relationship between sensory and analytical data using
multivariate statistics.



Citrus juices are becoming increasingly popular due to their unique flavor and

perceived health benefits. Flavor is a combination of both taste and aroma. In citrus

juices, flavor is affected by taste components like limonin, naringin, sugars and acid, and

by volatile aroma compounds. Considerable effort has been expended towards the

isolation, identification and quantitation of compounds influencing the taste of grapefruit

juice (Attaway, 1977; Rouseffet al., 1980; Fellers et al., 1986). The common consensus

of these studies is that a direct relationship exists between bitterness and flavor.

Extensive research has been conducted to identify and quantify volatile

components in grapefruit products (Moshonas and Shaw, 1971; N6fiez et al., 1985;

Cadwallader and Xu, 1994), but few workers have evaluated the relative sensory

significance of these compounds. Since 1989, a total of 264 volatile constituents have

been reported in grapefruit (Maarse and Visscher, 1989).

Nootkatone was suggested as the key flavor impact compound as early as 1970

(Stevens et al., 1970). However, the importance of nootkatone has been questioned as

Shaw and Wilson (1981) found that nootkatone when added to oil and juice, had a

significant flavor impact in oil, but very little impact in juice. They concluded that there

must be other components that affect the flavor of grapefruit juice. Subsequently, a

terpene-thiol, chemically known as I-p-menthene-8-thiol, was reported by Demole et al.

(1982), and is now considered one of the most potent flavor compounds found in nature.

The authors isolated 7.7 g of dried fraction from 100 L of canned grapefruit juice. A part

of this fraction (0.165 g) had sulfurous odor. There were eight compounds in this

fraction. One of the compounds wasp-menthene-8-thiol, with "a genuine, unmistakable

aroma of fresh grapefruit juice". The reported concentration of this compound in

grapefruit juice is 0.02 ppb (Maarse and Visscher, 1989), which is 200 times its threshold

level in the juice (Shaw, 1996). Its aroma threshold in water is 1x107 ppb (Demole et al.,

1982). However, until an analytical procedure is developed to quantify this compound at

the levels at which it exists in juice, it will not be possible to evaluate its relative

contribution to grapefruit juice flavor.

Statistical Correlations

Flavor is unquestionably one of the most important attributes of food and is

perceived as taste by the tongue and mouth and through the release of the volatile

components in the mouth which are sensed retronasally by the olfactory epithelium in the

nose (Ohloff, 1990).

Previous workers have developed models based on the correlations between

quantified volatile and sensory data (Jennings, 1977; Pino et al., 1986 a, b). A few orange

juice volatiles characterized using packed column gas chromatography and non-volatile

components and corresponding sensory hedonic scores were analyzed using multiple

regression (Attaway, 1972) or by using principal component analysis (Rouseff and Nagy,

1982). Multivariate statistical programs like principal component analysis (PCA)

investigate underlying relationships that exist between variables (Chien and Peppard,

1992). Pino (1982) used linear multiple regression to correlate the sensory and gas

chromatography (GC) data of grapefruit juice. Based on correlations, authors selected the

variables limonene, a-terpineol, linalool, and myrcene as the most significant in explaining

the sensory differences. Velez et al. (1993) classified orange juice samples stored at

different temperatures using PCA and GC-analysis. Increased temperature and storage

time generally reduced flavor quality. They observed that butanol, a-terpineol and furfural

correlated with increasing storage temperatures while linalool and terpin-4-ol correlated

best with storage time.

Even though Florida has been a world leader in the production of grapefruit juice,

no systematic study to determine the key flavor impact compounds from both aroma and

taste has been reported. Using canonical and cluster analysis, Pino et al. (1986a)

classified 24 commercial single strength grapefruit juices from different production days

and storage conditions. They concluded that nootkatone and an unknown component had

positive correlation to flavor while another unidentified component correlated negatively

with flavor.

In another experiment, Pino et al. (1986b) correlated sensory and chromatographic

measurements of grapefruit juice volatiles using multiple linear regression. Methyl

butyrate, ethyl butyrate, limonene, decanal and nootkatone correlated with positive

sensory perception while trans- and cis-epoxy dihydrolinalool and a-terpineol correlated

with unpleasantness of grapefruit juice. The statistical analysis used by the authors

identifies those components that change the most with the sensory measurements. In

other words, the compounds with high correlations may or may not be aroma active.


Excessive bitterness in grapefruit juice adversely affects flavor and marketability.

Compounds that are responsible for bitterness in grapefruit juice are limonin, nomilin, and

naringin. These compounds, in moderate quantities, provide the characteristic bite and

cleansing of the palette that is liked by most consumers of the juice (Fellers, 1991).

However, excessive quantities of these are also detrimental to consumer preference.

Maturity is one of the several factors influencing the content of these bitter

components (Berry and Tatum, 1986; Tatum et al., 1972). Albach et al. (1981a) observed

that naringin concentration in juice often increased in early spring (February, March, or

April) after the onset of rapid vegetative growth. In an other study, Albach et al. (1981b)

observed that limonin content was less than 6 ppm by March for most commercial

grapefruit varieties. In general, the authors concluded that limonin concentration

decreased rapidly as the season progressed, while naringin concentration remained steady

until spring, when it began to increase.

Rouseff(1982) reported that nomilin, a limonoid, is twice as bitter as limonin. The

authors quantified nomilin and limonin in commercial grapefruit juices produced in the

1978-79 season and observed low nomilin concentrations in all juices. Rouseffet al.

(1980) observed a consistent inverse relationship between bitterness and flavor during a

survey of canned single-strength grapefruit juice from 1977-1978 to 1979-1980. They

concluded that during a typical season bitterness decreased, flavor increased, limonin

decreased and naringin increased with fruit maturity.

Bitterness is one of the 4 basic tastes affecting the quality of juice. Fellers et al.

(1987) reported increased bitterness and tartness perception with increasing limonin

content, whereas sweetness perception decreased.

Naringin is present in the pulp, rag and albedo of the fruit (Attaway, 1977). The

presence of the bitter glycoside naringin in the juice depends upon extraction methods.

Therefore, hard squeezing of fruit and excess finishing of juice increases the naringin

content in juice.

To meet the requirements of Florida Department of Citrus (Fellers, 1990),

blending of different juices is done to keep the limonoids at a moderate concentrations.

Various techniques using insoluble polymers, enzymes, and immobilized bacteria (Wilson

et al., 1989) have been tried for reducing these compounds in citrus juices.

Immobilized bacterial cells were used by Hasegawa (1983) to reduce the limonin

content in orange juice. Carbon dioxide at pressures between 21 and 41 Mpa were used

by Kimball (1981) to reduce limonin by 25% from Washington navel orange juice. Ion

exchange and adsorbent resins are currently being used to reduce bitter components.

It was reported by Johnson and Chandler (1985) that juice with unacceptably high

bitterness can be debittered using IRA-68, S-861, and IRC-84 resin columns to produce

an acceptable Florida grapefruit juice. Residence time in the column bed and the

temperature of the bed was found to be critical in reducing the amount of limonin and

naringin (Wilson et al., 1989).


A hybrid technique has recently been developed that directly measures only those

components that are causative (that is, have aroma activity). It combines the resolving

power of a capillary gas chromatograph with modem sensory analysis. The technique is

called gas-chromatography olfactometry (GC-O). It utilizes a human assessor to

determine which of the many chromatographic peaks have aroma activity and

characterizes that odor. Some of the GC-O techniques available today are Charm

Analysis, Aroma Extraction Dilution Analysis (AEDA) and OSME which is a time

intensity method. Charm Analysis and AEDA are based on the determination of odor

detection thresholds of the compounds through a series of dilutions while OSME

determines intensities without dilutions.

Charm Analysis

Acree et al.(1984) developed the Charm analysis technique, and has used this

technique to evaluate a variety of products. Cunnigham et al. (1986) analyzed apple

volatiles and identified the 12 most odor active peaks. A generalized description of apple

odor produced by combining samples showed beta-damascenone, butyl, isoamyl, and

hexyl hexanoates, along with ethyl, propyl and hexyl butanoates, to be important to the

odor of most apple cultivars. Differences between fresh and pasteurized orange juices

were characterized by Marin et al. (1992) using this technique. The authors observed

large changes in odor activity for linalool, ethylbutyrate, vanillin and several unknown



Aroma extraction dilution analysis, developed by Schieberle and Grosch (1984), is

based on serial dilutions like the Charm analysis. In this method, serial dilutions (1:2)

are made and analyzed until the odor is perceived by human subjects. The resultant

intensities are plotted in an aromagram. Schieberle and Grosch (1988) used AEDA to

identify indicator substances for the assessment of the deterioration of lemon oil flavorings

in acidic foods. Fresh samples and samples stored for 30 days (at 37C) were compared.

The study suggested that p-methyl acetophenone, p-cresol, p-cymene, and fenchyl alcohol

are the most potent storage indicator components in the lemon oil.

Hinterholzer and Schieberle (1998) identified the most odor active volatiles in hand

squeezed juice of late Valencia oranges. The authors identified ethyl butyrate (fruity), Z-

hex-3-enal (green) and 3,4,5,7-tetrahydro-3,6-dimethyl-2(3H)-benzofuranone (sweet,

spicy) as the potent odorants with highest flavor dilution factor.


da Silva et al. (1994) claimed that the dilution techniques mentioned above would

not give accurate information, since the odorants have different intensity functions above

their threshold levels. The authors proposed and developed a new GC-O methodology

based on psycho-physical laws called OSME (Greek word meaning smell).

OSME is a time intensity procedure which determines the intensity of the

perceived odor without dilution. In this method, the trained subjects sniff the effluents

from GC mixed with humidified air, and directly records the odor intensity and duration of

each odor active component while describing its odor quality. Intensities of individual

components are plotted versus elution time and the resultant graphical representation is

known as an aromagram.

Orange aqueous essence was analyzed by Bazemore (1995) using OSME.

Octanal, linalool, and ethyl butanoate were found to have the strongest aroma in both

reflux and no reflux samples of aqueous orange essence.

OSME has also been used to differentiate Pinot Noir wines from grapes of

different maturities (Miranda-Lopez et al., 1992). Spicy (ethyl octanoate), vegetative,

herbal, and vanilla (ethyl vanillin) aroma's were detected in wines made from late maturity

grapes. The authors also found that 45 to 60% of odor active peaks found in GC-O were

not detected by an analytical detector (GC-FID).

One characteristic feature of GC-O methods is the occurrence of peaks in the

aromagram which might not match a corresponding FID peak. This occurs because the

human nose is much more sensitive to some of the compounds than are analytical

detectors. Mistry et al. (1997) detected a musty off-flavor in the extracts ofbeetsugar.

However, no FID peaks were detected in the region that produced the most aroma

activity. Upon enrichment of the extract by the authors, geosmin was identified as the

compound producing the musty odor.

Sample Preparation

Extraction and isolation of the representative aroma compounds in a food matrix is

one of the critical steps in flavor research. No single extraction method can be considered

universal, rather the extraction procedure employed depends on the needs of the

researcher and the nature of the sample. Various isolation procedures for volatile

components have been compared by many researchers. Weurman (1969) presented an in-

depth description of different isolation techniques used in odor research. In this study,

several different extraction techniques were evaluated for optimum odor recovery.

Nunez et al. (1984) compared five methods including solvent extraction (batch

wise and continuous), distillation and simultaneous distillation solvent extraction-SDE,

(Likens-Nickerson and Godefroot et al. apparatus) for volatile components of grapefruit

juice. The two SDE methods were reported to be most suitable for grapefruit juice in

terms of rapidity, reduced solvent removal and strong representative odor of the sample.

Jennings (1977) sampled peach volatiles with the Likens-Nickerson apparatus and

porous polymer traps. The polymer trap essence exhibited larger amounts of lower boiling

compounds than did the distillation extraction essence. When extended trapping periods

were utilized, higher boiling compounds were also present in the polymer trap essence


Moshonas and Shaw (1971 and 1982) isolated orange juice volatiles using

dichloromethane solvent extraction. Ethanol was not extracted by this method, which

aided in the analysis of other compounds normally masked by the large ethanol peak.

Moshanas and Shaw (1992) compared the static and dynamic head space methods

for orange juice volatiles. Acetaldehyde, methanol, methyl butyrate, a-pinene, y-terpinene

decanal and linalool were extracted in greater quantities by static head space, while ethyl

butyrate, hexanal, ethyl hexanoate and cis -3-hexenol were higher in dynamic head space.

Umano and Shibamoto (1988) described a new method in which head space

volatiles were purged into water in a gas washing bottle and simultaneously continuously

extracted with dichloromethane. An aqueous solution containing cysteaminee) was used

to trap aldehydes (as derivatives ofthiazolidine) and a phenylenediamine solution to trap

dicarbonyls (as quinozalines). GC revealed 22, 25 and 130 peaks in the whole grapefruit,

grapefruit juice and grapefruit peel extracts respectively, the predominant component

being limonene in all cases.


Solid-phase micro extraction (SPME) is a relatively new technique in which

analytes of interest partition from the sample matrix into a polymeric solid coating. SPME

was first reported by Zhang et al. (1994) and has been used in qualitative and quantitative

studies of citrus juices (Matich et al., 1996).

Comparisons between traditional head space Tenax adsorption/desorption and

head space SPME were made by Pelusio et al.(1995). According to the author, when

polydimethylsiloxane fiber coating was used, GC-MS analyses of the aromas showed that

the SPME technique was less suitable for quantitative analysis due to lower affinity of the

fiber for more polar and very volatile compounds.

Steffen and Pawliszyn (1996) reported 1- 20% relative standard deviation for most

components in orange and grapefruit juices analyzed by SPME. According to Xiaogen

and Peppard (1994), addition of salt enhanced the amount of volatiles absorbed using


SPME GC-MS enabled detection of more than 50 volatile compounds including

hydrocarbons, aldehydes, carboxylic acids, phenolic compounds, esters, ketones, lactones,

alcohols, N-containing compounds and S-containing compounds in the head space of milk

powder (Stevenson and Chen, 1996). Chin et al. (1996) observed that SPME fibers

extracted major cheese volatile components, but minor components such as volatile sulfur

compounds were not observed.

The principle behind SPME is the partitioning of analytes between sample matrix

and the extraction medium (Zhang et al., 1994). The amount absorbed by the coating at

equilibrium is directly related to the concentration of the component in the sample n =

KJVfCoV,/ (KrVf+V,) where n is the mass of the analyte absorbed by the coating; V, and

V, are volumes of coating and sample respectively; K, is the partition coefficient of the

analyte between the coating and the sample matrix; Co is the initial concentration of the

analyte in the sample. However, since V >> KfVf, in food analysis, the earlier equation

can be simplified as n = K.VC0 and hence is independent of the sample volume. This is

one feature that makes SPME suitable for food analysis.

Sulfur Compounds

Sulfur compounds play a major role in determining the flavor characteristics of

many food substances. Sulfur compounds are often formed as a result of the enzymatic

process when plants are cut or chewed, releasing flavor precursors and enzymes from

rupturing cells. Sulfur components are unusual since in low concentrations they are

responsible for many positive sensory qualities in foods and flavorings. However, higher

levels of the identical compound often result in off flavors (Tressl and Silwar, 1981). The

authors reported that furfurylmercaptan at 10-500 ng/L had a fresh roasted coffee aroma,

while at 1000 ng/L a sulfury stale coffee aroma was perceived.

Another aspect of organic sulfur compounds at low concentration is the influence

of functional groups (Boelens et al., 1993). The authors reported that the odor threshold

values of tertiary thiols are 300 3000 times lower than those of primary and secondary

thiols. The example they quoted for beer is 2-methyl-2-propanethiol which has a threshold

value of 80 units, while 2-methyl-l-propanethiol has a value of 2500 units. Although,

sulfur components are present only in trace quantities in most food materials, their

contribution to the overall flavor quality is significant due to their extremely low aroma


In spite of the significant role of sulfur compounds in the food matrix, there are

only a few reports regarding their affect in citrus juices. Shaw et al. (1980) detected

hydrogen sulfide, methyl sulfide, sulfur dioxide, methane thiol, and some higher alkyl

sulfides using a flame photometric detector in orange juice samples. Since concentrations

of H2S and methyl sulfide in orange juice were greater than their reported aroma

thresholds, these components may have a significant impact on overall juice quality. In

another study, Shaw and Nagy (1981) concluded that early season orange and grapefruit

juice had higher levels of H2S. When sensory analysis was conducted on these juices, the

panelists reported a harsher (pungent) aroma, and the authors attributed this to higher HS

levels. This attribute was not detected by the authors in late season orange and grapefruit


Demole et al. (1982) isolated and characterized p-menthene-8-thiol, which had the

"unmistakable aroma of fresh grapefruit." They found that when combined with

nootkatone, the mixture gave a "full bodied flavor" of fresh grapefruit. p-menthene-8-

thiol undergoes cyclization to form 2,8,-epithio cis-p-menthane, which also has a

characteristic grapefruit aroma. The odor threshold of this compound was 9 ppb (Maarse

and Visscher, 1989). The cyclization reaction takes place at room temperature in the

presence of light and these two compounds are reported to co-occur in grapefruit

(Demole et al., 1982).


A major goal of this project was to quantify and characterize the aroma impact

components in not-from-concentrate grapefruit juices. Non-volatile flavor attributes such

as sweetness, sourness and bitterness were also evaluated by measuring oBrix, titratable

acid, limonin and naringin separately in order to evaluate the relative contribution of taste

vs. aroma components. Sensory attributes were quantified and correlated with analytical

measurements. Experimental design and analytical techniques used to achieve this

objective are discussed in this chapter.

Grapefruit Juice Sample Collection

Survey Samples

Methylene chloride extracts

Twenty-nine not-from-concentrate (NFC) grapefruit juice samples were obtained

from processors with processing dates ranging from November, 1995, to June, 1996, and

stored at -8 C until analyzed. Both red/pink and white juices were used in this study.

Authentic solvents were purchased from Fisher Scientific (Pittsburgh, PA). Standards

used for quantifying volatiles and non-volatiles were purchased from Aldrich Chemical

Company Inc. (Milwaukee, WI). A few standards were obtained as gifts from SunPure,

Inc. (Lakeland, FL) or Givaudan Roure (Lakeland, FL).

Pentane-diethyl ether extracts

Forty not-from-concentrate (NFC) grapefruit juice samples (2 QT gable top

cartons) were purchased from a local supermarket with manufacturing dates ranging from

January, 1997, to June, 1997, and stored at -8 C until analyzed. Both red/pink and white

juices were used in this study. Sources for solvents and standards were the same as for the

methylene chloride extracts study.

GC-Olfactometry Samples

Early (November, 1996), mid (January, 1997) and late (May, 1997) season

grapefruit juice samples were obtained from the Florida Department of Citrus. Grove run

red grapefruit were purchased from a local packinghouse and processed in the pilot plant

located at the Citrus Research and Education Center, in Lake Alfred. Fruits were washed,

dried and sized for the extractors in the pilot plant. Extraction was accomplished using

commercial FMC model 391-B and 491 extractors with standard juice settings. An FMC

model 35 juice finisher was used with a moderate squeeze setting. The finished juice was

pumped to the holding tank prior to pasteurization. Pasteurization was done using a

Feldmeier tube-in-shell pasteurizer. The juice was heated to 90.60C at a flow rate of 1

gallon per minute. Samples were packaged in 32 oz clear glass bottles and stored at -8 C

until analyzed. Both unpasteurized and pasteurized red grapefruit juices were obtained.

Samples consisted of two bottles for each juice type.

Sample Preparation

Liquid-Liquid Extraction with Methylene Chloride

Extraction of volatiles was accomplished with methylene chloride using the

method described by Parliament (1986) and modified by Klim and Nagy (1992). Eight mL

of juice were added to 4 mL of methylene chloride and mixed using a Mixxor-like

apparatus. The apparatus consisted of two syringes : 50 cc and 30 cc capacity connected

with an 8 cm long, 3 mm outer diameter stainless steel connector. The mixture of juice

and solvent was poured in the larger syringe and, using forward and backward motion, the

mixture was pumped into and out of the smaller syringe. The juice and the solvent were

mixed for ca 2 minutes. The emulsion was broken by centrifuging for 10 min (15000 g).

The lower solvent layer of approximately 3 mL was collected for analysis. An internal

standard, 6 gL of propyl benzene, was added and the extract was concentrated to about

30 pL in a 100 pL graduated taper vial. Concentration was accomplished by blowing

nitrogen gas at a flow rate of 40 mL / min across the surface. Concentrated extracts were

prepared fresh every morning and analyzed the same day. Each juice sample was

extracted twice and each extract analyzed in duplicate.

Liquid-Liquid Extraction with Pentane-Diethyl ether (lI):

Extraction of the volatiles was accomplished according to the previously

described method except a 1:1 mixture of pentane and diethyl ether was used in the place

ofmethylene chloride. Two internal standards, propyl benzene (50 lL of 100 ppm) and 2-

heptadecanone (25 pL of 500 ppm), were added to 8 mL of juice and extracted. The

extracts were concentrated to 50 pL using the same procedure as that previously

described. Each sample was analyzed in duplicate.

Dynamic Head Space Purge and Trap Solvent Elution

Dynamic head space extraction was accomplished using a two necked 25 mL

round bottom flask. Ten mL of juice were added to the flask along with a stir bar.

Nitrogen was impringed upon the juice surface at a rate of 40 mL / min through one of the

flask necks. To the other opening, a 2 mm i.d. glass column comprising powdered

charcoal (Supelco, Bellefonte, PA) and Tenax (Supelco, Bellefonte, PA) in a 1:3 (v/v)

ratio was attached. Juice was heated to 37 C using a constant temperature water bath.

Volatiles were trapped in the column for 30 min. The column was removed and purged

with dry nitrogen (20 mL/min) for ca. 1 minute to reduce trapped moisture.

Three mL of(l: 1) pentane and diethyl ether mixture were used to elute volatiles

from trap materials. Extracts were concentrated in the same manner as with the methylene

chloride extracts. The column was cleaned both before and after extraction using 3 to 4

times the column volume of pentane.

Extraction Procedure for Sulfur Compounds

Extraction of the volatiles was accomplished using the same method as that

described for methylene chloride except ethyl acetate was used as the solvent. S-methyl

thio butanoate (15 VL of 10 ppm) was added to 8 mL ofjuice as an internal standard and

extracted. Extracts were concentrated to 50 pL using nitrogen with the procedure

described earlier for the methylene chloride extracts. All samples were analyzed in


Extraction Procedure for GC-Olfactometry Analysis

Extraction of juice volatiles was accomplished using the method described for

pentane-diethyl ether extractions. Two internal standards, benzaldehyde (25 pL of 5000

ppm) and methyl jasmonate (25 pL of 5000 ppm), were added to 8 mL of juice and

extracted. Extracts were concentrated to 50 pL using dry nitrogen as previously

described. Each sample was analyzed four times using three detectors (Flame Ionization

Detector (FID), Sulfur Chemiluminescence Detector (SCD) and OSME).

Instrumental Techniques

GC-Flame Ionization Detector

Individual volatile constituents were separated using an HP-5890 GC (Palo Alto,

CA) with a flame ionization detector and a 30 m x 0.25 mm i.d. x 0.5 pm film thickness

low bleed DB-5 column (J&W Scientific; Folsom, CA). The oven temperature was

programmed from 35 to 275 OC at 6 oC/min with helium at a flow rate of 2.19 mL/min

(34.6 cm/sec linear velocity). The injector temperature was maintained at 250 OC and

detector temperature at 320 OC. The nitrogen gas flow was maintained at 19 mL/min,

while air and hydrogen flows were maintained at 296 and 35 mL/min, respectively. The

injection volume was 1 pL for methylene chloride extracts and 0.5 pL for pentane-diethyl

ether and ethyl acetate extracts. Injection was split-less. Chromatograms were recorded

and integrated using Chrom Perfect (Justice Innovations, Mountain View, CA). The data

acquisition rate was 10 pt/sec. Chromatograms for methylene chloride extracts were

recorded and integrated using an APEX Chromotography Workstation (Autochrom Inc.,

Milford, MA) with a four channel data system. Data acquisition rate was 0.4 s/point.

GC-Sulfur Chemiluminescence Detector

Volatile constituents were separated using an HP-5890 GC (Palo Alto, CA)

equipped with a sulfur chemiluminescence detector (Seivers Instruments Inc., Boulder,

CO) and a 30 m x 0.25 mm i.d. x 0.5 lpm film thickness low bleed DB-5 column (J&W

Scientific; Folsom, CA). Oven temperature was programmed from 35 to 275 OC at 6

OC/min with helium at a flow rate of 2.19 mL/min. Injector temperature was maintained at

250 OC. Internal temperature of the SCD burner head was 780 oC. Air and hydrogen

were maintained at 114 and 9 mL/min respectively. Cell pressure was maintained at 5.5

torr and the ozone at 8.75 psi. The injection volume was 0.5 pL in split-less mode.

Chromatograms were recorded and integrated using Chrom Perfect (Justice Innovations,

Mountain View, CA). The data acquisition rate was 10 pt/sec.


The individual volatile constituents were separated using an HP-5890 GC (Palo

Alto, CA) with a sniff port (DATU, Geneva, New York) and a 30 mx 0.25 mm i.d. x 0.5

pm film thickness low bleed DB-5 column (J&W Scientific; Folsom, CA), with helium at a

flow rate of 1.55 mL/min. Oven temperature was programmed from 35 to 275 OC at 6

OC/min Injector temperature was maintained at 250 'C and detector temperature at

320 C.

Purified air was obtained by passing compressed air through drierite and a

molecular sieve 5A (Alltech, Deerfield, IL) and directed into a temperature controlled,

water filled round bottomed flask fitted with fitted glass impringers. Water temperature

was maintained at 35 C. Airflow through the sniff port was 11.2 L/min. The stainless

steel sniff port tube was 70 cm long and 1 cm in diameter.

Sniffing began after the solvent had eluted off the column (ca 3 minutes). Panelists

were requested to sit in a comfortable position and asked to indicate their responses using

a linear potentiometer (variable resistor). The device had a pointer which the subject

moved from left to right and back again across a 15 point structured scale (0=none, 7.5 =

moderate and 15 = extreme). Time and intensity were recorded by the OSME soft ware

system, installed on a 386-PC. The component odor was described by the panelist and

recorded by the researcher. Maximum sniffing time was 30 minutes.

GC-Mass Spectrometry

All GC-MS data were collected using a Finnigan GCQ Plus system (Finnigan

Corp, San Jose, CA) using helium (99.999%) for the GC carrier gas and the collision/bath

gas in the ion trap. Injector temperature was 250 C. Samples (0.2-1.0 pLL) were injected

using the split less mode with a purge time of 1.5 min. The initial column temperature was

held at 35 C for 3 min followed by a 4 "C/min temperature ramp to 221 C which was

followed by a 10 C/min ramp to 275 C which was held for 1.1 min to elute high boiling

components in extracts. Linear velocity was 31.9 cm/sec through a 30 m x 0.25 mm id,

0.25 m RTX5-MS column (Restek Corp, Bellefonte, PA). Transfer line and ion source

temperatures were 275 "C and 170 C. The mass spectrometer had a delay of 4 minutes

to avoid the solvent peak, and then scanned from m/z 40 to m/z 300 in order to achieve 7

spectra per second. Ionization energy was set at 70 eV.

Limonin and Naringin Analysis Using HPLC

Sample preparation

Limonin and naringin for grapefruit juice samples (extracted with methylene

chloride) were analyzed according to the method developed by Widmer and Martin

(1994). In a 10 mL volumetric flask, 5 mL ofjuice were equilibrated for 5 min at 90 C.

The sample was diluted to 10 mL with 40 % acetonitrile and filtered through a Whatman

GDX 0.45 p filter. About 2 mL of filtered sample were placed into 2.5 mL Snap-ItsT

(National Scientific Company, Quakertown, PA) glass vials and used for further analysis.

For grapefruit juice samples extracted with pentane-diethyl ether solvent mixture,

limonin and naringin analysis was similar to the procedure described above, except that

these were not heated prior to HPLC analysis.

HPLC instrumentation

A Thermo Separations (San Jose, CA) LC system (Spectra Focus Optical

Scanning detector and P4000 gradient pump) with a Spectra Physics AS 3000 (San Jose,

CA) auto sampler was used for the analysis of limonin. A Waters 6000A pump (Milford,

MA) with a Waters 440 (Milford, MA) two channel UV absorbance detector equipped

with a 280 nm filter was used to determine naringin. Chromatograms were recorded and

integrated with a Thermo Separations 4290 (San Jose, CA) integrator and Winner on

Windows 4290 (San Jose, CA). Separations were achieved using a 4.6 mm x 150 mm 5 p

CN analytical column (MacMod Analytical Inc., Chadds Ford, PA) for limonin and a 4.6

mm x 150 mm 5 p C-18 analytical column (Kromasil C-18, Higgins Analytical, Mountain

View, CA) for naringin. The mobile phase consisted of water acetonitrilee (80.5:19.5) for

naringin analysis, and water / acetonitrile (63:37) for limonin analysis. The injection

volume was 40 pL and flow rates of 1.0 mL/min were used.

Peak Identification and Ouantification

Chromatographic peaks were identified using their mass spectra and comparison of

their observed Kovat's index with published Kovat's retention indices (Kovats, 1965).

Calculation of retention indices for individual peaks was done using retention time data

from a series of alkane standards run under the same conditions. Alkane standards

(Supelco Inc. Bellefonte, PA) from C 6 to C18 were used for this. Kovat's Indices for

these standards were calculated by multiplying the corresponding carbon number by a

factor of 100. Retention time (seconds) for the standards were plotted against their

corresponding Kovat's Indices (Figure 1). The resulting plot was used to fit an equation,

which was then used to calculate the retention indices for individual grapefruit juice

volatile components.

Quantification of some of the GC-FID peaks from early, mid and late season

grapefruit juice was done by using authentic standards obtained from Sun Pure, Inc.

(Lakeland, FL). Solutions of ethylbutyrate, propyl benzene, sabinene, myrcene, octanal,

linalool, decanal, nerol, P-caryophyllene, nootkatone, 2-heptadecanone were prepared at

concentrations ranging from 22 to 227 ppm and injected in duplicate. Calibrations plots

were generated by plotting the peak areas versus sample concentration. Sample plots

generated for 4 components are shown in Figure 2. Equations for the rest are given in

Table 1. FID peak areas obtained for grapefruit juices were normalized using the peak

area of internal standard.

Quantification of peaks from GC-SCD was done by analyzing s-methyl

thiobutanaote at five concentrations (10, 5, 1, 0.01, 0.001 ppm) in duplicate. The

calibration curve for these is shown in Figure 3.


1200 +


500 1000 1500
Retention time (sec)

KI = a+ (b*t2 + c/t + d) ln(t)
Where KI Kovat's retention index
t Retention time in seconds
and constants
a =592.525
b =2.003e-5
c =1.703 and
d =-12.342

2000 2500

Figure 1. Standard curve used for calculation ofkovat's retention indices for volatile

u i







U 10000







y =13983x+ 132945
R2 = 0.9943

0 50 100 150 200 250
Concentration (ppm)


0v -

0 50 100 150 200 250 0 50 100 150 200 250
Concentration (ppm) Concentration (ppm)

(C) (D)
Figure 2. Calibration curves used for quantifying the volatiles. (A) propyl benzene, (B) myrcene, (C) linalool, (D) nootkatone.

y =9122.8x -83140
R2= 0.9964

0 50 100 150 200 250
Concentration (ppm)


y =8766.1x -71662
R2 = 0.9951

y = 9715.7x 56480
R2= 0.997


Table 1. Calibration equations used for calculating the concentrations of components
detected in GC-FID.

Component Linear regression equation r-squared

Ethyl butyrate 6128*x- 29184 0.997

Propyl benzene 13983*x+ 132945 0.994

Sabinene 7586*x- 61320 0.997

Myrcene 9123*x 83140 0.996

Ocatanal 9093*x- 108034 0.968

Linalool 8766*x- 71662 0.995

Decanal 4951*x 44666 0.999

Nerol 9178*x- 69883 0.990

Caryophellene 8257*x 80289 0.993

Nootkatone 9716*x- 56480 0.997

2-heptadecanone 13889*x 16990 0.994

Note: x in the linear regression equation represents the area of the peak to be quantified


y = 50873x 2369.3
R2= 0.9997



0 2 4 6 8 10
Concentration (ppm)

Figure 3. Calibration curve for s-methyl-thiobutanoate (sulfur compounds).

Sensory Analysis

DOC Preference Panel

Bitterness taste thresholds of individual taste panelists were determined using 5-50

ppm oflimonin and 150-950 ppm ofnaringin aqueous solutions. Twenty-four untrained

panelists were used. A nine-point hedonic scale (forced choice) was used with 0

indicating dislike extremely, 9 indicating like extremely and 5 indicating neither like nor

dislike. Panelists were presented with three samples under illumination with red light and

asked to rate their preference. Samples were coded with random three digit numbers

randomly arranged on serving trays, and then presented to panelists.

USDA Descriptive Panel

This panel consisted of 12 trained panelists. Taste threshold characteristics of

individual taste panelists were determined using 5-50 ppm oflimonin and 150-500 ppm of

naringin solutions. The attributes rated were grapefruit aroma intensity, grapefruit aroma

quality, bitterness, balance of sweetness/tartness and overall flavor quality. A 15 cm line

segment scale was used with 0 indicating least intensity, 15 indicating highest intensity. A

sample ballot given to the panelists is represented in Figure 4. Panelists were presented

with four samples and a reference juice.

The reference juice (10 gallons: pasteurized, not-from-concentrate) was obtained

from a local juice processor and stored in 2 L amber colored glass bottles at -8 C.

Grapefruit Juice Sensory Panel
Please Read the Instructions

Name: Sample Number: 321

Aroma Analysis: Uncover the sample, take a deep sniff, and rate the quality & intensity
for the grapefruity aroma.
Taste the juice and mark down the intensity for Bitterness and Sweetness.
Based on the above attributes rate the overall flavor quality of the juice

Grapefruit Aroma Intensity

Grapefruit Aroma Ouality



Sweet/Tart Balance

7 0 7
More Sour More Sweet
Than Sweet Than Sour

0 15

Overall Flavor Ouality




Comments: (If any Off Flavor is percieved describe the attribute and rate it as
None, Moderate or Strong.
Any additional comments are also welcome).

Figure 4. Sample ballot for the grapefruit juice descriptive sensory panel.


Panelists were given this juice 6 times over 3 weeks and the scores for individual attributes

were averaged. Average scores for all attributes were marked on the ballot sheet to serve

as reference points for other samples. Panelist's consistency was checked by giving the

reference sample after every 10 grapefruit juice samples.

Samples were coded with random three digit numbers randomly arranged on

serving trays, and then presented to panelists.

Training of Panelists

GC-Olfactometry Panel

The panel consisted of 2 males and 1 female. Training consisted of three practice

runs with grapefruit juice extracts to familiarize the panelists with the sliding scale,

optimum positioning and breathing technique, and to provide practice with verbal

descriptors. In addition, a mixture of standard components typically found in grapefruit

juice was injected to familiarize the panelists with these odors and to help standardize their

descriptors. The results from the standard mixture are presented in Results and

Discussion section. To condition the olfactory senses, individual standard solutions (20

mL at concentration of 4 ppm) were smelled by the panelists prior to OSME analysis of all

the grapefruit samples. The standards consisted ofhexanal, ethyl butyrate, myrcene,

linalool, decanal, a-terpineol, p-menthene-8-thiol, and nootkatone.

Descriptive Panel

Twelve members (6 women and 6 men) were recruited from the United States

Department of Agriculture, Winter Haven, FL for a descriptive taste panel. The panelists

were of varied age groups and ethnic backgrounds. All panelists had some prior citrus

taste panel experience. Minimum and maximum values for ratio of total soluble solids : %

acid from the United States grapefruit juice grading system were used to train panelists.

Brix:acid ratio of 14 : 1 (70 g of sucrose and 5 g of citric acid) and 8 : 1 (40 g of sucrose

and 5 g of citric acid) were prepared using food grade sucrose and citric acid in water.

Naringin solutions of 200, 100 and 25 ppm (in water) were used for a bitterness standard.

All solutions were prepared using double distilled water. Fresh squeezed grapefruit juice

and fresh grapefruit peel were used as standards for grapefruit aroma quality and intensity.

Statistical Analysis

Principal components analysis in SAS (Version 6.11, SAS Institute, Cary, NC) was

used to evaluate the data set from the preference sensory panel and GC-FID data.

Univariate statistics and step wise multiple regression (forward) with Wilks Lambda was

also employed to identify those components which would be most differentiating between

sensory classifications. Canonical discriminant analysis (STATISTICA version 5.0, Stat

Soft, Tulsa, OK) was used to identify the peaks which would help in differentiating the

juice preference groups. The cross-validation component in this section was employed to

determine the classification significance for each sample. Mahalonobis distances were


used to judge the distances between the juice groups. Posterior probabilities were used to

predict the juice quality.


Correlations Between Preference and Analytical Measurements

Initial attempts to determine the aroma impact components of grapefruit juice

utilized methylene chloride extracts. Methylene chloride extraction was chosen as a means

of extracting aroma volatiles as it had been used as a solvent by several authors for

isolating citrus volatiles (Moshanas and Shaw, 1971; Parliament, 1986; Klim and Nagy,

1992). Figure 5 represents a typical chromatogram from a grapefruit juice methylene

chloride extract. It is important to note the relative absence of early eluting (low boiling)

components. Over 125 chromatographic peaks were resolved in the chromatogram.

However, some peaks were too small to be accurately quantified. Of the original 125

peaks in the chromatogram, 52 were selected for further studies. Identification of these

peaks was based on Kovat's retention index values and mass spectral data.

Maximum, minimum and average area values for these peaks are given in Table 2.

All components identified in Table 2 were also identified by Nuifiez et al. (1985) and

Maarse and Visscher (1989).

-0 0

0s 'T
*0 d

4., J 0.. t

0 5 10 15 20 25
Retention Time (mmin)

Figure 5. Chromatogram of not-from-concentrate grapefruit juice methylene chloride extract.

30 35

Table 2. Maximum, minimum and average area percent for components extracted with
methylene chloride.

Component Name/
Retention Index Average Area % Minimum Area% Maximum Area%

a-thujene 0.51 0.30 0.78

a-pinene 0.40 0.00 1.21
mvrcene 2.40 1.60 3.82
octanal 0.09 0.01 0.65

a-phellandrene 0.36 0.00 1.13

RI-1008 0.19 0.00 0.88

P-E-ocimene 0.45 0.17 0.85
y-terpinene 0.08 0.00 0.18

cts-linalooloxide 0.18 0.04 2.35
trans-linalooloxide 2.00 0.28 4.49

linalool 1.11 0.41 2.58
RI-1100 0.13 0.00 0.35

allo-ocimene 0.40 0.00 1 76
RI-1153 0.06 0.00 0.34

P-pinene oxide 0.19 0.00 0.64

nonanol 0.10 0.00 0.40
terpin-4-ol 0 48 0.04 1.07

RI-1192 0.18 0.00 0.93
a-terpineol 0.61 0.02 3.14

decanal (n) 0.61 0.23 1,83
trans-carveol 0.31 0.00 1.57

carvone 0.15 0.00 0.35
RI-1270 0.17 0.00 0.76
RI-1282 0.29 0.00 1.13
RI-1299 0.08 0.00 0.36

undecanal 0.13 0.00 0.57
RI-1323 0.24 0.00 065
a-terpinyl acetate 0.16 0.00 0.39

Table 2. -- continued

Component Name/

Retention Index Average Area % Minimum Area% Maximum Area%

RI-1367 0.10 0.00 0.28

a-copaene 0.35 0.15 0.54

RI-1426 0.19 0.00 0.77

caryophyllene 7.60 0.88 15.11

a-humulene 0.66 0.09 1.29

germacrene 0.16 0.00 0.49

P-bisabolene 0.22 0.00 1.59

selinene 0.44 0.00 1.09

RI-1535 0.35 0.00 1.04

RI-1553 005 0.00 0.20

RI-1564 0.09 0.00 0.26

RI-1613 0.05 0.00 0.15

RI-1648 0.14 0.00 1.48

selin-1 -en4-a-ol 0.08 0.00 0.32

methyl jasmonate 0.08 0.00 0.29

RI-1677 0.12 0.00 0.29

cadinol 0.34 0 00 1.11

RI-1699 0.31 0.00 1.38

8,9-didehydronnootkatone 0.30 0.00 0.83

aristolene 0 17 0.00 0.74

RI-1796 0.12 0.00 0.63

RI-1810 0.13 0.00 0.44

nootkatone 6.70 1.68 17.82

Sensory Analysis

For comparison purposes all juices were ranked on the basis of average hedonic

preference score and divided into three approximately equal categories. There were ten

juices in the "low" category. Average hedonic scores were 4.75 or below. There were

nine juices in the "medium" category. They had preference scores between 4.75-5.75.

The 10 juices in the highly preferred category were rated above 5.75.

Sensory judgements of the panel were limited to a simple hedonic score based on

degree of like or dislike (preference). It should be kept in mind that the score for each

juice represents preference rather than defined flavor. This could cause some scatter in

sensory scores as some panelists might respond to different flavor aspects than others,

nevertheless the majority of the panel typically responded in a similar fashion. Some of the

scatter is reduced as the highest and lowest scores are typically discarded before the

remaining scores are averaged. This sensory approach was chosen as it more closely

reflects marketplace consumer attitudes.

Statistical Analysis

Univariate analysis

Table 3 shows the univariate correlations between preference scores of the

panelists and individual peak areas. Correlation coefficients for individual components

were low, ranging from 0.42 to -0.62. Myrcene, decanal, linalool, linalool oxides and

several unidentified peaks were found to correlate negatively with sensory preference.

Table 3. Univariate correlations of selected volatile and non-volatile data with preference

Variable Correlation (r)

allo-Ocimene 0.42

P-Caryophyllene 0.27

a-Humulene 0.22

RI-954* 0.21

Brix/Acid 0.18

Limonin -0.02

Nootkatone -0 14

trans-Linalool oxide -0.39

y-Terpinene -0,42

Naringin -0 47

Linalool -0.49

Acid -0.51

Decanal -0.53

RI-1796* -0.57

RI-935* -0.61

Myrcene -0.61

RI-1047* -0.62

Brix -0.67

*RI-Kovat's retention indices

Correlation coefficient for trans-linalool oxide was -0.39. Pino et al. (1986 a) also

reported that the linalool oxides correlated negatively towards grapefruit flavor


P-caryophyllene, a-humulene and several unidentified peaks correlated positively

with sensory preference. In contrast, Pino and co-workers reported that methyl butyrate,

ethyl butyrate, decanal, and nootkatone correlated positively with sensory preference.

Since the methylene chloride extraction and concentration procedure was used in our

study, methyl and ethyl butyrate were not adequately extracted. Therefore, a direct

comparison with Pino's results could not be made. Nootkatone, an important component

for grapefruit flavor (Stevens et al., 1970) did not correlate with sensory preference (r=-

0.14). This strengthens the argument that a multivariate approach should be taken for

flavor analysis, since flavor is a perception of a combination of many components.

Multivariate analysis takes several components into consideration at one time while

establishing the relationship of one component to the overall flavor.

In terms of non-volatiles, the bitter naringin and sour total acid correlated

negatively with preference (r=-0.47 and -0.51). However, there was no significant

correlation of bitter limonin (r=- 0.02) with preference. Similar findings were reported by

Pino and Cabrera (1988). However, earlier studies (Rouseffet al., 1980 ; Barros et al.,

1983) found significant negative relationship between limonin and preference.

Multivariate analysis

Principal component analysis (PCA). PCA can be used to determine the inherent

structure of the data and identifies the most differentiating variables within the data set as

a whole. Variables or measurements which help to separate the data points are given

more weight or emphasis. This weighting system is usually expressed as a loading factor.

The larger the loading factor, the more differentiating the measurement. The results of

the combined data set for the first three principal components are shown in Figure 6 a and

6 b. The first three eigenvectors accounted for 66 % of the total variance of the data. As

seen in these figures, the highly preferred juice samples were tightly clustered but not

completely separated from the low and medium preference juices. In general, the most

preferred juices had the lowest PC 1 eigenvector values. The second principal component

axis was not especially effective in separating the three categories of juices. In principal

component 3, the highest preferred juices had eigenvector values close to zero. The least

preferred juices had negative eigenvector values and the medium preference juices had

positive values. The loadings in PC3 are not easy to interpret. As indicated earlier, the

highly preferred juices had eigenvalues very close to zero. Thus the balance between

negative and positively loaded measurements will be associated with preference. For

example, a-humulene and acid have equal but opposite loadings and could contribute to

an eigenvalue of approximately zero.

Component analysis. PCAs are typically calculated in the correlation mode.

However, it is also possible to employ PCA in the covariance mode. In this mode, those

non-redundant measurements which can best account for the maximum variance in the



-2 L


A] E
r .





Prin 1

Figure 6a. Eigenvector values of PCl vs PC2 from principal component analysis of all 57 volatile and taste components:
( 0) high preference category, ( n) medium preference category, ( A ) low preference category.

* c

0 0%





8 C

A A f

el E




Prin 1
Figure 6b. Eigenvector values of PC2 vs PC3 from principal component analysis of all 57 volatile and taste components:
(9 ) high preference category, ( []) medium preference category, (A ) low preference category.


data, are given maximum loading. In the covariance mode, PCA 1 the loading is almost

exclusively in favor of nootkatone (0.95). This indicates that nootkatone is one variable

that can account for much of the variance in the data regardless of preference category.

PCA 2 most heavily loads 0-caryophyllene (0.94) whereas the loading in PCA 3 is

weighted between myrcene and linalool (0.88 and 0.30 respectively). Essentially 97 % of

the variance can be explained with these three eigenvectors. These compounds may be

highly effective in accounting for the variance in the total data set, but they may or may

not be effective in discriminating between samples in the three preference categories.

In order to determine if these four components might also discriminate with

respect to preference category, the univariate correlation coefficients were compared from

Table 3. It can be seen that nootkatone, which was effective in accounting for the total

variance in all samples, was almost completely ineffective in differentiating between juices

of various preference categories. On the other hand, myrcene which was also effective in

accounting for total variance between all samples, was reasonably effective, (r- 0.61) in

differentiating between juices of various preferences. Of the four measurements that

accounted for most of the variance in the total data set, myrcene, P-caryophyllene and

linalool were also effective in differentiating between juices of various preference. In

Figures 7 a and b, various combinations of the peak areas for these three components are

plotted against each other. It can be seen that essentially the same degree of separation

between juices of various flavor preference using peak areas from these three compounds

was achieved from the eigenvector value plots from all 57 components shown in Figures 6

a and b.








-10000 10000


A & AA

30000 50000 70000 90000 1.1e5

Figure 7a. Peak areas of linalool and caryophellene from 29 grapefruit juice extracts analyzed in triplicate:
( 0) high preference category, ( ]) medium preference category, (A ) low preference category.


S9 0


S 2.2e6

OR ] D ]

a M a A O1
Q 6e5 Ak


0 2e5 4e5 6e5 8e5 le6 1.2e6


Figure 7b. Peak areas ofmiyrcene and caryophellene from 29 grapefruit juice extracts analyzed in triplicate:
( ) high preference category, ( ]) medium preference category, (A ) low preference category.

Nootkatone was not a particularly discriminating variable in this study. Our

observed lack of nootkatone correlation agrees with the report of Shaw and Wilson

(1981) and Pino et al. (1986 a, b). The indication that a high "Brix (sweetness) was

strongly associated with the least preferred juices was unexpected. This suggests

however, that highly sweet juices were not preferred. Finally, in identifying the

components which correlate with highly preferred grapefruit juice, it is important to

acknowledge that these components only correlate with preference, but may or may not be


Discriminant analysis. In order to identify the variables which are most

differentiating with respect to preference, discriminant analysis was used (Table 4 and 5).

Discriminant analysis will load heavily those measurements which most effectively

distinguish between juices of different preference category. Figure 8 a illustrates the

results of discriminant analysis using just five components. All three preference category

juices are clustered but several highly preferred samples have overlapped with the mid

preference juices and four mid preference juices are found in the region of the low

preference juices. However, increased category separation can be achieved if additional

terms are used Figure 8 b illustrates the separation which can be achieved with 13

components. One of these components was the peak allo-ocimene, the others are noted

in the legend. This is the minimum number of components required to achieve 100%

separation between juices of different flavor preference.

Table 4. Forward stepwise discriminant analysis (methylene chloride extractions).

Variable Name Partial R**2 Wilk's lambda

Brix 0.46 0.54

RI-1677 0.29 0.38

a-Terpineol 0.23 0.29

P-Gujunene 0.17 0.24

Ratio 0.15 0.21

Limonin 0.13 0.18

cis-Linalool Oxide 0.14 0.15

Naringin 0.24 0.12

Nonanal 0.23 0.09

Acid 0.14 0.08

allo-Ocimene 0.14 0.07

a-Copaene 0.15 0.06

Table 5. Discriminant analysis classification results (methylene chloride extracts).

No. of Percent Correct
Group Compound Comp. T L M
Total Low Medium High

B Linalool +Myrcene
C B+*Brix

D C+P-Caryophyllene

E D+Nootkatone

F Linalool +Myrcene

G F+Brix

H G+RI-1677

I H+allo-Ocimene

Stepwise 16 components

Stepwise 19 components

100 100 100 100

100 100 100 100



d ; ? q n
0 00 0

-5 -4 -3 -2 -1 0 1 2
Figure 8a. Canonical discriminant analysis using myrcene, linalool, Brix, and peaks at RI-1677 and 1126:
(g ) high preference category, ( [) medium preference category, ( A ) low preference category.

3 4






0 0

0 *

.* *-o *

Figure 8b. Canonical discriminant analysis using 13 variables (ratio, RI-935, cis-linalooloxide, nonanal, allo-ocimene,
a-terpineol, decanal, RI-1299, a-copaene, b-gurjunene, RI-1762, and RI-1796):
( 0) high preference category, ( E) medium preference category, ( A) low preference category.

Identification of the Peak at RI-1126

The peak with a Kovat's index value of 1126 was the single highest positively

correlated component among the entire 57 components evaluated. GC-MS was

employed to identify this peak. It was noted that the mass spectra at the front of the peak

differed from that of the back half. Upon further examination, we found there was a major

ion mass of 121 which was evident only during the first portion of the peak and a second

major ion mass of 117 which could be seen only during the last half of the peak. This

strongly suggested the single peak at the retention index 1126 consisted of two co-eluting

compounds. When this peak was re-plotted as two single ion chromatograms, one

generating using only the mass of 117 and the second using only the mass of 121, two

distinct peaks were observed. By judiciously choosing the mass spectral scans spanning

the elution time of the second compound for averaging with the background chosen as the

mass spectral scans spanning the elution time of the first compound, it is possible to

achieve a mass spectrum that is essentially free from ions due to the co-eluting compound.

The same procedure can be repeated to produce library searchable spectra for both

compounds. For the second peak the following spectrum was observed: m/z 121, 100 %;

105, 53.32 %; 136, 49.03 %; 91, 35.55 %; 79, 27.92 %; 93, 20.65 %; 77, 15.36 %; 19,

11.91 %; 22, 9.73 %; 103, 8.88 %. A library search (Adams, 1995) produced a match for

the second peak that had a purity, fit, and rfit of 919, 944, and 954 respectively with

allo-ocimene (2,6-dimethyl 2,4,6-octatriene). Not only is the mass spectrum a good

match to the library spectrum, but the library spectrum has included with it a Kovat's

retention index (RI) for each compound. The library RI for allo-ocimene was 1129 which

very close to the observed 1126. Therefore, designation is based on two independent

means of identification.

The identification of the first eluting peak was more difficult. Its mass spectrum

consisted of: m/z 43, 100.00 %; 117, 96.22 %; 71, 67.73 %; 89, 44.04 %; 55, 41.18 %;

69, 28.59 %; 41, 22.63 % 42, 21.47%; 97, 21.00 %; 75, 18.76 %. The two best mass

spectral matches were hexyl n-hexanoate and butyl n-hexanoate. However, these two

compounds have RI values of 1383 and 1188 which were too high to be considered a

match. The mass spectra for these esters along with the unknown peak all have a m/z 117

ion as a base peak which is from the common hexanoic acid part of the ester. The

unknown spectrum contains a m/z peak of 43 which is indicative of a propyl fragment.

The unknown also contains a m/z 159 ion which could be from a protonated propyl

hexanoate ester. Also, the RI of 1126 would fit the pattern of decreasing RI's for

decreasing size of the alcohol portion of the ester. For these reasons, we have suggested

the first eluting compound might be propyl hexanoate (MW = 158).

This part of the study utilized the components which had highest correlations for

predicting the juice quality. However, these correlated components may or may not be

causative for the over all flavor quality of the juice. More-over, methylene chloride did

not efficiently extract the top note volatiles. Since the top notes were proven to contribute

to the aroma quality (Marin et al., 1992; Bazemore, 1995; Hinterholzer and Schieberle,

1998), further studies were done to investigate the optimum solvent and use of human

responses with GC-olfactometry.

Grapefruit Juice Aroma Extraction Methods

Isolating and analyzing the volatile components of a food product is essential due

to their significant contribution to overall flavor. Comparison of volatile component

isolation procedures have been reviewed by several researchers (Weurman, 1969; Nunez

et al., 1984; Moshanas and Shaw, 1982 &1992). The purpose of this portion of the study

was to establish the most representative extraction technique for grapefruit juice aroma

components. The three methods evaluated here are: liquid-liquid extraction, dynamic head

space purge and trap solvent elution, and static head space extraction using SPME. These

extraction methods have been used earlier in citrus juices. Moshanas and Shaw (1982)

and Nunez et al. (1984) have assessed liquid-liquid extraction in orange and grapefruit

juice respectively. Dynamic head space thermal desorption has been used in orange juice

by Moshonas and Shaw (1992) and in grapefruit juice by Cadwallader and Xu (1994).

Chromatographic Separation and Analysis

Capillary gas chromatography is the best technique to separate the volatile

components in grapefruit juice. In this technique, components are eluted based on their

boiling points and the peak areas are proportional to the components present in the

sample. Figure 9 represents a typical chromatogram for grapefruit juice. It can be roughly

divided into 4 regions:

1. top notes-- includes very volatile components such as ethanol, acetaldehyde,





3.) i-i

1 I

'I ; ,

i,, 1 li I
i ; "^yJ

Top Note



5 i0 15 20 25 30 3
Ret Time (min)

Figure 9. Chromatogram classification of pasteurized grapefruit juice (pentane-ether extracts).



2. terpene area-- includes components like limonene, myrcene, sabinene,

3. carbonyl region-- consists of octanal, nonanal, terpene alcohols and oxides,

4. sesquiterpene area-- includes components like caryophyllene and nootkatone.

Extraction Methods

There is no single extraction method which can extract all the aroma components

in the exact proportion they exist in the sample. Each procedure will concentrate some

components and to varying degrees discriminate against others. Since the aroma active

components in grapefruit juice range from low boiling top notes to high boiling

sesquiterpenes, one of the goals of this study was to optimize extraction procedures so as

to obtain the most representative aroma profile for grapefruit juice. Individual

components were quantified to facilitate comparison between extraction procedures.

Figure 10 compares the representative chromatograms obtained by different extraction

techniques. Table 6 summarizes analytical precision in terms of percent relative standard

deviations (% RSD)of the extraction methods for major juice components.

Liquid-liquid extractions

Pentane/diethyl ether (1:1) liquid-liquid extraction isolated a wide range of

components ranging from top notes to sesquiterpenes. In the earlier section, methylene

chloride was used as the solvent to extract aroma components. The relative absence of

low boiling early eluting components is shown in Figure 5. Table 7 compares the peak

areas obtained from the top note region. Pentane-diethyl ether extractions yielded 73 %

i A

I~ "_ __..... ... ..., ..

5 Time (min) 35
Figure 10. Aroma extraction methods in grapefruit juice: A)liquid liquid extraction (pentane ether 1:1), B) static headspace extraction
(solid phase microextraction-SPME), C) dynamic headspace purge and trap solvent elution (Tenax/charcoal trap).

Table 6. Percent relative standard deviation for different aroma extraction methods in
grapefruit juice.

Component % RSD

Pentane-Ether Dynamic HS

Hexanal 10 7

a-Pinene 13 8

Myrcene 3 3

a-phellandrene 9 18

cis-linalool oxide 8 5

trans-linalool oxide 4 16

allo-ocimene 10 ND

a-terpineol 16 ND

Terpin-4-ol 12 ND

Canrophyllene 10 3

a-Humulene 7 ND

Nootkatone 8 ND

Table 7. Topnote peak areas for different aroma extraction methods.

Peak Areas


Kovat's Indices MeCI P&E Dy-HS






















Total top note peak area




2,785 13,872


49,342 421,594






6,145 8,103

4,986 5,364



76,251 10,660

39,368 106,486

















31,574 8,307


24,217 22,148

299,882 780,279


more top note peak area than methylene chloride. Total top note peak area obtained from

dynamic head space analysis was 350 % more than the liquid-liquid methylene chloride

extracts. Preferential selectivity of methylene chloride for non-polar components in citrus

juices was also reported by Nunez et al. (1984) and Moshonas and Shaw (1982). Since

aroma active components in the top note area, like ethylbutyrate, hexanal, were efficiently

extracted by pentane-diethyl ether, it was utilized as the extraction solvent for this study.

Nunez et al. (1984) also used pentane-diethyl ether solvent mixture for extracting

grapefruit juice aroma components, but no quantitative data were presented in their study.

However, extraction of a wide range of components with a wide range of polarity by a

mixture of pentane-diethyl ether solvents for grapefruit juice has been reported by that

author. Lower percent relative standard deviations were observed for most components in

pentane-diethyl ether extractions (Table 6). To our knowledge, there are no previous

reports which provide extraction reproducibility utilizing liquid-liquid extraction for the

volatile components in grapefruit juice.

Dynamic head space extraction

Dynamic head space involves the continual movement of volatiles from the bulk of

the sample into the gaseous phase where it is swept into a trap (Wampler, 1997). The

sample volatiles are constantly swept by a flow of carrier gas and a state of equilibrium

between sample matrix and head space is never reached. This increases the volume of

head space gas beyond the limit of the head space in the sample vessel. Volatiles must be

collected on a trap and can be used for subsequent analysis. In this study, a mixture of

charcoal and Tenax sorbent materials were used as adsorbents. These absorbents are

commonly used for the isolation of volatiles (Buttery and Ling, 1996; Wampler, 1997).

Tenax is capable of trapping a wide range of organic volatiles but is not well suited for

low molecular weight hydrocarbons and smaller alcohols (C1-C4). Charcoal, on the other

hand, has affinity to collect small organic compounds and has higher retentive capacity.

Moisture can be a problem when trapping aroma volatiles. The use of sodium

sulfate or purging the absorbents with inert gases are common methods found in the

literature. Since grapefruit juice is approximately 90% water, the absorbents were purged

with dry nitrogen to remove any trapped moisture.

Dynamic head space purge and trap solvent elution was effective in extracting top

note volatiles (Figure 10). This method extracted 160 % more top note peak area than the

pentane-diethyl ether liquid-liquid extractions. However, higher vapor pressure

components like oxygenated mono and sesquiterpenes were not effectively purged from

the sample. This means components thought to be important to grapefruit flavor such as

nootkatone (Stevens et al., 1970) could not be quantified using this technique.

Cadwallader and Xu (1994) reported similar results for dynamic head space analysis of

grapefruit juice. However, they used cryotrapping and thermal desorption and were able

to detect early eluting components such as ethanol and acetaldehyde which are normally

obscured by the solvent peak. Percent RSD reported for our procedure was comparable

to those reported by Cadwallader and Xu (1994). Since this method did not effectively

extract the high boiling aroma active components, we did not use this method for further


Static head space extraction using SPME

Solid phase micro extraction is a rapid procedure to sample volatile components in

head space gases. It involves the adsorption of head space volatiles onto a coated fiber

which is exposed to the head space for a specific time. In the static head space method,

volatiles in the sample matrix are allowed to come to an equilibrium with the head space

before being sampled. The SPME technique is relatively new technique and has been used

for analyzing orange essence volatiles (Bazemore, 1995), orange juice volatiles (Steffen

and Pawliszyn, 1996), head space of milk powder (Stevenson and Chen, 1996), and

cheese volatiles (Chin et al., 1996).

The SPME method effectively extracted terpenes such as limonene and myrcene

(as they were the largest peaks in the resultant chromatogram) but was relatively

ineffective in extracting the top note volatiles. The SPME fibers adsorb components on a

competitive basis. Since terpenes (especially limonene) are in higher concentrations in

grapefruit juice and also due to their non-polar nature, distribution coefficients and affinity

of fiber to non-polar components, they tend to dominate the head space components

trapped by the fiber coating.

Steffen and Pawliszyn (1996) reported good reproducibility for the components in

orange juice. However, the authors centrifuged the samples prior to analysis, which

eliminated the juice pulp and suspended solids. Lower levels of precision values were

obtained when sampling was done on whole grapefruit juice (private communications -

Bazemore, 1998). Since SPME emphasizes terpenes, which are in high concentrations but

contribute little to aroma, this technique was not used for further analysis in this research.

GC-Olfactometry Studies

GC-olfactometry (GC-O) is an important analytical tool since it characterizes the

odors of individual compounds and identifies which GC peaks have aroma activity (Mistry

et al., 1997). A human nose is used to detect and evaluate the effluents from the column

instead of an analytical detector. It is a powerful and sensitive tool since the odor

detection limit of a human nose is 10 moles (Reineccius, 1994), which is considerably

more sensitive than most instrumental detectors.

Grapefruit juice is a complex matrix and not all volatile components have aroma

activity. Even among those components which have aroma activity, some will have more

impact than others. Therefore, GC-O has been utilized to identify and characterize the

odor active components in grapefruit juice extracts. Aroma active components in

grapefruit juice change with the fruit maturity and also from thermal processing. In this

study aroma extracts from unpasteurized and pasteurized juices from early, mid and late

season fruits were evaluated for individual aroma active components.

Instrumental Detectors vs. Human Response

GC-O detects only those components which have aroma activity. Some of these

aroma active components are very potent and are present in such small amounts that they

cannot be detected by typical GC detectors. Figure 11 compares the consensus

aromagram (aroma intensities of 3 panelists were averaged) produced by GC-O with

0 il
Ir .*>1 .., r




Figure 11. Comparison of aromagram from OSME and chromatograms from FID and SCD






0 5 10 15 20 25 30 3
Ret Time (min)

chromatograms produced by FID and SCD detectors. The instrumental detectors

responded to some components which the human nose did not recognize. Conversely, the

human nose detected some compounds which gave no instrumental response. Large peaks

in FID like limonene and caryophyllene seem to have little to no aroma activity. Panelists

described limonene as citrusy, medicine and minty with a moderate intensity, while they

could not detect any aroma activity for caryophyllene. Earlier work by Marin et al. (1992)

also reported a limited aroma activity of limonene in orange juice.

Among the small FID peaks, vanillin is notable. It was found to have intense

vanilla or white chocolate aroma (average aroma intensity = 13). Vanillin has been

reported for the first time in grapefruit juice by our group. A strong intense aroma peak

was obtained at a 25 min retention time that has the characteristic aroma of vanillin (see

Figure 11). The same grapefruit juice extract was analyzed using GC-MS for further

confirmation of the presence of vanillin. By comparing the mass spectrum of the sample

with the mass spectrum of the standard, it can be concluded that the peak with aroma

attribute vanilla was, in fact, vanillin. The total ion chromatogram and the mass spectra of

vanillin sample and the standard, are shown in Appendix A and B. Prior to this, vanillin

was identified in orange juice by Marin et al. (1992). Peleg et al. (1992) proposed the

path ways for formation of vanillin from ferulic acid in orange juice (Figure 12).

According to the authors (Peleg et al., 1992), vanillin can form from ferulic acid through

decarboxylation and oxidation or directly from free ferulic acid through retro aldol

reactions. Similar reaction pathways may also occur in grapefruit. Intense aroma activity

Ferulic Acid








H2C = CH

Figure 12. Formation of vanillin from ferulic acid.





of vanillin was also reported in oak aged wines (Aiken and Noble, 1984), Japanese green

tea (Acree and King, 1996) and in coffee (Akieda and Kato, 1987).

Maturity and Processing Changes

In this part of the study, effects of maturity (early, mid and late) and processing

(unpasteurized and pasteurized) are evaluated using GC-olfactometry. Fruit maturity as

well as thermal processing affect the aroma quality of grapefruit juice. This is reflected in

the differences in number and kinds of aroma active peaks detected in juices from different

maturities. Figure 13a and b compares aroma attributes in early, mid and late season

unpasteurized and pasteurized juices. A total of 37 49 aroma active peaks were found in

early, mid and late season grapefruit juices. Appendix C lists the attributes perceived in

juices of different maturities. Forty-one aroma components could be differentiated in early

season unpasteurized juices while 37 peaks were detected in pasteurized juices. As a

result of thermal treatment 11 aroma compounds were lost while 7 new components were

formed in early season juice. However, many compounds were unchanged. Table 8

shows the aroma attribute compounds formed or lost during thermal processing of early

season juices.

Mid season juices had 43 aroma active peaks in unpasteurized juice and 49 in

processed juice. Similarly, 43 aroma active peaks were detected in both unpasteurized and

pasteurized late season grapefruit juices. Eight components were lost in thermally treated

late season juices, while 8 new attributes were detected. The aroma active peaks lost due

800 1000

1200 1400
Retention Index

(attributes listed in Appendix C)

Figure 13a. Number of aroma active components at different maturities in unpasteurized grapefruit juice: A) early season,
B) mid season and C) late season.



800 1000 1200 1400 1600
Retention Index (attributes listed in Appendix C)
Figure 13b. Number of aroma active components at different maturities in pasteurized grapefruit juice: A) early season,
B) mid season and C) late season.

Table 8. Formation and loss of aroma attributes due to pasteurization in early season red
grapefruit juices.

Components/ Intensities
Retention Indices Pasteurized Unpasteurized Description
RI-896 7.0 Citrusy, Mediciny
RI-936 5.7 Floral. Smokey
a-Pinene 8.2 Greenish
a-phellandrene 12.3 Citrus
RI-1044 7.6 Rotten Fruit
RI-1095 10.3 Terpeney, Cucumber
RI-1116 12.7 Mediciney
RI-1166 10.4 Musty
RI-1217 7.3 Terpeney
RI-1223 9.3 Musty
RI-1227 7.8 Stinky fruit
RI-1318 10.8 Smokev, Rancid
RI-1374 6.8 Medicmey, Minty
RI-1381 92 Sweet
RI-1510 7 2 Spicey. perfumey
RI-1662 4.0 Peppery
RI-1684 7.2 Pungent
RI-1723 8.6 Rotten Grapefruit

to pasteurization had generally favorable sensory attributes like green, fruity while the

components formed as a result of heating had roasted, fruity, and spicey attributes.

Concentration of the components also changes due to maturity and thermal

processing. Total alcohols, aldehydes and hydrocarbons were higher in early season

unpasteurized juice (Figure 14a). Among the alcohols, a-terpineol, terpin 4-ol, trans

linalool oxide, and among the hydrocarbons, myrcene and y-terpinene were found to

correlate negatively with sensory preference of grapefruit juice (Jella et al., 1998). These

negatively correlated compounds were present in higher concentrations in early season

than in late season juices. Results of this are summarized in Table 9. The levels of these

components in grapefruit juice are in concurrence with those reported by Maarse and

Visscher (1989).

As a result of pasteurization, increased concentrations of alcohols, aldehydes and

hydrocarbons were observed (Figure 14b). Higher levels of alcohols are probably due to

acid catalyzed reactions of terpenes like limonene, P-pinene, myrcene and so on. These

components react in dilute aqueous acid and high temperatures to give several reaction

products, some of which are alcohols like a-terpineol, terpin-4-ol and linalool oxides

(Clark and Chamblee, 1992 and Shaw, 1991). Limonene is the major terpene in citrus

juices and readily forms several reaction products under the conditions present in citrus


a-terpineol in pure form and at low levels has a lilac aroma (Arctander 1994).

However, at higher concentrations it tends to have musty odor (Marcotte et al., 1998).

The level of this component in early, mid and late season pasteurized juices are 0.81, 1.97


Alcohols Aldehydes Hylrocar ons
(excluding Limonene)


Alcohols Aldehydes Hydrocarbons
(excluding Limonene)
Figure 14. Concentrations of components in grapefruit juice.
A) unpasteurized juices; B) pasteurized juices: (*) early season,
( ) mid season and (r-) late season












Table 9. Concentration levels (ppm) of components in early, mid and late season red grapefruit juices.

Early Season Mid Season Late Season

Component Unpasteurized Pasteurized Unpasteurized Pasteurized Unpasteurized Pasteurized

a-terpineol 0.343 0.809 0.228 1.967 0.242 0.614

Terpin-4-ol 0.174 0.171 0.126 0.194 0.168 0..215

trans-linalool oxide 0.426 0 450 0.420 1.461 0.295 0.311

Myrcene 1.790 1 371 1.788 2.942 1.166 1.049

y-terpinene 0.263 0.260 0.189 0.284 0.260 0.218

and 0.61 ppm respectively. Panelists in this study described it as having "stale church" or

"wet dog" smell. Limonene is reported to undergo acid catalyzed hydration to form a-

terpineol (Clark and Chamblee, 1992) (Figure 15). Mid season unpasteurized juice had

higher concentration oflimonene (37 ppm) than early and late season unpasteurized juices

(33 and 25 ppm respectively). Therefore, higher concentrations of a-terpineol can be

expected in mid season juices.

Standard Descriptors Vs. Panelist's Descriptors

Linalool is described as having a strong floral aroma (Arctander 1994), and is an

important contributor to the flavor and aroma of numerous products including lemon oil,

certain teas (Clark and Chamblee, 1992) and orange juice (Marin et al., 1992). Other

components having significant aroma contribution to orange juice are ethylbutyrate,

hexenal, vanillin, octanal and nonanal (Marin et al., 1992; Bazemore, 1995; da Silva et al.,

1994). Table 10 compares the aroma descriptors given by panelists for some of the

components present in grapefruit juice. Because there is no standard lexicon, free choice

descriptors were encouraged. Hence it was not surprising to see that for a single

component the descriptors given by the panelists differed. Also, multiple synonymous

terms were used by the panelists for one component. For example hexanal was described

by the panelists either as green, grassy or herbacious. However, by comparing the elution

times and Kovats indices for aroma active peaks, it can be concluded that the panelists

were describing the same peak using a different descriptor. Table 11 compares the

attributes described by the panelists with that of the standard descriptors given by


+ +

limonene a terpineol

Figure 15. Acid catalyzed hydration of limonene

Table 10. Aroma descriptors used by panelists from GC-O experiments of citrus standards.

SComponent Panelist 1 Panelist 2 Panelist 3


Ethyl Butyrate







p-menthene 8-thiol




Green, dead bug

Medicine, Piney

Unripe Mango


New cotton clothes

Cilantro, Musty

Rotten gft, Stinky terpeney

GFt stink, Rotten Gft


Fruity, Floral


Piney, Green



Stale Church

Moldy, Musty

Stinky Rotten GFT

Moldy GFT

Strong green

Fruity, Citrus

Smoked burnt roasted

Greenish, Vitamin C


Stinky floral


Green cilantro

Sweet grapefruit

Stinky grapefruit

Benzaldehyde (IS)

Methyl Jasmonate (IS)

Cherry, Almond

Floral, Jasmine

Cherry, Almond

Floral, Jasmine

Cherry, Almond

Floral, Jasmine

Table 11. Comparison of standard (Arctander lexicon) with panelist descriptors.

I Component Standard Descriptor Panelist's Descriptor

Hexenal, Ethyl Butyrate
cis-linalool oxide
trans-linalool oxide

Green/Warm sweet fruity
Green vegetable like
Warm resinous and herbacious
Bitter almond, sweet cherry
Warm peppery, herbacious
Balsamic resinous and citrusy
Citrusy, peppery, woody
Citrusy, kerosene like
Warm herbacious, sweet
Herbacious, citrusy
Sweet floral earthy
Sweet floral earthy
Fatty, floral
Floral woody
Lilac, piney
Warm herbaceous
Creamy, vanilla like
Fruity, citrusy

Green, dead bug, skunky
Cherry, almond
Unripe mango, piney
Unripe mango, citrus
Minty, citrusy
Citrusy, musty
Roasted cotton candy
Terpeney, cotton candy
Terpeney, cotton candy
Terpeney, cucumber, cotton candy
Musty, wet dog, cilantro
Floral, liquorice, mediceney
rotten nutty grapefruit fruit

Arctander (1994). Nootkatone was described by panelists as rotten fruity, sweet

grapefruity, stinky citrusy. Arctander's descriptor for nootkatone is citrusy. Comparison

of retention times, indices and the odor description given by the panelists for the standard

gives a good indication that same aroma active peak is being described.

Grapefruit Aroma

Nootkatone is considered by some scientists to be one of the important

contributors to the grapefruit flavor (Stevens et al., 1970; Pino et al., 1986a; Shaw and

Wilson, 1981). Maturity plays a significant role in determining the quantity of this

sesquiterpene ketone. Traditionally, late season juices are considered to be best quality.

Higher amounts of nootkatone were found in late (9.1 and 10.8 ppm) than in mid (3.2 and

3.9 ppm) and early (1.8 and 1.9 ppm) season unpasteurized and pasteurized juices.

However, nootkatone was a poor predictor for juice quality (r=-0.05) in this 30 juice

sample set of mid and late season juices.

Grapefruity aroma was also perceived by the panelists a few seconds before

nootkatone has eluted. This peak had a Kovats indices or retention index (RI) of 1754.

This peak has been tentatively identified as 8,9 didehydro nootkatone based on retention

index and aroma quality. This has been reported to be present at 0.001 ppm level in

grapefruit juice (Maarse and Visscher, 1989). Demole and Enggist (1986) reported its use

to augment or enhance the organoleptic properties of grapefruit or imitation grapefruit

beverages. This GC-O peak also occurs at the same time as one of the large sulfur peaks,

RI-1753 (retention time 32 min). Since both these components have similar retention

times, it is not currently resolved which component is responsible for the additional

grapefruit aroma peak. The question will have to be resolved with additional experiments

using chromatographic columns of different selectivity. Another important sulfur

component having a fresh grapefruity aroma isp-menthene-8-thiol. Discussion of this

component is included in a later section.

Dilution Analysis

Studies involving dilution analysis (AEDA, Charm) on grapefruit juice have not

been reported to date. However, orange juice has been extensively studied (Marin et al.,

1992; Hinterholzer and Schieberle, 1998) with both AEDA and Charm. Among the

components reported by the authors, hexenal, ethyl butyrate and vanillin were found to

have highest dilution values, while linalool, decanal were found at the lower end of the

dilution factors.

The peaks detected in our study were aroma active peaks from the juice extract

concentrated 160 times. This does not provide information about which of these peaks

have intense aroma activity at higher dilutions (lower concentrations). Since components

in juice are not present in concentrated form, dilution analysis was done to identify the

most aroma active, now referred to as aroma impact peaks. To assess the most intense

peaks, juice extract was concentrated 16 times instead of 160 times and analyzed using

GC-O as before. The list of peaks identified and their corresponding odors are given in

Table 12. Some of the components like cis and trans linalool oxides were not present in

the samples at 16 X concentration even though these components had intense aroma

activity (13 on a 15 point scale) in 160 X concentrated samples, da Silva et al. (1994)

stated that odorants have different intensities above their threshold values, that is, aroma

intensity may not be proportional to the concentration of the compound. According to

Meilgaard et al. (1991), a mathematical model proposed by Beidler works best for middle

and high range of sensory intensities. According to this model, there is a sigmoidal

relationship between the concentration of the product and the stimulus perceived. This

might be the reason for lack of odor perceptions of components like linalool oxides and

linalool at lower concentrations of grapefruit juice aroma extract.

The two attributes which had intense aroma activity in the 16 X grapefruit juice

extract were hexanal/ethylbutyrate and a-phellandrene (see Table 12). When these two

components were used for sensory correlations (discussed in section 5 of results and

discussion), they were found to have significant correlations (0.31 and -0.28 at p< 0.05)

with aroma intensity. This suggests that hexanal/ethylbutyrate and a-phellandrene are key

components in determining the quality of grapefruit juice.

Sulfur Compounds in Grapefruit


Organic sulfur compounds are present in a variety of food products and contribute

significantly to their odor and flavor profile (Mistry et al., 1994). These are often present

at sub-threshold levels and present a challenging task for chromatographers with respect

to their detection. Mistry et al. (1994) compared a flame photometric detector (FPD), an

Table 12. List of components present in 16x concentrated juice extract and their
intensities and aroma attributes.

Components Attribute Aroma Intensities

Hexenal, Ethyl Butyrate
Unknown Sulfur cmpd (RT 32min)

Sweet fruity
Unripe citrusy
Green, Citrusy
Green, Citrusy
Floral, Citrusy
Grainy, Mediciny
Green, Musty
Stinky Grapefruit
Apple Sauce
Apple Sauce


atomic emission detector (AED) and a sulfur chemiluminescence detector (SCD). The

authors reported best response in terms of sensitivity for AED. They rated FPD and SCD

comparable to each other; however, FPD was not linear with the concentration of sulfur.

SCD, on the other hand, had an equal molar response to all sulfur components.

The operation of sulfur chemiluminescence is based on the reaction of ozone with

sulfur monoxide which is produced from combustion ofanalyte (Figure 16). The excited

sulfur dioxide, upon collapse to the ground state, emits light with the maximum intensity

of 350 nm. This detector is very specific for sulfur compounds, has equi-molar response

and even the solvent peak was not detected.

Processing and Maturity Effects

The extraction solvent used for isolating sulfur compounds was ethyl acetate. This

was found to extract more sulfur compounds than the solvent mixture ofpentane-diethyl

ether. The specific reason for this is not known yet. To our knowledge very little work

has been done on sulfur compounds in citrus juices to make further comparisons and


Twenty-two sulfur compounds were isolated in early, mid and late season

pasteurized and unpasteurized grapefruit juice. This represents the most comprehensive

determination of sulfur compounds in citrus juices reported to date. Total number and

total sulfur peak areas decreased with increasing fruit maturity (Figure 17a) and increased

with processing (Figure 17b). Total peak area of early season pasteurized juices was 83

times more than early season unpasteurized grapefruit juices. Late season pasteurized

Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EW3SXRY0J_VN25C3 INGEST_TIME 2013-03-05T20:25:05Z PACKAGE AA00013521_00001