Citation
Aroma and taste impact components in grapefruit juice

Material Information

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

Subjects

Subjects / Keywords:
Bitterness ( jstor )
Chlorides ( jstor )
Flavors ( jstor )
Food ( jstor )
Grapefruit juice ( jstor )
Grapefruits ( jstor )
Gustatory perception ( jstor )
Juices ( jstor )
Odors ( jstor )
Solvents ( jstor )
Dissertations, Academic -- Food Science and Human Nutrition -- UF ( lcsh )
Food Science and Human Nutrition thesis, Ph.D ( lcsh )
Grapefruit juice -- Composition ( lcsh )
Grapefruit juice -- Odor ( lcsh )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

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

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
029544287 ( ALEPH )
40139348 ( OCLC )

Downloads

This item has the following downloads:


Full Text











AROMA AND TASTE IMPACT COMPONENTS IN
GRAPEFRUIT JUICE












By

PRASHANTHI JELLA


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.














ACKNOWLEDGMENTS


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

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

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

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

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

to work with him.

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

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

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

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

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

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

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

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

the problem.

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

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









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

while.

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

ROHINI!!!















TABLE OF CONTENTS




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

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

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

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

CHAPTERS

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

APPENDICES

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


Iabk page

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

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

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

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

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

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

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

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

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

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

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

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

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








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

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

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

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

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

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














LIST OF FIGURES


Figureage

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

xi









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

xii









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


By

Prashanthi Jella

August, 1998


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

This study represents the most comprehensive determination of aroma active

volatiles and sulfur compounds in grapefruit juice reported to date.

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

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

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

caryophyllene. Since concentrated methylene chloride extracts contained few highly

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

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

note peak area than methylene chloride liquid-liquid extraction.









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

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

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

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

and GC-MS.

A routine method for grapefruit juice sulfur compounds using sulfur

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

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

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

143% after pasteurization.

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

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

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

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

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

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

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

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

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

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

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









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

quality.














CHAPTER 1
INTRODUCTION


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









3

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














CHAPTER 2
LITERATURE REVIEW



Flavor



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

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

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

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

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

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

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

Extensive research has been conducted to identify and quantify volatile

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

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

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

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

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

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

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











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

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

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

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

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

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

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

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

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

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

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

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

contribution to grapefruit juice flavor.

Statistical Correlations


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

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

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

nose (Ohloff, 1990).

Previous workers have developed models based on the correlations between

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

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











components and corresponding sensory hedonic scores were analyzed using multiple

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

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

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

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

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

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

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

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

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

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

best with storage time.

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

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

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

classified 24 commercial single strength grapefruit juices from different production days

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

positive correlation to flavor while another unidentified component correlated negatively

with flavor.

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

measurements of grapefruit juice volatiles using multiple linear regression. Methyl

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











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

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

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

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

Bitterness


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).


GC-Olfactometry



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

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

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

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

determine which of the many chromatographic peaks have aroma activity and

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

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

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

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

determines intensities without dilutions.

Charm Analysis


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

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

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

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

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

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











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

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

components.

AEDA


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.


OSME


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

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











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

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

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

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

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

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

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

known as an aromagram.

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

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

reflux and no reflux samples of aqueous orange essence.

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

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

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

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

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

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

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

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

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

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











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

compound producing the musty odor.


Sample Preparation



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

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

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

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

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

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

several different extraction techniques were evaluated for optimum odor recovery.

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

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

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

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

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

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

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

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

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

extract.










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.

SPME


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









14
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.

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

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

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

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

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

compounds were not observed.

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

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

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

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

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

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

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

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

one feature that makes SPME suitable for food analysis.












Sulfur Compounds



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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

thresholds.

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









16
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

juices.

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).














CHAPTER 3
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



Survey Samples


Methylene chloride extracts

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

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

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

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

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










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

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


Pentane-diethyl ether extracts

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

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

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

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

methylene chloride extracts study.


GC-Olfactometry Samples


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

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

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

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

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

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

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

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

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

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










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

Samples consisted of two bottles for each juice type.




Sample Preparation



Liquid-Liquid Extraction with Methylene Chloride


Extraction of volatiles was accomplished with methylene chloride using the

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

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

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

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

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

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

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

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

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

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

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

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

extracted twice and each extract analyzed in duplicate.










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


Extraction of the volatiles was accomplished according to the previously

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

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

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

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

described. Each sample was analyzed in duplicate.


Dynamic Head Space Purge and Trap Solvent Elution


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

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

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

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

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

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

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

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

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

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

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

times the column volume of pentane.












Extraction Procedure for Sulfur Compounds


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

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

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

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

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

duplicate.


Extraction Procedure for GC-Olfactometry Analysis


Extraction of juice volatiles was accomplished using the method described for

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

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

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

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

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


Instrumental Techniques



GC-Flame Ionization Detector


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

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










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

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

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

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

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

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

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

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

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

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

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


GC-Sulfur Chemiluminescence Detector


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

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

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

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

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

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

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

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









23
Chromatograms were recorded and integrated using Chrom Perfect (Justice Innovations,

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


GC-OSME


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

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

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

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

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

320 C.

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

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

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

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

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

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

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

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

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

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

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

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











GC-Mass Spectrometry


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

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

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

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

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

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

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

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

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

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

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


Limonin and Naringin Analysis Using HPLC


Sample preparation

Limonin and naringin for grapefruit juice samples (extracted with methylene

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

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

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

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

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










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

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

these were not heated prior to HPLC analysis.


HPLC instrumentation

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

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

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

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

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

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

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

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

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

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

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

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



Peak Identification and Ouantification



Chromatographic peaks were identified using their mass spectra and comparison of

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











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

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

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

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

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

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

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

volatile components.

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

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

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

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

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

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

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

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

area of internal standard.

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

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

calibration curve for these is shown in Figure 3.













ZUV


1200 +


400+


500 1000 1500
Retention time (sec)



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


2000 2500


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


u i











4000000

3000000

200000

1000000

0









200000




U 10000


3000000


2000000


1000000


0


2000000


1000000
nW~


y =13983x+ 132945
R2 = 0.9943








0 50 100 150 200 250
Concentration (ppm)



(A)

0v -


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



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


y =9122.8x -83140
R2= 0.9964







0 50 100 150 200 250
Concentration (ppm)



(B)


y =8766.1x -71662
R2 = 0.9951


y = 9715.7x 56480
R2= 0.997


00











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










30






600000
y = 50873x 2369.3
R2= 0.9997

400000




200000




0
0 2 4 6 8 10
Concentration (ppm)


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












Sensory Analysis



DOC Preference Panel


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

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

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

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

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

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

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


USDA Descriptive Panel


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

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

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

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

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

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

with four samples and a reference juice.

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

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











Grapefruit Juice Sensory Panel
Please Read the Instructions


Name: Sample Number: 321
Date:

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


0
None
Grapefruit Aroma Ouality


15
Strong


15
V.Good


V.Poor
Sweet/Tart Balance


7 0 7
More Sour More Sweet
Than Sweet Than Sour


0 15


None
Overall Flavor Ouality


V.Poor


Strong


15
V.Good


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.









33

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

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

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

reference sample after every 10 grapefruit juice samples.

Samples were coded with random three digit numbers randomly arranged on

serving trays, and then presented to panelists.



Training of Panelists



GC-Olfactometry Panel


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

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

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

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

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

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

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

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

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

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










Descriptive Panel


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

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

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

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

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

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

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

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

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

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


Statistical Analysis



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

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

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

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

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

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

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

determine the classification significance for each sample. Mahalonobis distances were









35

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

predict the juice quality.














CHAPTER 4
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).















-0 0



0s 'T
*0 d
-0



4., J 0.. t


0 5 10 15 20 25
Retention Time (mmin)

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


30 35













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


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

a-thujene 0.51 0.30 0.78

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

a-phellandrene 0.36 0.00 1.13

RI-1008 0.19 0.00 0.88

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

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

linalool 1.11 0.41 2.58
RI-1100 0.13 0.00 0.35

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

P-pinene oxide 0.19 0.00 0.64

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

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

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

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

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












Table 2. -- continued


Component Name/

Retention Index Average Area % Minimum Area% Maximum Area%

RI-1367 0.10 0.00 0.28

a-copaene 0.35 0.15 0.54

RI-1426 0.19 0.00 0.77

caryophyllene 7.60 0.88 15.11

a-humulene 0.66 0.09 1.29

germacrene 0.16 0.00 0.49

P-bisabolene 0.22 0.00 1.59

selinene 0.44 0.00 1.09

RI-1535 0.35 0.00 1.04

RI-1553 005 0.00 0.20

RI-1564 0.09 0.00 0.26

RI-1613 0.05 0.00 0.15

RI-1648 0.14 0.00 1.48

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

methyl jasmonate 0.08 0.00 0.29

RI-1677 0.12 0.00 0.29

cadinol 0.34 0 00 1.11

RI-1699 0.31 0.00 1.38

8,9-didehydronnootkatone 0.30 0.00 0.83

aristolene 0 17 0.00 0.74

RI-1796 0.12 0.00 0.63

RI-1810 0.13 0.00 0.44

nootkatone 6.70 1.68 17.82











Sensory Analysis


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

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

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

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

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

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

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

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

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

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

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

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

reflects marketplace consumer attitudes.


Statistical Analysis


Univariate analysis


Table 3 shows the univariate correlations between preference scores of the

panelists and individual peak areas. Correlation coefficients for individual components

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

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












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


Variable Correlation (r)

allo-Ocimene 0.42

P-Caryophyllene 0.27

a-Humulene 0.22

RI-954* 0.21

Brix/Acid 0.18

Limonin -0.02

Nootkatone -0 14

trans-Linalool oxide -0.39

y-Terpinene -0,42

Naringin -0 47

Linalool -0.49

Acid -0.51

Decanal -0.53

RI-1796* -0.57

RI-935* -0.61

Myrcene -0.61

RI-1047* -0.62

Brix -0.67

*RI-Kovat's retention indices











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

reported that the linalool oxides correlated negatively towards grapefruit flavor

preference.

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

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

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

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

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

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

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

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

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

Multivariate analysis takes several components into consideration at one time while

establishing the relationship of one component to the overall flavor.

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

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

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

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

1983) found significant negative relationship between limonin and preference.











Multivariate analysis

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

an eigenvalue of approximately zero.

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

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

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














A A



*1^


-2 L
-1.5


A

A] E
r .


EC
A


kA


A
A AA AD
MI


-0.5


Prin 1


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















* c

0 0%
AA AA

Eai3]


A
tA


-0.5


0.5


8 C


A A
A A f


O C
el E


1.5


2.5


3.5


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


-31.
-1.5










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.










2.6e6

2.2e6

1.8e6

1.4e6

le6

6e5


2e5

-2e5
-10000 10000


A AA


A & AA
OD [ID


30000 50000 70000 90000 1.1e5


Linalool
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.


3e6


S9 0









ItVn.


S 2.2e6



1.4e6
OR ] D ]



a M a A O1
Q 6e5 Ak

6M AA

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

Myrcene

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.
00











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

causative.

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

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

Discriminant analysis will load heavily those measurements which most effectively

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

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

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

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

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

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

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

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

separation between juices of different flavor preference.











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

Variable Name Partial R**2 Wilk's lambda

Brix 0.46 0.54

RI-1677 0.29 0.38

a-Terpineol 0.23 0.29

P-Gujunene 0.17 0.24

Ratio 0.15 0.21

Limonin 0.13 0.18

cis-Linalool Oxide 0.14 0.15

Naringin 0.24 0.12

Nonanal 0.23 0.09

Acid 0.14 0.08

allo-Ocimene 0.14 0.07

a-Copaene 0.15 0.06











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

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


B Linalool +Myrcene
C B+*Brix

D C+P-Caryophyllene

E D+Nootkatone


F Linalool +Myrcene

G F+Brix

H G+RI-1677

I H+allo-Ocimene


Stepwise 16 components
(backward)

Stepwise 19 components
(forward)


100 100 100 100


100 100 100 100
















AA


AAA
A DA A


d ; ? q n
0 00 0


-61
-5 -4 -3 -2 -1 0 1 2
CAN 1
Figure 8a. Canonical discriminant analysis using myrcene, linalool, Brix, and peaks at RI-1677 and 1126:
(g ) high preference category, ( [) medium preference category, ( A ) low preference category.


3 4













0
[]
[]


AA
A

A


A
A
A
A A


O DO
O

E O
0 0

0 *




.* *-o *


CAN 1
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









55
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,
hexanal,










Carbonyl


H

C)


Sesquiterpene


3.) i-i
0

1 I



'I ; ,


i,, 1 li I
i ; "^yJ


Top Note


Terpene


*1)
r2J


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

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


I'


-












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

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

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


Extraction Methods


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

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

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

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

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

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

components were quantified to facilitate comparison between extraction procedures.

Figure 10 compares the representative chromatograms obtained by different extraction

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

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


Liquid-liquid extractions

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

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

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

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

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













i A


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







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











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


Component % RSD

Pentane-Ether Dynamic HS

Hexanal 10 7

a-Pinene 13 8

Myrcene 3 3

a-phellandrene 9 18

cis-linalool oxide 8 5

trans-linalool oxide 4 16

allo-ocimene 10 ND

a-terpineol 16 ND

Terpin-4-ol 12 ND

Canrophyllene 10 3

a-Humulene 7 ND

Nootkatone 8 ND











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


Peak Areas

Liquid-liquid

Kovat's Indices MeCI P&E Dy-HS


RI-801

RI-805

RI-814

RI-821

RI-834

RI-840

RI-844

RI-854

RI-872

RI-877

RI-891

RI-897

RI-909

RI-915

RI-924

RI-936

RI-941

RI-944

RI-965

RI-971

RI-982

Total top note peak area


913

6,149

2,318


2,785 13,872

10,925

49,342 421,594


3,929

16,318

7,199

15,610


10,612




6,145 8,103

4,986 5,364

8,167

10,128

76,251 10,660

39,368 106,486


10,737

1,809

6,800

7,087

173,175


34,206


8,495

94,108

13,347

44,674

3,256

12,443

5,434

20,843

10,472

43,638


31,574 8,307

12,518

24,217 22,148

299,882 780,279









62

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

analyses.










Static head space extraction using SPME

Solid phase micro extraction is a rapid procedure to sample volatile components in

head space gases. It involves the adsorption of head space volatiles onto a coated fiber

which is exposed to the head space for a specific time. In the static head space method,

volatiles in the sample matrix are allowed to come to an equilibrium with the head space

before being sampled. The SPME technique is relatively new technique and has been used

for analyzing orange essence volatiles (Bazemore, 1995), orange juice volatiles (Steffen

and Pawliszyn, 1996), head space of milk powder (Stevenson and Chen, 1996), and

cheese volatiles (Chin et al., 1996).

The SPME method effectively extracted terpenes such as limonene and myrcene

(as they were the largest peaks in the resultant chromatogram) but was relatively

ineffective in extracting the top note volatiles. The SPME fibers adsorb components on a

competitive basis. Since terpenes (especially limonene) are in higher concentrations in

grapefruit juice and also due to their non-polar nature, distribution coefficients and affinity

of fiber to non-polar components, they tend to dominate the head space components

trapped by the fiber coating.

Steffen and Pawliszyn (1996) reported good reproducibility for the components in

orange juice. However, the authors centrifuged the samples prior to analysis, which

eliminated the juice pulp and suspended solids. Lower levels of precision values were

obtained when sampling was done on whole grapefruit juice (private communications -

Bazemore, 1998). Since SPME emphasizes terpenes, which are in high concentrations but

contribute little to aroma, this technique was not used for further analysis in this research.












GC-Olfactometry Studies



GC-olfactometry (GC-O) is an important analytical tool since it characterizes the

odors of individual compounds and identifies which GC peaks have aroma activity (Mistry

et al., 1997). A human nose is used to detect and evaluate the effluents from the column

instead of an analytical detector. It is a powerful and sensitive tool since the odor

detection limit of a human nose is 10 moles (Reineccius, 1994), which is considerably

more sensitive than most instrumental detectors.

Grapefruit juice is a complex matrix and not all volatile components have aroma

activity. Even among those components which have aroma activity, some will have more

impact than others. Therefore, GC-O has been utilized to identify and characterize the

odor active components in grapefruit juice extracts. Aroma active components in

grapefruit juice change with the fruit maturity and also from thermal processing. In this

study aroma extracts from unpasteurized and pasteurized juices from early, mid and late

season fruits were evaluated for individual aroma active components.


Instrumental Detectors vs. Human Response


GC-O detects only those components which have aroma activity. Some of these

aroma active components are very potent and are present in such small amounts that they

cannot be detected by typical GC detectors. Figure 11 compares the consensus

aromagram (aroma intensities of 3 panelists were averaged) produced by GC-O with













0 il
,FID
FID
Ir .*>1 .., r


*


I-I
i
z?


IL,


Figure 11. Comparison of aromagram from OSME and chromatograms from FID and SCD
*p-menenthene-8-thiol


OSNIE


i

II C


SCDm

I


0 5 10 15 20 25 30 3
Ret Time (min)











chromatograms produced by FID and SCD detectors. The instrumental detectors

responded to some components which the human nose did not recognize. Conversely, the

human nose detected some compounds which gave no instrumental response. Large peaks

in FID like limonene and caryophyllene seem to have little to no aroma activity. Panelists

described limonene as citrusy, medicine and minty with a moderate intensity, while they

could not detect any aroma activity for caryophyllene. Earlier work by Marin et al. (1992)

also reported a limited aroma activity of limonene in orange juice.

Among the small FID peaks, vanillin is notable. It was found to have intense

vanilla or white chocolate aroma (average aroma intensity = 13). Vanillin has been

reported for the first time in grapefruit juice by our group. A strong intense aroma peak

was obtained at a 25 min retention time that has the characteristic aroma of vanillin (see

Figure 11). The same grapefruit juice extract was analyzed using GC-MS for further

confirmation of the presence of vanillin. By comparing the mass spectrum of the sample

with the mass spectrum of the standard, it can be concluded that the peak with aroma

attribute vanilla was, in fact, vanillin. The total ion chromatogram and the mass spectra of

vanillin sample and the standard, are shown in Appendix A and B. Prior to this, vanillin

was identified in orange juice by Marin et al. (1992). Peleg et al. (1992) proposed the

path ways for formation of vanillin from ferulic acid in orange juice (Figure 12).

According to the authors (Peleg et al., 1992), vanillin can form from ferulic acid through

decarboxylation and oxidation or directly from free ferulic acid through retro aldol

reactions. Similar reaction pathways may also occur in grapefruit. Intense aroma activity










Ferulic Acid




-C02
A


OCH3





-CH3COOH
H20


COOH


OC3


Oxidation


p-Vinylguaiacol

H2C = CH

Figure 12. Formation of vanillin from ferulic acid.


Vanillin


OCH3


O=CH









69

of vanillin was also reported in oak aged wines (Aiken and Noble, 1984), Japanese green

tea (Acree and King, 1996) and in coffee (Akieda and Kato, 1987).


Maturity and Processing Changes


In this part of the study, effects of maturity (early, mid and late) and processing

(unpasteurized and pasteurized) are evaluated using GC-olfactometry. Fruit maturity as

well as thermal processing affect the aroma quality of grapefruit juice. This is reflected in

the differences in number and kinds of aroma active peaks detected in juices from different

maturities. Figure 13a and b compares aroma attributes in early, mid and late season

unpasteurized and pasteurized juices. A total of 37 49 aroma active peaks were found in

early, mid and late season grapefruit juices. Appendix C lists the attributes perceived in

juices of different maturities. Forty-one aroma components could be differentiated in early

season unpasteurized juices while 37 peaks were detected in pasteurized juices. As a

result of thermal treatment 11 aroma compounds were lost while 7 new components were

formed in early season juice. However, many compounds were unchanged. Table 8

shows the aroma attribute compounds formed or lost during thermal processing of early

season juices.

Mid season juices had 43 aroma active peaks in unpasteurized juice and 49 in

processed juice. Similarly, 43 aroma active peaks were detected in both unpasteurized and

pasteurized late season grapefruit juices. Eight components were lost in thermally treated

late season juices, while 8 new attributes were detected. The aroma active peaks lost due




























800 1000


1200 1400
Retention Index


1600
(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.

















49



J




800 1000 1200 1400 1600
Retention Index (attributes listed in Appendix C)
Figure 13b. Number of aroma active components at different maturities in pasteurized grapefruit juice: A) early season,
B) mid season and C) late season.











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

Components/ Intensities
Retention Indices Pasteurized Unpasteurized Description
RI-896 7.0 Citrusy, Mediciny
RI-936 5.7 Floral. Smokey
a-Pinene 8.2 Greenish
a-phellandrene 12.3 Citrus
RI-1044 7.6 Rotten Fruit
RI-1095 10.3 Terpeney, Cucumber
RI-1116 12.7 Mediciney
RI-1166 10.4 Musty
RI-1217 7.3 Terpeney
RI-1223 9.3 Musty
RI-1227 7.8 Stinky fruit
RI-1318 10.8 Smokev, Rancid
RI-1374 6.8 Medicmey, Minty
RI-1381 92 Sweet
RI-1510 7 2 Spicey. perfumey
RI-1662 4.0 Peppery
RI-1684 7.2 Pungent
RI-1723 8.6 Rotten Grapefruit










to pasteurization had generally favorable sensory attributes like green, fruity while the

components formed as a result of heating had roasted, fruity, and spicey attributes.

Concentration of the components also changes due to maturity and thermal

processing. Total alcohols, aldehydes and hydrocarbons were higher in early season

unpasteurized juice (Figure 14a). Among the alcohols, a-terpineol, terpin 4-ol, trans

linalool oxide, and among the hydrocarbons, myrcene and y-terpinene were found to

correlate negatively with sensory preference of grapefruit juice (Jella et al., 1998). These

negatively correlated compounds were present in higher concentrations in early season

than in late season juices. Results of this are summarized in Table 9. The levels of these

components in grapefruit juice are in concurrence with those reported by Maarse and

Visscher (1989).

As a result of pasteurization, increased concentrations of alcohols, aldehydes and

hydrocarbons were observed (Figure 14b). Higher levels of alcohols are probably due to

acid catalyzed reactions of terpenes like limonene, P-pinene, myrcene and so on. These

components react in dilute aqueous acid and high temperatures to give several reaction

products, some of which are alcohols like a-terpineol, terpin-4-ol and linalool oxides

(Clark and Chamblee, 1992 and Shaw, 1991). Limonene is the major terpene in citrus

juices and readily forms several reaction products under the conditions present in citrus

juices.

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












A














Alcohols Aldehydes Hylrocar ons
(excluding Limonene)





B


Alcohols Aldehydes Hydrocarbons
(excluding Limonene)
Figure 14. Concentrations of components in grapefruit juice.
A) unpasteurized juices; B) pasteurized juices: (*) early season,
( ) mid season and (r-) late season


Ketones


Ketones


2

0


12

10

S8


6

4

2


0









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









76
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


Hexenal

Ethyl Butyrate

t-2-hexenal

a-Pinene

Myrcene

Linalool

Terpin-4-ol

a-terpineol

p-menthene 8-thiol

Nootkatone


Green

Fruity

Green, dead bug

Medicine, Piney

Unripe Mango

Floral

New cotton clothes

Cilantro, Musty

Rotten gft, Stinky terpeney

GFt stink, Rotten Gft


Green

Fruity, Floral

Skunky

Piney, Green

Melon

Linalool

Stale Church

Moldy, Musty

Stinky Rotten GFT

Moldy GFT


Strong green

Fruity, Citrus

Smoked burnt roasted

Greenish, Vitamin C

Citrus

Stinky floral

Vinyl

Green cilantro

Sweet grapefruit

Stinky grapefruit


Benzaldehyde (IS)

Methyl Jasmonate (IS)


Cherry, Almond

Floral, Jasmine


Cherry, Almond

Floral, Jasmine


Cherry, Almond

Floral, Jasmine








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

I Component Standard Descriptor Panelist's Descriptor


Hexenal, Ethyl Butyrate
t-2-hexenal
a-Pinene
Benzaldehyde
Sabinene
Myrcene
a-phellandrene
para-cymene
trans-p-ocimene
y-terpinene
cis-linalool oxide
trans-linalool oxide
Nonanal
Linalool
a-terpineol
carvone
p-menthene-8-thiol
vanillin
Nootkatone


Green/Warm sweet fruity
Green vegetable like
Warm resinous and herbacious
Bitter almond, sweet cherry
Warm peppery, herbacious
Balsamic resinous and citrusy
Citrusy, peppery, woody
Citrusy, kerosene like
Warm herbacious, sweet
Herbacious, citrusy
Sweet floral earthy
Sweet floral earthy
Fatty, floral
Floral woody
Lilac, piney
Warm herbaceous
Grapefruity
Creamy, vanilla like
Fruity, citrusy


Green/Fruity
Green, dead bug, skunky
Medicine
Cherry, almond
Unripe mango, piney
Unripe mango, citrus
Citrus
Minty, citrusy
Citrusy, musty
Roasted cotton candy
Terpeney, cotton candy
Terpeney, cotton candy
Terpeney, cucumber, cotton candy
Floral
Musty, wet dog, cilantro
Floral, liquorice, mediceney
rotten nutty grapefruit fruit
Vanilla
Grapefruit










Arctander (1994). Nootkatone was described by panelists as rotten fruity, sweet

grapefruity, stinky citrusy. Arctander's descriptor for nootkatone is citrusy. Comparison

of retention times, indices and the odor description given by the panelists for the standard

gives a good indication that same aroma active peak is being described.


Grapefruit Aroma


Nootkatone is considered by some scientists to be one of the important

contributors to the grapefruit flavor (Stevens et al., 1970; Pino et al., 1986a; Shaw and

Wilson, 1981). Maturity plays a significant role in determining the quantity of this

sesquiterpene ketone. Traditionally, late season juices are considered to be best quality.

Higher amounts of nootkatone were found in late (9.1 and 10.8 ppm) than in mid (3.2 and

3.9 ppm) and early (1.8 and 1.9 ppm) season unpasteurized and pasteurized juices.

However, nootkatone was a poor predictor for juice quality (r=-0.05) in this 30 juice

sample set of mid and late season juices.

Grapefruity aroma was also perceived by the panelists a few seconds before

nootkatone has eluted. This peak had a Kovats indices or retention index (RI) of 1754.

This peak has been tentatively identified as 8,9 didehydro nootkatone based on retention

index and aroma quality. This has been reported to be present at 0.001 ppm level in

grapefruit juice (Maarse and Visscher, 1989). Demole and Enggist (1986) reported its use

to augment or enhance the organoleptic properties of grapefruit or imitation grapefruit

beverages. This GC-O peak also occurs at the same time as one of the large sulfur peaks,

RI-1753 (retention time 32 min). Since both these components have similar retention











times, it is not currently resolved which component is responsible for the additional

grapefruit aroma peak. The question will have to be resolved with additional experiments

using chromatographic columns of different selectivity. Another important sulfur

component having a fresh grapefruity aroma isp-menthene-8-thiol. Discussion of this

component is included in a later section.


Dilution Analysis


Studies involving dilution analysis (AEDA, Charm) on grapefruit juice have not

been reported to date. However, orange juice has been extensively studied (Marin et al.,

1992; Hinterholzer and Schieberle, 1998) with both AEDA and Charm. Among the

components reported by the authors, hexenal, ethyl butyrate and vanillin were found to

have highest dilution values, while linalool, decanal were found at the lower end of the

dilution factors.

The peaks detected in our study were aroma active peaks from the juice extract

concentrated 160 times. This does not provide information about which of these peaks

have intense aroma activity at higher dilutions (lower concentrations). Since components

in juice are not present in concentrated form, dilution analysis was done to identify the

most aroma active, now referred to as aroma impact peaks. To assess the most intense

peaks, juice extract was concentrated 16 times instead of 160 times and analyzed using

GC-O as before. The list of peaks identified and their corresponding odors are given in

Table 12. Some of the components like cis and trans linalool oxides were not present in

the samples at 16 X concentration even though these components had intense aroma











activity (13 on a 15 point scale) in 160 X concentrated samples, da Silva et al. (1994)

stated that odorants have different intensities above their threshold values, that is, aroma

intensity may not be proportional to the concentration of the compound. According to

Meilgaard et al. (1991), a mathematical model proposed by Beidler works best for middle

and high range of sensory intensities. According to this model, there is a sigmoidal

relationship between the concentration of the product and the stimulus perceived. This

might be the reason for lack of odor perceptions of components like linalool oxides and

linalool at lower concentrations of grapefruit juice aroma extract.

The two attributes which had intense aroma activity in the 16 X grapefruit juice

extract were hexanal/ethylbutyrate and a-phellandrene (see Table 12). When these two

components were used for sensory correlations (discussed in section 5 of results and

discussion), they were found to have significant correlations (0.31 and -0.28 at p< 0.05)

with aroma intensity. This suggests that hexanal/ethylbutyrate and a-phellandrene are key

components in determining the quality of grapefruit juice.


Sulfur Compounds in Grapefruit



Detection


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
t-2-hexenal
RI-863
RI-936
Sabinene
Myrcene
c-phellandrene
para-cymene
trans-P-ocimene
y-terpinene
RI-1116
allo-Ocimene
RI-1141
RI-1177
a-terpineol
p-menthene-8-thiol
RI-1349
RI-1374
RI-1381
Vanillin
RI-1464
RI-1510
RI-1723
RI-1754
Unknown Sulfur cmpd (RT 32min)
Nootkatone


Green/Fruity
Green
Mushroom
Sweet fruity
Unripe
Unripe citrusy
Green, Citrusy
Citrusy
Green, Citrusy
Floral, Citrusy
Grainy, Mediciny
Sweet
Nutty
Green, Musty
Musty
Stinky Grapefruit
Vinyl
Apple Sauce
Apple Sauce
Vanilla
Burnt
Perfumy
Incense
Grapefruity
Grapefruity
Grapefruity


7.61
1.59
3.12
3.22
3.06
4.38
7.20
5.26
2.83
6.89
3.02
4.89
4.15
3.33
4.88
4.76
6.39
3.76
6.03
4.91
4.24
2.18
4.59
3.08
2.29
6.76











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

conclusions.

Twenty-two sulfur compounds were isolated in early, mid and late season

pasteurized and unpasteurized grapefruit juice. This represents the most comprehensive

determination of sulfur compounds in citrus juices reported to date. Total number and

total sulfur peak areas decreased with increasing fruit maturity (Figure 17a) and increased

with processing (Figure 17b). Total peak area of early season pasteurized juices was 83

times more than early season unpasteurized grapefruit juices. Late season pasteurized




Full Text

PAGE 1

AROMA AND TASTE Th1P ACT COtv!PONENTS IN GRAPEFRUIT JUICE By PRASHANTHI JELLA A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILL:MENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1998

PAGE 2

This thesis is dedicated to Shirdi Sai

PAGE 3

ACKNOWLEDG11ENTS I would like to expres s m y gratitude to my advisor Dr Russell Rouseff for his a s sistance and support during the cour s e of my research His constant encouragement promoted independent thought and hi s words '' You Can Do It ',, '' Go Get Them '' challenged me at e v ery 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 o ne 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 the y were one of the most cherished e x periences for me I will never forget the definition of ''a thi x otropic fluid '' and the way he demonstrated it in the class He is one of the best teachers I had who never took no for an answer but helped the students to work through the problem There are no words to e x pre s s my thanks to friends and room mates from Te x as A&M who are like a second famil y to me The bonding we ha v e is a s pecial one and I ... ill

PAGE 4

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

PAGE 5

My fatherand 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 ROHINI! ! V

PAGE 6

TABLE OF CONTENTS ACKNOWLEDG1\1ENTS . . . . . . . . . . . . . . . . . . . . . . . iii LIST OF TABLES ........... ......... ................. ........... ix LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . xi ABSTRACT CHAPTERS XIV 1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . 1 2. LITERATURE REVIEW . . . . . . . . . . . . . . . . . . . . . . 4 Flavor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Statistical Correlations . . . . . . . . . . . . . . . . . . . 5 B ittemess . . . . . . . . . . . . . . . . . . . . . . . . 7 GC-0 If acto me try . . . . . . . . . . . . . . . . . . . . . . . . 9 Charm Analysis . . . . . . . . . . . . . . . . . . . . . 9 AEDA . . . . . . . . . . . . . .. . . . . . . . . . . . 10 0 Sl\1E . . . . . . . . . . . . . . . . . . . . . . . . 10 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . 12 SP1\1E . . . . . . . . . . . . . . . . . . . . . . . . 13 Sulfur Compounds . . . . . . . . . . . . . . . . . . . . . . . 15 3 MATERIALS AND 1-ffiTHODS . . . . . . . . . . . . . . . . . . . 17 Grapefruit Juice Sample Collection . . . . . . . . . . . . . . . . . 17 Survey Samples . . . . . . . . . . . . . . . . . . . . . 17 Methylene chloride extracts . . . . . . . . . . . . . . 17 Pentane-diethyl ether ex.tracts . . . . . . . . . . . . . 18 GC-Olfactometry Samples . . . . . . . . . . . . . . . . . 18 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . 19 Liquid-Liquid Extraction with Methylene Chloride . . . . . . . . 19 Liquid-Liquid Extraction with Pentane-Diethyl ether (1 : 1 ) : ......... 20 V l

PAGE 7

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-Mass Spectrometry . . . . . . . . . . . . . . . . . . 24 Limonin and N aringin Analysis Using HPLC . . . . . . . . . . 24 Sample preparation . . . . . . . . . . . . . . . . . 24 HPLC instrumentation . . . . . . . . . . . . . . . 25 Peak Identification and Quantification . . . . . . . . . . . . . . . 25 Sensory Analysis . . . . . . . . . . . . . . . . . . . . . . . 2 7 DOC Preference Panel . . . . . . . . . . . . . . . . . . 27 USDA Descriptive Panel ..... .. ........... .......... ..... 31 Training of Panelists . . . . . . . . . . . . . . . . . . . . . . 3 3 GC-Olfactometry Panel . . . . . . . . . . . . . . . . . . 33 Descriptive Panel . . . . . . . . . . . . . . . . . . . . 34 Statistical Analysis . . . . . . . . . . . . . . . . . . . . . . . 3 4 4 RESULTS AND DISCUSSION . ....... ............................. 36 Correlations Between Preference and Analytical Measurements ............ 36 Sensory Analysis . . . . . . . . . . . . . . . . . . . . 40 Statistical Analysi s . . . . . . . . . . . . . . . . . . . . 40 Uni v ariate analysis . . . . . . . . . . . . . . . . . 40 Multivariate analysis . . . . . . . . . . . . . . . . 4 3 Identification of the Peak at RI-1126 . . . . . . . . . . . . . 54 Grapefruit Juice Aroma Extraction Methods . . . . . . . . . . . . . 5 6 Chromatographic Separation and Analysis . . . . . . . . . . . 56 Extraction Methods . . . . . . . . . . . . . . . . . . . 5 8 Liquid-liquid extractions . . . . . . . . . . . . . . . 5 8 Dynamic head space extraction . . . . . . . . . . . . 62 Static head space extraction using SP1\1E . . . . . . . . 6 4 GC-Olfactometry Studies ............. ................ .......... 65 Instrumental Detectors vs Human Response . . . . . . . . . . 65 Maturity and Processing Changes . . . . . . . . . . . . . . 69 Standard Descriptors Vs Panelist s Descriptors . . . . . . . . . 76 Grapefruit Aroma . . . . . . . . . . . . . . . . . . . . 8 0 Dilution Analysis . . . . . . . . . . . . . . . . . . . . 81 Sulfur Compounds in Grapefruit . . . . . . . . . . . . . . . . . . 82 Detection . . . . . . . . . . . . . . . . . . . . . . . 82 Processing and Maturity Effect s . . . . . . . . . . . . . . . 84 p-menthene-8-thiol . . . . . . . . . . . . . . . . . . . . 88 Correlation Between Aroma Component s and Sensory Measurement s ....... 8 9 Vll

PAGE 8

Juice Classification . . . . . . . . . . . . . . . . . . . . 8 9 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 . . . . . . . . I 07 Flavor models using aroma and taste components . . . . . 111 5 CONCLUSIONS 119 Correlation Between Preference and Analytical Measurements . . . . . . 119 Aroma Extraction Methods . . . . . . . . . . . . . . . . . . . 120 GC-Olfactometry .......... ........... . .. ..... ..... ...... 121 Sulfur Compounds in Grapefruit Juice . . . . . . . . . . . . . . . 122 Correlations Between Aroma Components and Sensory Measurements .. .. 123 APPENDICES A TOTAL ION CHROMATOGRAM OF LATE SEASON GRAPEFRUIT JUICE 125 B MASS SPECTRUM OF V ANILLIN ..... . ........... ............. 127 C LIST OF DESCRIPTORS AND THEIR RELATIVE INTENSITIES (GC-0) ... 129 D. COMPOUNDS IDENTIFIED IN NOT-FROM-CONCENTRATE GRAPEFRUIT JUICE ....... ... ......... ...................... ......... 132 LIST OF REFERENCES ........... .......... .... . ............... 136 BIOGRAPHICAL SKETCH . . . . . . . . . . . . . . . . . . . . . 144 Vlli

PAGE 9

LIST OF TABLES Table page 1 Calibration equations used for calculating tl1e concentrations of components detected in GC-FID ... ....... ...... .. .. . . . ........ ..... 29 2 Maxim11m minimum and a v erage area percent for components extracted with methylene chloride. . . . . . . . . . . . . . . . . . . . . . . 3 8 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 fru t . 75 grape 1 Juices . . . . . . . . . . . . . . . . . . . . . . . . 10 Aroma descriptors used by panelists from GC-0 experiments of citrus standards 78 11 Comparison of standard (Arctander lexicon) with panelist descriptors ... ... 79 12 List of components present in l 6x concentrated juice extract and their intensities and aroma attributes . . . . . . . . . . . . . . . . . . . . . . 83 13 Minim,1m and maximum descriptive sensory panel scores for grapefruit juices . 92 IX

PAGE 10

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 distance s 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 Discrlmi nant Analysis) . . . . . . . . . . . . . . . . . . . . . 112 18 Forward step wise discriminant analysis-volatiles and taste components (Number of steps and corresponding component). . . . . . . . . . . . . . . 116 19. Comparison of sensory and statistica l classification of grapefruit juices. (Model has been tested using I 7 aroma components and 5 taste components) . . . 117 X

PAGE 11

LIST OF FIGURES Bwre page I Standard curve used for calculation of Kovat s retention indices for volatile components . . . . . . . . . . . . . . . . . . . . . . . . . . 2 7 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 pane] . . . . . . 3 2 5 Chromatogram of methylene chloride extract of pasteurized (NFC) grapefruit juice on a DB-5 column . . . . . . . . . . . . . . . . . . . . . 3 7 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 = medium preference category and b. = low preference category. .... . ..... 44 6b Eigenvector values of PC I vs PC 3 from principal component analysis of all 57 volatile and non-volatile components where = high preference category = medium preference category and b. = low preference category. . ........... 45 7a Peak Areas oflinalool and caryophyllene from 29 grapefruit juice extracts analyzed in triplicate where = high preference category = medium preference category and b. = low preference category . ... .... . .......... ..... . .... 47 7b Peak Areas of myrcene and caryophyllene from 29 grapefruit juice extracts analyzed in triplicate where = high preference category preference category and b. = low preference category. . . . . . . . . . . . . . . 48 8a Canonical Discriminant Analysis of using myrcene linalool 0 Brix, and the peaks at RI 1677 and 1126 where = high preference medium preference category and b. = low preference category . . . . . . . . . . . . . . 5 2 X1

PAGE 12

8b Canonical discriminant analysis using thirteen variables (Brix/Acid ratio RI93 5 cis linalool oxide Nonanal a//o-ocimene cx. terpineol Decanal RI-1299 cx. copaene P-gurjunene, RI-1762 RI-1796) where = high preference category = medium preference category and 6. = 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) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 9 11 Comparison of aromagram from OSrvIB 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 13 b 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 : (~) early season ( ) mid season, ( o) late season ........ 74 15 Acid catalyzed l1ydration of limonene . . . . . . . . . . . . . . . . 77 16 Sulfur chemiluminescence reactions . . . . . . . . . . . . . . . . . 8 5 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 fla vo r score and sweet/tart balance . . . . . . 97 20 Correlation between aroma quality score and overall fla v or score. . . . . . 99 21 Correlation between aroma quality and nootkatone peak area . . . . . . 103 X1J

PAGE 13

22 Standard discriminant analysis using 5 taste components: ( + ) worst category ( e ) fair category ( ..._ ) good category ( ) best category juices . . . . . . . 106 23 Forward stepwise discrimin ant analysis using 17 aroma and 4 taste components : ( + ) worst category ( e ) fair category ( ..._ ) good category ( ) best category juices . . . . . . . . . . . . . . . . . . . . . . . . . . . 114 Xlll

PAGE 14

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 IMP ACT COMPONENTS IN GRAPEFRUIT JUICE Chairperson : Russell L Rouseff By Prashanthi J ella August 1998 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 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 XlV

PAGE 15

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 Vani11in was one of the aroma active peaks detected in grapefruit juice for the first time using GC-olfactometry and GC-MS. A routine method for grapefruit juice sulfur compounds using sulfur chemiluminescence detection was developed for the first time. Twenty two sulfur compounds were detected. Total peak area increased with pasteurization but decreased with maturity p-menthene-8 thiol a key aroma impact component increased as much as 143% after pasteurization FID peak areas of 26 aroma active volatile components extracted with pentane diethyl ether and 5 taste components were correlated with sensory descriptive panel results Myrcene and a-terpineol corre l ated 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 xv

PAGE 16

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

PAGE 17

CHAPTER 1 INTRODUCTION 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 ( Citrus paradisi Mcfadyen) has a highly distinctive flavor with slight bitterness and tanginess. Florida is the wor ld s leading producer of this fruit with a record production of 2 4 million tons in the year 1996-97 However the val ue of the crop for this season was$ 68 436 000 which is the l owest 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 Nie l sen numbers for supermarkets the demand for grapefruit juices has decreased (50 million gallons from 1991 to 42 2 million gallons in 1996) whi l e production (61 million boxes of fruits for 1996) of grapefruit increased (Stinson and Barros 1997) Decreasing popularity of this 1

PAGE 18

2 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 ; Rouseff et 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

PAGE 19

7 Determining the relationship between sensory and analytical data using multivariate statistics. 3

PAGE 20

CHAPTER2 LITERATURE REVIEW Flavor 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 ; Rouseff et al ., 1980 ; Fellers et al ., l 986). 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 ; Nunez 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 4

PAGE 21

5 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 l-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 was p-menthene-8-thio~ 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 lxl 01 ppb (Demole et al ., 1982) However until an analytical procedure is developed to quantify this compound at the levels at which it exists in juice it will not be possible to evaluate its relative contribution to grapefruit juice flavor Statistical Correlations Flavor is unquestionably one of the most important attributes of food and is perceived as taste by the tongue and mouth and through the release of the volatile components in the mouth which are sensed retronasally by the olfactory epithelium in the nose (Ohloff, 1990) Previous workers have developed models based on the correlations between quantified volatile and sensory data (Jennings 1977 ; Pino et al ., 1986 a b ) A few orange juice volatiles characterized using packed column gas chromatography and non-volatile

PAGE 22

I 6 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 s elected 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 butane~ 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 wor ld leader in the production of grapefruit juice, no systematic study to deterruine 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 negati v ely with flavor In another experiment Pino et al ( 1986b) correlated sensory and chromatographic measurements of grapefruit juice v olatiles using multiple linear regression Methyl butyrate ethyl butyrate limonene decanal and nootkatone correlated with positi v e

PAGE 23

sensory perception while transand 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 Bitterness 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 7 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 (1981 a) 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 (1981 b) observed that limonin content was less than 6 ppm by March for most commercial grapefruit varieties. In genera~ 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 limooin The authors quantified nomilin and limonin in commercial grapefruit juices produced in the 197879 season and observed low nomilin concentrations in all juices Rouseff et al

PAGE 24

(I 980) 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 (I 981) 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 8

PAGE 25

temperature of the bed was found to be critical in reducing the amount of limonin and naringin (Wilson et al ., 1989) GC-Olfactometry 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-0) It utilizes a human assessor to deterrnine which of the many chromatographic peaks have aroma activity and characterizes that odor. Some of the GC-0 techniques available today are Charm Analysis Aroma Extraction Dilution Analysis (AEDA) and OS:NtE 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 OS:ME determines intensities without dilutions Charm Analysis Acree et al (1984) developed the Charm analysis technique, and has used this technique to evaluate a variety of products Cunnigham et al (I 986) 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 juice s 9

PAGE 26

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 components AEDA 10 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 37 C) 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) 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 0Sl\1E da Silva et al. (I 994) claimed that the dilution techniques mentioned above would not give accurate information since the odorants have different intensity functions above

PAGE 27

11 their threshold levels The authors proposed and developed a new GC 0 methodology based on psycho-physical laws called OS:ME (Greek word meaning smell) OS:ME is a time intensity procedure which deter111ines 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 OS:ME. 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-0 were not detected by an analytical detector (GC-FID) One characteristic feature of GC-0 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

PAGE 28

activity. Upon enrichment of the extract by the authors geosmio was identified as the compound producing the musty odor Sample Preparation 12 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 poly1ner 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 extract.

PAGE 29

13 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 Umana 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 ( cysteamine) was used to trap aldehydes ( as derivatives of thiazolidine) 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 respecti v ely the predominant component being limonene in all cases SPME 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 quantitati v e 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

PAGE 30

14 polydimethylsiloxane fiber coating was used GC-MS analyses of the aromas showed that the SPNIE 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 120% 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 SPNIE SPNIE 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-containin g compounds in the head space of milk powder (Stevenson and Chen 1996) Chin et al (I 996) observed that SPNIE 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 sa mple 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 = K fs V tC o V s/ (Kr s Vr + V s ) where n is the mass of the analyte absorbed by the coating ; V r and V s are vo lumes of coating and sam ple respectively ; K fs is the partition coefficient of the analyte between the coating and the sample matrix ; C 0 is the initial concentration of the analyte in the sample However since V s > > !
PAGE 31

15 Sulfur Compounds Sulfur compounds play a major role in determining the flavor characteristics of many food substances. Sulfur compounds are often form ed as a result of the enzymatic process when plants are cut or chewed releasing fla vo r 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 hi g her levels of the identical compound often resu lt in off flavors (Tress! 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-1-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 thresholds 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

PAGE 32

16 sulfides using a flame photometric detector in orange juice sa mples Since concentrations ofH 2 S 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 ofH 2 S When sensory analysis was conducted on these juices the panelists reported a harsher (pungent) aroma, and the authors attributed this to higher H 2 S levels This attribute was not detected by the authors in late season orange and grapefruit Jwces Demo le et al ( 1982) isolated and characterized p-menthene-8-thiol which bad 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-8thiol 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 compo unds are reported to co-occur in grapefruit (Demole et al ., 1982).

PAGE 33

CHAPTER3 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 0 Brix titratable acid limonin and naringin separately in order to evaluate the relative contribution of taste vs aroma components. Sensory attributes were quantified and correlated with analytical measurements Experimental design and analytical techniques used to achieve this objective are discussed in this chapter. Grapefruit Juice Sample Collection Survey Samples Methylene chloride extracts Twenty-nine not-from-concentrate {NFC) grapefruit juice samples were obtained from processors with processing dates ranging from November 1995 to June 1996 and stored at -8 C until analyzed. Both red / pink and white juices were used in this study Authentic solvents were purchased from Fisher Scientific (Pittsburgh PA) Standards used for quantifying volatiles and non-volatiles were purchased from Aldrich Chemical 17

PAGE 34

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 18 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 so l vents and standards were the same as for the methy l ene chloride extracts study GC Olfactometry Samples Early (November 1996) mid (January 1997) and late (May 1997) season grapefruit juice samples were obtained from the Florida Department of Citrus Grove run red grapefruit were purchased from a local packinghouse and processed in the pilot plant located at the Citrus Research and Education Center in Lake Alfred Fruits were washed dried and sized for the extractors in the pilot plant Extraction was accomplished using commercial FMC model 391-B and 491 extractors with standard juice settings An FMC model 3 5 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 6 C at a flow rate of 1 gallon per minute Samples were packaged in 32 oz clear glass bottles and stored at -8 C

PAGE 35

until analyzed Both unpasteurized and pasteurized red grapefruit juices were obtained. Samples consisted of two bottles for each juice type Sample Preparation Liquid-Liquid Extraction with Methylene Chloride 19 Extraction of vo latile s 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 l ong 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 L of propyl benzene was added and the extract was concentrated to about 30 Lin a 100 L graduated taper via l 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

PAGE 36

20 Liquid-Liquid Extraction with Pentane-Diethyl ether (I : I) : Extraction of the volatiles was accomplished according to the previously described method except a 1 : I mixture of pentane and diethyl ether was used in the place of methylene chloride Two internal standards propyl benzene (50 L of 100 ppm) and 2heptadecanone (25 L of 500 ppm) were added to 8 mL of juice and extracted The extracts were concentrated to 50 L 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 I : 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 m.L / min) for ca I minute to reduce trapped moisture Three mL of (1 : I) 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 wa s cleaned both before and after extraction using 3 to 4 times the column v olume of pentane

PAGE 37

21 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 L of 10 ppm) was added to 8 mL of juice as an internal standard and extracted. Extracts were concentrated to 50 L using nitrogen with the procedure described earlier for the methylene chloride extracts All samples were analyzed in duplicate 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 L of 5000 ppm) and methyl jasmonate (25 L of 5000 ppm) were added to 8 mL of juice and extracted Extracts were concentrated to 50 L using dry nitrogen as previously described Each sample was analyzed four times using three detectors (Flame Ionization Detector (FID) Sulfur Chemiluminescence Detector (SCD) and OS:ME) Instrumental Techniques GC-F]ame 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 m film thickness

PAGE 38

22 low bleed DB-5 column (J&W Scientific ; Folsom CA) The oven temperature was programmed from 3 5 to 275 C at 6 C / 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 C and detector temperature at 320 C The nitrogen gas flow was maintained at 19 mL / min while air and hydrogen flows were maintained at 296 and 3 5 mL/min respectively The injection volume was I L for methylene chloride extracts and O 5 L 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 IO pt/sec Chromatograms for methylene chloride extracts were recorded and integrated using an APEX Chromatography Workstation (Autochrom Inc ., Milford MA) with a four channel data system Data acquisition rate was O. 4 s / point GC-Su1fur 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 m fi]m thickness low bleed DB-5 column (J&W Scientific ; Folsom CA) Oven temperature was programmed from 35 to 275 Cat 6 C / min with helium at a flow rate of 2 19 mL / min Injector temperature was maintained at 250 C Internal temperature of the SCD burner head was 780 C. 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 Lin split-less mode

PAGE 39

23 Chromatograms were recorded and integrated using Chrom Perfect (Justice Innovations Mountain View CA) The data acquisition rate was 10 pt / sec GC-OSME 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 m x 0.25 mm i.d x 0.5 m film thickness low bleed DB-5 column (J&W Scientific ; Folsom, CA), with helium at a fl.ow rate of 1.55 m.L/min Oven temperature was programmed from 35 to 275 Cat 6 C / min Injector temperature was maintained at 250 C and detector temperature at 320 C Purified air was obtained by passing compressed air through drierite and a molecular sieve 5A (Alltech, Deerfield IL) and directed into a temperature controlled water filled round bottomed fl.ask fitted with fritted glass impringers. Water temperature was maintained at 3 5 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 (O = 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.

PAGE 40

24 GC-Mass Spectrometry All GC-MS data were collected using a Finnigan GCQ Plus system (Finnigan Corp, San Jose CA) using helium (99 999%) for the GC carrier gas and the collision/bath gas in the ion trap Injector temperature was 250 C Samples (0 2-1.0 L) were injected using the split less mode with a purge time of 1 5 min The initial column temperature was held at 3 5 C for 3 min followed by a 4 C/min temperature ramp to 221 C which was followed by a IO C/min ramp to 27 5 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 27 5 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 3 00 in order to achieve 7 spectra per second Ionization energy was set at 70 e V Limonin and Naringin Analysis Using HPLC Sample preparation Limonin and naringin for grapefruit juice samples ( extracted with methylene chloride) were analyzed according to the method developed by Widmer and Martin (1994) In a 10 mL volumetric flask, 5 mL of juice 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 filter About 2 mL of filtered sample were placed into 2 5 mL Snap-Its (National Scientific Company Quakertown PA) glass vials and used for further analysis.

PAGE 41

25 For grapefruit juice samples extracted with pentane-diethyl ether solvent mixture limooin and naringin analysis was similar to the procedure described above except that these were not heated prior to HPLC analysis HPLC instrumentation A Thermo Separations (San Jose CA) LC system (Spectra Focus Optical Scanning detector and P4 000 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) inte grator and Winner on Windows 4290 (San Jose CA) Separations were achieved using a 4 6 mm x 150 mm 5 CN analytical column (MacMod Analytical Inc ., Chadds Ford PA) for Iim onin and a 4 6 mm x 150 mm 5 C -18 analytical column (Kromasil C-18 Higgins Analytical Mountain View CA) for naringin The mobile phase consisted of water / acetonitrile (80 5 : 19 5) for naringin analysis and water / acetonitrile (63 : 37) for limonin analysis The injection volume was 40 L and flow rates of 1 0 mL / min were used Peak Identification and Quantification Chromatographic peaks were identified using their mass spectra and comparison of their observed Kovat s index with published Kovat s retention indices (Kovats 1965)

PAGE 42

26 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 C 18 were used for this Kovat s Indices for these standards were calculated by multiplying the corresponding carbon number by a factor of 100 Retention time (seconds) for the standards were plotted against their corresponding Kovat's Indices (Figure 1) The resulting plot was used to fit an equation, which was then used to calculate the retention indices for individual grapefruit juice volatile components. Quantification of some of the GC-FID peaks from early mid and late season grapefruit juice was done by using authentic standards obtained from Sun Pure Inc (Lakeland FL). Solutions of ethylbutyrate propyl benzene sabinene myrcene octanal linalool decanaJ 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

PAGE 43

27 1600 )( a, 1200 'O C: II) nJ > 0 800 X: 400 0 -1-------+------+---------1------+---------1 0 500 1000 1500 2000 R!tention time (sec ) Kl = a+ (b*t 2 + c..f t + d) ln(t) Where Kl Kovat s retention index t Reten t ion time in seconds and constants a = 592 525 b = 2 003e-5 c = 1 703 and d = -12 342 Figure 1 Standard curve used for calculation of kovat s retention indices for volatile components 2500

PAGE 44

ns Q) ... 4000000 y = 13983x + 1 32945 3000000 R 2 = 0 9943 2000000 1000000 0 ----+-----1----1--------+--------i 0 50 100 150 200 250 Concentration (ppm) (A) 2000000 ....-----------------__...,_ --, y = 8766 1x 71662 R 2 = 0 9951 m 1000000 .. < O -l-----.1-----1------+-------1----------1 0 50 100 150 200 250 Concentration (ppm) flJ a, ... < 3000000 y = 9122 8x 83140 2000000 R 2 = 0 9964 1000000 0 -----+------1--t------+-------; 0 50 100 150 200 250 Concentration (ppm) (B) 3000000 ~------------2000000 1000000 y = 9715 7x 56480 R 2 = 0 997 0 ------------------0 50 100 150 200 250 Concentration (ppm) (C) (D) Figure 2 Calibration curves used for quantifying the volatiles (A) propyl benzene (B) myrcene (C) linalool (D) nootkatone N 00

PAGE 45

Table 1 Calibration equations used for calculating the concentrations of components detected in GC-FID. Component Ethyl butyrate Propyl benzene Sabinene Myrcene Ocatanal Linalool Decanal Nerol Caryophellene Nootkatone 2-heptadecanone Linear regression equation 6128*x-29184 13983*x + 132945 7586*x 61320 9123*x 83140 9093*x 108034 8766*x 71662 4951 *x 44666 9178*x 69883 8257*x 80289 9716*x56480 13889*x 16990 r-squared 0 997 0.994 0 997 0 996 0.968 0.995 0.999 0.990 0 .993 0.997 0 .99 4 Note : x in the linear regression equation represents the area of the peak to be quantified 29

PAGE 46

600000 -------------------, 400000 200000 0 0 2 4 y = 50873x 2369.3 R 2 = 0 9997 6 Concentration (ppm) 8 10 Figure 3 Calibration curve for s-methyl-thiobutanoate (sulfur compounds) 30

PAGE 47

31 Sensory Analysis DOC Preference Panel Bitterness taste thresholds of individual taste panelists were determined using 5-50 ppm of limonin and 150-950 ppm of naringin 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 present ed 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 O 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.

PAGE 48

Grapefruit Juice Sensory Panel Please Read the Instructions 32 Name: Sample Number : 321 Date: 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 0 None Grapefruit Aroma Quality 0 V Poor Sweet/Tart Balance Bitterness 7 More Sour Than Sweet 0 None Overall Flavor Quality 0 V Poor 0 15 Strong 15 V Good 7 More Sweet Than Sour 15 Strong 15 V Good 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

PAGE 49

33 Panelists were given this juice 6 times over 3 weeks and the scores for individual attributes were averaged. Average scores for all attributes were marked on the ballot sheet to serve as reference points for other samples Panelist s consistency was checked by giving the reference sample after every 10 grapefruit juice samples Samples were coded with random three digit numbers randomly arranged on serving trays and then presented to panelists Training of Panelists GC-Olfactometry Panel The panel consisted of 2 males and 1 female. Training consisted of three practice runs with grapefruit juice extracts to familiarize the panelists with the sliding scale optimum positioning and breathing technique and to provide practice with verbal descriptors In addition a mixture of standard components typically found in grapefruit juice was injected to familiarize the panelists with these odors and to help standardize their descriptors. The results from the standard mixture are presented in Results and Discussion section To condition the olfactory senses, individual standard solutions (20 mL at concentration of 4 ppm) were smelled by the panelists prior to OS:rvffi analysis of all the grapefruit samples The standards consisted of hexanal ethyl butyrate myrcene linalool decanal a-terpineol p-menthene 8-thiol and nootkatone

PAGE 50

34 Descriptive Panel Twelve members (6 women and 6 men) were recruited from the United States Department of Agriculture Winter Haven FL for a descripti v e taste panel. The panelists were of varied age groups and ethnic backgrounds All panelists had some prior citrus taste panel experience. Minimum and maximum values for ratio of total soluble solids : % acid from the United States grapefruit juice grading system were used to train panelists Brix:acid ratio of 14 : 1 (70 g of sucrose and 5 g of citric acid) and 8 : 1 ( 40 g of sucrose and 5 g of citric acid) were prepared using food grade sucrose and citric acid in water Naringin solutions of 200 100 and 25 ppm (in water) were used for a bitterness standard All solutions were prepared using double distilled water Fresh squeezed grapefruit juice and fresh grapefruit peel were used as standards for grapefruit aroma quality and intensity Statistical Analysis Principal components analysis in SAS (Version 6 11 SAS Institute Cary NC ) was used to evaluate the data set from the preference sensory panel and GC-FID data Univariate statistics and step wise multiple regression (forward) with Wilks Lambda was also employed to identify those components which would be most differentiating between sensory classifications Canonical discriminant analysis (STATISTICAversion 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

PAGE 51

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

PAGE 52

CHAPTER4 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 resol v ed 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 gi v en in Table 2 All components identified in Table 2 were also identified by Nuiiez et al (1985) and Maarse and Visscher (1989). 36

PAGE 53

l 1-I'j 1 0 V'I (; [ 8 0 OI s 0 H-) 5 7" 3 I (") 0 g (l) [ t.-. E. (') CD s CD CD =::s CD (') g: 0 ::l. CD f"""+ (D 0 t:$ ( t ,__. U1 0 t:$ ,__. s
PAGE 54

Ta bl e 2 Maximum, minimum and average area percent for components extracted with methylene chloride Component Name / Retention Inde x a -th u j e ne a p1 ne n e m yrce n e o ctanal a ph e llandr e n e Rl-1 008 PE -o c imen e y -t e rpin e n e c 1 s J i nal o ol oxi d e t r ans lin aloo l oxidc linal oo l RI 11 0 0 all o oc im e n e RI-1 153 P-pin ene oxi d e non ano l te rpin-4o l RI-1 192 a -t e rpin eo l decanal ( n ) tra n s -carv eo l c arv o n e Rl-1 270 RI-1 2 8 2 Rl-1 299 unde c anal Rl-1 323 a t e rpin y l ac e ta t e A ve ra g e Area % 0 5 1 0. 4 0 2. 4 0 0 09 0.36 0. 1 9 0. 4 5 0.08 0. 1 8 2.00 1 1 1 0. 1 3 0. 4 0 0 06 0. 1 9 0. 1 0 0 4 8 0 1 8 0 6 1 0.6 1 0.3 1 0 15 0 1 7 0 29 0.0 8 0 1 3 0. 24 0. 1 6 Minimum Ar e a o/ o 0 30 0.00 1 .6 0 0. 01 0.00 0.00 0. 1 7 0.00 0 0 4 0. 28 0 41 0 00 0.00 0.00 0.0 0 0.0( ) 0 0 4 0.00 0 02 0.23 0.00 0 .0 0 0.00 0 .0 0 0.00 0 00 0.00 0 00 Ma.ximum Area % 0 7 8 l 2 1 3.82 0.6 5 1 1 3 0. 8 8 0 85 0. 1 8 2 .3 5 4 4 9 2 58 0 .3 5 1 7 6 0 3 4 0 6 4 l). 4 0 1 .0 7 0. 9 3 3. 14 1 83 1 57 0 35 0.76 1 1 3 0.36 0 57 0 65 0.39 38

PAGE 55

Tab l e 2 -continued Component Nam e/ Retention Index Rl 1 367 aco p ae n e RI1 4 26 caryo ph y ll e n e a humul e n e ge nna cre n e ~ bi sa bol e n e se lin ene RI 1 535 Rr 1 553 Rr 1 56 4 RI-1 613 RI 1 6 4 8 se lin 11e n -4 a o l m et11y l jas m o n ate Rl1677 ca din ol RI 1 699 8 9 -ctid e h y d ro nn oo tk ato n e ari s t o lene RI 1 796 RI18 1 0 n oo tkat o n e Average Area % Minimum Area% Maximum Are ao/ o 0. 1 0 0.35 0. 1 9 7.60 0 66 0 16 0 22 0 44 0.35 0 05 0.09 0.05 0. 14 0.08 0.08 0.12 0 .3 1 0 .3 0 0 1 7 0 1 2 0. 1 3 6.70 0 00 0. 1 5 0.00 0.88 0.09 0.00 0.00 0.00 0.00 0 .00 0.00 0.()0 0.00 0.00 0.00 0.00 0.00 0.00 0 .0 0 0 00 0.00 0 00 1 .68 0 .2 8 0. 5. i 0.77 1 5 .11 1 .29 0 ..J 9 1 59 1 .09 1 .0 4 0.20 0 26 0. 1 5 1 4 8 0 .32 0.29 0 29 1 11 1 .3 8 0.83 0.7 4 0.63 0 ..i...i. 17 82 3 9

PAGE 56

Sensory Analysis For comparison purposes all juices were ranked on the basis of average hedonic preference score and divided into three approximately equal categories There were ten juices in the '' low' category Average hedonic scores were 4 75 or below There were nine juices in the '' medium '' category They had preference scores between 4 75-5 75 The 10 juices in the highly pref erred category were rated above 5 7 5 40 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 simi lar fashion. Some of the scatter is reduced as the highest and lowest scores are typically discarded before the remaining scores are averaged This sensory approach was chosen as it more closely reflects marketplace consumer attitudes. Statistical Analysis Univariate analysis Table 3 shows the univariate correlations between preference scores of the panelists and individual peak areas Correlation coefficients for individual components were low ranging from 0.42 to -0.62 Myrcene, decanal linalooi linalool oxides and several unidentified peaks were found to correlate negatively with sensory preference

PAGE 57

1 I 41 Table 3. Univaria te correlations of se lected volatile and non-volatile data with preference category. Variable allo -O cimene P-Caryophyllene a-Humulene Rl-954* Brix/Acid Limonin Nootkatone tran s -Linalool oxide y -T erpinene Naring i n Linalool Acid Decanal RI-1796* RI-935* Myrcene RI-1047* Brix *RI-Kovat's retention indices Correlation (r) 0 42 0 2 7 0. 22 0. 21 0 18 0 0 2 0 14 0 3 9 0 42 0 47 0 4 9 0 51 0. 53 -0 57 -0 61 0.6 1 0. 62 -0 67

PAGE 58

42 Correlatjon coefficient for trans-Iinalool oxide was -0 39 Pino et al (1986 a) also reported that the linalool oxides correlated negatively towards grapefruit flavor preference. P-caryophyllene o:-humulene and several urudentified peaks correlated posjtj v ely 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 Pine 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 (I 988) However earlier studies (Rouseff et al. 1980 ; Barros et al ., 1983) found significant negative relationship between limonin and preference

PAGE 59

43 Multivariate analysis Principal component analysis (PCA) PCA can be used to deter1nine 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 i s 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 highl y pref erred 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 val ue s 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 wi ll 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

PAGE 60

N I I 4 2 A al A L[J 8 []] 0 ., ~A A fl rn A A& A mJ -2 -1.5 -0.5 0.5 1.5 2.5 3.5 Prin 1 Figure 6a. Eigenvector values of PC 1 vs PC2 from principal component analysis of all 57 volatile and taste components: ( e) high preference category preference category, (A) low preference category.

PAGE 61

('() ...... 3 ---------------2 1 0 -1 -2 t -0.5 B [bI I B 0.5 Prin 1 OlJ 1.5 2.5 3.5 Figure 6b. Eigenvector va l ues of PC2 vs PC3 from principal component analysis of al l 57 volatile and taste components: (e ) high preference category, ( medium preference category, l ow preference category.

PAGE 62

46 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 P-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 uni variate 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 e:ff ective in accounting for total variance between all samples was reasonably effecti ve, (r= 0 .6 1) 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 vario us 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 vario us flavor preference using peak areas from these three compounds was achieved from the eigenvector val ue plots from all 57 components shown in Figures 6 a and b

PAGE 63

3e6 ,--------------------, 2.6e6 C) 2.2e6 C) 1.8e6 ...0 1.4e6 0 le6 u 6e5 []] ut] lit A A 2e5 A A DI] -2e5-------------------~ -10000 10000 30000 50000 70000 90000 1.le5 Linalool F i gure 7a. Peak areas of linaloo l and caryophe l lene from 29 grapefruit juice extracts ana l yzed in triplicate: ( e) high preference category ( medium preference category (A) low preference category.

PAGE 64

3e6r----------------------(1) 2.2e A (1) ; I I I E l.4e6 0 u 6e5 -2e5 0 2e5 .t i] M Al 4e5 .. 6e5 Myrcene ODIi 8e5 le6 F igur e 7b. Peak areas ofmyrcene and caryophellene from 29 grapefruit juice extracts analyzed in triplicate: ( e) high preference category, ( medium preference category ( ) low preference category. l.2e6

PAGE 65

49 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 0 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 causative 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 minim1un number of components required to achieve 100% separation between juices of different flavor preference

PAGE 66

Table 4. Forward stepwise discriminant analysis (methylene chloride extractions) Variable Na m e Brix Rl-1 677 aT erpineol P Gurjunene Rat io Limonin cis-Linaloo l Oxide Naringin Nonanal Acid allo Ocimene cx-Copaene Partial R **2 0.46 0 29 0 23 0 17 0 15 0 13 0 14 0 24 0 23 0 14 0 14 0 15 Wilk's lambda 0.54 0 38 0 29 0 24 0 21 0. 1 8 0 1 5 0 1 2 0 09 0 08 0 0 7 0 0 6 50

PAGE 67

51 Table 5 Discriminant analysis classification results (methylene chloride extracts) No. of Percent Correct Group Compo und Comp Total Low Medium High B Linalool +Myrcene 2 68 50 50 95 C B + 0 Brix 3 78 83 53 9 8 D C+~ Caryophyllene 4 78 90 58 89 E D +Noo tkatone 5 82 90 75 82 F Linalool +Myrcene 2 68 50 50 95 G F + 0 Brix .... 78 83 53 98 .) rl G+ RI-1 677 4 86 77 83 95 I H + allo-Ocimene 5 90 97 80 95 Stepwise 16 co mponent s 1 6 1 00 100 100 100 (backward) Stepwise 1 9 components 19 100 100 100 100 (forward)

PAGE 68

N u 4 2 0 -2 -4 -6...__ ________________ -5 -4 -3 -2 -1 0 1 2 3 4 C ,............... 1 Figure 8a Canonical discriminant analysis using myrcene linalool Brix, and peaks at RI-1677 and 1126 : ( e ) high preference category, preference category, (A) low preference category.

PAGE 69

N u 3 0 llli 4 A A -3 -6 -4 IP a Ii t:,. AA -2 0 CANl 2 Figure 8b. Canonical discriminant analysis using 13 variables (ratio, RI-935, cislinal oo l oxide, nonanal a ll o-ocimene, a -t e rpin eol, decanal, RI-1299 a copaene, b-gurjunene, RI-1762 and RI-1 796): ( e) high preference category, ( o) medium preference category, ( A) low preference category. 4

PAGE 70

I dentification of the Peak at RI-1126 The peak with a Kovat 's inde x value of 1126 was the sing le highest positively correlated component among the entire 57 components evaluated GC-MS was 54 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 la st 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 a s the mass spectral scans span nin g 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 -elutin g 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 .6 5 % ; 77, 15 36 %; 19 11 9 1 % ; 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 9 54 respectively with allo -ocimene (2,6-dimet hyl 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 Ko v at 's

PAGE 71

55 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

PAGE 72

56 Grapefruit Juice Aroma Extraction Methods Isolating and analyzing the vo latile components of a food product is essential due to their significant cont ributi on to overall flavor Comparison of vo latile component iso lation procedures have been reviewed by several researchers (W eurman, 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 SP!vfE 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 Mosbonas and Shaw (1992) and in grapefruit juice by Cadwallader and Xu (I 994) 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 samp le 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 hexanal,

PAGE 73

11 Top Note rferpene j t / 5 10 15 0 0 A ") Carbonyl I 20 Ret Time (min) /l_ ___ Sesquiterpene Q) Q) r./'J. C C 4 2 u fl r I \A ,.,l,. u ri,JV' ~, 25 30 Figure 9 Chromatogram classificat i on of pasteurized grapefruit juice (pentane ether extracts). Q) 0 c ? 35

PAGE 74

58 2 terpene area-includes components like limonene myrcene sabinene 3 carbonyl region-consists of octanal ) nonanal terpene alcohols and oxides 4 sesquiterpene area-includes components like caryophyllene and nootkatone Extraction Methods There is no single extraction method which can extract all the aroma components in the exact proportion they exist in the sample Each procedure will concentrate some components and to varying degrees discriminate against ot her s 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)ofthe extraction methods for major juice components Liquid-liquid extractions Pentane / diethyl ether (I : l) 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 sol v ent 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 yie ld ed 73 %

PAGE 75

Q) Q) A A Q) Q) 0 A ' t Q) E A 0 0 0 0 z u A 5 Time (min) 35 Figure 10. Aroma extraction methods in grapefruit juice: A)liquid liquid extraction (pentane ether 1: 1), B) static headspace extraction Vl (solid phase microextraction-SPME), C) dynamic headspace purge and trap solvent elution (Tenax/charcoal trap).

PAGE 76

Table 6. Percent relative standard deviation for different aroma extraction methods in grapefruit juice Component He xanal a-P inene Myrcene a-phellandrene cis-linalool oxide tra,zs-linalool oxide a//o-ocimene a-terpineol Terpin-4-ol C~ ophyllene a-Humulene Nootkatone RSD Pentane-Ether 1 0 13 3 9 8 4 1 0 1 6 12 1 0 7 8 Dynamic HS 7 8 3 18 5 16 ND ND ND 3 ND ND 60

PAGE 77

61 Table 7 Topnote peak areas for different aroma extraction methods Peak Areas 1 iquid -liquid Kovat's Indices MeCI P&E Dy-HS RI-801 2,785 13 872 RI-805 I 0,925 RI-814 9 13 49 342 421 594 RI-821 6 14 9 3 ,9 29 RI-834 2 318 16 3 I 8 34 206 RI-840 7 1 99 RI-844 15 6 1 0 8 4 95 RI-854 9 4 1 0 8 RI-872 10 612 13 ,3 47 RI-8 77 44 674 RI-8 91 3,256 Rl-8 97 6 145 8 I 03 12 443 RI909 4 ,986 5,364 5 434 Rl9 1 5 8, 1 67 20,843 RI924 10 128 1 0,472 RI-936 76 251 10 660 43 ,638 RI9 41 39 368 1 06, 486 RI-944 1 0,737 RI-965 1 8 09 31 574 8,307 RI-971 6,800 12 518 RI-982 7 087 24 217 22 148 Total top note peak area 173 175 299 882 780 279

PAGE 78

62 more top note peak area than methylene chloride Total top note peak area obtained from dynamic head space analysis was 3 50 o/o 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 bexanal 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 stan dard 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 vo latiles 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

PAGE 79

63 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 (C l-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 simi lar 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 analyses

PAGE 80

64 Static head space extraction using SP:ME 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 SP:ME 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 SPl\ffi 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 b y the fiber coating Steffen and Pawliszyn (1996) reported good reproducibility for the component s 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 SP11E emphasizes terpenes which are in high concentrations but contribute little to aroma this technique was not used for further analysis in this research

PAGE 81

65 GC-Olfactornetry Studies GC-olfactometry (GC-0) 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 eflluents 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 101 9 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-0 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 Instrum ental Detectors vs. Human Response GC-0 detects on l y 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-0 with

PAGE 82

s:== ro ro Q) Q) 0 r/J. .-4 c, R~ r/J. > OSME LJVL IL...IL.. Q) Q) Q) A A r:r:i ro r:r:i Q) Q) r,~ A '._:_ 0 >-:: 0 Q) s 0 =t FID ,.., .-4 0 0 r:r:i Q) ro A .-4 u Jul UlLJ A I\I v.. .A A ....._A A __ J._ UvJ" I.J A ... ......., . SCD~ _...... ~----,_...__ _____ '-"---.-Ji __ ...___,, \____,..__,.,___. ...__,1 -~ '\-..I~ ________ _,.._,.....___.Jl,,__,J .._, i ~. 00 5 10 15 20 25 Ret Time (min) Figure 11. C omparison of aroma g ram from OSME and chromatogram s from FID and S C D p-menenthene-8-thiol 30 I~
PAGE 83

67 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 mediciny 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 ( se e 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 va nillin The total ion chromatogram and the mass spectra of vanillin sample and the standard are shown in Appendix A and B Prior to this v anillin was identified in orange juice by Marin et al (1992). Peleg et al (I 992) proposed the path ways for formation ofvanillin from ferulic acid in orange juice (Figure 12). According to the authors (Peleg et al ., 1992) vani l 1in can form from ferulic acid through decarboxylation and oxidation or directly from free ferulic acid through retro aldol reactions Similar reaction pathways may a l so occur in grapefruit Intense aroma acti vi ty

PAGE 84

OH Ferulic Acid CO2 -CIDCOOH H20 OH OH COOR OC8:3 Oxidation p-Vinylguaiacol Vanillin H2C = CH Figure 12. Formation of vanillin from ferulic acid.

PAGE 85

69 ofvanillin 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 3 7 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 3 7 peaks were detected in pasteurized juices As a result of thermal treatment 11 aroma compounds were lost while 7 new components were formed in early season juice However many compounds were unchanged Table 8 shows the aroma attribute compounds formed or lost during thermal processing of early season Juices Mid season juices had 43 aroma active peaks in unpasteurized juice and 49 in processed juice Similarly 43 aroma active peaks were detected in both unpasteurized and pasteurized late season grapefruit juices Eight components were lost in thermally treated late season juices while 8 new attributes were detected The aroma active peaks lost due

PAGE 86

r./1 ,-,c 0.) ro 0.) d cd s 0 800 1000 ro ,-,c ,-,c -~ > 1200 1400 A 41 B 43 C 43 r./1 1600 0.) d Retention Index (attributes listed in Appendix C) Figure 13a. Number of aroma active components at different maturitie s in unpasteuri ze d grapefruit juice: A) early seaso n B) mid season and C) lat e season. -..) 0

PAGE 87

0 A ..ft ..0 r:/) ...-4 0 M M 1---4 0 0 M > 0 A r:/) 37 """ L...., ... "" .,_ .... ,B .... 49 .. .. ,. ..... ....... C Ill~ '. 43 Ill IL __ IUI I I I I 800 1000 1200 1400 1600 Retention Index ( attributes li sted in Appendix C) F i g ur e 13b. Num b er of aroma active co mpon ents at different maturities in pa ste uri zed grapefrui t juice: A) ear l y season, B ) mid season and C) lat e season.

PAGE 88

72 Table 8 Forma t ion and loss of aroma attributes due to pasteurization in early season red grapefruit j u ices Components / R etention Indices Rl-8 96 RI936 a-P ine ne a-phellandrene RI-1 0 44 RI 1 09 5 RI 111 6 RI1166 RI -12 1 7 RI -1223 RI1227 RI13 1 8 RI1374 RI 138 1 RI15 1 0 RI1662 RI-1 6 84 Rl-1 723 Intensities Pasteurized Unpasteurized Description 5 7 7 3 9 3 7 8 1 0 8 7 2 4 0 7 0 8 2 1 2 3 7 6 I 0 3 12 7 1 0 4 6 8 9 2 7 2 8 6 Citru sy M e di c in y Fl o ral Sm okey Gree nis h Citru s Rotten Frui t Terp eney C u c u m ber Medj ci ne y Mu s t\ T e rp eney Mu sns tinb fi, Jit Sm o k ev R a n c id Med icin e y M in ty S weet Spi ce~ perfu m ey Pepp ery Pun g ent Rotten Grapefruit

PAGE 89

73 to pasteurization had generally favorable sensory attributes like green, fruity while the components farmed 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 cx-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 (Je lla et al ., 1998) These negatively correlated compounds were present in higher concentrations in early season than in late season juices. Results of this are summarized in Table 9 The levels of these components in grapefruit juice are in concurrence with those reported by Maarse and Visscher (1989). As a result of pasteurization increased concentrations of alcohols aldehydes and hydrocarbons were observed (Figure 14b) Higher levels of alcohols are probably due to acid catalyzed reactions of terpenes like limonene P-pinene myrcene and so on. These components react in dilute aqueous acid and high temperatures to give several reaction products some of which are alcohols like cx-terpineol terpin-4-ol and linalool oxides (Clark and Chamblee, 1992 and Shaw 1991 ). Limonene is the major terpene in citrus juices and readily farms several reaction products under the conditions present in citrus Juices cx-terpineol in pure form and at low levels has a Wac aroma (Arctander 1994) However at higher concentrations it tends to ha v e musty odor (Marcotte et al. 1998) The level of this component in early mid and late season pasteurized juices are O 81 1 9 7

PAGE 90

74 12 r A a 10 ",, -,I = 0 8 ,-.4 ....,._ = I 6 = CJ = 4 0 u 2 Alcohols Aldehydes Hy rocar ons ( exc ludin g Limonene) Ketones 1 2 10 r' a B 5 8 = 0 Jllllli4 I 6 = I = 4 CJ = 0 u 2 Alcohols Aldehydes Hydrocarbons Ketones ( excluding Lin1onene) Figure 14. Concentrations of components in grapefruit juice. A) unpasteurized juices; B) pasteurizedjuices: C o ) ea rl y season, ( ) mid season and ( D ) late season

PAGE 91

Table 9 Conce ntrati on levels ( ppm ) of co mp o n e nt s i_n early, mid and l ate seaso n red g r apefruit juice s Ear l y Sea so n Mid Season Late Season Component U11pasteurized ( >asteurized Unpastet1rized P as teuri ze d U 11 pasteurized Pasteurized a-terpineol 0 343 0 809 0 228 1 967 0 242 0.6 14 Terpin-4-ol 0 174 0 1 7 1 0 1 26 0 1 9 4 0 1 68 0 . 215 tran sli_nal oo l oxide 0 42 6 0 450 0.420 1 46 1 0 295 0.311 Myrcene 1 790 1 3 7 1 1 788 2 9 42 1 1 66 1 .0 49 0 263 0 260 0 1 89 0 284 0 260 0 218 y-terp1nene

PAGE 92

76 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 terpineol (Clark and Chamblee 1992) (Figure 15) Mid season unpasteurized juice had higher concentration of limonene (3 7 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 jt can be concluded that the panelists were describing the sa me peak using a different descriptor. Table 11 compares the attributes described by the panelists with that of the standard descriptors given by

PAGE 93

+ + + limonene a terpineol Figure 15. Acid cata l y z ed hydration of limonene

PAGE 94

Table 10. Aroma descriptors used by panelists from GC-0 experiments of citrus standards Component Hexenal Ethyl Butyrate t-2-hexenal cx-Pir1ene Myrcene Linalool Terpi11-4-ol cx-terpineol p-menthene 8-thiol Nootkatone Benzaldehyde (IS) Methyl Jasmor1ate (IS) Panelist I Green Fruity Green, dead bug Medicir1e l'i11 cy Unripe Mango Flora l New cotto11 cl <. 1thes Cilantro Musty Rotten gft Stinky terpeney GFt stink Rotten Gft Cherry Almond Floral Jasrnine Panelist 2 Green Fruity Floral Skunky 1>i11ey Gree11 Melon Linalool Stale Ct1L1rch Moldy Musty Stinky Rotten GFT Moldy GFT Cherry Almond Floral Jasmine Panelist 3 Strong green Fruity Citrus Smoked burnt roasted Gree11isl1 Vitamin C Citrus Stinky floral Vinyl Green cilantro Sweet grapefruit Stinky grapefruit Cherry Almond Floral Jasmine

PAGE 95

Table 11 Comparison of standard (Arctander lexi con) with panelist descriptors Component Hexenal, Ethyl Butyrate t-2 hexenal a-Pinene Benzaldehyde Sabinene Myrcene a-phellandrene para-cymene trans-P-ocimene y-terpinene cis -linaJ ool oxide 1,a,1 \'-linalool ox id e Nor1anal Linalool a-terpineol carvone p-menthene-8-thiol vani llin Nootkatone Standard Descriptor Green/Warm sweet fruity Green vegetable lik e Warm resinou s and t1erbacious Bitter almond sweet cherry Warm peppery herbacious Balsamic resinous and citrusy Citrusy peppery woody Citrusy kerosene like Warn1 herbacious, sweet Herbacious citrusy Sweet floral eartl1y Sweet floral eart l1 y Fatty tloral Floral woody Lilac piney Warm herbace o us Grapefruity Creamy vani ll a like Fruity citrusy Panelist s Descriptor Gree11/F rui ty Green dead bug skunky Medicine Cherry almond Unripe mango piney Unripe mango citrus Citrus Mir1ty citrusy Citrusy musty Roasted cotto n candy Terpeney cotton candy Terpeney cotton candy Terpeney cucumber cotton candy Floral Musty wet dog cilantro Floral liquorice mediceney rotten nutty grapefruit fn1it Vanilla Grapefruit

PAGE 96

80 Arctander (1994). Nootkatone was described by panelists as rotten fruity sweet grapefruity stinky citrusy Arctander s descriptor for nootkatone is citrusy Comparison of retention times indices and the odor description given by the panelists for the standard gives a good indication that same aroma active peak is being described Grapefruit Aroma Nootkatone is considered by some scientists to be one of the important contributors to the grapefruit flavor (Stevens et al. 1970 ; Pino et al. 1986a ; Shaw and Wilson 1981 ) Maturity plays a significant role in deterrnioing 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 bas 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-0 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

PAGE 97

81 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 is p-menthene-8-thiol. Discussion of this component is included in a later section Dilution Analysis Studies involving dilution analysis (AEDA, Charm) on grapefruit juice have not been reported to date However orange juice has been extensively studied (Marin et al ., 1992 ; Hinterholzer and Schieberle 1998) with both AEDA and Charm Among the components reported by the authors hexenal ethyl butyrate and vanillin were found to have highest dilution values while linalool de canal were found at the lower end of the dilution factors The peaks detected in our study were aroma active peaks from the juice extract concentrated 160 times This does not provide information about which of these peaks have intense aroma activity at higher dilutions (lower concentrations) Since components in juice are not present in concentrated form, dilution analysis was done to identify the most aroma active now referred to as aroma impact peaks. To assess the most intense peaks, juice extract was concentrated 16 times instead of 160 times and analyzed using GC-0 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

PAGE 98

82 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 ma y 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 ha v e 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 Detection 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 (19 9 4) compared a flame photometric detector (FPD ), an

PAGE 99

Table 12. List of components present in 16x concentrated juice extract and their intensities and aroma attributes Components Attribute Aroma Intensities Hexenal Ethyl Butyrate Green/Fruity t-2-hexenal Green RI-863 Mushroom RI-936 Sweet fruity Sabinene Unripe Myrcene Unripe citrusy a-phellandrene Green Citrusy para-cymene Citrusy trans-P-ocimene Green Citrusy y-terpmene Floral Citrusy Rl-1116 Grainy Medicin y allo-Ocimene Sweet RI-1141 Nutty Rl-1177 Green Must y a-terpineol Musty p-menthene-8-thiol Stinky Grapefruit RI-1349 Vinyl RI-1374 Apple Sauce RI-1381 Apple Sauce Vanillin Vanilla RI-1464 Burnt RI-1510 Perfumy RI-1723 Incense RI-1754 Grapefruity l J nknown Sulfur cmpd (RT 32min) Grapefruity Nootkatone Grapefruity 7 .6 1 1 59 3 12 3 22 3 06 4 38 7.20 5 26 2 83 6 89 3 02 4 .89 4 15 3 33 4 88 4 76 6 39 3.'76 6 03 4 91 4.24 2.18 4 59 3.08 2 29 6 76 83

PAGE 100

84 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 of analyte (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 of pentane-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 conclusions Twenty-two sulfur compounds were isolated in early mid and late season pasteurized and unpasteurized grapefruit juice This represents the most comprehensi v e 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 ,.

PAGE 101

Sulfur Compound ( analyte) S0 2 + 0 2 + hv ( < 400 nm) Figure 16 Sulfur chemiluminiscence reactions 00 VI

PAGE 102

A (1.) > QJ 0 = > Q ..D 0 """"4 B QJ f1 I C E u I > (1.) 00 I r.n '_.-' rjJ I j C 0 Time (min) 56 Figure 17a. Total number of sulfur peaks at different maturities in pa ste ur i zed grapefruit juice: A) early season, B) mid seaso n C) late season.

PAGE 103

0 ...... -;3 I I 00 I (1) A A (1) -;3 Q) (1) rfJ s I 0 rfJ Q) Q u B r./J o 5 10 15 20 25 30 35 Time (min) Figure 17b. Effect of pasteurization on sulfur compounds in early season grapefruit juice: A) unpasteurized and B) pasteurized 00 -...)

PAGE 104

88 juices had I 03 times more total peak area than the corresponding unpasteurized juices We suspect that increase in these components might be due to thermal reactions occurring during pasteurization Sulfur containing amino acids such as cysteine glutathione (Miller and Rockland 1952) and methionine (Burroughs 1970) has been reported in fresh grapefruit juice Thermal decomposition of these non-volatile sulfur containing amino acids might also be responsible for the increase in sulfur compounds due to thermal treatment Two of the SCD peaks had the same retention times (20 and 32 min) as aroma active compounds detected in OSME Both peaks were described as having '' grapefruity '' aroma by the panelists One of the peaks eluting at 32 min has the same retention index as didehydro nootkatone (RI-1754) which also has grapefruity aroma Efforts to characterize this sulfur compound have not yet been successful Therefore the exact identity of this OSME peak could not be determi1Jed at this time p-menthene 8-thiol The SCD peak eluting at 20 min has been identified as p-menthene-8-thiol Thi s was first reported by Demole et al (1982) in grapefruit juice and had a '' fresh grapefruity' aroma They found this thiol to have the lowest flavor threshold (107 ppb) in water of any substance yet reported. The level of this component in grapefruit juice was reported at 0 002 ppb by the authors The concentration of this component was found at 4 6 ppb in early and late season pasteurized juice while the level in pasteurized mid season juice was at 11 2 ppb This terpene thiol was reported to be a reaction product ofH 2 S and limonene

PAGE 105

89 (Demole et al 1982) and the reaction is accelerated by higher temperatures Since ]imonene is present at higher concentration in mid season juice higher concentration of this component could be expected in this juice type A cyclization product of p-menthene-8-thiol 2 8-epithio cis-p-menthane was described by Demole et al (1982) to have grapefruity aroma with a much higher odor threshold (9 ppb ) The presence of this compound in our sample s could not be established at this time although our samples had a peak which closely matched with that of author s samples The panelists did not detect any grapefruity aroma at the retention time of this compound (19 min) therefore this peak was not considered in this study Correlation Between Aroma Components and Sensory Measurements The goal for this part of the study was to use the aroma active components (identified by GC-0) and taste components and assess their contribution to the o v erall juice quality For isolating the volatile aroma components a more representative aroma extraction technique (pentane-diethyl ether) was employed Juice Classification Juices were classified using natural clustering of average overall flavor scores of the 40 juices from the descriptive panel. There were three juices in the '' worst '' category Average flavor scores were 6.3 or below There were seven juices in the '' fair'' category They had flavor scores between 7 3 17 83 Nine juices were classified as '' good '' and had

PAGE 106

90 an average flavor score of 8-8 83 The 11 juices in the best category were rated above 9 12. In addition to rating the overall flavor for each juice panelists evaluated grapefruit aroma & intensity sweet/tart balance and bitterness level. Analysis of variance (STATISTICA 5 0) was conducted to judge the significance between the juice groups classified by the sensory scores Duncan s multiple range test at 95% level was used All the groups were significantly different from each other at p < 0 05 The probabilities obtained in Duncan s multiple range post hoc tests with group as main effect are as follows : 0 0019 0 000061 0 000054 between worst and fair / good / best juices ; 0 02 0 000061 between fair and good/best juices ; 0 00093 between good and best JUlCes Sensory Analysis Sensory attributes of a food product are a combination of odor taste visual impression and mouth feel Volatile components contribute to the odor while non volatiles contribute to taste (sugars acid etc ) and color (carotenoids xanthophylls etc ). Insoluble material contributes to the mouthfeel A descriptive taste panel was trained to quantify aroma attributes such as aroma intensity and aroma quality and taste attributes such as bitterness and sweet / tart balance The panelists were also asked to rate the overall flavor score using these attributes In our earlier study preference scores were used to evaluate juice quality The limitation of that approach was that it was not possible to determine why a particular juice was rated high

PAGE 107

91 or low A descriptive panel overcomes this problem by requiring the panelists to evaluate several flavor attributes in addition to overall flavor quality Table 13 shows the average minimum and maximum sensory scores produced by the panelists for grapefruit juice samples As expected grapefruit aroma intensity quality and overall flavor score values were lowest in worst quality juices The worst juices also had higher bitterness scores when compared to other juice types (see Table 13) However the highest bitterness score for the worst juice types was only 8 66 on a 15 point scale This can be attributed to the fact that this juice data set did not contain any early season juices Since bitterness decreases rapidly with increasing fruit maturity it is not surprising that the juice set contained few excessively bitter juices. Sweet / tart balance scale ranged from -7 to + 7 (see Figure 4) When the panelists perceived more tartness than sweetness they rated the juices on negative scale The value 0 represents equal balance between sweetness and tartness. All of the best juices were rated on the positive side ranging from 0 34 to 2 50 Thus all the best juices were slightly more sweet than sour The range of sweet / tart values for all juice types can be seen in Table 13 There seems to be a trend in wruch higher ranked juices are considered more sweet than sour Many of the worst juices also had higher bitterness score and lower grapefruit juice aroma intensity and aroma quality which would have contributed to lower overall flavor score for these juices

PAGE 108

92 Table 13 Minimum and maximum descriptive sensory panel scores for grapefruit juices Worst Fair Good Best Attribute Low High Low High Low High Low High GFT Intensity 6.29 8 69 7 37 8 43 6.67 10 08 6.72 9 10 GFT Quality 6 48 8 76 7 44 9 30 7 58 l 0 15 7 88 10 20 Balance 0 12 0 89 0 59 1 74 -0 27 1 74 0 34 ? -o ... ) Bitterness 4.00 8 66 3 1 0 6 62 3 29 7 59 3.28 6 53 Overall Flav score 6 49 6 85 7 29 7 83 8 00 8 93 9.12 I 0 43

PAGE 109

93 Univariate Analysis Taste components Univariate correlations of sensory data and taste components are shown in Table 14 Marked correlations are significant at p < 0 05 The sensory attributes evaluated here are : grapefruit aroma intensity aroma quality sweet / tart balance bitterness and overall flavor score. The taste components which were evaluated include 0 Brix (sugars) acid (sourness) ratio of 0 Brix/acid (sweet/tart) limonin and naringin (bitterness) The bitterness perceived by human subjects is a combination of two components limonin and naringin Since the taste threshold of naringin is 20 times greater than that of limonin (Guadagni et al ., 1973) it can be assumed as a first approximation that equibitter solutions of naringin and limonin would maintain this proportion When naringin and limonin were correlated individually with panel bitterness scores the correlations were significant (r = 0 85 0 83 respectively) Linear regression for limonin can be seen in Figure 18 and the corresponding plot for naringin is very similar Rouseff et al (1980) and Barros et al (1983) observed similar behavior in earlier grapefruit juice studies However there was no significant correlations between limonin or naringin with bitterness in our earlier study employing a preference panel (Jella et al ., 1998) To account for the total bitterness evaluated by the descriptive panel limonin and I / 20th of naringin were combined and the correlation with bitterness score improved slightly (r = 0.86 at p < 0 05)

PAGE 110

Table 14. Univariate correlations between sensory and taste components (Brix, acid ratio, limonin and naringin) Aroma 1nte1is1ty Aroma quality Sweet/tart Balance Bitterness Overall flav score BRIX Acid Ratio Li1nonin Naringin Li111onin+ l /2 2 I Aroma quality 0.25 0 05 0.54* () 11 -0 44* 0 50* -0 11 -0 .0 1 -() ()5 Co rrelation s are sig1lificant at p < 0 05 Sweet/tart Bitter11e ss Overall Balance flav score -0 .68* 0.66* -0 J( ) 0 08 -0 19 -0 1-t -0.22 0 1 3 -0 .20 0 .3 0 () 32 () 05 -0 60* () 85* -() 27 -0 .5 8* 0.8 3 -0 28 -0. 61 () 8(,* -() 2 8 BRIX Acid 0 42* () 4 7* -() 6 1 -0 09 0 20 0 08 0 20 () (l I < l 2 I Ratio -0. 3 1 -0 1 6 -() 21 Limonin Naringin Limonin+ 0 88* () 96* 0.98* 1120th of Naringin

PAGE 111

... 8.5 5.5 2.5 0 6 12 Limonin ~pm Figure 18. Co rr elation of l imonin concentration with bitterness score 18 24

PAGE 112

96 When these bitterness factors (naringin limonin and limonin + I / 20 th naringin) were correlated with overall flavor score the correlation was not significant at p < 0 05 (r = 0.27 0 28 0 28) Similar findings were reported by Pino and Cabrera (1988) and Jella et al. (1998) Low correlations between bitter components and overall flavor score might be due to the lack of early season, highly bitter juices in the sample set As was noted earlier the highest bitterness score recorded by the panel was only 8.66 out of 15 Thus none of the juices were deemed to be excessively bitter. It should also be noted that some bitterness is expected and desired in grapefruit ju i ce Only when bitterness becomes excessive are the overall flavor of the juices downgraded. Brix ( total sugars) and acid (titratable acid) are factors responsible for sweetness and sourness in grapefruit juice The ratio between these two is important since it is one of the major quality standards used by the grapefruit industry. When panelists evaluated sweet/tart balance it did not significantly correlate with Brix/acid ratio (r = 0 30 at p < 0 05) Analytical measurements of Brix and acid individually also did not correlate significantly with overall flavor score from the sensory panel (r=-0 14 and -0 2 respectively) However as shown in Figure 19, when the sweet/tart balance sensory scores were compared with the overall flavor score rated by the panelists there was a significant positive correlation (r=0 66) The reasons for differences in correlations within sensory ( sweet/tart balance and overall flavor score) and between analytical vs sensory (ratio and sweet/tart balance) is explained below Perceived sweetness in citrus juices is due to sugars like sucrose fructose and glucose which are 68-80% of total solub l e solids. The analytical measurement Brix on

PAGE 113

3--------------------2 1 ........... ... 0 .... o ~ ... 0 ... ... .... 0 0 0 ........ .. . .. 0 _,. .. 'Cl .... ...... ... .. ... 0 ... 0 .. 0 ..... .... .. ... .,. 0 O .. ..... ..... ....... ...... ...... ........ .. .. .. ... 0 0 .............. .. '--'I e ... .. .... ... .. ... -.. ...... .,... Q .. o .. o o .. 9 Q ..... 0 .... ... . .... 0 0 0 0 -1 L-----------------------------l 6 7 8 9 10 11 Flavor Score Figure 19. Correlation between overall flavor score and sweet / tart balance.

PAGE 114

98 the other hand measures not only sugars but other soluble solids present in the citrus juices nor does it differentiate between the relative sweetness of different sugars. Since panelists rate sweetness in juices based on the contribution of all sugars and not the total soluble solids differences in the analytical and sensory measurements can be expected Acidity (Sourness/tartness), as perceived by panelists is due to free hydrogen ions in the juice which is a direct measurement of pH However acid titrations with sodium hydroxide measures the total amount of available acid both free and undissociated The acidity values reported in this study are titratable acidity rather than free acid hence the analytical and sensory values in this study might be expected to have different correlatjons Grapefruit aroma quality and intensity are influenced by many components as discussed in section 3. Aroma intensity and aroma quality are affected by several factors such as maturity processing history and juice extraction method These two sensory attributes provide a first impression to consumers about the flavor quality of the juice When grapefruit aroma intensity and quality were evaluated by the pane~ the correlation with overall flavor score (r = 0 36 and 0 54 respectively) was significant at p < 0 05 These values suggest that aroma quality is more important than aroma strength in determination of overall juice flavor Figure 20 shows the correlation between grapefruit aroma quality and overall flavor score. The correlation of aroma quality (r=O. 54 at p < 0 05) is almost rated on a equal basis with that of sweet/tart balance and overall flavor score (r=0.66) This suggests that grapefruit aroma quality plays an important role in determining the overall quality of grapefruit juice This is an important conclusion from

PAGE 115

0 10 0 0 0 0 0 0 0 l \ (l) 0 0 0 0 0 u VJ. 0 f)" 0 8 0 o0 00 > r-5" g 0 ( 0 .... 0 0 0 6 6 8 10 Aroma Quality Figure 20. Correlation between aroma quality score and overa l l flavor score.

PAGE 116

100 this study as the current industry standards are based solely on taste factors and no aroma components are considered. Overall sensory quality judgements are made based on the interaction of many attributes A balance between sensory components is required to create a positive overall sensory perception The reason for non-significant sensory correlations with individual taste components might be due to a requirement of balance between several components Panelists made additional comments when unusual flavor attributes were encountered Comments for some of the lo west rated juices include : watery bland mouth feel and off flavor The best juices were considered to have an optimum flavor balance Aroma components Table 15 shows the univariate correlations between grapefruit aroma intensity aroma quality and overa ll flavor scores and volatile aroma component FID peak area s. Myrcene cx-terpineol vanillin cx-phellandrene and several unidentified peaks were found to have significant negative correlations (p < 0.05) with aroma intensity scores cx terpineol was reported to cause off-flavors in citrus juices (Clark et al ., 1992) It tends to have a musty aroma at higher concentrations which might be objectionable to panelists Hence it is negatively correlated with aroma intensity Hexenal and ethylbutyrate had s ignificant positive correlations with aroma quality (r=O. 31 and r=O 31 respectively) The other component having significant positi ve correlation with aroma quality is the peak at RI -1754 Hexenal and ethylbutyrate were described as having green and fruity smells in GC-0 studies with a combined intensity of

PAGE 117

101 Table 15 Univariate correlations between 26 aroma active volatiles and sensory scores Correlation (r) Component Aroma intensity Aroma quality Flavor score Hexenal, Ethyl Butyrate 0 31 0 31 0 03 t-2-hexenal 0 13 0.04 -0 19 RI-862 0 09 0 12 -0 08 RI-936 -0 56* -0 42* 0 33 Sabinene -0 17 -0 15 -0 14 Myrcene -0 37* -0 31 -0 13 a-phellandrene -0 28* 0 20 -0 07 para-cymene -0 16 -0 28* 0 08 trans-P-ocimene -0 13 -0 11 -0 03 y-terpmene 0 09 0.00 -0 14 RI-1116 -0 24 -0 21 0 04 allo-Ocimene -0 15 -0 16 -0 20 RI-1140 -0 25 -0 05 -0 14 Rl-1177 -0 17 -0 07 -0 17 a-terpineol -0 41 -0 20 -0 29* RI-1349 -0 l 1 0 18 -0 12 RI-1374 -0.22 0 05 -0.24 I RI-1381 -0.28* -0 07 -0 27* RI-1464 -0 08 -0 09 -0 01 Rl-1510 -0 21 -0 15 -0 23 RI-1722 0 15 0 14 -0.08 RI-1754 0 29* 0.31 0.04 Nootkatone 0 02 -0 06 -0 05 p-menthene -8-thiol 0.08 0.00 0 11 Unknown Sulfur (32 min) 0.09 0. I 6 0 31 Vanilla -0 37* -0 25 -0 13 Correlations are significant at p < 0.05

PAGE 118

102 7 61 on a scale of 15 The peak at R1 17 54 was described to have grapefruity aroma (intensity = 3 08) Since green/fruity and grapefiuity attributes are perceived as positive qualities these components can be expected to have significant positive correlations with the sensory scores Pino et al (1986a) also reported positive correlations of hexanal and ethylbutyrate with sensory scores Myrcene, p-cymene and an unknown compound were found to have negative correlations with aroma quality The description given by the panelists for these components was green unripe citrus smell of moderate intensity. The negative correlations with the sensory score (aroma quality) might be due to this green unripe smell It should be noted that myrcene was one of the three most differentiating components in our earlier study (Jella et al ., 1998) Myrcene was highly negatively correlated with overall flavor These more recent results suggest that it is also strongly negatively correlated with aroma quality When the aroma components were correlated with overall flavor score cx-terpineol and several unknowns exhibited negative correlations (Table 15) The effect of cx terpineol on sensory perception has been discussed in earlier sections None of the most intense aroma components exhibited significant positive correlations with overall flavor score. Nootkatone has been reported to have significant positive correlation with the sensory quality (Pino et al ., 1986a) However nootkatone failed to exhibit a significant correlation in our studies (r= -0 05 at p < 0 05) The correlations between aroma quality intensity and nootkatone were also not significant The lack of correlation between nootkatone and overall flavor score is clearly shown in Figure 21. Earlier studies

PAGE 119

3e5 .----------------------2.4e5 l.8e5 l.2e5 .... ........ ..... 0 0 8 0 0 0 0 0 0 o o ............ ___ a o 0 .. .... e, __ . ......... --r----=~----~---~ o . 0 0 0 ,_, (~ 0 0 e o, A 0~ 60000 ----------------------0 0 ----------------c2 0 .... 0 ....... -... __ i 0 g O O O ~t:, : 0 0 0 OL--------------------___..J 6 7 8 9 10 Aroma Quality Figure 21. Correlation between aroma quality score and nootkatone peak area.

PAGE 120

104 conducted in our lab also suggested a lack of significant correlation between nootkatone and sensory preference (Jella et al ., 1998) Surprisingly another important aroma component in grapefruit juice p-menthene-8-thiol (Demole et al. 1982) also did not significantly correlate with overall juice quality (r=O 08 at p < 0 05) Therefore components which have been suggested to be key factors for grapefruit flavor (nootkatone and p-menthene-8-thiol) had essentially no correlation individually with overall flavor Of the 26 most intense aroma volatiles the component with retention index Rl-936 exhibited significant correlations with aroma intensity (-0 56) aroma quality (-0 42) and overall flavor score (-0 33) This compound was described by the panelists as having sweet fruity aroma When the retention index was compared with the literature value the match suggested that the compound was a-thujene However the Arctander descriptor for this compound was green piney Further conclusions regarding the exact identity of this compound could not be made at this time However in this study this compound seemed to be effective in differentiating the juices based on sensory quality Multivariate Statistical Analysis Discriminant analysis can be used to determine which variables discriminate between two or more naturally occurring groups (Stat Soft 1997) In this part of the study juices were evaluated using values for the major taste components (limooin naringin Brix acid and ratio) aroma active components and a combination of both taste and aroma components

PAGE 121

105 Twenty-six odor active peaks (24 from FID and two from SCD) chosen from GC0 dilution studies and five taste attributes (Brix acid ratio limonin and naringin) were utilized for analyzing and predicting the juice quality using discriminant analysis Of the 3 9 juices analyzed for aroma components taste components and sensory attributes data from 3 0 juices were used for the training set Results of nine juices ( chosen at random)were held back from training set and were used for testing the accuracy of the fmal flavor model Flavor models using taste components None of the individual taste measurement compounds correlated significantly with the overall sensory scores (see Table 14) Therefore multivariate statistical analysis was used to determine if a combination of taste components could be used to model and predict flavor quality Taste components alone could not adequately discriminate the 30 juices in the training set Whereas only 27% of the '' good '' juices 50% of '' fair '' juices and 68% of '' Best '' juices were successfully grouped 100% of the '' worst '' juices were successfully grouped This can be seen by the grouping of worst juices in the lower left hand side of Figure 22 This suggests that taste components best discriminate '' worst '' juice types When cumulative canonical coefficients (Root 1 and 2) of the taste components data were plotted in a 2D plot positive Root 1 canonical coefficient scores were observed for all the worst juices while the other three juice classes had negative coefficients (Figure 22) The first and second discriminant functions (root 1 & 2) are weighted most heavily by

PAGE 122

N ) 0 0 4.--------------------, 2 0 -2 0 Oo e 0 0 oo~~ cit {b o'b 0 CJ) + + 0 + [jJ + + Root 1 Figure 22. Standard discriminant analysis u s ing 5 taste components: ( + ) worst category, ( Q ) fair category, ( lJ.) good category, category juices.

PAGE 123

ratio and acid on the positi v e side (where worst juices are clustered) and by Brix on negative side (where good best and fair juices are clustered) Their standard canonical coefficients were 10 1 9 2 and -8 9 respectively In comparison, the coefficients for naringin and limonin were surprisingly low (-0 7 and 1 2) 107 The squared Mahanalobis Distances (MD) is the distance computed between each case and the center of each group(( centroid ''. The larger the value between groups the better they are discriminated Comparing the group clusters in Figure 22 and also values in Table 16 it can be concluded that the 5 taste components only adequately separated the worst juices from other groups Since MD value is a direct measure of how far apart the juices are from each other lower values between fair good and best juice classes (e g .: distance between fair and best is 2 40) indicate that there is an overlap between them Data from nine of the 3 9 juices were used for testing the model These nine juices were not included in the training set Fifty-five percent of statistical classification agreed with actual sensory classification Flavor models using aroma components Twenty-six v olatile components (24 from FID and 2 from SCD) were selected based on their aroma activity for analysis Based on their Kovats indices it was assumed that the 24 volatile components quantified from their FID peaks were the aroma active peaks identified by GC-olfactometry However some peaks may not be the aroma acti v e peaks but another component that elutes from the GC at the same Ko v at s index (retention time) It is known that some fla v or impact components such as p-menthene-8

PAGE 124

108 Table 16 Squared mahalanobis distances for groups separated by taste components (Brix, acid, ratio, limonin, and naringin) Fair Best Worst Good Fair 0 2 396 10.222 1 592 Best 2 396 0 11 63 3 0 870 Worst 10 222 11 633 0 11 880 Good 1 592 0 870 11 880 0

PAGE 125

thiol cannot be measured by FID or MS hence the need for sulfur detector Similarly there might be other potent components which can be detected only with specific detectors 109 Earlier flavor models by Pino et al (1986a b) and Jella et al (1998) have employed components that correlated highly with juice quality or acceptance The components used in these models may or may not have had aroma activity For example P-caryophyllene was found to have high correlation with sensory preference (Jella et al ., 1998). However none of the sniff panelists perceived P-caryophyllene in the grapefruit juice extracts even though it exhibited a large FID peak Thus even though it was highly correlated P-caryophyllene was not considered in the flavor model because no aroma activity was observed at concentrations found in grapefruit juice Standard discriminant analyses using aroma components Standard discriminant analyses takes into consideration all the variab le s specified in the data set to build the model The data set of 26 aroma active components had successfully grouped 100% of the 30 training juices classified as ''worst '', '' fair '' and '' best ''. '' Good '' juices were 72% successfully grouped When the mahalanobis distances for the good juices were calculated the distances between good and best was least (11 5 units) The distance between good and fair was 21 4 units and between good and worst was 23 7 units Lower the mahalanobis values closer are the groups to each other These values suggest there was some overlap between the '' good '' and '' best '' juices When the canonical coefficients were calculated myrcene and RI-1510 had the largest negative root 3 canonical coefficient values (-2 9 and -2 5 respectively) When

PAGE 126

110 canonical coefficients were plotted these components were effective in separating the worst juices from the other 3 juice classes Univariate correlations also suggest that myrcene was significantly negatively correlated with sensory scores (r= 0 37 at p < 0 05) These findings about myrcene concurred with our earlier studies (Jella et al ., 1998) Forward step wise discriminant analysis using aroma components To identify the most discriminating variab les forward step wise discriminant analysis was performed on the volatile data set (26 components). In this analysis the program reviews all the variables evaluate s which one discriminates the most and includes that in the model The program then proceeds to review the remaining variables evaluates the second most discriminating variable and includes that in the model This procedure continues until all the components which help discriminate the data are included In other words the first component selected by the program is the most discriminating second component is next best and so on. When aroma active volatiles were analyzed with forward step wise discriminant analysis 17 steps (tolerance level = 0.01 and F to enter = 1) were performed by the statistical program (i e. these components were most efficient in separating the data set into 4 classes) The components which were selected (in order of most to least discrimination) are : Rl-1381 Unknown sulfur (32 min) RI 936 1116 1127 858 Hexanal/ethylbutyrate sabinene RI -1464 p-cymene nootkatone a-terpineol RI-1510 vani llin trans-P-ocimene myrcene and RI -1754 The component p-menthene-8-thiol considered by some as flavor impact component in grapefruit juice (Demole et al ., 1982) was not in the list of discriminating components

PAGE 127

111 The above 17 compounds were utilized for grouping the training set data to see if statistical classification concurs with sensory classification Best and fair juices were 100% successfully regrouped while 83 % and 72% of the worst and good juices were regrouped successfully These results suggest that volatiles are best at differentiating the better juices and taste components best discriminate the worst juices. Component RI -1381 and unknown sulfur compound eluting at (32 min) were most discriminating variables in the forward step wise discriminant analysis When the peak areas of these two were plotted against each other tight grouping of the best juices was observed. However these two components by themselves were not effective in separating the other juice types. F l avor models using aroma and taste components Grapefruit has a unique flavor which is a result of a combination of aroma components and a few taste components. In this part of the study 5 taste components and 26 aroma active volatiles were selected for use in developing a flavor model Standard discriminan t analysis Data from the 26 aroma active volatiles and 5 taste components were used in testing the accuracy of sensory classification with statistical classification In the training set, best fair and worst juices were 100% successfully regrouped. Good juices were 94% correctly segregated The mahalanobis distances for these four groups was highly significant (Table 17). The higher the values for mahalanobis distances the farther apart are the juices. The distance between the best and the worst

PAGE 128

112 Table 17 Squared mahalanobis distances for 26 aroma and 5 taste components (Standard Discriminant Analysis) Fair Best Worst Good Fair 0 .000 36 391 60 356 38 847 Best 36 391 0 000 72 0 58 34 783 Worst 60.356 72 058 0 000 40.622 Good 38 847 34 783 40 .6 22 0 000

PAGE 129

113 juices was highest (72 units) This suggests that these two juice types were farthest from each other The mahalanobis values for good juices suggest that some of these were closer to the best and fair juices and hence it resulted in lower percentage (94%) for grouping good juice types Mahalanobis distances between good and best juices was 34 units between good and fair was 3 8 units and between good and worst juices was 41 units. The mahalanobis distances described above can be clearly pictured in Figure 23 This figure shows the cumulative canonical coefficients for different juice types using 26 volatiles plus 5 taste components Root 1 was very efficient in separating best juices from worst juices Best juices were on negative axis of Root 1 while worst juices were on positive scale The ellipses around the juice types are at 95% confidence level One outlier was observed in both best and good juice classes Eighty-one percent of the variance was explained in Roots I and 2 When canonical coefficients were calculated myrcene weighed heavily on the positive side (7 74) in Root I where worst juices are clustered (see Figure 23) trans ocimene (-6 8) and nootkatone (-1 3) were on negative scale of Root 1 where best juice s are grouped Acid (17 43) Ratio (14 06) and Brix (-15 34) were instrumental in separating fair group from the other juice classes in Root 2. Forward step wise discriminant analysis. Forward step wise discriminant analysis of the volatile and taste components was accomplished in 21 steps There were 1 7 aroma

PAGE 130

4 2 N I 0 0 0 Q] 0:: -2 -4 -6 -4 -2 ] 11 11 1111 0 Root 1 oO G 0 0 0 111111 fP 11 0 0 0 0 0 0 11 11 + + + 2 4 Figure 23. Forward stepwise discriminant analysis using 17 aroma and 4 taste components: ( + )worst category, ( Q) fair category ( [j_) good category ( best category juices + + 6

PAGE 131

115 components and 4 taste components The volatile components selected steps and Wilk s lambda values are given in the Table 18 Wilk s lambda is a measure of the discriminatory power of a group of variables The smaller the Wilk s lambda the better the overall discrimination is Perfect discrimination would occur at a Wilk s lambda of 0 Wilk s lambda value reported here represents the overall variance after the corresponding variable was included in the model. Since the discriminatory power of the function increases by adding a new variable, the corresponding overall Wilk s lambda value decreases This is shown in Table 18 When variables selected by step wise discriminant analysis (17 volatiles and 4 taste components) were analyzed and plotted I OOo/o separation of the juice classes was observed in training set The groupings were similar to the ones observed in Figure 23 The model was tested by evaluating 9 juices not used in the training set There were two replications for each juice (i e 18 samples) The flavor scores were known but the juices were treated as unknowns Sixteen of 18 juices were correctly classified within one fla v or category (Table 19) To explain some of the mis classifications one must consider that some of the aroma active peaks detected in GC-0 could be below the detectable range for analytical instruments These components can be perceived by human nose since the sensitivity of it is much higher than analytical detectors The FID peak area used in the model may not correspond to the aroma active component which occurs at the same retention time Thus quantitative values may not reflect actual concentrations of the aroma active component and therefore reduces the effectiveness of the model However it should be

PAGE 132

Table 18 Forward step v1ise discriminant analysis la tiles and taste components (Number of steps and corresponding component) Component Acid RI-1374 Unknown Sulfur (32 min) RI-1127 RI-1510 a-terpineol Limonin RI-1116 RI-1464 p-cymene Brix Rati o RI-1754 RI-1141 RI-8 63 RI-1723 Nootkatone RI-1381 RI-985 Myrcene RI-1047 Step 1 2 3 4 5 6 7 8 9 1 0 I I 1 2 1 3 14 1 5 16 1 7 18 1 9 20 21 Wilk s lambda 0 .6 42 0 45 6 0 340 0 250 0.2 10 0. 178 0 145 0.121 0 1 04 0.0 8 2 0 065 0 050 0 0 4 0 0. 035 0 026 0 022 0 017 0 015 0 014 0 013 0 011 116

PAGE 133

117 Table 19. Comparison of sensory and statistical classification of grapefu.tit juices ( Model has been tested using 17 aroma components and 5 taste con1ponents) Juice Classification Batch Replication Sensory Statistical A 1 Best Good A 2 Be s t Good B I Good Good B 2 Good Good C 1 Good Best C 2 Good Fair D l Best Best D 2 Be s t Best E l Fair Fair E 2 Fair Fair F 1 Best Best F 2 Best Best G l Be s t Be st G 2 Be s t Best H l Good Good H 2 Good G oo d I l Worst Good I 2 Worst Good

PAGE 134

118 emphasized that while the model is not 100% accurate it is quite good at predicting fla v or quality It represents the most accurate flavor model developed to date for grapefruit JU1Ce

PAGE 135

CHAPTERS CONCLUSIONS Aroma and taste are the predominant factors influencing the flavor of grapefruit juice One of the goals of this study was to identify the key aroma impact volatile components in grapefruit juice These aroma active volatile components along with non volatile taste components were utilized to accurately predict grapefruit juice quality Correlation Between Preference and Analytical Measurements Multivariate statistics helped identify which analytical measurements best correlated with sensory preference measurements Nootkatone was effective in accounting for the variance in the total data set but relatively ineffective in differentiating between juices of various flavor preference Myrcene P caryophyllene and linalool could be used to differentiate juices of various flavor preference Using discriminant analysis they could correctly predict preference category for 7 4 % of the samples At least 19 components were required to correctly predict the preference category for 100% of the samples using forward step wise discriminant analysis. However using backward step wise discriminant analysis it was possible to construct a predictive model using only 16 components It should be emphasized that the models developed to predict flavor preference are based on statistical correlation only and may or may not be causative 119

PAGE 136

120 These models may not include all the flavor impact compounds as it is not necessary to have a statistical model which includes all (or any) causative components as long as other components consistently correlate Taste components were included in almost all predictive models despite the fact that there were 10 times as many volatile components measured This strongly suggests that taste is an important component in any grapefruit juice flavor model However there were no successful models which contained only taste components Therefore the most successful models must contain both taste (non volatile) and aroma (volatile) components Aroma Extraction Methods Isolation of fla v or volat il es is one of the most important and limiting step in food aroma analysis 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 vary in g degrees discriminate against others Three methods of extracting the flavor volatiles were compared Liquid-liquid (pentane : diethyl ether-I : 1 ) static head space (solid phase micro extraction-SP1\1E) and dynamic head space analysis (Tenax/charcoal trap) with solvent elution were used to extract volatiles from a single juice Pentane-diethyl ether extracts produced chromatograms with the best balance of top notes terpenes and sesquiterpenes However, it also extracted undesirable high boiling compounds which increased GC analysis time When compared with methylene

PAGE 137

121 chloride liquid liquid extraction, pentane-diethyl ether solvent extraction gave 73 % more top note peak area Dynamic head space extracted primarily top note volatiles The top note area was 160 % times more than pentane-diethyl ether liquid-liquid extractions and 3 50 % more than methylene chloride liquid liquid extractions However few of the higher boiling compounds cou l d not be analyzed using dynamic head space analysis Static head space SPME extracts contained relatively small amounts of top note volatiles compared to other extraction methods It preferentially concentrated terpenes The main advantage of this is that it is simple rapid does not use an extracting solvent which would obscure highly volatile compounds However it exhibited the poorest analytical reproducibility Of the above three extraction methods evaluated pentane-diethyl ether extracts gave the most representative chromatographic profile Hence, this procedure was chosen to complete the remainder of the study GC-Olfactometry Maturity and heat treatment effects were deterrnined using pasteurized and unpasteurized ( early mid and late season) juices Of the 150 components separated from grapefruit juice aroma extracts 80 were identified and are listed in Appendix D Of these 80 components approximately 37 49 exhibited aroma activity as identified by GC olfactometry There were 25 aroma active components with intensities high enough that

PAGE 138

122 they should be considered key aroma impact components Increased fruity attributes were observed in late season juices when compared to early season juices Several aroma components were lost due to processing Earthy skunky and liquorice type odors were found in pasteurized but not in unpasteurized juices l -p-menthene-8-thiol a sulfur character impact compound produced an intense aroma response but was not detected using FID A strong vanilla aroma peak was detected in all the juices It has been identified as vanillin and is reported for the first time in grapefruit juice Sulfur Compounds in Grapefruit Juice A routine method for grapefruit juice sulfur compounds employing ethyl acetate extraction and sulfur chemilumine s cence detection was developed for the first time Using this process 22 compounds were quantified in early mid and late season pasteurized and unpasteurized grapefruit juices This represents the most comprehensive determination of sulfur compounds in citrus juices reported to date Previous procedures were limited to low molecular weight sulfur compounds such as hydrogen sulfide or dimethyl sulfide Total number and total sulfur peak areas decreased with increasing fruit maturity Thermal processing increased the sulfur peak areas Peak area in early season pasteurized juices was 83 times more than the early season unpasteurized grapefruit juices Late season pasteurized juices had 1 03 times more peak area than corresponding unpasteurized juices The reason for this increase in peak area might be due to reactions occurring at high temperatures and also degradation of nonv olatile sulfur components such as cysteine

PAGE 139

and glutathione. At least two SCD peaks had the same retention times as aroma active compounds detected using gas-chromatography olfactometry (OSME). One of these peaks p menthene-8-thiol produced a small SCD peak but an intense OSME peak Correlations Between Aro ma Components and Sensory Measurements 123 The relative correlation of 26 aroma impact components extracted by pentane diethyl ether and identified by GC-0 and 5 taste components with overall sensory scores were determined. The worst quality juices had the lowest sensory scores for grapefruit aroma intensity and quality (6 29 and 6 48 respectively) Significant correlations were observed between bitter limonin/naringin concentrations and bitterness score (r=O 8 5 and 0 83 respectively at p < 0 05) However there were no correlations between these bitter components and overall flavor score The best juices were rated as being more sweet than sour on sweet / tart balance scale (see Table 14). When correlated with overall sensory score both grapefruit aroma quality and sweet / tart balance had relatively equal significant values (r=0 54 and 0 66 at p < 0 05) This indicates that the contribution of aroma to the overall flavor perception is as important as taste components. This is an important conclusion from this study as many of the previous studies and the current industry standards are based solely on taste components When sweet/balance sensory scores were correlated with the corresponding analytical value (Ratio), the correlation was not significant

PAGE 140

124 Among the volatile components myrcene a-terpineol and several unidentified peaks were found to have significant negative correlations (see Table 15) with aroma intensity and quality Hexanal ethylbutyrate were found to correlate positively with aroma quality Nootkatone was not a significant factor in univariate analysis but in multivariate analysis along with trans-P-ocimene and several unknown compounds it was effective in discriminating the best juices Discriminant analysis was used to classify the juices based on sensory scores Data from taste components was efficient in differentiating the worst juice class Tight clustering of best juice types was observed when volatiles were used to separate the juice types This suggests that taste components are important in separating worst juice types and volatile components for the best juice types Finally 17 aroma active v olatiles and 4 taste components were successfully employed in a grapefruit juice flavor model based only on components with flavor activity The model could separate training juices with 100% accuracy This model was further tested by evaluating nine juices (2 replications each) not used in the training set. The flavor scores were known but the samples were treated as unknowns. Sixteen of the 18 samples were correctly classified within one flavor category. This model represents the most accurate and comprehensive flavor model developed to date It could be used by the grapefruit processing industry to improve both flavor consistency and quality

PAGE 141

APPENDIX A TOTAL ION CHROMATOGRAM OF LATE SEASON GRAPEFRUIT JUICE

PAGE 142

i,,,-a, N 0\ 13-Limonene* 17 00 > 15E+05 0 u 18 '..._;' C 15 00 10E+05 A Q.) > A ) ( 1 16 500000 18 21 19 20 20 27 26 28 22 24 23 25 2 33 34 30 35 3132 36 Vanillin 46 39 40 37 38 22 24 26 Retention Time (min) Appendix A. Total ion chromatogram of late season grapefruit juice highlighting vanillin peak Components Identified in Appendix D 28 49Caryophyllene 43 45 4 4142 \/ 30

PAGE 143

APPENDIXB MASS SPECTRUM OF V ~LIN

PAGE 144

,.... N 00 Adam's Library Spectrum I I I 11 I I I' I I I I I I Vanillin Standard I I Purity = 833 Fit = 932 R.fit = 857 I .. Purity = 878 Fit = 961 R.fit = 911 .,,,, .. ... .... ... . 1 .... . . .. . 1 I ... .. .. ... .. I Sample ti,, ' I I I. I . . . . I ... ... I . . . I .. . '. . I 40 60 80 100 120 140 Mass m/z) Appendix B. Comparison of mass spectrum ofvanillin (from grapefruit juice extra c t ) with library and reference spectra 160

PAGE 145

APPENDIXC LIST OF DESCRIPTORS AND THEIR RELATIVE INTENSITIES (GC-0 ) ..

PAGE 146

Appendix C. Average attributes perceived by the panelists in juice extracts of different maturities Aroma intensities rated bv 12aneli st.s tGC-Ol Kovats Early l J nn. 722 9 0 728 7.6 801 9 0 812 817 9 6 842 6. 4 850 862 9 6 882 7.0 885 7 9 901 934 8 2 965* 13.7 981 10 4 98 4 12.0 989 6.5 997 100 3 12.3 102 6 10 30 9 9 1040 7 6 1047 9.5 1 0 50 106 0 10 2 1 0 72 1089 12.0 1095 1098 10 .3 1101 11 .3 1112 12 0 1120 12 7 l 130 11 1 1148 1165 10 4 1176 7.4 Unp = Unpasteurized, Past = Paste1rrized Early P:l~t 7.8 5 4 8.8 8.3 6.5 7.4 7.3 5.7 12 4 1 2.0 7 1 11 .3 6.8 6.2 1 0.0 12.0 10 .0 10 1 9 8 8.6 Mid Mid Late l T nn P::i~t Unn 7 1 9 2 8 .6 8 9 7 7 9.3 9.6 9 1 10 .7 5 8 6 8 7 .9 8 9 11 6 11 1 4 .0 9 1 6.9 10. l 7.3 8 5 7 6 14 2 14 8 1 3.8 9 9 10 .8 11 2 10 2 11 6 10 8 8 5 9.1 9 7 6.9 12 .3 11 .0 12 .7 7 7 8.3 6 9 5 8 7 0 11 8 8.5 .) .) 8 .6 11 5 12 l 9 8 6.6 6.3 11 3 13. 7 13.5 11 .3 12 .6 12. l 1 3. l 11 .3 12.0 12 .4 12 .0 8 9 9 8 11.3 10 .2 10. 6 9.0 11.5 9 .3 9.9 ... 1 l > 11 1 5 .9 7.2 7.8 130 Late Attributes 9.7 Fruity Fruity 10 .2 Green 10.8 Chemical Stink Green. dead bug 6. 1 Fruitv 10 3 Roasted Grain 11 .2 Oatmeal Citrus 9.5 Cooked Oat 7.9 Floral 11 9 Vitamin C Cat Urine 1 3.7 Cherry Almond 11 0 Unripe Mango Sweet Fruity 6 8 Mintv 11.4 Citrus Lemony 12 4 Citrus, Lemon Grass 7 4 Minty Citrusv Citrus\ 7 8 Floral Unripe Cucumber 14 0 Cotton can dy 6.0 Burnt Sugar 13 9 Cotton candy Citrusy Cotton Candy 10.6 Burnt Sugar, fruity 12 4 floral 8.8 Cooked Oat 8.8 Cooked Rice Medicin y 9.8 Terpeney Greerush Floral 11. 8 Rubber Vinyl 8.2 Temenev

PAGE 147

131 Appendix C. Continued .. Aroma intensities rated by p anelists (GC-0) Kovats Early Early Mid Mid Late Late Attribute Uno. Past Unu. Past Unp. Past 1187 7 5 Alcohol Distelleri 1201 11 2 8 9 1 0 6 12 6 9 3 9 9 Cilantro 1225 7 .3 7 2 6.8 10 4 10 7 Dead Bug Terpeney 1230 9.3 7 8 Earthy Musty Cooked 1251 7 8 7 .0 stinky feet ~ stinky fruit 126 9 7 9 6 9 4 0 Floral 1272 8 0 6 9 Savory 1299 8 6 8.7 9. l 9 1 9 5 8 6 p-menthene thiol 1312 10 .8 10 3 8.5 Rancidoi l 1318 12 2 10 8 10.3 11 7 Spicey Oil y 1 350 8.0 10 8 10 5 1 3 0 Coal Smokey 1360 5.0 7 8 Greenish. Viny l 1 375 6 8 7 l 10 2 5 .6 Fru.iy Peach apricot 1 38 4 9 2 10 4 9 .9 9 2 Cooked Caramalized 1404 1-l .6 11 8 12 5 14 0 10 5 1 2 9 Vanilla 1440 4 .6 Fruity watery Fumes 14-68 12 .6 7.0 Musty 14 92 9 l 6 1 9 0 11 0 11 2 10 -l Floral 15 0 9 7.2 6 7 8 .6 floral fresh 1638* l3 2 5 8 11 -t 1 3 5 13 0 1 3.2 Jasmine 165 9 5 3 Mushroom 1 1 665 () 7 4 8 8 peppery 1 670 7.0 Stinky Bad Breath 168 0 7 2 6 7 Pungent 1690 9 8 5 .8 Stinky Oft 1708 6 3 7 4 5.7 10 .0 10 4 peppery 1717 8.7 8.5 Pepper 1722 8 6 7.2 7 1 10 2 Rotten GFf 173 0 6.6 Stinky GFr Citrusy 1740 1 0.3 Grapefruit peel Oil 1754 10 .0 6 2 6 7 7 .3 11 9 1 3. 1 Floral 1790 8.5 8 9 13 4 10 8 12 5 11.0 GFt Total 41 37 ~Ire 43 49 43 4 3 Internal standards.

PAGE 148

APPENDIXD COMPOUNDS IDENTIFIED IN NOT-FROM-CONCENTRATE GRAPEFRUIT JUICE

PAGE 149

Appendix D Co 1np oU11ds identified in 1101-fron1c on ce ntrate g rap e fruitjuic e. No Co n 1 p one 11t na1ne Retentio11 Index Aldehydes Es t e r s Alcol1ols Ketones Hydrocarbon (Adams R P 1 9 95 ) 1 Pro p yl Acetate + 2 Hexanal 8()() + 3 Etl1yl Butyr a te 80l> + 4 E tl1 y l Ace t ate 807 + 5 Furfura l 8 3( ) + 6 cx-Thujen e 9 3 1 + 7 Thuja-2 4( I 0 ) -Diene 95 6 + 8 Benzaldeltyde ( IS) 96 l + 9 Myrcene 9 9 1 + 10 Et h y l h exa1 1 oate 99 6 + 11 cx-Phellandre11e 1 0 0 5 + 12 o:-Terpinene 1018 + 1 3 Limonene 10 3 I + __. 14 c iJ -P-o c in1ene l C> -l<> + w 15 tran s -P-Oci,nene 1 ()5() + w 16 y-Terpinene l ( )62 0 17 c i s -lin aloo l oxide 1 () 7 4 18 tran s lin alool oxide 1 088 1 9 LinaJool 1 0 9 8 + 20 Nona n a l 1098 + 2 1 Iso-propyl hexanoate + 22 a// o ocirnene 11 2 9 + 2 3 c i J -lim o n ene oxide 11 3 4 24 tran s -p-men tl1 -2-enl ol + 25 }-Pinene oxide 115 6 *private comn1uni c ation s Kevin L Goodner

PAGE 150

Appendix D. Continued . No Co111ponent name Retention lndex Aldeh y de s Esters Alcohols Ketones Hydrocarbon 1 Adan1 s R P 19951 26 Nonariol 1171 + 27 Terpi11-4-ol 11 77 + 28 a:-Terpineol 118 9 + 29 Et li y l Octanoate 11 95 + 30 Decanal 1 20 4 + 3 1 Dih y dr o c itron e ll ol 11 96 + 32 tr ans -carv eo l 1 2 17 + 33 Nerol 12 28 + 34 Citronellol 1 228 + 35 cis ca rveol 1 229 + 36 Ncral 1 2 4() + 37 Carvo 11 e 1 2 ..J 2 + 38 Geranio l 1 255 + 39 Geranial 1 27() + 4 0 Perilla aldehyde 1 27 1 + 41 a: Terpinyl acetate 1 350 + ..J2 a-Cubebene 1 35 1 + ..J 3 a:-Copa11e 1 3 76 4-t4 GeranyJ acetate 1 3 8 3 + 4 5 PC ub e b e n e 1 3 9() + 46 Va11illin 1 39 1 + + 47 Decanol acetate 14 09 + 48 a-Cedrene 14( )9 + .i9 C arvool\ e l l c11e 141 8 +

PAGE 151

Appendix D. Co11tinued .. No Co1 11p o1 1 ent 11a111 e Rc1c11tio11 Ind ex Alde l1 yde s Es t ers Alcoliols Keto11es H ydrocarbo 11 (Ada1 1 1s R P 1995) 50 P-Gurjunene 14 3 2 + 51 a-Guaiene 14 3 9 + 52 Geranyl aceto11 e l-l5 3 + 53 a-Hu111ulene l-l 5-l + 54 P-Cadinene 1 47 3 + 55 gennacre11e 1480 + 56 o-Cadinene 1524 + 57 Caryophyllene oxide 1581 58 Methyl jas 1n onate (IS) 1647 + + 59 Aristolone 1767 + 60 Nootkato11e 180() +

PAGE 152

LIST OF REFERENCES Acree T E. ; Barnard J .; Cunningham, D G A procedure for the sensory analysis of gas chromatographic effluents Food Chem. 1984 14 273-286 Acree T E .; King K. M Identification of aroma-impact compounds in three Japanese teas Institute of Food Technologists Annual meeting. 1996 147 l 082-1236 Adams RP Ion trap mass spectra In Jdentifi.cation of essential oil components by gas chromatography I mass spectJ oscopy ; Allured Publishers : Carol Stream IL 1 9 95 ; pp 45-449 Aiken J W .; Noble A C Comparison of the aromas of oakand g la ss-aged wines American Journal of Enology and Viticulture 1984 35 196-199 Akieda T ; Kato T [Analysis of aroma components in coffee flavour ] Reports of th e C entral Customs Laboratory [Kanzei Chuo Bunsekishohoj 1987 27 17-23 Albach R F .; Redman G H .; Cruse R R Annual and seasonal changes in naringin concentration of Ruby Red grapefruit juice . ] Agric. Food C hem. 1981a 29 808 811 Albach, R F .; Redman G H .; Cruse R R .; Petersen H D Seasonal variation of bitterness components pulp and vitamin C in Texas commercial citrus juices. J Agric. Food Chem 198th 29 805-808 Arctander S Perfume and flavor chemicals (Aroma chemicals) ; Vols I and II ; Allured Publishers : Carol Stream, IL 1994 Attaway J A. Factors influencing the flavor of grapefruit juice Proc e edings of International Society of C itriculture 1977 816-820 Attaway J. A Some new analytical indicators of processed orange juice quality Pro c. Fla State Horti So c. 1972 85 192-203 136

PAGE 153

137 Barros S M. ; Da v is J E .; Dougherty M H .; Griffiths J T Interrelationships of Brix Brix-acid ratio Naringin and Limonin and their effect on flavor of commercial canned single st rength grapefruit juice Proc. Fla. State Ho,ti. Soc 1983 96 316-318 Bazemore RA OS:ME and sensory analysis of aqueous orange essence ; M S Thesis Oregon State University Corvallis OR 1995 Berry R E. ; Tatum J H Bitterness and immature fla v or in grapefruit analysis and improvement of quality J. Food S c i 1986 51 1368-1369 Boelens M H .; Gernert L J van Volatile character-impact sulfur compounds and their sensory properties Perfum e r & Flavorist. 1993 1 8, 19-34 Burroughs L F Amino acids In Th e biochemistry of fruits and their pr o ducts ; Hulme A C ., Ed. ; Academic Press : New York 1970 ; pp 119-146 Buttery R G .; Ling L C. Methods for isolating food and plant volatiles In Biotechnology for improved foods and flavors ACS Symposium Series 19 96; pp 240-247 Cadwallader K R ; Xu Y Anal ys is of vo latile components in fresh grapefruit juice b y purge and trap / gas chromatography J Agric. Food C hem 1994 4 2, 316-318 Chien M .; Peppard T Use of statistical methods to better understand gas chromatographic data obtained from complex fla v or systems In Flavor Measur e m e nt ; Ho C T ., Manley C H ., Eds .; Marcel Dekker Inc : New York 1 99 2 ; pp 135 Chin H W .; Bernhard R A .; Rosenberg M Solid phase micro extraction for cheese volatile compound analysis J Food Sci. 1996 61 1118-1122 Citrus Summary Florida Agricultural Statistics Service : Orlando FL 1996-97 Clark B C .; Chamblee T S Acid-catalyzed reactions of citrus oils and other terpene containing flavors In Off-flavors in foods and bev e rages ; Charalambous G ., Ed. ; Elsevier : New York, 1992 ; pp 229-286. Cunnigbam 7 D G .; Acree T E .; Barnard J Charm Analy s is of Apple Volatiles Food C h e m 1986 19 137-147

PAGE 154

138 da Silva M A.A P .; Lundahl D S .; McDaniel M R The capabilities and psychophy s ics of Osme : a new GC-Olfactometry technique In Trends in flavor research ; Maarse H ., Van Der Heij Eds. ; Elsevier : New York 1994 ; pp 191-210 Demole E P. ; Enggist P [Utilization of a sesquiterpene ketone as perfume / fla vo ur compound.] European Pat en t 1986 Demole E. ; Enggist P .; OhloH: G 1-p-menthene-8-thiol : A powerful flavor impact constituent of grapefruit juice (Citrus paradisi Macfadyen). Helvetica C himi ca Acta. 1982 65, 1785-1794 Fellers P J Florida 's citrus juice standards for grades and their differences from United States standards for grades and united states food and drug administration standards of identity. Proc Fla. State Hort Soc. 1990 103 260-265 Fellers P J The relationship between ratio of degrees Brix to percent acid and sensory fla v or in grapefruit juice Food Te c hnol 1991 45 68 70 7275 Fellers, P J .; Carter R. D .; Jager G de Influence of limonin on consumer preference of processed grapefruit juice J Food Sci 1987 52 741-743 Fellers P J .; Jager G De .; Poole M J .; Hill E C .; Mittal P Quality of Florida-packed retail grapefruit juices as determined by consumer sensory panels and chemical and physical anal ys es J Food Sci 1986 51 417-420 Guadagni D G .; Maier V P .; Turnbaugh J G Effect of some citrus juice constituents on taste thresholds of limonin and naringin bitterness J. Sci Fd Agric. 1973 24 1277-1288 Hasegawa S .; Pelton V A .; Bennett R D Metabolism of limonoids by Arthrobacter globiformis II : Basis for a practical means of reducing the limonin content of orange juice by immobilized cells. J Agric. Food C hem 1983 31 1002-1004 Hinterholzer A ; Schieber le P Identification of most odour active volatiles in fresh hand extracted juice of Valencia late oranges by odour dilution techniques Flavor and F,agrance Journal. 1998 13 49-55 Jella P. ; Rouseff, R .; Goodner K .; Widmer W Determination of key flavor components in methylene chloride extracts from processed grapefruit juice J. Agric. Food C h e m. 1998 46 242-24 7 Jennings W G Objective measurements of flavor quality : general approaches problems pitfalls and accomplishments. A CS. Symp. Se r. 1977 51 1-10

PAGE 155

139 Johnson R. L .; Chan dler B V. Ion exchange and adsorbent resins for removal of acids and bitter principles from citrus juices J. Sci. Food Agric. 1985 36 480-484 Kimball D A Debittering of citrus juices using supercritical carbon dioxide J. Food Sc i 1981 52 481-482. Kimball D A Citrus processingquality control and technology ; Van Nostrand Reinhold : New York, 1991 K1im M ; Nagy S Analysis of orange juice volat ile : Comparison of extraction with Freon 113 and ethyl acetate Proc. Fla. State Hort. Soc 1992 105 110-112 Kovats E Advances in Ch1omatography vol 1 Giddings J C ., Keller R A ., Eds .; Marcel Dekker : New York 1965 ; 229-247 Maarse H .; Visscher C A Volatile co mpounds in Food-qualitative and quantitative data, vol 1 TNO-Civo Food Analysis Institute : The Netherlands, 1989 Marcotte M .; Stewart B. ; Fustier P Abused thermal treatment impact on degradation products of chilled pasteurized orange juice J. Ag,ic. Food C h em. 1998 46 1 991-1996 Marin, A B .; Acree T E .; Hotchkiss J. H .; Nagy S Gas chromatography-olfactometry of orange juice to assess the effects of plastic polymers on aroma character J Agric Food C h e m 1992 40 : 650 654 Matich A J .; R owan D. D .; Banks N H Solid phase micro extraction for quantitati v e headspace samp ling of apple volatiles Analytical C hemistry 1996 68 4114-4118 Meilgaard M .; Civille G V .; Carr B T Sensory Evaluation Techniqz 1 es CRC Pres s, Inc .: Boca Raton FL 199 I ; pp 49-50 Miller J M .; Rockland L B Determination of cysteine and glutathione in citrus juices by filter paper chromatography Archs Bio c h em Biophys. 1952 40 416-423 Miranda-Lopez R .; McDaniel M R .; Watson B T .; Libbey L M Odor analysis of Pinot Nair wines from grapes of different maturities by a gas chromatography olfactometry technique (OSME). J. Food Sci. 1992 5 7, 985-993 Mistry B S .; Reineccius G A .; Jasper B L Comparison of gas chromatographic detectors for the analysis of volatile sulfur compounds in foods In Sulfur compounds in Foods ; Mussinan C J ., Keelan, M.E ., Eds. ; ACS Symposium Series-564 : 1994 ; pp 8-21

PAGE 156

140 Mistry B.S .; Reineccius T .; Olson, L K Gas chromatography-olfactometry for the determination of key odorants in foods In Techniques for analyzing food aroma ; Marsili R ., Ed .; Marcel Dekker : New York, 1997 ; pp 265-292 Moshonas M G .; Shaw P E Analysis of volatile flavor constituents from grapefruit essence J. Agric. Food C hem 1971 19 119-120 Moshonas M G .; Shaw P E Irradiation and fumigation effects on flavor aroma and composition of grapefruit products J. Food Sci. 1982 4 7, 958-960 Moshonas M G. ; Shaw P E Comparison of Static and Dynamic Headspace Gas Chromatography for Quantitative Deter1nination of Volatile Orange Juice Constituents Food Science and Technology Lebensmittel Wissenschaft & Technologi e. 1992 25 236-239 Nunez A J .; Bemelmans J M H .; Maarse H [Isolation methods for the volatile components of grapefruit juice Distillation and solvent extraction methods ] C hromatographia 1984 18 153-158 Nufiez, A J .; Maarse H .; Bemelmans M H Volatile flavor components of grapefruit juice (Citrus paradisi Macfadyen) J. Sci Food Agric 1985 36 757763 Ohloff, G The chemical senses In Scent and Fragrances ; Springer-Verlag : New York 1990 ; pp 1-9 Parliament T H A new technique for GLC sample preparation using a novel extraction de vice. Perfumer and Flavorist. 1986 11 1-8 Peleg H .; Nairn M .; Zehavi U .; Rouseff, R L .; Nagy S Pathways of 4-vinylguaiacol formation from ferulic acid in model solutions of orange juice J. Agric. Food C hem 1992 40 764-767 Pelusio F .; Nilsson T. ; Montanarella L. ; Tilio R .; Larsen B. ; Facchetti S .; Madsen J 0 Headspace solid-phase micro extraction analysis of volatile organic sulfur compounds in black and white truffle aroma J. Agric. Food C hem. 1995 43 2138-2143 Pino J A Correlation between s ensory and gas chromatographic measurements on orange volatiles Acta Alim e ntaria 1982 11 1-9 Pino J A .; Cabrera M Quality of canned grapefruit juice produced in Cuba for four seasons. Nahrung 1988 32 875-879

PAGE 157

141 Pino J. A .; Torricella, R. ; Orsi F. Correlation between sensory and gas chromatographic measurements on grapefruit juice volatiles Acta Alimentaria 1986a 15 237246 Pino J A .; Torricella R ; Orsi F. ; Figueras L. Application of multivariate statistics for the quality classification of single-strength grapefruit juice J Food Qual 1986b 9 205-216. Reneccius, G.A Source book of flavors 2 nd edition ; Chapman and Hall : New York, 1994 Rouseff, R L. Nomilin a new bitter compound in grapefruit juice J. Agric. Food Chem. 1982 30 504-507 Rou seff, R. L .; Barr os S. M .; Dougherty M H .; Martin, S F A survey of quality factors found in Florida canned single strength grapefruit juice from 1977-78 1978-79 and 1979-80 seaso ns Proc Fla. State Hort. Soc 1980 93 286-289 Rouseff, R .; Nagy S A multivariate pattern recognition approach for identifying quality factors in Florida orange juice In Flavor Science and Technology. Martens M ., Dalen G.A ., Russwur~ H ., Jr Eds Wiley: New York 1982 Schieberle P. ; Grosch W. Identificierung von Aromastoffen ausder Kruste von Roggenbrot Verleich mit den Aromastoffen der Krome Z Lebensm. Unters. Forsch. 1984 479-483 Schieberle P. ; Grosch. W Identification of potent flavor compounds formed in an aqueous lemon oil / citric acid emulsion J. Ag,ic Food Chem 1988 36 797 -8 00 Shaw P .E Fruits II In Volatile compounds in foods and beverages ; Maarse H ., Ed .; Marcel Dekker : New York, 1 99 1 Shaw P E. Volatile components important to citrus flavors In Quality control manual for citrus processing plants ; Red J B ., Shaw P E ., Hendricks C M ., Hendricks D .L .: Eds. ; Ag Science : Florida 1 996 134-172 Shaw P E .; Ammons J M .; Braman R S Volatile sulfur compounds in fresh orange and grapefruit juices : identification quantitation and possible importance to juice flavor. J. Agric. Food Chem. 1980 28 778781

PAGE 158

142 Shaw P E .; Nagy S Analysis and flavor effects of volatile sulfur compounds in citrus juices. In The quality of foods and beverages ; Academic Press: New York, 1981 ; pp 361-376 Shaw, P E. ; Wilson C W ill Importance of nootkatone to the aroma of grapefruit oil and the flavor of grapefruit juice J Agric. Food Chem 1981 29 677 679 Steffen A. ; Pawliszyn J Analysis of flavor volatiles using headspace solid-phase micro extraction J Agric. Food Chem. 1996 44 2187-2193 Stevens K L. ; Guadgni D C .; Stern, D J Odour character and threshold values of nootkatone and related compounds. J Agric Food Chem 1970 21 590-593 Stevenson R J .; Chen X D Solid-phase micro extraction: a useful sample preparation technique for gas chromatography-mass spectrometry analysis of volatile components in food products Food Technol. 1996 26 24-28 Stinson W.S. ; Barros S Grapefruit juice improvement program Proceedings of citrus short course meeting Clearwater FL 1997 pp 113-126 Tatum J H .; Lastinger Jc Jr .; Berry R E Naringin isomers and limonin in canned Florida grapefruit juice Proce e dings of the Florida State Horticultural Society 1972 85 210-213 Tressl, R .; Silwar, R. Investigation of sulfur-containing components in roasted coffee J Agric Food C hem. 1981 29 1078-1082 Umano K .; Shibamoto T Aia : United States of America F M. A A new method of headspace sampling : grapefruit volatiles (In Flavors and fragrances : a world perspective' edited by B M Lawrence et al Conference Washington DC USA 16-20 Nov 1986 Amsterdam Netherlands ; Elsevier Science Publishers BV ISBN 0-444-42964-6 ) Developments in Food Science 1988 18 981-998 Velez C. ; Costell, E .; Orlando L ; Nadal M. I .; Sendra, J M .; Izquierdo L Multidimensional scaling as a method to correlate sensory and instrumental data of orange juice aromas J Sci. Food Agric 1993 61 41-46. Wampler T P Analysis of food volatiles using headspace-gas chromatographic techniques. In Techniques for analyzing food aroma ; Marsili R ., Ed .; Marcel Dekker: New York 1997 ; pp 27-58 Weurman C Isolation and concentration of vo latile s in food odor research J Agric. Food Chem 1969 17 370-384

PAGE 159

143 Widmer W W .; Martin S F Analysis of limonin in citrus juices by direct injection and online sample clean up Abstracts of Citrus Processors Meeting October 1994 Lake Alfred FL 1994 Wilson, C W .; Wagner C J .; Shaw P E Reduction of bitter components in grapefruit and navel orange juices with b-cyclodextrin polymers or XAD resins in a fluidized bed process J Agric Food Chem 1989 37 : 14-18 Xiaogen Y .; Peppard T Solid-phase micro extraction for fla vo r analysis Journal of Agricultural and Food C h emistry 1994 42 1925-1930 Zhang Z .; Yang M J .; Pawliszyn, J Solid-phase micro extraction Anal. Che m. 1994 66 844-853

PAGE 160

BIOGRAPHICAL SKETCH Prashanthi J ella received her undergraduate degree in horticulture from Andhra Pradesh Agricultural University in 1991 She came to United States in 1 992 to pursue her master s degree at Texas A&M University College Station, Texas She received her master s degree in food science and technology in 19 94 Her major profe ssor at T AMU was Dr. Luke Howard who is now at the University of Arkansas Fayetville AK Prashanthi went to the University of Florida in 1 995 to pursue her doctoral degree Her research under guidance of Dr Russell Rouseff at the Citrus Research and Education Center Lake Alfred was selected as best graduate research at the National ACS meeting (ACS / AGFD) 1998 Upon graduation Prashanthi will work as an Associate Scientist in the Flavor Development Group at Coca-Cola Company Atlanta, Georgia 144

PAGE 161

I certify that I have read this study and that in my opinion it conforms to acceptable standards of schoJarly presentation and is fully adequate in scope and qualjt y, as a dissertation for the degree of Doctor of Philosophy '==~ ). "4-c.t Russell L Rouseff, Chairman P1 ofessor of Food Science and Human Nutrition I certify that I have read tllis study and that in my opiruon it conforms to acceptable standards of scholarly presentation and is fully adequate in scope and qualit y, as a dissertation for the degree of Doctor of Philo so hy 7 J sse F. Gregory rn Professor ofFood Science and Human Nutrition I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly pre s entation and is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy Sean F. O Keefe Associate Professor of Food Science and Human Nutrition I certify that I have read this study and that in my opinion it conforms to acceptabJe standards of scholarly presentation and is fully adequate i11 scope and quality as a dissertation for the degree of Doctor of Philosophy David H Powell Associate Scientist of Chemistry

PAGE 162

I ce1tify that I have read this study and that in 1ny opinion it conforms to acceptable standards of scholarly presentation and is fully adequate in scope and quali ty, as a dissertation for the degree of Doctor of Philosoph Charles A Sims Associate Professor of Food Science and Human Nutrition I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate in scope and quali ty, as a dissertation for the degree of Doctor of Philosophy Arthur A '" f eixeira Professor of Agricultural and Biological Engineering This dissertation was submitted to the Graduate Faculty of the College of Agriculture and to the Graduate School and was accepted as pa1tial fulfillment of the requirements for the degree of Doctor of Philosophy 7 < August 1998 Dean College f Agriculture Dean Graduate School

PAGE 163

LD \l'& o ,ctc1<2 'J 'lf UNIVERSITY OF FLO R ID A II I 1111111 11 I 11 I 3 1262 08554 9466


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
FILES