This item is only available as the following downloads:
AROMA AND TASTE IMPACT COMPONENTS IN
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
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
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
TABLE OF CONTENTS
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
A. TOTAL ION CHROMATOGRAM OF LATE SEASON GRAPEFRUIT JUICE 125
B. MASS SPECTRUM OF VANILLIN .......... ..... ........... 127
C. LIST OF DESCRIPTORS AND THEIR RELATIVE IN ENS IT IES (GC-O)... 129
D. COMPOUNDS IDENTIFIED IN NOT-FROM-CONCENTRATE GRAPEFRUIT
JUICE ................ ............................... 132
LIST OF REFERENCES ........ .......................... 136
BIOGRAPHICAL SKETCH .. .......... .............. ... .. 144
LIST OF TABLES
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
LIST OF FIGURES
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
AROMA AND TASTE IMPACT COMPONENTS
IN GRAPEFRUIT JUICE
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
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
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.
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
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.
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.
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 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).
MATERIALS AND METHODS
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
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.
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.
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).
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
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.
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
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.
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
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.
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
c =1.703 and
Figure 1. Standard curve used for calculation ofkovat's retention indices for volatile
y =13983x+ 132945
R2 = 0.9943
0 50 100 150 200 250
0 50 100 150 200 250 0 50 100 150 200 250
Concentration (ppm) Concentration (ppm)
Figure 2. Calibration curves used for quantifying the volatiles. (A) propyl benzene, (B) myrcene, (C) linalool, (D) nootkatone.
y =9122.8x -83140
0 50 100 150 200 250
y =8766.1x -71662
R2 = 0.9951
y = 9715.7x 56480
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
0 2 4 6 8 10
Figure 3. Calibration curve for s-methyl-thiobutanoate (sulfur compounds).
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
7 0 7
More Sour More Sweet
Than Sweet Than Sour
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
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.
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.
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.
RESULTS AND DISCUSSION
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).
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.
Table 2. Maximum, minimum and average area percent for components extracted with
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
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
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.
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)
Nootkatone -0 14
trans-Linalool oxide -0.39
Naringin -0 47
*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.
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
A AA AD
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.
A A f
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.
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.
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
F Linalool +Myrcene
Stepwise 16 components
Stepwise 19 components
100 100 100 100
100 100 100 100
A DA A
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.
.* *-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,
'I ; ,
i,, 1 li I
i ; "^yJ
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.
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.
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~ "_ __..... ... ..., ..
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
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.
Kovat's Indices MeCI P&E Dy-HS
Total top note peak area
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 (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
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
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
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
(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
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
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
Alcohols Aldehydes Hydrocarbons
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
Green, dead bug
New cotton clothes
Rotten gft, Stinky terpeney
GFt stink, Rotten Gft
Stinky Rotten GFT
Smoked burnt roasted
Greenish, Vitamin C
Methyl Jasmonate (IS)
Table 11. Comparison of standard (Arctander lexicon) with panelist descriptors.
I Component Standard Descriptor Panelist's Descriptor
Hexenal, Ethyl Butyrate
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
Sweet floral earthy
Sweet floral earthy
Creamy, vanilla like
Green, dead bug, skunky
Unripe mango, piney
Unripe mango, citrus
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.
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.
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
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)
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
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
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC