1

CHLOROGENIC ACID CONTENT

OF GREEN COFFEE BEANS

by

Omozoje Ohiokpehai, M. Sc. AIFST

A thesis submitted in accordance with the

requirements of the University of Surrey for
the degree of Doctor of Philosophy.

Dept. of Biochemistry,
Faculty of Biological and Chemical Sciences,
University of Surrey,
Guildford, Surrey. September, 1982.
 11

ABSTRACT

Coffee history, production, processing and composition have been

reviewed with particular emphasis upon factors known, or thought to affect

beverage quality. In this respect attention has been focused upon chlorogenic

acid (CGA) content. Four different conventional methods of analysis for

chlorogenic acids have been studied and their repeatability and reproducibility
established.

An attempt was made to assess the possible interference of CGA-

quinone with these techniques. A chemical and biochemical method were
employed to synthesize the quinones. It was found that some oxidation
products were detected by the periodate reagent.

Results from these methods were integrated and the concept of

Analytical Ratios introduced. Values for these ratios were predicted and
tested on eight samples of commercial green coffee, forty-two immature

green coffees, thirty-four peculiarly coloured green coffees, and five

samples of roasted coffee. It was found that these Analytical Ratios lacked

discrimination and could not be recommended for commercial quality assessment.

Studies on the progressive accumulation of CGA during maturation showed

that there existed phenolic compounds other than CGA in very immature beans.

Two types of peculiarly coloured beans were examined. One type

appeared to be produced by an enzymic browning mechanism: evidence is

presented which suggests that immature beans may be more likely to discolour

in this way than mature ones. The second type has a tendency to retain their

silverskin: it was shown that some contained chlorophyll-like pigments and

carotenoids, some carotenoids only and others may have lost both by bleaching

reactions.

Objective and subjective studies established that dicaffeoylquinic acid

(DCQA) is astringent. This sensation is a function not only of DCQA, but

of the caffeoylquinic acid (CQA): DCQA ratios.

Two unconventional methods of analysis utilising protein precipitation

were tested for detecting beans rich in DCQA but neither proved satisfactory.
 111

ACKNOWLEDGEMENTS

The author wishes to express her sincere appreciation to her

supervisor, Dr. M. N. Clifford for continued guidance and encouragement

during the course of this study.

The constructive criticism, suggestions and patient editing of this

manuscript by Professor J. W T. Dickerson are gratefully appreciated.

Gratitude is also extended to the members of the Chemistry Department

for their co-operation, Dr. A. E. J. McGill of Home Economics Department for

the Taste Panel facilities and Dr. M. Crowther for his valued statistical

advice, my fellow research workers and the technical staff of the Department
of Biochemistry especially Mrs. B. Smith for their help and encouragement.

Thanks are due to Dr. J. W. Corse of U. S. D. A., Albany, Calif., U. S. A.

for samples of feruloylquinic acid (5-FQA), 3,4-, and 3,5-dicaffeoylquinic

acid (DCQA), Dr. A. Lea of Long Ashton Research Station, Long Ashton,

Bristol, England, for the supply of 3-, 4-, and 5-p-coumarylquinic acid,
Dr. S. Martino of the Department of Plant Biochemistry, University of
Buenos Aires, Buenos Aires, Argentina for the supply of 4,5-DCQA. Also to

Fed. Nat. Cafeteros in Colombia, IFCC in the Ivory Coast, KIRDI in Kenya,
The Royal Botanic Gardens at Kew, for the supply of coffee samples.

Finally, the author is particularly grateful to her friends for moral

support, to Dr. E. Illy and Illycaffe for practical advice and part financial

support.
 iv

CONTENT

Title Page i
Abstract ii
Acknowledgements iii

Content iv

Abbreviations vii
CHAPTER 1 INTRODUCTION AND LITERATURE SURVEY
I Introduction 2
II The Composition of Green and Roasted
Coffee Beans 6

a. Moisture content 8

b. Carbohydrate content of coffee beans 8

(i) Low molecular mass carbohydrates 9

(ii) Polysaccharides 9

(iia) Reserve polysaccharides 10

(iib) Structural polysaccharides 10

c. Lipid content of coffee beans 12

d. Mineral content of coffee beans 14

e. Nitrogen content 16

(i) Amino acids 16

(ii) Proteins 16

(iii) Non -protein nitrogen 19

f. The phenolic compounds of coffee beans 21

g. Non-phenolic acids 26

h. The volatiles of coffee beans 27

III Green Coffee Processing 29

IV Theories about Coffee Quality 34

V Analysis Chlorogenic Acids in Coffee 40
of
a. Extraction of chlorogenic acids from
coffee beans 40

b. Identification of chlorogenic acids in

coffee beans 40

c. Estimation of chlorogenic acid fraction 44

VI Aims of The Present Investigation 45

 V

CHAPTER 2 METHODOLOGY
I Introduction 48

II Colorimetric Methods 48

III Chromatographic Method 55

IV Integration of Colorimetric Data with
Chromatographic Data 65

EXPERIMENTAL, RESULTS AND DISCUSSION

CHAPTER 3 ESTABLISHMENT OF BASELINE DATA

I The CGA content of Normal Green Coffee Beans 70
II Effect of Roasting on CGA content of Tanzanian
A rabica Coffee 73
III Studies of Quinones as Potential Interfering
Substances 82

CHAPTER 4 Q-IAN,. S IN CGA OJN'INT DURING 'filE

DEVElOPMENTOF GRMI C01 I BEISTS
I Introduction 95

II Origin and Nature of Coffee Samples 96

III Treatment and Analysis of Beans 103

CHAPTER 5 CHARACTERISATION OF PECULIARLY COLOURED

COFFEE BEANS
I Introduction 117

II Extraction and characterization of chlorophyll 118

III Chlorogenic Acid Analysis of Peculiarly

Coloured Beans 125

CHAPTER 6 COFFEE ASTRINGENCY

I Introduction 144

II Organoleptic Investigation 146

III Objective Investigation 153

CHAPTER 7 GENERAL SUMMARY OF RESULTS 159

Future Work 162

 vi

FINAL CONCLUSIONS 163

BIBLIOGRAPHY 165

APPENDIX A

C
D
E

F
G
H
 vii

ABBREVIATIONS

BD QA - Bound quinic acid

CA - Caffeic acid
CFQA - Caffeoylferuloulquinic acid
CGA - Chlorogenic acid
cm - Centimetre
°C
- Degree centigrade
CoQA - Coumarylquinic acid
5-CQA - Caffeoylquinic acid
DCoQA Dicoumarylquinic acid
-
DCQA Dicaffeoylquinic acid
-
dmb Dry matter basis
--
Est. - Estimated
FA - Ferulic acid
FQA - Feruloylquinic acid
FDCQA - Feruloyldicaffeoylquinic acid
GC - Gas chromatography

g - Gram
HPLC - High Pressure Liquid Chromatography
HPLC CGA HWLC CQA + HPLC FQA + 1.37 HPLC DCQA

I. D. - Internal diameter
IR - Infra red
L - Length
M. Wt. - Molecular weight
M - Mole

mg - Milligram
mm - Millimole

mm - Millmetre

mL - Millilitre

MV - Molybdate value

Ail, - Microlitre

gum - micron
 viii

NCS - N-chlorosuccinimide
NMR - Nuclear magnetic resonance

nm - nanometre
O. D. - Optical density
PC - Paper chromatography
PV - Periodate value
PPO - Polyphenol oxidase
PVP - Poly-N-vinylpyrrolidone
QA - Quinic acid

r- Correlation coefficient
r. p. m. - Revolution per minute
SD - Standard deviation
TBA - Thiobarbituric acid
TBA QA - Thiobarbituric quinic acid
TBA BD QA - Thiobarbituric acid bound quinic acid
TMV - Tobacco mosaic virus
TEA - Triethylamine

TLC - Thin layer chromatography

Total HPLC CGA = HPLC CQA + HPLC FQA + HPLC DCQA
UV - Ultra violet
% Percentage
AA - Difference in absorbance

- Sigma

Beta
a- Delta
7- Gamma
(max) - Maximum wavelength

var. - Variety

w/v - Weight per volume

* All data reported in this study are calculated on dry matter basis.
 -1 -

CHAPTER ONE

INTRODUCTION AND LITERATURE SURVEY

 -2-
I Introduction

Coffee was used as a beverage by the Arabs as long ago as 600 A. D.

and they introduced it to the Mediterranean countries about 1500 A. D.

From here it found its way into Western Europe about 1630 A. D.
(Hartman et al., 1981).

Coffee plants were brought to Brazil in 1727 A. D., and after 40 years,
it became, and still is, the world's leading coffee grower and exporter, now

producing one-third of the world's coffee supply - about 1.5 million tons

annually.

Coffee plants are evergreen shrubs within the family Rubiaceae and

genus Coffea (Haarer, 1963). Arabica coffee is self-pollinated, and

Robustas are largely cross-pollinated. Consequently, the variety of forms

among Robusta is much greater.

The major species of commerce are known as Arabicas and Robustas.

Purseglove (1968) and Williams (1975) suggest that Arabica coffees originated

in the mountain forests of Ethiopia. There are two varieties, Coffea Arabica

var. Arabica and C. Arabica var. Bourbon, the latter variety being considered

the better. These are cultivated in elevated sites in tropical and subtropical

regions of America and East Africa.

The Robustas, prepared from C. Canephora, yield a beverage of inferior

flavour but the plant is vigorous, with disease resistance (Williams, 197 5).

Another species of minor commercial importance is C. Liberica which is grown

in West Africa and Malaysia. Recently a hybrid C. Arabusta-Capot and Assi
has been developed and is grown commercially in the Ivory Coast.

There are many wild species, Mascarocoffea being one. According to

Assi (1977) there are at least 32 species in Africa. Among them are

C. Congensis, C. Eugenoides and C. Zanguebarrie and C. Excelsa. It is

recommended that these be retained for use in hybridisation, the Mascarocoffea

being of particular interest because of a low caffeine content.
 3_
-

Coffee trees are planted about 2.4m (8 ft ) apart and are kept low -
about 1.8m (6 ft ) by pruning to facilitate harvesting. They start bearing
full crops on the lateral branches at about 5 years and reach full production

at 15. The tree produces mature fruits (cherries) 8- 11 months after

flowering in a single-bearing region (Williams, 1975). The red fruit,

which is a drupe, remains at its prime for about one week. For this

reason, harvesting is one of the most labour intensive tasks in coffee

growing. The traditional method is to pick periodically or when necessary,

collecting loose fruit from the ground at the same time.

Colombia ranks second in coffee production, with about 0.5 million

tons annually. Mexico, El Salvador, and Guatemala each produce about
0.1 million tons per year. Coffee is, in fact, the mainstay of the economy

of the Central American countries.

Production in Africa is low, for example Angola, Uganda, Ivory Coast

and Madagascar together produce about 0.8 million tons annually (Ass i, 1977).
In these countries the second species (C. Canephora) is mainly grown. This

yields a cheaper, less flavourful coffee, but it is now in great demand for use
in blending with C. Arabica coffees and in the manufacture of instant coffees.

The coffee fruit contains two seeds, coffee beans (Fig. 2) which are

separately enclosed in a thin membrane or silverskin. These are in turn

surrounded by a layer of mucilage immediately inside the 'fleshy' part of the

cherry. After harvesting, the seeds are classified and then separated from
the pulp either by wet or dry processing (Clarke, 1976; Fig. 1).

Coffee beans in a dry state are exported to the manufacturing countries

(Germany, Italy, U. S. A. and Great Britain etc. ). On reaching these

countries, they are critically sorted, graded and roasted.

Roasting develops the characteristic odour, taste and colour of the

coffee bean which we know as coffee. Roasted coffee beans are ground or
 -4-

Harvesting of coffee cherry

Classification

Drying Pulping

Mucilage removal

Drying parchment coffee

Dehusking Hulling

Sorting and grading Sorting and grading

The Dry Process The Wet Process

Fig. 1. Green coffee processing in the producing country.

(Clarke, 1976).
 -5-

sold whole to consumers. Ground coffee is used for making instant coffees

without the removal of caffeine. Decaffeination is the process by which

the caffeine is selectively extracted from green coffee beans by the use of
water or other solvents.
 -6-

II Composition of Green and Roasted Coffee Beans.

Table 1. Chemical Composition of Green Coffee Beans.

(Clifford, 1975)

Contents (%admb)

Components Arabica Robusta

Total polysaccharides 50.0 - 55.0 37.0 - 47.0

Amino acids 2.0 2.0

Oligosaccharides 6.0 - 8.0 5.0-7.0

Lipids 12.0-18.0 9.0-13.0

Proteins 11.0 - 13.0 11.0 -13.0

Total chlorogenic acids 5.5 - 8.0 9.0 - 10.0

Aliphatic acids 1.5 -2.0 1.5 -2.0

Caffeine 0.9 - 1.2 1.6 -2.4

Minerals 3.0 -4.2 4.0 -4.5

Trigonelline 1.0 -1.2 0.6 - 0.8

 -7-

Disc

Mesocarp (pulp)

Exocarp (skin-red)

Endocarp (patchment)

Mucilage

Spermoderm (Testa)
-' silverskin' when dry
- 'chaff" when roasted

Seeds called 'Beans' when dry

Fig. 2 Cross-section through the coffee cherry.

 -8-

The quality of coffee beverage depends on the selection of the raw beans

and the roasting process - particularly the roasting time.

a) Moisture Content

The moisture content of green coffee beans would affect the storage
stability and thus the flavour qualities of the roasted beans (Rao et at., 1969).
A standardised method of moisture determination and the expression of the
analytical data on dry matter basis are essential for accurate comparisons
of samples (Smith, 1965).

The moisture content of green coffee was defined as the loss in mass
undergone by coffee when it is brought to true equilibrium with an atmosphere
having zero water vapour pressure, under conditions such that interfering
reactions are avoided (International Standards Organisation, 1978). However,

most research workers prefer to define it as the weight loss after heating the

ground bean to a constant weight at 105-110°C at atmospheric pressure or at

°C
100 and a pressure no greater than 100mm Hg (Clifford, 1979, Rees and

Theaker, 1977).

For green coffee beans a moisture content between 10 and 12970

is

considered to be ideal (Clark, 1967; Sivetz and Foote, 1963).

b) Carbohydrate content of green coffee beans

Some 50 to 60% of green coffee consists of carbohydrates (Goldoni, 1979).

These may be subdivided into water-soluble and water-insoluble fractions, the
latter being much more difficult to study. Such a subdivision however bears

little relationship to the functional properties of these substances. In this

respect, they may be subdivided into:

low molecular mass carbohydrates
reserve polysaccharides, and
structural polysaccharides,

some of which are intimately associated with non-carbohydrate material such

lignin or proteins (Taylor, 1975).

as
 -9-

i. Low molecular mass carbohydrate

Clifford (197 5) reviewing the data available said that green coffee beans
contain 5 to 8570
sucrose. There was more sucrose in Robusta coffees than
in Arabicas. Small quantities of free glucose (0.5 to 1.0%) are also
present. He indicated that raffinose (1-alpha-6-galactosylsucrose) and
stachyose (1-alpha -6-galactosylraffinose) are also present, these being
higher in Robustas than in A rabicas
.

More recently, Solov'eva (1976), Nakabayashi (1977) obtained similar

values for sucrose but found no monosaccharides.

Roasting caused a progressive and ultimately severe loss of these

carbohydrates. Sometimes small quantities of sucrose survive and usually
some glucose and fructose. Traces of arabinose, galactose, and in one
case, maltose, suggested that polysaccharides had been degraded (Calzolari

and Lokar, 1967; Nakabayashi, 1977).

It is generally accepted that sugar degradation, with or without

reaction with other components, yields many of the flavour producing
substances of roasted coffee (Feldman et al., 1969; Amorim et al., 1974).
However, Amorim et al. concluded that the differences in beverage quality

could not be explained by differences in the sugar content in the green beans.

ii. Polysaccharides

The water-soluble polysaccharides may have been isolated unchanged, but

the water - insoluble polysaccharides can only be studied after suffering some
chemical treatment. The data so obtained cannot easily be related to the

rntive polysaccharides.

The acid hydrolysis of green coffee beans from which the mono- and

oli gosaccharides have been removed yields the constituent sugars of the
polysaccharide, i. e. galactose, glucose, mannose, arabinose, xylose,
rhamnose and at least one uronic acid.
 - 10 -

iia. Reserve Polysaccharides

It was not until 1977 that the presence of starch was confirmed
(Dentan, 1977). Previously a galactomannan was considered the major
reserve polysaccharide (Clifford, 1975).

iib. Structural Polysaccharides

Most structural polysaccharides are either poorly or totally insoluble

in water and the severe extraction procedures commonly employed must be
expected to produce artefacts. One would expect celluloses, hemicelluloses
and pectins to be present and in his review Clifford attempted to relate the
fractions obtained by various workers to these three groups of substances.

Cellulose is a major component of plant cell walls; Wolfrom et al.

(1964) and Robinson (1980) have unequivocally demonstrated its presence
in coffee beans at a level of approximately 5%.

Pectins have been extracted using hot water, dilute acid (primary cell
wall pectins) and ammonium oxalate (middle lamella pectins). Depending

on the bean and method, yields of approximately 1 to 4% have been obtained

(Plunkett, 1956; Thaler and Arneth, 1967). As one would expect, galactose

and arabinose were always present - in some cases glucose and mannose
were also present although these are not considered typical components of
pectins. Amorim et at. (1974) reported that the content of oxalate-soluble
pectins bears no relationship to coffee bean quality.

Extraction into alkali has yielded several hemicelluloses of variable

composition but with mannose always present at a significant level.

The most recent investigations by Thaler (197 5) and Ara and Thaler
(1976) made extensive use of chlorine dioxide to solubilise lignin. Unfortunately,

this also causes considerable changes in the polysaccharides, and the results

so obtained cannot easily be related to native polysaccharides. Nevertheless,

 - 11 -

these workers have observed some marked differences between Arabicas

and Robustas.

Although not a polysaccharide, lignin is a structural element that is

intimately associated with them. Thus lignin could be present in coffee

beans; however, the measurement of it has proved difficult. Traditional

methods for measuring lignin are designed for the analysis of woody tissues
that have a very low protein content. Coffee has a high protein content and
interference occurs (Clifford, 1972).

Thaler and Arneth (1967) indicated that during roasting cellulose was
very stable, the mannan slightly affected, the galactan partially destroyed,

and the araban was greatly reduced in proportion to the severity of roast.

Ara and Thaler (1976) reported that the hot water-soluble fraction

increases during roasting as a result of thermal degradation. These

products doubtless contribute significantly to the yield of commercial

instant coffees; however, they are poorly characterised. In some

instances these fractions have probably been described as humic acids.

In this context the term humic acids refers to the water-soluble high

molecular mass fraction of roasted coffee beans from which sugars and/or

amino acids and/or phenols can be released by hydrolysis.

Clifford (1972) indicated that after water-soluble humic acids were

purified by dialysis, yields of 12 to 15% for these compounds (mass range

5000 to 50,000) were observed. When products of acid, alkaline and
enzymic hydrolysis were examined chromatographically it was found that

protein -polysaccharide and protein; chlorogenic complexes were present.

On analysis, 1
nitrogen contents of .5 to 5.0% were found and up to 18

ninhydrin -reactive substances could be released by hydrolysis.

The polysaccharide-containing fractions form 9.0 to 12.0% of roasted

Arabica and 10.0 to 15.0% of similarly roasted Robusta beans, and galactose,

mannose and arabinose could be released by hydrolysis.

 - 12-

The content of these phenol-containing fractions decreased with

severity of roasting from mildly roasted beans (4.07) to severely roasted

(about 1.070). This decrease was probably due to the insolubilisation of
the proteins caused by phenols produced during roasting.

c) Lipid content of coffee beans

Early data relate primarily to crude lipid contents. On this basis

Coffea sp. may be divided into three groups. Arabicas contain 12 to 18%,
Robusta 9 to 13570
and Liberica 11 to 12% (Carisano and Gariboldi, 1964).

This crude lipid contains neutral triglycerides and a complex

unsaponifiable fraction. The early data have been summarised by Clifford
(1975). More recent investigations have shown that some of the early

reports would be misleading. Folstar and colleagues (1975,1976) looked

at the effect of bean particle size on the yield and composition of the crude
lipid. They ultimately recommended a particle size of less than 0.5mm
and reported that larger particle sizes led to preferential extraction of
certain components of the oil and thus produced misleading data.

It is clear that the major component of coffee oil is neutral

triglycerides, in which the major fatty acids are palmitic, stearic, oleic,
and linoleic. The oil has an unusally high unsaponifiables content which
consists of sterols, sterol esters, diterpenes, wax, pigments, vitamins
and hydrocarbons (Folstar et al., 1975, Tiscornia, 1979).

The sterols require little comment, being typical of vegetable oil.

The diterpenes however are unusual and appear to be unique to coffee. The

oil extracted from green coffee contains two diterpenoid alcohols, cafestol
and kahweol (Fig. 3). They are present both in the free state and as

monoesters of long chain aliphatic (Alcaide 1971). Kulaba (1981)

acids, et al,
reported a possible connection between the content of kahweol and beverage
quality. He concluded that low kahweol and high oil content was associated

with high beverage quality. High kahweol and medium oil content was
 - 13 -

:u OH
OH

(i)
(

20H
H

(ii)

Fig. 3

Two Green Coffee Diterpenoid Alcohols (i) Kahweol and

(ii) Cafestol.

(Gibson, 1971)
 - 14 -

associated with medium quality and low kahweol and low oil content was
associated with low quality.

The waxes have been studied in some detail, because they are
derivatives of serotonin and have been blamed for digestive disorders
(Van der Stegen, 1979). The waxes are N-beta-alkanoyl-5-hydroxy-

tryptamine (C-5-HT) containing primarily arachidic, behenic and

lignoceric acids (Folstar et al ., 197 5; Van der Stegen and Noomen, 1977).

d) Mineral content of coffee beans

The data on mineral content were summarised by Clifford (1975) and

his summary table is reproduced as Table 2. A report from Quijano-Rico
and Spettel (1975) provides additional data including values for twelve

more elements. They noted that Arabicas were richer in bromide and

chloride and poorer in potassium, copper, strontium and barium than

Robusta coffees.

There has been considerable interest in the possible relationship

between the total ash, sulphated ash and potassium content as an index
of instant coffee extraction. However, Clarke and Walker (1974) have
indicated that determinations of these parameters in instant coffee can only
give an arbitrary measure of extraction rate, due to the variability of
mineral composition of different green coffees, and also to the variable
processing conditions that are used commercially in the extraction of
roasted coffees. In general terms, the higher the yield of soluble material
extracted from roasted coffee the lower the mineral or potassium content.

There have been suggestions that certain minerals predispose to

higher quality and others to lower quality in the beverage. Such suggestions
do not seem to have been confirmed but are of considerable interest since the
mineral content of coffee beans is significantly influenced by soil composition

and agricultural practices (fertilizers or antifungal sprays). One can

hypothesize that minerals, particularly the transition metals, might influence

quality by acting as catalysts during roasting.

 - 15 -

Table 2: Typical Mineral Content of Green Coffee Beans (Clifford, 1975)

(Values expressed per 100g air-dried beans)

Major Components Minor Components

(range, mg) (range, /Lg)

K 1350-1712 Cr 74-1327
Mg 142-176 V 70-110
Ca 76-120 Ba 100-615
Na 2.3-17 Ni 11-388
Fe 2.1-10.5 Co 10-93
Mn 1.1-9.8 Pb 1827

Rb 0.6-4.2. Mo 11-27

Zn 0.5-3.2 Ti 4-20

Cu 0.5-2.3 Cd 3

Sr 0.4-1.3
 - 16 -

It is believed that most of the minerals are associated chemically with

the main constituents of coffee beans; part of the potassium is present as
the salt of chlorogenic acids, and a further part as the salt of a loose

caffeine-chlorogenic acid complex (Smith, 1963). A magnesium- chlorogenic

acid complex has been reported to be responsible for the colour of good

quality green coffee beans (Northmore, 1965 and 1967). The magnesium

content was the quality-limiting factor.

e) Nitrogen content

(i) Amino acids

Arabica coffee contains 2% of free amino acids (Walter, 1970). Campos

and Rodrigues (1971) using chromatographic and electrophoretic methods

indicated that most of the amino acids commonly found in plant tissues were

present in green coffee However, there were traces of three

..
unidentified amino acids. These authors observed that there were varietal
differences in the amino acid content. These differences were mainly in

the arginine, beta-alanine and pipecolic acid of Arabica or Robusta.

Table 3 (p. 17), taken from Poisson (1977) summarises the data from-
Wallter, and Barbi rol li (1965).
__(1,9a70)
Amino acids are destroyed by heat; however they were detected in

roasted coffee and coffee brew in small quantities (Pereira and Pereira, 1971).

Also an uncharacterised peptide was reported to be of importance in roasted

coffee aroma production (Russworm, 1969).

(ii) Proteins

The crude protein content (estimated as Nx6.25) of coffee was reported

(Ba.birolti, 1965; Streuli, 1975) to be 11 to 15.8g per 100g. However, the

water-soluble protein, (amounting to 3% of the bean) has a nitrogen content of

15% (Underwood and Deatherage, 1952). They also reported that the water-

soluble proteins of Santos and Colombia coffees had an iso-electric point at

pH 4.6 to 4.7 indicating that there was a low content of basic amino acids.
 - 17 -

Table 3. Free Amino Acid Content of Green Coffee Beans

m /100 ) (Poisson, 1977)

Amino Acids Walter Barbiroli

Alanine 24 40-80

Arginine 4 0

A sparagine 30 0

Aspartic acid 33 10-28*

Glutamic acid 49 32-50

Glycine 2 0

Histidine 4 0

Is oleu cine 3 )
3-12
Leucine 3

Lysine 4 4 -1 O*

Methionine 8-28

Phenylalanine 8 2 -10*

Pipecolic acid 3 0

Proline 14 0

Serine 12 18-40

Threonine 0 3-8*

Tyrosine 4 0
Valine 2 14 -22*

* Only in some varieties.

 18 -

Amorim and Josephson (1975) studying the water-soluble nitrogen

component of green coffee of Brazilian origin (Rio and Soft) with different

methods of gel filtration and dialysis with membranes of different molecular

mass cut-offs, showed that there were no significant differences in the

protein content of Rio or Soft (good quality) coffees. They observed that

there were numerous isoelectric proteins in the pH range of 5.7 to 6.3

and a few in the range of 4.4 to 4.7 using non-urea gel contrary to previous

reports. They were able to show that green coffee contains protein-

chlorogenic acid complexes similar to those which had been confirmed in

sunflower seeds (Sabir et al., 1974).

Poisson (1977) reviewing the work on the composition of proteins in

coffee observed that when proteins are degraded during roasting, a number

of substances are produced concurrently with the formation of coffee aroma

and colour. The degradation of proteins causes loss of protein and the
more severe the roast the lower the protein content. It was reported that

when Santos coffee was roasted to six different degrees of roast, that is,
10,12,14,16,18 and 20 based on percentage loss in weight, there was a

loss in protein of 6% at the lowest weight loss and 14% at the highest (Fobe

et al., 1967).

When extracts of protein from green and roasted coffee beans were
hydrolysed they yielded 18 amino acids (Underwood and Deatherage, 1952;

Thaler and Gaigl, 1963). Supporting this evidence, Centi-Grossi et al.,

1969) showed that there was no difference between the species Arabica and
Robusta. Their amino acid analyses showed that individual protein-bound

amino acids have a different roasting sensitivity. Glutamic acid, leucine,

valine, phenylalanine and proline increase on roasting (Roffi et al., 1969),

whereas cystine, cysteine, methionine, lysine, serine and threonine were

destroyed (Thaler, 1963).

Clifford (197 5) commented that those amino acids that are destroyed
during roasting do not have simple aliphatic side chains, whereas those that

increase (with the exception of glutamic acid) have an aliphatic side chain.
 - 19 -

He suggested that those amino acids that appeared to increase were least

reactive and thus least destroyed. The increase was not necessarily an
indication of synthesis, more likely interference from other roasting

products or a failure to correct the results for dry matter loss. There
is little doubt that protein-bound amino acids with a non-aliphatic side
chain do react and are partially destroyed during roasting.

(iii) Non-protein nitrogen

In coffee, there are two major alkaloids - caffeine and trigonelline;

xanthine, threobromine and theophylline are also present (Fig. 4: Katz,

1980,

Caffeine, 1,3,7 -trimethyl-2,6-dioxopurine, not only exists in coffee,

but also in tea, mate, guarana, cacao and cola nuts (Bothe and Cammenga,

1980). Caffeine is the stimulant in coffee beverage. Its content varies

with coffee species: wild Coffea from Madagascar does not contain caffeine;
instead cafamarine was isolated (D'ornano et al., 1965). Arabica contains

approximately 1.07 caffeine, Robusta 2.0% and Liberica 1.5%; extreme

values of 3.0970have also been reported (Chassevent et al., 1969). Roasting

gradually reduces the caffeine content by sublimation (Navellier and Brunin,

1962).

Trigonelline is a methyl betaine of pyridine (Fig. 5 ). It is of particular

interest because it is related to niacin and shows some potency as a vitamin
(Dyke, 1965). Adrian et al. (1969) reported that when coffee was prepared
by the dry method, there was a higher content of niacin after roasting than

when wet processes were used.

Trigonelline accounts for 1% of green coffee (Arabica and Robusta). It

decomposes during roasting giving rise to niacin by demethylation, and in

severe roasts to pyridine (Viani and Horman, 197 5). The decomposition is

responsible for the increase in niacin content of the brew. Other degradation

products participate in coffee aroma (Adrian et al., 1967).

 -20-

R3
0
Rl,,, N

o' N N

R2

R1 R2 R3

CH3 CH3 CH3 Caffeine

H CH3 CH3 Threobromine

CH3 CH3 H Theophylline

Fig. 4 Purine Skeleton.

COON

i(.-H3
Fig. 5. Trigonelline
 - 21 -

Apart from trigonelline -a non-purinic base in green coffee, some

others (ammonia, betaine, choline, serotonin amide) are present in small

quantities (Viani and Horman, 1975). However, they are important in

respect of their breakdown products, during roasting which play a major

part in the aroma of coffee. Ammonia, betaine and choline are quite stable

at coffee roasting temperatures. Tepley and Prior (1957) reported that

0
betaine isomerized at 310 C to methyl dimethylaminoacetate, while choline
increased during roasting from 0.06 to 1.0% and forms trimethylamine only
at higher temperatures. Serotonin on roasting is easily broken down to a

series of compounds, alkylindoles and alkylindanes (Kato et al., 1971), which

probably contribute to the flavour of coffee.

Part of the physiological effect of coffee is due to its caffeine content.

Because of its ability to increase mental activity and wakefulness, refined

caffeine is used to help prolong wakefulness (Katz, 1980).

f) The phenolic compounds of coffee beans

The term chlorogenic acid (CGA) was first introduced in 1846 to describe

a major component of green coffee beans (Payen, 1846). This was later

characterised as 3-caffeoylquinic acid (3-CQA) ,

but in the modified numbering

system (NPAC, 1974), this has become 5-CQA. The literature on CGA

contains many confusing trivial names. Many of these were explained by

Clifford (1975) and Table 4 has been modified to comply with the latest IUPAC

terminology. So far as possible, all other references have been similarly

amended.

It has been confirmed that coffee beans generally contain quinic acid

esterified with p-coumaric, caffeic and ferulic acid (see Fig. 6). Esterifi-

cation occurs via one or more of three vicinal hydroxyl groups and thus gives
rise to at least 13 individual compounds.

Clifford (1979) recommended that collectively these are referred to as

CGA and suggests the following simple and unambiguous abbreviations for the

sub groups:
 - 22 -

Table 4. Chlorogenic Acid Nomenclature

(Modified from Clifford, 1975)

Trivial Name Systematic Name

Band 510 (1958) 4 -CQA

Chlorogenic acid (1846) 5-CQA (1932)

Cryptochlorogenic acid 4-CQA (1963)

Haus child's substance 3-CQA lactone (1963)

Isochlorogenic acid, Postulated 3-CQA in 1950. This was

Barnes (1950) refuted but 4-CQA was postulated in
1955. In 1964 it was reported to
consist mainly of 3-DCQA

Isochlorogenic acid A 4,5-DCQA

Isochlorogenic acid B 3,5-DCQA

Isochlorogenic acid C 3,4-DCQA

Neochlorogenic acid 3-CQA (1963)

(1963)

Pseudochlorogenic Mixture of caffeoyl and/or DCQA

acid (1955)
 - 23 -

Rl

Ho 0

RZ OH

Fig. 6 The Structure of the phenolic residue commonly present in

chlorogenic acids.

If Rl = R2 = H, p-coumaric acid

Rl = H, R2 = OH, caffeic acid

Rl = H, R2 = OCHS, Ferulic acid

4
H0

1 6 IUPAC System
......

6 Old System
GOON

OH

Fig. 7 The structure of quinic acid.

 - 24 -

1. caffeoylquinic acids (CQA) which are esters of caffeic acid with quinic

acid;

2. dicaffeoylquinic acids (DCQA) which are esters involving two residues

of caffeic acid attached to the same residue of quinic acid;

3. total caffeoylquinic acids (total CQA) - CQA and DCQA;

4. feruloylquinic acids (FQA) which are esters of ferulic acid with quinic acid

monoethyl ethers of CQA;

5. p-coumaroylquinic acids (CoQA) are esters of p-coumaric acid with

quinic acid;

6. caffeoylferuloylquinic acids (CFQA) - esters with one residue of

caffeic acid plus one residue of ferulic acid attached to the same residue of
quinic acid, monomethyl ethers of DCQA.

These substances are thought to influence the colour (Northmore, 1977),

flavour, odour (Tressl, 1977) and beverage quality (Amorim et al., 1974).
These hypotheses are discussed further on page. 37. -

Chlorogenic acids are almost universal in higher plants, and 5-CQA is

usually the major component. Few tissues are as rich as coffee seeds,

where the CQA level may exceed 107 dry matter basis (dmb). In some fruits,

4-p-coumaroylquinic acid seems predominant, and artichoke contains several

isomers, 1,3-dicaffeoylquinic acid, (1,3-DCQA) and 1,5-dicaffeoylquinic acid
(1,5-DCQA) (Nichiforesco, 1970), not found in coffee. Pineapple contains

1,4 -DCoQA (Sutherland and Gortner, 1959).

There is some evidence that artefactorial isomerisation (chemical, enzymic

or both) may occur when extracting plant tissues (Nichifores co, 197 0). It has

been said that 5-isomers are normally the 'parent'; it could be so, since to

date no enzyme has been isolated that will synthesise the 3- or 4-isomers.
However, neither has the enzyme responsible for synthesising DCQA been

isolated, but DCQA undoubtedly exist. Indeed the only CGA isomer isolated

from the South American plant Pterocaulon virgatum is 4,5-DCQA (Martino

et at,, 1979), and since this was under conditions that some workers say would
 - 25 -

cause isomerisation to at least 3,5-DCQA one must question whether

isomerisation does occur under such conditions. Clifford (Personal

Communication) has not been able to demonstrate such isomerisation during

extraction.

Green Robusta coffee beans contain 7 to 10% of chlorogenic acids

(Kunget al., 1965; Clifford, 1972; Rees and Theaker, 1977), whereas
Arabica coffee contains 5.5 to 8% (Kung et al., 1965, Clifford, 1972; Rees

and Theaker, 1977). The caffeoylquinic acids form the major fraction
(Arabicas 5.5 to 7%, Robustas 8%), followed by DCQA (Arabicas 0.6%,
Robustas 1.8%), and feruloylquinic acids (Arabicas 0.3%, Robustas 0.6 to
1.2%).

The effects of decaffeination and steaming were studied by high

performance liquid chromatography (HPLC) by Van der Stegen and Van Duijn
(1980). They reported that the total amount of mono-CQA was slightly

reduced without an increase of the fret. caffeic acid and that within the group

of mono-CQA the 5-isomer was clearly reduced and the 3- and 4-isomers were

increased; similar effects were observed for FQA.

These observations led these authors to believe that probably under high

temperatures isomerisation of CQA and FQA was possible.

Roasting causes progressive loss of the mono- and dicaffeoylquinic acids,

but it appears that the FQA are more heat stable and become the major CGA

fraction of heavily roasted coffee beans (Clifford, 1972; Rees and Theaker, 1977;

Van der Stegen and Van Duijn, 1980).

Published data suggest that some CGA degradation products are incorporated

into humic acids (see p 11 ). However, some degradation products are not bound

and these may be expected to influence the flavour and aroma.

Tressl (1977) reported that the level of relatively non-volatile dihydroxy

and trihydroxy phenols were Robusta 35. lppm, Arabica 34.6ppm and Arabusta
 - 26 -

20.7ppm. The relative position of Arabica is somewhat unexpected, but some

workers e. g. Rees and Theaker have commented that these three species have

very similar CQA contents. A complicating factor however is that the

trihydroxy phenols, e. g. 1,3,4, -trihydroxybenzene, 1,2,3 -trihydroxybenzene,

and some of the dihydroxy phenols (e. g. 1,4 -dihydroxy benzene) may be

produced from the quinic acid residue rather than caffeic acid residue, even
the quinic acid residue of FQA.

The major volatile phenols are guaiacol, 4-methoyl guaiacol and 4-vinyl
guaiacol and that these were present above their threshold values. Using

model systems, Clifford (1972) indicated that the guaiacols were almost

certainly derived from the degradation of the ferulic acid residue of the CQA.

Tressl (1977) reported that Robustas had a higher content of these

volatile phenols (56. Oppm) than Arabusta (38. Oppm) and Arabica (31.1ppm),

and that the concentration rose significantly with increased severity of

roasting. These relative concentrations parallel the observed levels of the

FQA precursors.

Fujimaki et al., (1974) attributed the 'sweetish smoket'' organoleptic

properties to this type of guaiacol.

g) Non-phenolic acids

The non-phenolic acids have received less attention than CGA. Clifford

(1975) reviewed the information available and since there have been no further

significant reports his comments are summarised here.

Green coffee beans contain acetic, butyric, citric, malic, oxalic,

propionic, pyruvic, quinic, tartaric and valeric acids. The concentrations of

malic, oxalic, pyruvic and tartaric acids are between 0.2 and 0.5%
citric,
(dmb) and total about 1.5% (dmb).

Roasted coffee has been reported to contain in addition citraconic, formic,

fumaric, 2-furoic, iso-valeric, itaconic, lactic, maleic and mesaconic acids.

 - 27 -

Formic and acetic acids increase during roasting, whereas citric and

malic acids were reduced by 33 to 56%and 16 to 407 respectively.

h) The volatiles of coffee beans

In 1969 Merritt and colleagues commented that developments in

methodology - extraction, separation, characterisation, had permitted

accelerating progress to be made despite the great complexity of coffee aroma.

After a short pause this progress has continued in the late 1970's and early
1980's with valuable contributions by Vitzhum and colleagues, and Tressl.

The volatiles of green coffee have received some attention from Merritt

et El . (1969) and Vitzhum et al. (197 5) . Over 100 volatiles have been
identified and Vitzhum considers methoxy pyrazines primarily to be responsible
for green coffee aroma. Although some green coffee volatiles may contribute

to roasted coffee aroma, it is generally accepted that roasted coffee aroma

develops as a result of complex reactions during roasting. Much more is known

of the products than of the precise reactions leading to their formation.

In 1968 Walter and Weidemann reported that 363 volatiles had been

identified in roasted coffees. In his review, Clifford (1975) indicated that the

total had risen above 400 and in 1981 the total exceeds 515 (Smith, 1980; Tressl

et al., 1981). However, not all of these contribute equally to coffee aroma
(Clifford, 1975). The threshold value (the concentration at which it is perceived)

and the odour value (the ratio between its concentration in the food and its threshold

value) are not always available for coffee volatiles. Thus, as Clifford points out,
it is difficult to give the potency of coffee aroma constituents quantitatively.

Clifford reported in 1975 that moderately volatile components were thought

to be particularly important. This fraction is divided into phenolic compounds,

carbonyls, pyrazines and sulphur-containing volatiles.

In 1970 only 43 S-containing volatiles were known (Weidemann and Molir)

but by 1981 Tressl and Silwar reported the total exceeded 100. These recent

the importance in coffee

studies have confirmed of sulphur containing volatiles
 - 28 -

aroma and coffee staling. Other groups of coffee volatiles to have received

attention are lactones (see Maga, 1976), simple phenols and phenolic

compounds (Tressl, 1977, see Maga, 1978), pyridine (see Maga, 1981a) and

oxazoles (see Maga, 1981b).

The data obtained from the investigation on the phenolic volatiles of

coffee by Tressl have been discussed in detail on page 26 in the section
dealing with CGA.

Gutman et al. (1977) compared the sensory properties and GC finger-

prints of roasted Arabica, Arabusta and Robusta coffees. Both methods
indicated that Arabica and Arabusta were similar and quite distinct from
Robusta. The Robustas contained a large number of sulphur compounds but

at lower concentrations than in Arabicas and Arabustas. The furans also

were less concentrated in Robustas. Tressl et al. (1981) reported

furfurylpyroles and more alkylated pyrroles in Robusta compared to Arabica.

Some of these pyrroles increased with ageing and deterioration of roasted

coffee.

Radtke-Granzer and Piringer (1981) demonstrated decreases in the

content of several volatiles during storage of instant coffee which they

associated with decline in sensory quality.

Formation of volatiles

The quality of the aroma of a coffee beverage depends on the type of

coffee variety used for preparation of the beverage and is also determined
essentially by time and temperature of roasting process (Baltes, 1975).

Roasting coffee causes a multitude of complex reactions. The reactants

and products as a whole are known, but rarely have the precise pathways been
determined. However, much useful data have been obtained by the use of

roasting-stimulating model systems and one can postulate three major types

of reaction:
 - 29 -

i. the thermal degradation of green bean components, e. g. polysaccharides

to yield, e. g. aliphatic carbonyls, alcohols, acids, furans and cyclic
diketones (Gautschi, 1967). Work by Clifford (1972) and Tressl (1977) suggest

that volatile phenols arise from the degradation of CGA. The possible

relationship between relative levels of volatile guaiacol in roasted beans and

the levels of their probable precursors have been discussed in page 26.
However, in his review, Clifford (1975) comments that other precursors, e. g.
phenolic amino acids, sugars, celluloses and lignin are possible.

ii. The interaction of green bean components to yield larger and smaller
molecular mass products, only some of which are volatile, e. g. sugars and
amino acids in the classic Maillard Reaction and Strecker Degradation, both

of which have been recently reviewed by Nursten (1981). The volatile products
include carbonyls and a host of simple heterocyclics such as pyrazines and
furans. Variations in sugar content between Arabicas and Robustas (see
Table 1) could well account for the lower furan content of roasted Robustas

(see page 28).

iii.. Reactions involving the degradation products arising from the reactions

mentioned above. It would seem that such reactions are important routes for
the formation of N- and S-containing heterocyclics, e. g. see Schutte (1974),

Tressl et al (1981) and Nursten (1981), Maga (1981b) such as pyrroles, furfuryl

pyrroles, thiophenes and oxazoles. The lower furfuryl pyrroles content in

roasted Robustas compared to roasted Arabicas might once again be related to
the lower content of sugars in green Robustas.

III Green Coffee Processing

Coffee processing can be divided into two major operations:

1. Those processes that take place in the producing countries (Fig. 1 p. 4)

2. Those processes that take place in the manufacturing (consuming)

countries (Fig. 8).

 - 30 -

C)

0
U

cn u
P,
1-4 cd
ö
a) `ao

cd ri)
a)

O
U

1. bo
E

r.
0

.o U

a)
a)
w
Ö
bo
U
w
O
m
8
94 U)
Q)
U
0

a
03
0 C)

cý
a)
w w
w
cd
U
c)
Q

bO
0
b
Ü a /
cd v]
 31

The green coffee beans are harvested when red-ripe. After

harvesting the outer layers are separated from the seeds either by the

wet or dry processes.

The wet process: This involves the following stages:

(i) Pulping removes the outer layer of the cherry from the bean. This
takes place in the presence of water, and involves a mechanical tearing and
squeezing operation. The machines leave a mucilagenous layer which is
,
removed traditionally by fermentation.

(ii) Fermentation of the mucilage takes place in large tanks over 24 to 40

hours. However, the fermentation in the best
procedure which results
quality, as well as allowing a reasonably convenient and rapid factory routine
is the two-stage 'dry' fermentation process (Wootton 1971). During the
first stage the mucilage is degraded and during the second stage it is soaked
in water for 24 to 48 hours.

A major drawback of the two-stage fermentation method, which is also

a feature of underwater fermentation generally, is that the coffee parchment

shows increased tendency to crack during subsequent drying (Kulaba, 1979).

Nevertheless, this process enhances the final coffee quality, through

modification of liquor characteristics. The raw bean has a greatly improved

appearance, which becomes more evident during the drying process.
Beverage quality is also improved. This effect has been attributed to the loss
by diffusion of some otherwise deleterious water-soluble chemical components
possibly polyphenols or caffeine (Wootton, 1979).

Fermentation can be accelerated by the use of enzymes (Butty, 1973;

Ehlers, 1980) without adverse effects upon the final coffee quality.

After fermentation, the coffee is known as 'parchment coffee', since the

seed retains its endocarp layer. It must be dried to about 12% moisture

content to ensure stability.

 - 32 -

(iii) Drying : Parchment coffee is generally sun-dried. Mechanical driers

are, however, used (Gibson, 1971) to hasten the process and ensure that the

drying is more uniform.

(iv) Hulling : The dry parchment coffee is hulled to remove the dried
parchment layer and also the testa or silverskin layer.

The dry process : The cherry is dried from about 709voto 10% moisture in

the sun, or more rapidly by hot air driers. Microbial spoilage is at a

minimum when the latter method is used, and it is therefore preferred.

Rolz et al. (1969) reported on the use of fluidized beds in the drying of

coffee cherries. They found that the best quality coffee was obtained when
drying was done in two stages, an initial period at a low temperature (20°C)

followed by a longer one at a higher termperature (60°C).

After drying, the cherry is dehusked to separate the seeds from the
outer layers. With the dry process, particularly applied to Robusta, the

silverskin is difficult to remove and the bean may have a distinct and less

acceptable appearance. If necessary, polishing in the presence of added

moisture may be used to remove it, hence the expression, 'washed and

cleaned' is used to describe such beans. Green coffee beans from either of

these processes are graded and sorted.

The coffee beans are passed along a rotating horizontal sieve, with

holes or bars varying to permit the recognised sizes of beans to fall through

as they proceed from one end to the other. Defective beans are also removed.

Sorting takes place both in the producing and manufacturing countries

to achieve high quality coffee beans, and is carried out by either mechanical

or optical means.

In the mechanical method, defective beans are hand-picked and fed into

air classifiers (catadors) where they enter an adjustable rising current of

 - 33 -

air. The dense beans fall through the air current whilst those that are less

dense are carried upwards to be released when the air speed is reduced.
Alternatively, gravity classifiers may be used, where the beans are passed
over a vibrating table which has a porous woven wire cover. A current of

air floats the beans. Separation is achieved by adjustment of the table

angle, by air pressure and direction of air flow, and by amplitude and
frequency of vibration. If the machine is correctly adjusted, separation is

usually good for removing shells, unhulled cherry or parchment, and

withered beans.

In the optical method, sorting is carried out by electronic assessment

of colour under visible and/or ultraviolet light. The electronic colour
sorters are either monochromatic or bichromatic. The monochromatic

machines are the simplest sorters, making measurements at a single wave-

length.

In the bichromatic machines, the ratio of reflectivity is measured at

two separate wavelengths. This is necessary when there are subtle colour
differences in the case of sorting green Arabica coffee (Maughan et al., 1980).
as

Roasting: this is essentially a two-step process. As the temperature

of the coffee beans is raised by hot gases, it first dries, then roasts. Beans

are roasted in batches at an air temperature of 180°C-200°C for approximately

20 minutes. However, being an exothermic process, the temperature within

the bean may be considerably higher as roasting takes place.

At first, free and bound water are driven off. The green coffee bean

colour slowly changes to buff, then light brown as drying continues. As the

bean temperature approaches 200°C, pyrolysis, the second step occurs. The

bean expands due to internal pressure and the chemical changes thus occur

above atmospheric pressure. As soon as the desired bean colour is readied,

the beans are removed from the heated gases and promptly cooled by ambient

air or a water spray. Most of the water sprayed evaporates off, cooling the
beans, with hardly any water being absorbed by the beans. Cooling of the

roasted beans stops the pyrolysis.

 - 34 -

In general, light roasting leaves more acidity in the bean and the

roasting weight loss may be only 14%. 'Dark roasting' leaves little acidity

or aroma and has more extensive cell destruction making the extraction of

soluble coffee easier.

Roasting loss is defined as the percentage loss in weight from wet

green bean to unquenched roast bean. Pyrolysis loss is defined as the

percentage loss in weight from dry green bean to dry roasted bean.

Instant coffee is produced by grinding the roasted beans, followed by

brewing and drying of the soluble solids and aromatic components.

IV Theories about Coffee Quality

On the basis of experience subjectively assessed characteristics are

linked with beverage quality. The criteria chosen vary geographically and
the descriptive terms adopted often have meanings different from those

placed upon them by the layman.

One can argue that certain chemical, physical and physiological

characteristics must have a more direct connection with beverage quality.

There have been attempts to link such characteristics retrospectively to

quality judged initially by subjective means.

Coffee beverage is consumed for the pleasure, satisfaction and stimulation

it gives to the consumer, through its flavour, aroma and desirable physiological

and psychological effects. Good quality is therefore the 'key' to this intangible

experience of pleasure that a cup of coffee arouses (Stirling and Jackson, 1979;
Illy, 1980, Personal Communication). The ultimate beverage quality is

influenced by the practices of the producer (planter), buyer, processor and

consumer, for assessment of green bean quality varies from one producing
country to another.

In Kenya coffee quality may be associated with the bean size (Munene, 1973).

The following descriptions of coffee standards are used by the Liquoring Department
 - 35 -

of the Coffee Board of Kenya.

Bean Size Standard

Arabica. 'AA' 1 largest
Arabica 'A' 2
Arabica 'AB' 3
Arabica. 'C' 4 smallest
Arabica 'C/TT' 5 mixture of large and small
beans, broken and shells;
however no blacks.
Arabica. 'TT' 6 Brokens, defects and blacks.

It has been reported that quality in Kenya coffee is associated also with
the colour of the green bean. The colours normally found in the green beans

are blue, green, yellow and brown (Kulaba, 1978). The best quality beans

are predominantly bluish in colour, whilst the yellow or brown ones have
poor liquoring characteristics. Munene observed that should there be
difficulty in classifying an out turn, the taste assessment by professional
liquorers is considered as important and the coffee classified accordingly.

In Brazil, as in the other South American countries, coffee quality is

classified with the beverage taste assessment. The scale is soft (best

quality), hard, rioy and rio (poor quality). Rio coffee is less expensive.
However, some consumers in Brazil as well as Latin America, United States

of America (U. S. A. ) and Europe prefer this kind of coffee because of its
strong medicinal or phenolic flavour (Amorim et al., 1974). This

characteristic flavour is produced when coffee is harvested in wet climatic

conditions and so fermentation takes place before drying.

Guatemalan coffees are graded according to the elevation and areas

where the coffee is grown (Basu, 1977):
 - 36 -

Antiguas (altitude of/from 4500 to 5000 ft)

Strictly hard beans (5000 ft and higher)

Hard beans (4000 to 4500 ft)

Semi-hard beans (average altitude of 3800 ft)
Extra Prime Washed (3000 to 3500 ft)
Prime Washed (2500 to 3500 ft)
Good Washed (low grown coffees)

These grades are presented in declining order of quality.

Green coffee is exported in bags and therefore any classification or

standardisation must relate to a sample taken from these bags. Most

consuming countries issue a set of regulations for controlling the import

of green coffee beans (International Coffee Organisation, 1962).

In general, the importer must define what he has in the bag, i. e.

commercial coffee beans of the coffee plant. Bags must be labelled with the

country of origin, grade or type and the species (e. g. Arabica or Robusta).
The composition of the beans must also be given and particularly the moisture

content (12970
maximum). Sometimes the caffeine content is also given

according to the species. 'Defects' must be given. This is used to describe

the amount of defective beans and foreign matter present in a sample. Usually

coffees from Kenya and Colombia have a very small amount of defective beans

(Clarke, 1979). The main types are listed below:

Dried coffee cherry

Broken beans
Bean fragment
Black bean
Semi-black bean
Insect-damaged bean
Stinker
Immature bean
White bean
Withered bean
Sour bean
 - 37 -

Sometimes physical counting of the number of defects in a sample is used to

assess the grade (e. g. U. S. A in llb, or in 300g weight in countries with a
metric system). The total number of equivalent defects is determined and

maxima-minima established for each grade, (i. e. NY4 means 30 maximum

defects present and Brazil, 26 defects in 300g sample). The Portuguese

and French have a similar system with grades known as Extra Prima (15),
Prima (30), Superior (60), Courante (120), and Limite (180). Furthermore,
Clark (1979) indicated that bean size, colour and residue from pesticide also
contribute to the grading of coffee beans.

Ehlers (1980) indicated that when enzymes were used to depulp coffee

there was an improvement in the quality. She produced heavier coffee beans

which were clean, and without the risk of 'tainting' and quality loss through

microbial spoilage.

Amorim and Silva (1968) reported a relationship between the polyphenol

oxidase activity of green coffee and the quality of the beverage. Sanint and

Valencia (1972) confirmed this observation and explained that the higher the

green bean polyphenol oxidase activity, the higher the beverage quality.
Furthermore it has been suggested that acids such as chlorogenic acids and
caffeic acid act as antioxidants for aldehydes.

The higher the aldehyde content of the green bean, the higher the

beverage quality (Forsyth, 1964). This seems to suggest that when quinones

are formed by enzyme activity in the bean, this renders the aldehydes

unprotected and they are lost causing a reduction in bean quality. The quinones
inhibit the enzyme and cause the low activity associated with low quality beans.
(Amorim et al., 1977)

Also the chlorogenoquinones and caffeoquinones formed by the action of

this enzyme on chlorogenic and caffeic acids react with amino acids (except
lysine and cysteine) primarily through their alpha-amino group to give red or
brown products (Pierpoint, 1969; Synge, 1978).
 - 38 -

It has been suspected that high storage temperatures and high bean

moisture contents have been the major factors influencing quality loss in

stored coffee. Stirling (1980) was able to show that for coffee to be preserved,

it must be kept cool and dry in storage.

Mechanical and chemical injuries caused by micro-organisms affect

the plant metabolism, inducing the production of more phenolic compounds
(Uritani, 1964; Kuc, 1964).

Amorim et al. (1977) observed that coffee deterioration was caused by

an effect of the environment on the membranes of the seeds. They indicated

that membranes must be the first place in the green coffee beans to undergo
chemical and structural changes leading to quality deterioration.

Wurziger (1977) reported that green coffee beans infested by 'Coffee

Berry Borer' have a higher than normal CCA content. This was associated

with low quality in green beans. Marigo and Boudet (1979) confirmed that the

subjection of plants to stress generally led to the accumulation of caffeoylquinic

acid (CQA). Also, Legrand et al. (1978) reported that tobacco infected with
'Tobacco Mosaic Virus' showed an increase of activity in the phenol methylating

enzymes (i. e. increased potential for the conversion of CQA to FQA) probably as

an early step in cell wall lignification. This lignification is looked upon as a

means of isolating the virus and preventing its spreading through the plant. It

would also increase the potential for the formation of volatile guaicols during

roasting.

Fujimaki et al. (1974) indicated that lightly roasted coffee should not be

particularly smokey and a high content of such guaiacols therefore appears to

be undesirable and this in turn suggests that green beans with high FQA contents

will be of lower quality.

Northmore (1965) related the colour of the green bean to the quality of

the beverage. He reported that the oxidation of alkaline magnesium

chlorogenate produced a blue colour similar to that of the good quality green
bean. The green bean colour has been linked also with the presence of
 - 39 -

cafesto 1, kahweol and chlorophyll (Gibson, 1971). Wurziger and Harms

(1969) reported that the browner the green bean, the lower the 5-hydroxy-
tryptamides content. Possibly the brown colour arose from CGA oxidation

which was facilitated by the lower content of the anti-oxidant, 5-hydroxy-

tryptamide.

Amorim et al. (1977) reported that phenolic acids and lipids were
found throughout the bean, but at a greater concentration in the outer layer

of the endosperm. They concluded that these two components were linked

with the colour of the green bean.

When coffee beans are roasted, gases are formed. These are initially

retained and normally the bean expands, i. e. the roasting reactions can be

said to occur in a pressure vessel. If the bean is physically damaged, it is

thought that the normal reactions do not take place since the normal internal

temperatures and pressure are not achieved. It is also thought that structural
differences at the cellular level may render the Arabica pressure vessel
different from the Robusta's. If this hypothesis is correct, it could, in part,

account for the differences in quality between Arabicas and Robustas.

However, differences in the nature and quantity, particularly in the

ratios of reactants, i. e. green bean components, must also be important in

accounting for such differences in quality. The major compositional

differences have been summarised in Table 1; composition and the manner
in which these differences might influence the aroma composition and the

coffee quality have been discussed on pages 28 and 29.

Villar and Ferreira (1971) found that Brazilian soft coffee (good quality)
had lower percentages of total CGA as compared to other classes of coffee.
However, Amorim et al. (1973) stated that Robusta coffee had more CGA than

the Arabica. This fact alone was not sufficient to account for the difference in

taste between the two species.

 - 40 -

V Analysis of Chlorogenic Acids (CGA) in coffee

Analysis of CGA is preceded by the extraction of ground coffee beans

into a suitable solvent (water or organic solvent).

(a) Extraction of CGA from coffee beans

Coffee beans must be ground to pass through a sieve having at most a

0.7mm aperture following the method used by Clifford (1976) and Lyons
Central Laboratories (Report No. D. 1852,1977). This ground coffee is
soaked and extracted with 70% 2-propanol for 30 minutes as recommended
by Clifford (1972). The resulting suspension is allowed to settle and is
decanted. The residue is re-extracted four times, the extracts bulked and
diluted as necessary with 7 0J 2 -propanol. Lyons have shown this system

to be exhaustive and to recover 98 to 101% of added 5-CQA.

(b) Identification of CGA in coffee beans

Many procedures for determining CGA and other phenolic compounds

in coffee and other natural products use separation of these compounds by
paper chromatography (PC) (Harbourne, 1967; Walker and Lee, 1968; Schulz

and Hermann, 1980). Paper chromatography lacks the resolution, speed and
accuracy needed for fast reliable analysis of complex mixtures. Thin layer

chromatography (TLC) offers greater resolution and, speed than PC, but lacks
the quantitative accuracy (Harbourne, 1973). Clifford (1974) reported on the
use of TLC with poly-N-vinyl-pyrrolidone (PVP) as the absorbent. He was

able to separate sub groups of CGA, but did not quantify them. Column and

gas chromatography (GC) have also been used (Kung et al., 1967; Andersen

and Vaughan, 1972; Nagels et al., 1979) in the analysis of these complex

mixtures. Gas chromatography is a fast, efficient and accurate technique

but CGA requires a quantitatively reproducible derivatization step. High

molecular mass compounds generally require high injection point and column
temperatures which increase the risk of thermal degradation.
 - 41 -

High pressure liquid chromatography uses columns packed with small-

diameter (maximum 51An) particles, coupled with high-pressure pumping

systems and sensitive, accurate detectors. The HPLC technique rivals GC

and in addition, can be applied to non-volatile substances, thus largely

avoiding the limitations of molecular mass and thermal degradation referred

to above.

High pressure liquid chromatography has been used extensively for

separation of phenolic compounds including CGA (Murphy and Stutte, 1978;

Wulf and Nagel, 1976; Proksch et al., 1981). Separation and quantification

of CGA was reported by Rees and Theaker, 1977 and Van der Stegen and Van

Duijn, 1980. The individual CQA, DCQA and at least one FQA have been

resolved.

The drawback of these procedures is the time to analyse each sample

(20-50 minutes) and the difficulty in characterising peaks because of the

absence of commercially available standards. However, the method has a

good degree of reproducibility.

Apart from chromatographic procedures, quantification of CGA in

coffee has been carried out based either upon their absorbence maxima in

the ultraviolet (UV) region, or on the formation of coloured complexes or

derivatives (Moore et al., 1948; Weiss, 1953; Clifford and Staniforth, 1977).

Clifford and Staniforth (1977) discussed the use of six reagents in the

spectrometric measurement of the CGA content of coffee beans. Out of the

six reagents, three were rejected and periodate (0.25% aqueous sodium

periodate) was most successful. This reagent reacts with the CQA, FQA,
DCQA and caffeoylferuloylquinic acids (CFQA), i. e. some 987 of the , total CGA.
Best results were obtained by controlling the time and temperature of the colour
producing reaction, 10 minutes at 27°C being the optimum.

Clifford and Wight (1973) explained that aqueous periodate oxidises the o-
 - 42 -

or p-dihydroxyphenols to their corresponding benzoquionone and water.

The benzoquinone may then react further to give a naphthoquinone, and

this mixture of products yields a yellow-orange colour. The treatment

of monomethyl ethers with periodate removes the methyl group as methanol

and produces the corresponding dihydroxyphenol which enters the reaction
described above. FQA are the monomethyl ethers of CQA and thus this

method can be used to measure both groups simultaneously.

A limitation of this method is that it overestimates the DCQA. Each

molecule of DCQA (M. Wt . 516) is measured as two molecules of CQA

(M. Wt. 2x 354 = 708). Therefore overestimation is by:

(2 x 354) - 516
_ 37%
516

The periodate reagent gives an equal response on the basis of the

caffeic acid and/or the ferulic acid content to individual isomers of the CQA,

FQA, DCQA and CFQA. Free caffeic and ferulic acids are detected by this

reagent, but the molar response is only 25% that of CQA and FQA, and

interference is slight because their content in coffee is low compared to CGA.

Another promising reagent used by these authors was the molybdate

reagent. This was based upon the reagent reported by Swain and Hillis (1959).
At pH 6.5 molybdate yields a yellow colour (A max 37 0mm) with those CGA

having a caffeic acid residue (i. e. CQA and DCQA). Normally, 5-CQA is used

as a reference standard, but because the colorimetric response is independent

of the caffeic to quinic acids ratio, this practice also leads to an overestimation

of the DCQA fraction. If caffeic acid is present it also interferes, but since

the molar response is only 60% of that of CGA' and its presence is small in most

extracts, interference from this source is not serious.

Finally, the use of a thiobarbituric acid (TBA) reagent was reported.

This reagent had poor reproducibility in the determination of quinic acid in

coffee beans (see Table 5 ). If modifications and standardization of conditions

 - 43 -

0
...
Cd
b Cl, co Cl)
V N41-4
NX
b
Cd

O
0 i5
0-%

U
X ' ýp tf) Cl)

IC di Cl) 9 A
118 r-
00

aý cd O
E
'A Q V

'b O
if) to b0
crý cl
0 M O
O
O O O H

1
O
m
Qi
U) -:4
C4, cd
.0
ö
b0

o `d
b ýc
U w
`ý LO C14 to

Ö L() N 1
öý N
",1
R

b
Cd
41
Cl C.) H
04

cd Ö Ö
U Ö
O cd ý4
O O O O H
rý
.cl O

Cd
a)
cd
O
a w

0-1
'd
4-1 a)
If) aý

aý .g b
ý. U
Cd
H 04 P64
 - 44 -

of analysis allow this problem to be overcome, then this method has considerable

potential because the TBA reagent measures total quinic acid after hydrolysing

the CGA overnight with sodium hydroxide. Since all CGA contain quinic acid

one assumes that all CGA will give an equal response on the basis of their
quinic acid content. Clifford and Staniforth reported that to measure quinic

acid released from CGA by saponification, it was necessary to examine

extracts before and after saponification to correct for the presence of free

quinic acid.

(c) Estimation of chlorogenic acid fractions

FQA : Clifford (1976) reported that the molybdate and periodate

reagents differ in that the periodate reagent detects FQA, whereas the molyb-
date does not. This specificity may be expressed:

Molybdate (MV) = CQA + 1.37DCQA (1)

Periodate (PV) = CQA + 1.37DCQA + FQA (2)

and thus FQA = Periodate - Molybdate.

This relationship has been borne out by studies using model systems
(Clifford and Wight, 1976), and by comparisons with HPLC methods (Rees

and Theaker, 1977) when analysing coffee beans that are not discoloured.
However, where there is discolouration which might have arisen via polyphenol

oxidase (PPO) oxidation of CGA this relationship might no longer be valid

because benzoquinones produced from CQA by polyphenol oxidase are probably

similar to, if not identical to, the benzoquinones produced by the periodate

reagent. Thus these quinones might react with periodate but having lost their

1,2-dihydroxy structure would not react with molybdate. In discoloured beans

the difference between the two reagents may not be just FQA, but FQA and

quinones.

DCQA : Clifford and Staniforth (1977) have reported that the TBA reagent
detects DCQA as one molecule of CQA, whereas periodate reagent detects DCQA

as two molecules of CQA. This specificity may be expressed as:

 - 45 -

TBA = CQA + o. 69DCQA + FQA + CoQA (3)

from equations (2) and (3)

PV TBA 1.47 (-D CC? A) 1.47CdQA

- = -

If the CcQA content is very low (Rubach, 1969), then:

DCQA = 1.47 (PV - TBA) (4)

CQA : Substituting the estimated DCQA content (equation 4) into

equation 1 permits an estimate of the CQA content, i. e. CQA = Molybdate -

1.37DCQA. One must remember that the estimates of FQA, DCQA and CQA
quoted above are subject to progressively increasing accumulative errors of
the methods used (Clifford and Wight, 1976). Therefore, the search for

other more accurate and simple methods of estimating these CGA isomers is

appropriate.

VI Aims of the Present Investigation

It is clear from the literature survey that in the field of coffee chemistry,
many fascinating problems await investigation. However, bearing in mind
the facilities and expertise available, two areas stand out:

1. there are numerous experimental data and some hypotheses that

suggest a link between CGA and several aspects of green bean, roasted bean
or beverage quality;

2. there is a need to investigate critically, refine and integrate

several colorimetric and chromatographic methods of CGA analysis.

Accordingly, these two areas were selected for investigation and it was
proposed to approach these objectives by:

a) applying consecutively these various analytical methods to defined

CGA and potential interfering substances;
 - 46 -

b) establishing analytical norms for the CGA content of typical good

quality commercial green beans; and

c) applying these same methods to atypical green beans that appear

in commercial batches.

In this respect, immature green beans and peculiarly coloured green

beans were chosen for investigation.
 - 47 -

CHAPTER TWO

METHODOLOGY
 - 48 -

Introduction

Most of the colorimetric methods used were those reported by

Clifford and Staniforth (1977). Some of the more promising methods of
analysing CGA were practiced, repeated and, using the method of Steiner (1967),

their repeatability and reproducibility were calculated. The methods were not

accepted until these values were equal to or better than those previously
published (Table 5). Any new method would have to perform comparably or

offer some particular advantage before being adopted.

Extraction of CGA from the coffee beans

The seeds are dipped in liquid nitrogen until boiling ceased and then

ground to pass through a sieve with a 0.7mm aperture. The ground material

is extracted by shaking a 1. Og sample in a screw-capped tube with 70%

2-propanol (BDH, England) for a minimum of 30 minutes. The suspension is

allowed to settle and is decanted onto a Whatman No. 1 filter paper. The

residue is re-extracted five times, the filtered extracts are bulked and diluted
to lOOmL with 70% 2-propanol.

I. Colorimetric Methods

Molybdate reagent (pH 6.5)

The reagent is prepared by dissolving (1) sodium molybdate -

Na2MoO4 ' 2H20 (16.5 (2) disodium hydrogen phosphate - 2H20

Na2HPO40
g),
(8.04g) and (3) sodium dihydrogen phosphate - NaH2PO4-2H20 (7.93g) (BDH,
England) in about 800mL distilled water and diluting to 1 litre. The pH value
is checked and adjusted if necessary.

An aliquot of each sample (0.2mL) is added to 10. OmL of molybdate

reagent in a stoppered test tube and mixed thoroughly. This solution is

examined spectrophotometrically against a blank (0.2mL sample + 10mL

buffer, pH 6.5) at 370nm. The buffer is prepared by dissolving sodium
hydrogen phosphate and sodium dihydrogen phosphate at 8.04g and 7.93g

respectively in one litre of distilled water. Most samples were examined in

triplicate. The repeatability and reproducibility of molybdate reagent are

 - 49 -

shown in Table 8. They are slightly better than those previously reported.

Periodate reagent
Aqueous sodium metaperiodate (BDH, England) is prepared at 0.25%
(w/v). The stoppered test tubes containing 10. OmL of the reagent are placed
in a waterbath (270C) and the reagent allowed to stabilise at 270 C. This is

confirmed by using a thermometer and an additional tube of reagent.

The extracts (1. OmL) to be analysed are added to these tubes while

they are still in the waterbath, the contents mixed immediately and thoroughly
by shaking and the test tubes returned to the waterbath. The absorbance
(> 406) is read ten minutes after mixing against a blank of distilled water
max
containing 1. OmL of the coffee extract. All samples were examined in

triplicate. The repeatability and reproducibility of the periodate reagent is

shown in Table 8. They are better than previously reported.

Thiobarbituric acid (TBA) reagent

The extract (1. OmL) is saponified overnight with 1. OmL of 1M NaOH
(BDH, England) and 1. OmL HC1 (BDH, England) is added to neutralise the reaction
in the morning.

One millilitre of pH 1.4 buffer (iM toluene sulphonic acid -i- 1M toluene

sulphonic acid sodium salt ; BDH, England) is added, then shaken. Then

1. OmL of 0.8M periodic acid (Sigma, U. S.A. ) in sulphuric acid is added, and
held for 20 minutes at room temperature. This step was later replaced by
holding at 37°C (see p. 54). Sodium arsenite (1. OmL of 47; BDH, England) is

added and the contents shaken to destroy excess periodic acid.

This is followed by the addition of 1. OmL of 2% TBA solution with

shaking and, the colour is allowed to develop for 20 minutes in a boiling

,
waterbath. The mixture is allowed to cool and is made up to 25. OmL with
distilled water. Absorbance is read at 549nm against a blank which is
max
put through all the stages of analysis except for the addition of TBA solution.
Water is used to replace TBA solution in the blank test tubes. All samples

were examined in triplicate.

 -50-

to

ce

1o ý tf)
ö ö 0 0
`d ö ö ö C
to a
W +1 +1 I 1 +1 1 +1 U
O
N
°w l d+ w N
a " . "
w
N O O O O
Qi
CSi

O di
r-
M

0 O
-4 m
N
4-1
cd O
'C7 41

0 O
-14
O O O O 0 O
.U b Ö Ö Ö Ö C
O O
x
+1 +1 +1 +1 +1 +1 +1
W
U . -I r-1 r-i r-1 u ýf r-1
;,8 4-1 d4 dl d4 dI d+ 14 ce
)
oO O O. O O O O" O" b
,
H c3, u2
U e
4)
C) 4-1 PCJ

OH

3 E ' ö ö ö *
'U ý
".C) O O Cl O wO
0 ö ö ö ö w
U W
+1 +1 +1
.O
, 1 1 1 +1

ö Co 00,
Lf) LI) O
0 Co Co Co
N ö ö ö ö
a
ö
ºa Cd CSi

(U CL c)
C)
C)
a H
U
.v
Ü
E
0
0N
"" O N
0
U q ý 41
O
+' V O O O O O O
"ýý Ö Ö
,N O O O C O
Ä V :: +t +1 +1 +1 +1 +1 +1
0 N C N N N C
a.+ C14
U.. cri
Ö Ö Ö Ö Ö Ö Ö
. "0
I

0 4
ýÖ blo

5 N M 'o N
Ä C,
to a)

E"+ 0 +
 - 51 -

Figure 9. Molybdate Reagent Calibration Curve for CGA

E370

0.

0.

0. '

0. E

0.5

0.4

0.3

0.2

0.1

25 50 75 100 12 5 175
mg/250mL CQA
 - 52 -

Figure 10. Periodate Reagent Calibration Curve for CGA.

E406

0.40

0.35

0.30

0.25

0.20

0.15

0.10

0.05

ý0 40 60 80 100 120 141)

Concentration (mg/250rL CQA)

 - 53 -

Figure 11. TBA Reagent Calibration Curve for CGA.

E549

0.8

0. i

0., 6

0.5

0.4

0ý
.3

0. cl

0.1

25 50 7' IO jet- I1 75

Concentration (mg/250mL CQA)

 - 54 -

When the TBA reagent was used as reported by Clifford and Staniforth
(1977) it was noted that the results were not reproducible. They had indicated

that the TBA reagent method needed improvement. The cause of this variation

was traced to inadequate temperature control of the periodic acid oxidation

step; therefore an experiment was carried out to find the optimum condition
for this step.

An experiment to find the optimum conditions for periodic acid oxidation step

Method

The TB.A reagent method was carried out using a synthetic quinic acid

at 0.4mg/mL 700 2-propanol concentration as previously reported except for

the oxidation step with periodic acid. This step was held at room temperature
37°C for various times (20,30,40,50,60,120 minutes and overnight).
and
The temperature of 37°C was chosen, because it was easy to operate at this

temperature.

This experiment was repeated a week later to see if the results could

be reproduced.

Results
In the first experiment (Table 6) at room temperature almost half of

the quinic acid was not accounted for, compared with the recovery of quinic

acid oxidised with periodic acid at 37°C. This was so at each time studied.

The repeat of the experiment (Table 6) shows that when room

temperature increased to approximately 22°C, there was an improvement in

the recovery of the product of oxidation of quinic acid over that at approximately
19°C. The data from both experiments show that the oxidation product of
quinic acid recovered at room temperature increased with time while that at
37°C remained constant. This implies that oxidation of periodic acid is

temperature dependent and 37°C for 20 minutes was chosen, because this

combination was highly reproducible (Table 8) and convenient.

The calibration curves were constructed using several concentrations

of 5-CQA. The results are presented graphically in Figures 9, lCand 11

 - 55 -

Coffee bean extracts were interpreted by reference to these calibration

curves.

Moisture Content
The moisture content was taken to be the weight loss after heating the

coffee bean powder to constant weight at 105°C.

III HPLC Method

Rees and Zheaker (1977) and Van der Stegen and Van Duijn (1980)

demonstrated the value of HPLC for the analysis of CGA in coffee. Accordingly

the latter method, with minor modifications, was adopted as an approach that

was distinct from and yet complementary to colorimetry.

Instrumentation

Liquid chromatographic separations were performed on a Waters

Associates (Milford, Mass, U. S. A. ) liquid chromatograph consisting of two
Model 6000A solvent delivering pumps. The absorbance detector was Model

440 used at 313nm, with a solvent programmer Model 660 equipped with a U6K

injector. A stainless steel (L. 15cm, I. D. 5mm) column Spherisorb 5 ODS

(HPLC Technology, England) was used.

The mobile phase was a gradient elution as summarised thus:

eluent A- 75% water (0.025M potassium citrate, pH 2.5) and 25970

methanol
(AR, BDH, England); eluent B- 1007 methanol. The amount of eluent B was

gradually increased from 0 tD 30970over 40 minutes at a flow rate of 1. OmL per

minute(tor41).

All organic solvents, standard solutions and distilled water used for

the HPLC were filtered through 0.22um millipore filter paper (Millipore Ltd.,

England).

Reagents
Methanol (AR, BDH, England)
5-CQA (Sigma Chemical Inc., U. S. A. )
 - 56 -

Ferulic Acid (Sigma Chemical Inc., U. S. A. )

Caffeine (Sigma Chemical Inc., U. S. A. )
Caffeic Acid (Sigma Chemical Inc., U. S. A. )
Quinic Acid (Sigma Chemical Inc., U. S. A. )
DCQA (Roth GMBH, Karlsruhe, W. Germany)
5-FQA (see acknowledgements)
3,4-DCQA (see acknowledgements)

3,5-DCQA (see acknowledgements)

4,5-DCQA (see acknowledgements)
3-, 4-, 5-p-CoQ (see acknowledgements)
3-CQA (see acknowledgements)

Several standards were injected into the HPLC in order that their tR

values could be determined. These standards and their rR values are listed
in Table 7 (page 60).

Standard Curve (HPLC)

Calibration curves were construced using triplicate injections of
each standard of as many concentrations as was practicable bearing in mind

the amount of material available. The results are shown graphically in Fig.
12,13 and 14. Most standards proved to be almost homogeneous (CQA, FA).

The crude DCQA (Roth, W. Germany) was recrystallised until the three DCQA

isomers (3,4-, 3,5-, and 4,5-DCQA) accounted for some 95% of the material
detected by HPLC. The peak heights of the three isomers were summed and

related to the total amount injected. This is possible since the three isomers
have virtually identical molar absorbtion coefficients (Rubach, 1969).

Calculations

Peak retention time (tR) is measured as the time from the injection of
the sample until the time at which the elution peak is at its maximum. Peak

height measurements are used to quantify the standard solutions and components

of coffee extracts. The repeatability and reproducibility are shown in Table 8.

 - 57 -

Figure 12. Calibration for HPLC determination of CQA..

Peak
Height
(cm)

1.4

1.2

1.0

0.8

0.6

0.4

().'D2

0.1 0.2 0.3 0.4 0.5

rc/rL
 - 58-

Figure 13 Cal ibration for HPLC determination of FA-

.

0.35

0.30

Peak
Height
(Cm)

0.20

0.10

Q. O2

0.1 0.2 0.3 0.4 0.5

mg/mL
 - 59 -

Figure 14. Calibration for HPLC determination of DCQA.

Peak
Heigh-
( cm)

1.5

1.0

0`. 5

0.1

0.1 0.2 0.3 0.4 0.5

rrg/mL
 - 60 -

Table 7. tR Values of Standards

t
Standards tR values (minutes)
Gallic acid 2.5
3-CQA 3.0
3-CoQA 5.0
4-CoQA' 7.5
5-CQA 8.6
4-CoQA 11.0
CA 13.5
5-CoQA 14.0
5-FQA 16.0
FA 16.0
Caffeoyl glucose +
16.5
Coumaric acid 21.0
5-CQA isomers (a) 3.0,6.5 and 9.0
(b)
5-FQA isomers 9.0,12.5,16.0 and 16.5
3,4-DCQA 24.5
3,5-DCQA 26.0
4,5-DCQA 32.0

* Impure standard
(a) (b)
and - isomerised by heating with ammonia
+- recovered after ion-exchange purification method
(Griffiths, 1982)

Coff ee extract

The green and roasted coffee extracts used for colorimetry were
filtered through a 0.22pm millipore filter, and diluted whenever necessary.

Later a 10µL sample was injected for test separation.

 -61 -

HPLC Data

The method of Van der Stegen and Van Duijn (1980) with sample
detection at 313nm, was used. Though the HPLC was equipped with a high

resolution reversed phase system, it could still be limited with respect to peak
identification and peak purity.

Peak Identification

In an extract up to 20 peaks have been separated (see Chapter 3, Fig. 16, ?,'IS)
Peaks identified with a high degree of certainty:
These are peaks matching a homogeneous commercial standard or
homogeneous sample provided as a gift and matching closely with the behaviour

reported by Van der Stegen and Van Duijn.

Peak Assignment
1 3-CQA
3 5-CQA
6 5-FQA

8 p-coumaric acid
10 3,4 -DCQA
12 3,5-DCQA
16 4,5-DCQA

Peaks probably identified:

Those peaks which correspond to a peak identified by Van der Stegen

and Van Duijn or which match a less well defined standard, e. g. isomerised
FQA, CQA or 4-p-CoQA.

Peak Assigment
24 -CQA Van der Stegen & Van Duijn;
is omeris ed 5-CQA
4 4-FQA/4 -CcQA Van der Stegen & Van Duijn;
isomerised 5-FQA

5 CA/5-CoQA Caffeic acid/5-CdQA

By comparing the relative retention times with those of Van der Stegen it
 - 62 -

that 3-FQA is approximately at 9 minutes, probably before caffeic

was suspected
5-CQA, the 4-FQA is 12.5 before
acid, and after about minutes, also caffeic

acid and after 5-CQA. However 4-CoQA gave two peaks, one at 7.5 and the

other at 11 minutes, i. e. one before 5-CQA and the other after.

Unknowns:
Peaks nos. 7,9,11,13,14,15,17,18,19 and 20.
When the total peak heights of the unknowns of Tanzanian Arabica

sample was measured and compared to 5-CQA standard, it was shown that

those peaks contributed 0.064 mg/mL. However when the unknown peaks for

the Indonesian Robusta coffee were measured, the unknowns contributed

0.12 mg/mL.

Peak Purity

The chromatographic system employed in this investigation has good

resolution separating at least 12 CGA satisfactorily. However one cannot be

certain that all peaks in all samples are homogeneous. This problem is
highlighted by considering ferulic acid (tR =16 minutes) and 5-FQA (tR = 16

minutes); these do not resolve when present in mixtures. Van der Stegen and
Van Duijn found that their peak 4 (tR = 12.5 minutes) was impure and could,

using acetonitrile, be resolved into 4 -FQA and what may have been 4 -CoQA .

One can never completely rule out the possibility of co-chromatography and

simultaneous detection, but for simplicity, it has been assumed that the peaks

are homogeneous until proven otherwise. The likelihood of non-homogeneity is

possibly greater with immature or peculiarly coloured beans than with typical

commercial beans for which the system was developed and of which the

composition is relatively well known.

One can rule out interferences from certain substances known to be in

the extracts, e. g. quinic acid, caffeine and other methylxanthines, since they

are not detected at 313nm. Similarly one can hypothesise about substances

which might be present and thus which might account for one of the unknowns
or which might co-chromatograph.
 - 63 -

Previous workers (Corse et al., 1965) have reported at least one

CFQA in green coffee beans; since FQA follows CQA, such a compound(s)
might well chromatograph behind the DCQA. Van der Stegen and Van Duijn
(1980) reported two peaks peculiar to Robustas, one of which they tentatively
identified as FDCQA.

The cinnamic acid esters of glucose are known now to be widespread

in occurrence and in some plants to be intermediates in CGA biosynthesis

(Molderez et al., 1978). Accordingly some glucose esters might be present
in coffee. After the work reported here was completed, ' Griffiths (1982) using

an ion exchange purification procedure recommended for isolating such glucose

esters recovered from coffee a molybdate-positive neutral phenol with'tRof 16.5

minutes. This appears to elute as peak 7 and could be caffeoyl glucose.

Ulbrich and Zenk (1980) reported that a coffee cell culture contained

the enzyme p-coumaroyl shikimate transferase which implies that coffee beans

could contain p-coumaroyl shikimic acid (CSA) and possibly other shikimic

acid esters. Even if such esters do occur they may be only non-accumulating

intermediates.

Quinonoid CGA oxidation products are another possibility, particularly

in discoloured green beans.

It should be noted in passing that some of the above mentioned

substances would be detected by one or more of the colorimetric reagents and

that none is readily available for study.

 -64-

4-4

0 z z z z z En
z En
z
H \ \ \ CY) \ \ Clq
N ý1
r-I r-I 00 1-1 00 00
P--4

Cd
94
U Cl) N LO O\ If) Cl) 00

p "-ý --ý "-i N N

Cd

cd

co! N
LO r-1 CY) LO r-1 -i M LO 6R
* N 2 O
to
Cl, - rn ý
'O Ü

V)
ýz
0 0
w
-4
cn a
a) a)
H 0 0 co CO ö ö
CLr m
o o O 0 0 0 0
a) ö °)
10 C; 0 0 0 o O C;
zco
II

En
I)
C
M n 00 O G% N
O
Cl)
O
z.
N M 0
L
a)

Cd
Qýy) cd
a)
O v) O 00 0
Ö Ö
I
Ö
O I O
21
O
V-q
O O O O O O O

Ö
"0
co
ö'00
bßä ýý, ý1l -<to to
b
,a IL) C)
ý,
4) C) 40
g-.
CY CY
cý r A U
o ,
:ý
Q
..Cd Mý
10

a) 0
U
a4 cd

cc
cd U
O V] ß+
x ä w E-4 W A x
Ei

Cd
 - 65 -

Notes to Table 8.

1. Standard Deviation ( 60) expressed as an

absorbance.

2. Standard Deviation ( 6x) expressed as a

percentage of an appropriate typical value for

e. g. MV.

3. Calculated using values for MV and PV.

4. Calculated using values for MV PV and

TBA BD QA.

5. Calculated using values for HPLC CQA, FQA and

DCQA. The basis of calculation is:

,ý222
(J1 -+..... (In
TOTAL=

VI Integration of Colorimetric Data with Chromatographic Data

The attricbutes and limitations of the colorimetric nE thods were

discussed in Chapter 1. p. 41 and those of the HPLC method were similarly
 - 66 -

discussed on p. 61 All methods are of high precision (see Table 8). Thus
.
one would expect that:

Pv
HPLC CGA = 1.0 Ratio 1

where PV = CQA + FQA + 1.37 DCQA,

and HPLC CGA = HPLC CQA + HPLC FQA + 1.37 I[PLC DMA

In practice a value slightly above unity can be expected since periodate

detects e. g. caffeic acid, ferulic acid and glucose esters. Similarly there

should be a close correlation between colorimetric and chromatographic

estimates of BDQA, i. e.

HPLC BDQA
_ 1.0 Ratio 2
TBA BDQA
,

In practice the quantitatively minor CGA (e. g. CoQA, CFQA) are

dif f.isu It : to identify and quantify reliably and thus HPLC BDQA is calculated

only. from the quantitatively major CGA such that: HPLC BDQA = HPLC ODA + 1PLC FQA +

0.69 HPLC DOQA.

Thus in practice this ratio will be less than unity. Examination of

typical chromatograms suggest that for Arabicas the uncharacterised

quantitatively minor components represent approximately (calculated
10570 as
CQA) of the total phenolics content, and for Robustas some 20%. If these

unknowns all contain quinic acid the ratio would fall to approximately 0.90 for

Arabicas and approximately 0.80 for Robustas. It is unlikely that all the

unknownswill contain quinic acid, thus one can predict that this ratio will fall

within the range 0.8 to 1.0.

The TBA and periodate reagents both approximate to measuring total

CGA. One would expect a correlation between these methods, although the

ratio will not approximate to unity because of the different responses to DCQA.
This may be explained as follows. Assume a typical green bean has the

composition:
CQA 57, FQA 1%, DCQA 1.597o

then PV =5+1+ (1.5 x 1.37) = 8.05

and TBA BDQA =5+1+ (1.5 x 0.69) = 7.04

and PV 1,14 Ratio 3
=
113ABLYgA
 - 67 -

There may be further small deviations because periodate does not

detect the CoQA and TBA does not detect e. g. caffeic acid.

These ratios are based upon certain assumptions about the nature of
the phenols in typical green coffee beans. If these ratios can be confirmed for

such beans, then it may be deduced that for any sample which yields peculiar
ratios there must be a peculiar composition. The manner in which the ratio

changes could give some pointers to the nature of this change in composition.
In the following chapter these hypotheses are tested.
 - 68 -

EXPERIMENTAL,

RESULTS AND DISCUSSION

 - 69 -

CHAPTER THREE

ESTABLISHMENT OF BASELINE DATA

 -70-

I Chlorogenic Acid (CGA) Content of Normal Green Coffee Beans

Introduction

For the sake of comparison, a large number of typical commercial

green and some roasted coffee beans were analyzed. The samples were

supplied by Lyons and Illycaffe. The data from these samples served as a
baseline for the interpretation of all other analyses carried out in the course

of the study.

The samples were of diverse origin (Table 9) and included Arabica,

Robusta and Arabusta species.

The Tanzanian Arabica is discussed separately because Tanzanian

green coffee beans were roasted in my presence in Illycaffe, a coffee brew

made, and the residue after extraction dried and packed for analysis.

Method

The samples were extracted and analysed by colorimetry and HPLC

as previously described.

Results and Discussion

Table 9 shows that for Robusta coffees the periodate value, molybdate

value, and TBA BDQA value are highest, Arabusta as intermediate, and
Arabicas the lowest. The estimated FQA follow a similar trend although the

Arabusta has a lower estimated DCQA content than some of the Arabicas. The

Robusta-Arabica differences are particularly marked for the beans from

Indonesia.

The HPLC data (Table 10)essentially confirm these trends but total
HPLC CGAwere lower than the periodate values indicating that at least some of

the HPLC unknowns contributed to the colorimetric data. Rees and Theaker

(1977) reported that higher CGA in Robusta coffee arose primarily from higher

FQA and DCQA contents rather than differences in CQA content. However for

the samples analysed during this study the Robustas also had higher CQA

contents.

Table 11 summarises the hypothetical analytical ratios and the values

 -71 -

b
a)
d+ 00 M IN P-I It) Oý ýO
U N44 Cl) eta NO0
N-01
"" W
IM4
cd

U U %0 Itto N %0 kC [N %0 LO

VR O O O O O O O O
E
Ü 0 A
0 0 ýr
ca t. ui "ö
O' er v

.o Cd op %0 w m to w N cY)
N N
Nl: .d N N ýi N

E-,
E
C) .o
i a)
' 4-0 O'. N
N
cd O O O - 1 O O O
U) Ea
U W

Ü
d+ 110 %lo m %0 00 LN [N
U ? N 'O oÖ N 1f) 'O
'O
4-4

g
0

U) 4-j fý.
C C%j
Cä 'L 00 00 O 144 U) O N 00
0) 0
Cý c:

ä ý
0
U

0
aý Q,
Q

a% M 't CO 1
ö m 00 '0
"
D1 (: OÖ OÖ OÖ CO CO 00

w cn
'Ö

0
U
-< a Cd
0 cd cd cd 8 8 8
U U) U)
-1
rn cl 0 U) :d U c Ä0
ö 19
Q vý ' cd ,q
cd
C% 0) a tý Cd IQ U cd .8
0 ..ý 0 S-4
.
° 0

E H N U CO
toZ0 r-1 N 1-4 CO dý :i to ýO .
 -72

'0 .-ý tf) 1ý0 N %0

N N ý4 Cl) v

N N 01% N 0
CO CO N N
N
O O O O O O O O
Cd
U

11-1

cs3 00 M M C "I Cl ) 00
'd N M N co

öý "
U
a -,
ä
(
U
A U+ N Co 0% in to 1.0
tn
co v %lO O 144
as N 10
.
4-4 Ill
to CM dý ýD lO M

Cd
ý .r

,a E° U 1- 01

10-1
Cd
U_
G) -i
',. 4
i-1 U

co r-4 r- 0 0
N ii)) N
U U ýn N
pQ N r-I
V r-( . -i r-i . -1
U)
N
`a+ Cam)

Cl) U
"ý d N N N
`N I%D
v d1+ l-
m 0 0 O O C; O O
4) D+
41 0
0
U j

U)
p 10
U
O Q'
U

Q, '0 N O N N
Omi Cl)
4-4 N
O N M Cl)
N C4 M Cl)
0

ö C
U
U

Cd Cd cd
Ü Cd
r-i
Cýl 14-4 Cd Cd ".-
H
Cd
.ý as Cd
.ý v v
°a) ö ) r.
P-4 10 U) 10
b
90
H 0 :3 0
q
A .ý.. ý cd .--ý
'a 0
6V 5 P4 U
1-4 .
 - 73 -

obtained experimentally.

Table 11 Hypothetical and Experimental Analytical Ratios of

Green Coffee Beans

Hypothetical Experimental
Mean + SD

Ratio 1 A little above 1.0 1.29 ± 0.09

Ratio2 0.8-1.0 0.8 ± 0.08

Ratio 3 Approximately 1.14 1.19 ± 0.16

Ratio 1 is higher than predicted, indicating substantial periodate

inflation. This could be
caused by any periodate-sensitive substance that does

not cochromatograph with the quantitatively major CGA, e. g. caffeic acid,

ferulic acid, possibly some glucose esters. Whatever the cause of the significant
deviation from unity, this value may be taken provisionally as a norm and any
subsequent deviation be taken as indication of peculiar composition.

Ratio 2 is at the lower end of the predicted range, suggesting that

most of the uncharacterized minor phenolic compounds do contain quinic acid.

This is consistent with Van der Stegen and Van Duijn who commented that all

the unknowns were acidic. However as discussed in p. 63, Griffiths produced

some evidence for the presence of a neutral phenol.

The prediction for Ratio 3 was confirmed.

Conclusion

Analyses have provisionally established norms for the Analytical Ratios

that are in good agreement with the hypotheses.

II Effect of Roasting on CGA Content of Tanzanian Arabica Coffee

Introduction

Clifford (1972,1979) using TLC observed that roasting caused a

progressive and ultimately severe loss of the mono- and dicaffeoylquinic acids.

This observation was supported by the results obtained with the molybdate and
 - 74 -

b
G)
U,
E U Ö
A

0 C ON N N N
C)
C O O O O O
0 v r4
U

V 00 11 N %O
to if)
Ch. t- %0 Lo Cf)

ä12 O O O O O

U v) CO tr) 0
r-+
O O C%
4
O
1 . --, . -4

O
0
U 14
.O'
W
0 0 Ö 0
O
cad Ö Ö
O O O

-9
to
4
A
G)
Co

u r-I oo O so as.
r-1 'D tf) ßi4 i-4
ct! O O O Ö
0 . -i
ci Iti
U
A
ctS
H

cd a)
Cd cat
2
tti
-
. -1
N
IN
0%
to
co
to
'0
N
H
cd a
v O O O O
b
O
Cý
a
w
0
Co
U
U)
Wý
ý-1
N w
ul 1 O tN O . -4 5O

. ta 11
0
E4 P4
 - 75 -

periodate reagents. However, he reported that it appeared that FQA as

judged by colorimetric methods were more heat stable and could become the

major CGA of heavily roasted Robusta beans.

Rees and Theaker (1977), and Van der Stegen and Van Duijn (1980)

using HPLC methods reported that roasting caused a loss in the CQA, FQA

and DCQA. 'Also, Rees and Theaker observed that with increased roasting

there appeared two other peaks in the chromatogram which increased in height

with roasting.

It was thought that these apparent differences in behaviour during

roasting might arise from the use of different methods of analysis by these
three groups. To clarify this point, and to provide a base-line in this study, a
high quality Tanzanian Arabica was roasted.

Method

The roasted coffee beans were analyzed by colorimetry and HPLC as

previously described. Also the HPLC chromatograms are presented. (Fig 16 to 20)

Results and Discussion

All methods of analysis indicated (Table 12) an essentially linear

disappearance of CGA and the HPLC data (Table 13; Fig. 15) showed that the

relative rates of disappearance were CQA 1.0, DCQA 0.56 and FQA 0.09 as
judged by the relevant slopes in Fig. 15. These data essentially confirm
Clifford's (1972) previous observation of the relative stability of the small

amount of FQA.

The CGA are not volatile and these substantial losses imply the

formation of new products. One of these appear to be quinic acid (QA) which
increased progressively during roasting. Up to the third degree of roast the
increase in free QA could have accounted for much of the BDQA that had been
destroyed. Beyond this stage the loss of BD QA far exceeds the production of
free QA which suggested that the BD QA was rapidly converted to a 713A-insensitive
product(s).

During roasting Ratio 1 (Table 13), the ratio of the PV to calculated

 - 76 -

Cf) rn to rn N
M 4
4 ö ö ö Ö

0
ca
a
O oo m r- 'I'D
00 N 144 14 M
N Ö Ö Ö Ö
O

cý
a
6
U, C% IN "3+ M
A 110 t- O
"-i v
N
A

U
a

cU IP) %0 M 00 Cl)
00 di M N . -1
O O O O O

0) co
ÄH
rj)

Cd
U ö, N
0 O O O
ýU Ö
oA O O O O
Cd
U

cý
H

Ö Ö Ö Ö O
Ca
'C Ö Ö Ö Ö
w O

q 00 ei N co N
to to Cl) N - -
O Ö Ö Ö Ö
U
O
P4
w
0
to
U
to
to
0
0-4
to O
M ... i lý O --i '.O
"-ý . -1
1>1
O
H
 - 77 -
Figure 15: Effect of Roasting Tanzania Arabica on CGA
as Determined by HPLC (g/100 beans)

U)
cd

1.0

O.S

o.F

0.7

0.6

0.5

0.4

0.3

0,2

0,1

02468 10 12 14 16

per cent Pyrolysis loss

 -7s -

Fig. 16 HPLC separation of phenolic constituents of Green

Tanzanian Arabica Coffee
1Z

p 3o Mý ýUi%% Ko
to ao

Fig. 17 HPLC separation of phenolic constituents of the Ist Roast

of Tanzanian Arabica Coffee.

1.

3 to
6 12
2. to
it 1113
RNI
O to so 0,0-h, U 40
Fig. 18 HPLC separation of phenolic constituents of the 2nd Roast
of Tanzania Arabica Coffee
11

IIt

0iD 10 3o -'
'to
Mºýuýas
 - 79 -
Fig. 19 HPLC separation of phenolic constituents of the 3rd Roast
of Tanzanian Arabica Coffee

),

O10 äo 30 ýºrt.n ý1-aQ yý0

Fig. 20 HPLC separation of phenolic constituents of the 4th Roast

of Tanzanian Arabica Coffee

0 10 3.0 10 10
1.1iw4tts
 - 80 -

HPLC CGA rose. This observation may be explained by the production during

roasting of periodate-sensitive degradation products that cannot contribute to

the HPLC estimate of CGA. Previous workers have reported a number of

degradation products either in model system studies of CGA degradation or in

roasted coffee beans. These include catechol, guaiacol, 4-ethylguaiacol and

4-vinylguaiacol (Pypker and Brouwer, 1969; Clifford, 1972 and Tressl, 1977).

The guaiacols are volatile and may be lost during roasting, but the

di- and trihydroxy phenols are unlikely to be lost in this way. The majority of
these react with the molybdate reagent and/or the periodate reagent and could
thus account for the observed changes in Ratios 1 and 3, but few have significant

absorbance at 313nm and are unlikely to be detected by HPLC. Nevertheless

Fig. 20 shows that at least one degradation product could be detected.. Peak 4

with a retention time of approximately 12.5 minutes showed an increase with

roasting. This peak is almost certainly a mixture but could contain 4-FQAk12.5m')
and caffeic acid (tß = 13m) particularly in view of the propable release of

QA by hydrolysis of FQA. However, other possibilities cannot be ruled out.

Van der Stegen and Van Duijn (1980) reported the production of a 'new' peak at a

similar position on the chromatogram.

Ratio 2 monitors the relative changes in the HPLC and TBA estimates

of BD QA. This ratio declined as roasting progressed, particularly after 11%

pyrolysis loss, ' and indicated greater loss of BDQA as judged by HPLC (free QA

is not detected at 313nm). This could be explained by damage to the caffeic

and/or ferulic acid residue of the CGA such that the new product(s) chromato-

graphed separately (and/or were not detected at 313nm) but on saponification

still released a substance that reacted with TBA in such a way that it was
indistinguishable from QA.

Ratio 3 compares the PV and TBA estimates of the CGA. This ratio

declined at 7970
pyrolysis loss, remained essentially constant until 117 pyrolysis
loss, and then declined rapidly at 167 pyrolysis loss. Such a decline may be

explained by a more rapid destruction of the site responsible for periodate

sensitivity compared to the site reacting with TBA. This is consistent with the

preceding statement.

Chlorpgenic acid and/or its roasting degradation products are also found
 - 81 -

M M N
O O O O
4,,;
U)

C) CV N er) N
1 Ö 0
H O 0
U
H
'Lb
+.+
Q)
.'Q "0
M M M M
U
H O O O O
O
O
CY
0

't7 CCýý O O O O

aý
A
a
(ý N N Cl) Cl)
O O O
Co
R7

0
0
H Cl) Cl) N r-+
C7 0 Ö O 0
aý
aý
w
0 11) 1! ) t1)
U
a ä ö 0 0
10
aý
Ha3i
Cý
Co
w
O U)
O Or O 'o
44 "-+
O
Cob - - -
U

C)

H
 _82_

in high molecular mass complexes (Clifford, 1972,1975; Nakabayashi and

Watanabe, 1975). Some of these are ethanol soluble (Clifford, 1972) and

might be extracted by propanol. Although chromatographic behaviour and

responses to molybdate, periodate and TBA reagents are unknown, it seems

quite feasible that data reported
such substances would produce the analytical
here.

Analysis of brewed grounds (Table 14) indicated that propanol

extracted small quantities of substances that behaved identically to free QA,

and to BDQA and which responded to the molybdate and periodate reagents.
However, FQA and DCQA estimates are probably invalid as degradation products

may interfere and invalidate the assumptions used in these calculations.

Conclusion

The major CGA are progressively destroyed during roasting. The

CQA are more labile than the DCQA and much more labile than the FQA.

Early in roasting the phenolic residue of the CGA appears to be labile,

then the quinic acid residue, although some quinic acid appears to be released
by hydrolysis. Later in roasting quinic acid may also be destroyed.

III Studies on Quinones as Potential Interfering Substances

There are three reasons for choosing to investigate CGA quinones:

1) As discussed on p. 92, such quinones, if present, would probably be
interpreted as FQA;
2) such quinones might interfere with the interpretation of the chromatograms;
3) as discussed on p. 37, such quinones might well Indicate deterioration of the
green bean.

Furthermore, Oliveira et al (1977) reported that Robusta coffee beans

have PPOactivity approximately five times than of Arabica coffees. These

same workers reported that PPOactivity of dry processed Brazilian green
coffee beans correlates closely with the beverage quality as judged by professional

coffee graders. The highest quality have the highest PPO activity.

Similar studies on washed Colombia green beans revealed a positive

correlation between PPO activity and beverage body, and a negative correlation
 - 83 -

with beverage acidity (Arcila and Valencia, 1975).

In general terms it would seem that high PPO activity indicates high

quality in the final beverage, but the inherently higher PPO activity in Robusta

compared to Arabicas is not easily reconcilable with this theory.

It was this possible connection with quality that led to the selection

of quinones for priority of attention rather than the other potential interfering

substances.

Quinones

Introduction
Quinones are widespread in the plant kingdom (Thomson, 1971). In

general they are yellow, red or brown in colour, but salts of hydroxyquinones are
purple, blue or green (Ikan, 1969). The naturally occurring quinones may be

subdivided (Bernstein et al, 1974) as follows :

1.1,2-benzoquinones, e. g. chlorogenoquinone;
2.1,4 -benz oquinones ;
3. naphthoquinones;
4. anthraquinones;
5. quinones with larger fused rings.

Thomson states that anthraquinones are found in Rubiaccae, the family which
includes the genus Coffea. Accordingly this investigation concerned itself
primarily with 1,2-benzoquinones while recognising the possibility that anthra-

quinones might also be present.

Thomson (1976) stated that very few 1,2-benzoquinones are found in

nature - probably he excluded the almost ubiquitous but poorly characterised

enzymic browning quinones. Whatever the reason, it is clear that the 1,2-
benzoquinones have been studied relatively little and this fact hampered this
investigation.

Many texts and original papers, e. g. Shibata et al, (1950); Fiegl and

Anger (1966) ; Simatupang and Hausen (1970); Thomson (1976); Robinson (1980),

and Minard et al (1981) were consulted.

Synthesis of 1,2-quinone by enzymic or chemical oxidation of 1,2-

 - 84 -
diphenols has been reported by several authors (Mason, 1955; Pierpoint, 1969

and Durst et al, 1975). However the identification of the quinones has proved
difficult because of their high reactivity and instability which led to their
incorporation into complex, high molecular mass polymers (Minard et al, 1981).

Attempts were made to synthesise 1,2-quinones from 3,5-di-tert-butyl-

catechol, catechol, ferulic acid and caffeic acid and 5-C A, using the method

of Durst et al, (1975), using the enzyme PPO.

0

O Polymer
Reaction:
.... _.. _ ... ý (brown)
NO
0.2. OH 1. NCS -TEA
1,2-quinone

Catechol oxidase
H2O
1,2 dihydroxyphenol
Fig. 21: Basic Reaction of Quinone Synthesis

Chemical Synthesis

Materials

N-chlorosuccinimide (NCS) (BDH, England)

triethylamine (TEA) (BDH, England)

methylene chloride (BDH, England)

Hexane (BDH, England)
Catechol (BDH, England)
Ferulic acid (Sigma Chem. Inc., USA)
Caffeic acid (Sigma Chem. Inc., USA)

5-CQA (Sigma Chem. Inc., USA)

3,5-di-tert-butyl. 1,2-benzoquinone
(Aldrich Ltd., England)
3,5-di-tert-butyl- catechol
(Aldrich Ltd. England)
,
1,4-benzoquinone - technical grade
(Sigma Chem. Inc., USA)

Methods
A solution of NCS (400mg) in fifteen mL methylene chloride was made
with stirring and allowed to cool at -10°C. The 3,5-di-tert-butyl -
catechol (445mg, 2mM) or the molar equivalent of the acids was added. This

mixture was held in the cold for 10 minutes, then 0.3mL TEA added dropwise.
Stirring was resumed for another 10 minutes, the solution filtered and evaporated
to dryness. The red residue was dissolved in hot hexane and then filtered,
 - 85 -

evaporated down to a few mL, then cooled to produce crystals.

Starting material and product were characterised by IR spectrum in

nujol, reaction with molybdate, UV spectra and NMR.

Commercial 3,5-di tert-butyl 1,2-benzoquinone was similarly

examined. In addition IR spectra were made for hexane and nujol, so that
their absorbance peaks could be eliminated from the test substances' own peaks.

Nuclear magnetic resonance (NMR) spectra was carried out on the

starting materials, then the products and the commercially available reference
compounds.

The crystalline quinone synthesized from 3,5-di-tert-butyl -

catechol (0. lg) was dissolved in 25 mL of chloroform. However the products

from catechol and the acids would not crystallise so the whole residue was
dissolved in 5OmL chloroform. These solutions were used in the W and

molybdate reagent analysis.

Results

The reaction products were extracted into chloroform. Durst et al,

(1975) reported that the quinone products would be red. This was do for the

products from catechol 3,5-di-tert-butyl cate"chol and ferulic acid,

but caffeic acid and 5-CQA yielded yellow-brown products. These appeared

similar to the products of PPO attack on these substrates.

The 3,5-di-tert-butyl 1,2-benzoquinone was reasonably stable but the

other quinones were very labile. Any attempt to concentrate the chloroformic

solutions, ' or even storage in the dark at low temperature, resulted in, a
darkening of colour and reduced solubility. This instability hindered precise

characterisation of the products.

Nevertheless a body of data was collected that strongly imply that the

products contained the desired quinones along with more complex, possibly

polymeric products. The evidence supporting this statement may be summarised

as follows :

1. IR Spectra

The IR Spectra in nujol for commercial and synthetic 3,5-di-tert-butyl

1,2-benzoquinone were very similar (see Appendix A) showing a carbonyl peak

 - 86 -

at 1680 cm-1. This peak was absent from the starting materials and from the
quinones synthesized from catechol, caffeic and ferulic acids and 5-CQA.
However these quinones showed peaks near 1700 cm-1, presumably due to the

quinone carbonyl groups. A sample of commercial 1,4-benzoquinone yielded

a carbonyl peak at 1590 cm-1, thus the values of 1680 and 1700 cm-1 mentioned
above are consistent with Thomson's (1971) report that 1,2-benzoquinones
have carbonyl peaks at higher wavelengths than 1,4 -benzoquinones.

2. NMR
The NMR spectra (see Appendix B) show that the synthesized 3,5-
di-tert-butyl 1,2-benzoquinone was very similar to the standard, thus

confirming the success of this synthesis.

The products derived from the acids yielded NMR spectra that defied
interpretation, even after attempts to purify the products by TLC. This

difficulty was ascribed to the impurity of the products arising from rapid
polymerisation after the initial purification.

3. UV Visible Spectra
The relevant data are summarised in Table 15. In all cases the

products had a spectrum that was distinct from that of the starting material.
This is evidence of reaction.

The spectra of commercial and synthetic 3,5-di-tert-butyl 1,2-benzo-

quinone were close and showed two UV and one visible peak. This is in
keeping with the data of Berger and Rieker (1974) who reported that 1,2-
benzoquinones had three ranges of absorbance with strongest max in the
250 to 290nm, an intermediate in 37 0 to 47 Onm and a weak one in the 500

to 580mm range.
However, the product from catechol showed only two peaks and the

products from the acids only one diffuse peak which suggests the presence of
impurities.
4. Reaction with Molybdate

The response is interpreted in CQA equivalents (Table 16). In all

cases the response is low, indicating almost complete loss of the 1,2-dihydroxy

structure of the starting material.

 - 87 -

Table 15 W( Xmax) of 2-Propanol Soluble Residue of Synthesized Quinones

x max (nm) of Range of Absorbance of

starting material Synthesized Quinones (nm)

3,5-di-tert-butyl'
1,2-benzocatechol 280 260-280 450-470 560-580

Catechol 280 390-428 550-570

5-CQA 340 390-423
FA 340 390-423
CA 340 390-423

Table 16 Colorimetric Analysis of Synthesized Quinones

Quinones Molybdate
(mg/lOOmL as
(CQA)

3,5-di-tert 1,2-
benzoquinone 15.0
Catechol 12.0
5-CQA 0.0
Ferulic acid (FA) 3.2

Caffeic acid (CA) 2.5

 _88-
The results from these four characterisation methods are consistent

with the successful syntheses of 3,5-di-tent-butyl 1,2-benzoquinone. One

cannot be so certain that the other synthesis were successful, but there is no

doubt that a reaction occurred. The overall impression is that the products

were a mixture in which the desired quinones were present at least transiently.
Because of the difficulty of synthesis and purification of the products, further
investigation was considered unprofitable. Accordingly, no further attempt was
made to:
(a) examine 5-CQA quinone for interference in the analytical methods;
(b) isolate such quinones from green coffee beans.

Conclusion

Chlorogenic acid quinone could not be synthesized satisfactorily by

this method, although it was successful in preparing 3,5-di-tert-butyl 1,2-

benzoquinone.
 - 89 -

3
3

I. Stock solution of CQA fixture

(conc. 0.106g/L ' in 70%
oloxidase
propanol diluted ?: 20) 'QA at

II
1I 10
. n. -I ;'a
2 to M.

ction mixture of
and CQA at 1 hour.
3 4. Reaction mixture
of PPO and CQA
at 2 hours.

H 8 lo Mo
6 ,0M.

Figure 22: Breakdown of 5-CQA by polyphcnoloxidase (PPO)as monitored

by HP LC.
 -90-

Enzyme Synthesis

Material
Mushroom tyrosinase (EC No. 1.14.18.1, Sigma Chem. Inc. , U. S. A. )
Activity - 800 units/mg (1 unit = 0.001 absorbancy increase per

min. at 280nm in the phosphate buffer pH 6.5 at 25°C,

containing 3x 10-4 ML- Tyrosine). 0.0025g tyrosinase was
dissolved in 10mL.

Chlorogenic Acid. CQA: 0.106g/L in 707 2-propanol =3x 10-4M

DCQA: 0.155g/L in 70970

2-propanol.
Phosphate buffer. pH 6.5 (see Chapter 2).

Method
A mixture of enzyme, phosphate buffer and CQA or DCQA was prepared

and allowed to react for different times (0,60 and 120 minutes). One millilitre

of phosphate buffer was added to 1. OmL of substrate solution (5-CQA or DCQA),

0. lg sodium dodecyl sulphate (SDS) and 1. OmL of enzyme solution was also

added. The SDSwas added to enhance activity of the enzyme with CGA (Walker

and McCallion, 1980).

At 0,60 and 120 minutes the reaction was stopped with 17mL of 99%

2-propanol. The total reaction mixture of 20mL was analysed by molybdate,

periodate and TBA reagents, and by HPLC.

Results and Discussion

When the reaction mixtures were examined by HPLC, it was apparent

that 5-CQA was progressively destroyed with destruction ultimately reaching

nearly 9070(Fig. 22). In contrast the DCQA vanished completely from the

chromatogram within one hour, and the 5-CQA impurity declined nearly 10°70
(Fig. 23) over two hours. This implies that the DCQA bound to the PPO, inhibited
it, and thus effectively protected the 5-CQA from attack. The ability of DCQA
to bind to proteins is discussed further in Chapter 6.

No degradation products were detected at 313nm from either substrate

and at least in the case of the two hour samples, this cannot be attributed to

co-chromatography.
 -91-

1. Stock solution of DCQA (conc. 0.155g/L in 707 propanol diluted 1: 24.

-I lI .10 30 "-1 Sºo

to
2. Reaction mixture of PPOand DCQA at 0 hour. nntvN
.nkcs
3

3. Reaction mixture of PPO and DCQA at 1 hour.

3

4. Reaction mixture of PPO and DCQA at 2 hours.

Figure 23: Breakdown of DCQA by PPOas monitored by HPLC

 - 92 -

Colorimetric analysis indicated tjiat there are no losses of bound

quinic acid. Pierpoint -et al (1977) made a similar observation. The molybdate

reagent and periodate reagent (Table 17) indicated progressive loss of 5-CQA,

thus supporting the chromatographic data. However, the periodate reagent

indicated a slower and less extensive destruction than the molybdate reagent.
This difference, which was more prpnounce4 at one hour than at two hours,
would normally be interpreted as FQA, but clearly is better explained as due

to PPO oxidation products, possibly 5-CQA quinone. With DCQA as substrate,

periodate and molybdate gave identical results, within experimental error,

suggesting no quinone has been found. This observation is consistent with the
inhibition of PPO by DCQA.

These data indicate that PPOoxidation products may if present interfere

in the colorimetric estimation of FQA, but not in the HPLC determination.
Data from the previous section on chemical synthesis indicated how labile such

compounds are and it is thus unlikely that such compounds would be extracted

unchanged, even if present in the green beans. Accordingly, interference from

this source would be small and any attempt to continue this investigation by

isolating such quinones from green beans was considered unprofitable.

Conclusions
1. In vitro 5-CQA is a substrate for PPO, but DCQA cause rapid
inactivation of the enzyme.
2. With 5-CQA as substrate PPOproduces a transient oxidation product(s),

possibly 5-CQA quinone.

3. These product(s) might cause some slight interference in colorimetric

estimation of FQA, but apparently none in the HPLC determination.

 - 93 -

Table 17 Effect of PPO on CQA or DCQA

Substrate Time PV MV FQA/Quinone

(min. ) mg/100mL mg/100mL

5-CQA 0 10.6 10.6 0

60 3.5 2.1 1.4
120 1.6 1.1 0.5

DCQA 0 15.5 15.5 0

60 7.8 8.1 -
120 7.0 7.8 -
 - 94 -

CHAPTER FOUR

CHANGESIN OGA OONTIIVT DURING T

DEVELOPMENTOF GREEN COFFEE BEANS
 - 95 -

I Introduction

It is reasonable to assume that the composition of green coffee beans

will change during the development of the seed.

The CGA content of sunflower seed increases during fruit ripening

and decreases during seed germination (Ruckenbrod, 1955). Dorrel (1976,1978)

demonstrated that the CGA content of sunflower seeds was influenced by the

time of year at which they were formed. Late seeding by one month resulted in

a CGA content some 21 per cent lower than normal.

Similar changes have been reported in other tissues, Hillis and Swain
(1959) reported that there was a correlation between the accumulation of CGA

and maturity of fruit and the ultimate seed development. They found that there

was an increase in phenolic compounds of Prunus domestica as the fruit increased

in age.

There is a progressive and marked synthesis (20-fold increase) of

5-CQA in the fruit of tobacco as it matures up to the stage just short of full

maturity. At this stage there is an almost complete loss which is accompanied

by severe browning (Sheen, 1973).

In ara s'colymus, Lattanzio and Morone (1979) showed that the CGA

content decreased progressively as the plant grew. They observed that the
CGA content decreased from 4.4 to 1.670' as the plant gradually developed to the

stage of differentiation of the capitulum. From this moment onward it remained

at a constant level.

The only experiment on coffee found in the literature is that of Hamidi

and Wanner (1964). They reported that the highest CGA content was found in

immature coffee beans. This fell to 0.17% in half-ripe fruits, and then rose to
0.37% in seeds from the ripe fruit. These values seem extremely low compared

to, for example, the value 7.67 for mature Arabica beans from Kenya and 10.1%

for mature Robusta coffee beans from Indonesia as reported by Rees and Theaker
(1977). The difference could be explained by differences in the methods of

analysis used.
 - 96 -

II Origin and Nature of Coffee Samples

Green coffee beans were sent from:

Ivory Coast (IVC), West Africa

Colombia (CL), South America
Kenya (KA), East Africa
England, Royal Botanical Gardens, Kew (Kew)

The participating Research Institutes were sent a guideline which they

were asked to follow in the collection of samples. The criteria to be followed

are set out in the format (Format 1). However, not all participating bodies
kept precisely to the guideline. In addition, many samples contained discoloured
beans. The beans were examined by Illycaffe using reflectance spectrophotometry
and were converted to homogeneous samples on the basis of this criterion.
typical reflectance spectra are drawn in Appendix C, D and E.

Kew, England

The coffee variety from Kew was grown at a controlled temperature

(28°C). The variety studied was Coffea Arabica var Marogogipe (Fig. 24), from

the tropical plants collection, Palm House. Marogogipe is a variety of Arabica

coffee discovered near Bahia, Brazil, and has very large cherries of heavy
flavour. This variety is mostly sold in Europe at a premium price (Sivetz and

Desrosier, 1979).

Flower clusters were marked by labels indicating their time of flowering.

Collection of fruit samples started twelve weeks after flowering and continued at

regular intervals up to the stage at which ripe fruit were falling. Normal

commercial maturity occurs just prior to this stage for wet processing, and at
fruit fall for dry processing.

The cherries were depulped by hand, weighed and immediately dipped

in liquid nitrogen, freeze-dried and weighed again.

Ivory Coast, West Africa

These were two Robusta clones (IVC 182 and IVC 503) and one of

Indenie. The collection of coffee cherries started twenty-four weeks after

flowering and six harvests were made.
 - 97 -

FORMAT I

SPECIFICATIONS FOR GREEN BEAN SAMPLES

If bean composition-bean quality relationships are to be established

it is ESSENTIAL that all samples are precisely described with reference to

the variables set out below:

1. Green bean maturity - days after flowering (± 7).

2. Specie and variety.
3. Altitude at which grown.
4. Soil characteristics, especially if peculiar in any way.
5. Soil treatments, e. g. fertilisers.
6. Cherry treatments, e. g. use of ethrel.
7. Details of any unusual weather during the period from
setting of fruit to harvesting; average day length and
light intensity.
8. Details of method of processing the cherry:

a) Wet process or dry process;

b) Time and temperature of process;
c) Use of sun or artificial drying.

It is essential that samples should be in a biochemically stable form,

i, e. dried to 107 moisture content, before dispatch.

For semi-mature cherries it is desirable that samples are subdivided

after harvesting and subjected to wet and dry processing, and possibly further

subdivision to permit the effect of variations in these processes to be

investigated.

It is desirable that sample size should be 100g minimum.

 - 98 -

Figure 24: Coffea Arabica var. Marogogipe

showing some leaves and seeds.
 - 99 -

All the coffee cherries were sun dried and dehusked. The samples

were sent by Institut Francais du Cafe et du Cacao Abidjan, Cote d'Ivoire.

Colombia, South America

Samples of Arabica green coffee beans were sent from the Federacion
National de Caf eteros, Bogota, Colombia.

All cherries were hand-picked from one tree at different stages of

ripening through to normal commercial maturity. They were dried using two

methods -wet, and dry (60°C) processing. The cherries were then packed in

pouches under carbon dioxide and sent to Italy.

At Illycaffe, the cherries were carefully dehusked and the silverskins

removed. In this form the green beans were examined by reflectance spectro-

photometry as previously stated for uniformity in colour. The homogeneous

samples are listed below in order of increasing maturity (Appendix D).

Dry Processed Wet Processed
1. Verde (green)

2. Amarillo (yellow) 2a. Amarillo

3. Pinton (under-ripe) 3a. Pinton
4. Maduro (mature) 4a. Maduro

Kenya, East Africa

Three samples of Arabica green coffee beans were sent from Kenya
Industrial Research and Development Institute, Nairobi. These cultivars are
French Mission (FM), SL28 and SL34, which are widely grown in Kenya.

Flowering of these samples was in August, November and February

respectively and collection started twelve weeks later and continued at regular
intervals.

The cherries were either wet processed and mechanically dried (30°C)

or dry processed (sun dried).

The flow diagram (Fig. 25) shows the relationship between samples,

method of processing and maturity of bean. Each coffee bean was examined for
homogeneity of colour by Illycaffe (Appendix E)
.
 - 100 -

Variety Processing Method Maturity

Immature

FM dry Under ripe

Mature

Immature
SL28 dry
'-', Mature

Immature
dry Under ripe
SL34 ý ýý Mature

wet .
afore

Figure 25: Flow Diagram of the Relationship between Coffee

Varieties, method of Processing, and Maturity of
Kenya Arabica Green Coffee.
 - 101 -

a
U
Ö r-ý 00 N "0 N %D
4 O ý4 O
Ö Ö Ö Ö
r-1 O O O
Cd
E
41
Ell

Cl) p N 0 M -4
rM-1
~ Ö Ö Ö Ö Ö Ö Ö
w
U

00 e-1
O O d+ ý ý d' mot'

.,., 4
-. 00 cy) in 00
ý 0 N c -3+ to --1 "O
cis
lr
OO O O O ' -1 N ý4

U4
".
H
i. +
O
ce
14
N CO M Cl) CO
O 'Z7 O O O O O O
0
0
4-1
O O O O O O O Zý
I
E
to
T

cy) P-4 cy) ' -4

O Cl) N O IN
O O O.
ö
9
C13 -

cd .. ý
,a Cl) r-1 \O O N N
0 Q) O CO I--. r-q dI
d Ö Ö O

U)

+
i>11

S
. -4
ti
ýO d4 N O
.41 ýl1
Cý7 1 1 1 t
0 oA

CO G)

N M dý to ',O N
C) r-1
cl
U)
E-+
 - 102 -

N :° ö ° ö
c'
m
0
4-
cd
C4
C t rn rn 00 N
N
O O O O O

. -r

Qi

tf) O r--1 'O ef+ M N

%0 0 r- 0 01%

O r+ N ri O '-+ --ý

V Ö N N
Cd 0'% eM -4
4
O O O C; ý--1
U

U
Ö
a CY O Cf) 'd' N
x
6
b
a)
cd

aý
H
aý 0
'C
ý, ö ö ö ö ö 0 ö 0
A
ö ö ö ö ö ö ö W
1) P4 U
x
aý
Cd v
Q cd
Cý1 0
+

0)
N N O\ to N C M
ö 00 00
0 '-1 . --i 4 [N
cd c ö ö ö ö ö ö
E v

A
8

rn -4
'-11 CV O
. -I 9-4
( 1 1 1 1
0)

H
 - 103 -

Treatment and Analysis of Beans

The coffee bean samples were extracted and analysed colorimetrically

and by HPLC as previously described.

Results and Discussion

Kew - C. Arabica var. Marogogipe

Although these beans were cultivated under somewhat atypical
conditions, the mature samples have a composition typical of good commercial
Arabicas (see Table 18).

The most striking feature of this set of data is the apparent absence

of TBA BDQA in the -19 and -16 week samples. The small amounts of CQA
FQA and DCQA detected by HPLC (Table 19) at -19 weeks are consistent with

this within the limits of experimental error, which are larger than usual because

of the unusually dilute extract. The low PV and MV are similarly consistent.
However the very low TBA BDQA at -16 weeks is not in such good agreement

with the PV, MV and HPLC data.

Thereafter progressive rises occur up to -2 weeks followed by a decline,

generally modest, but pronounced for Estimated and HPLC DCQA.

The formation of phenolic compounds was sufficiently rapid to ensure

normal experimental errors for the data from -16 weeks onwards.

There are clear and substantial deviations in the Analytical Ratios

(Table 19) at -16 and -11 weeks. There is clear evidence of periodate-inflation

(Ratios 1 and 3) and of cochromatography of a substance(s) that does not yield

quinic acid (Ratio 2).

Ivory Coast
(a) Robustas
These IVC Robusta cultivars have a significant bound quinic acid

content, even when immature (Table 20). But in neither case are they sufficient

to account for the periodate value or the total HPLC CGA value (Table 21).

This is illustrated by the extreme Analytical Ratios in Table 21, and is evidence

for the presence of periodate-sensitive, quinic acid-free substances that

 -104-

U
ö w ° co 914
ö ö ö
0 w c'l 1n
"ci LO

4.1 ö ö ö ö ö ö ö ö ö ö ö ö

CD
c> 00 ö ö ö ö co
CD ö .
. . . . .
0 0 . 0 . 0 . 0 . 0
U" 0. 0. 0 0 0 0
N oo o O N . -1 Oý N )
CY00
. --i ri N oho
0 N CO u-) \o N N L/)
Ö Ö Ö O O O O O O
O O O

0 ö ö ö ö ö ö ö ö ö ö '
rID r
-4

cd

U c4 C CO N
m N In Co
,. ý O O O
O O O O O
a3 O O O O O O
O O O O O O

0 rn
0
0 6

cd
'd
'0 N 0 N cd
ým c4 r-i 0% N
N0 CO Oý in
05
4.1 _' in CO 'O 10
a O O O O O O O
.; O O
N N

C) 0

U) U)
ca
Co c)
rl a)
ca
a)
O O '0 '"4 CO N +
\O M N N N 0 CO
N N CO C5 N
4.1 to M 'O 'O N
U) >
ci ö ö ö ö ö ö ö O O ".1 r, '-+
0 -v a
U

C)
II
N-1
ýý O O% to N O
rl r1 01 to c4 rl
1 1 1 1 1
1 1 1 1 1
yý

ý~ Ö
N CO
Ü 00
a0 M
c O
U
tr)
O C)
N
SK 0
0
A tý .
ßi rl N M dý N 'O Ri
... N M d'

Ei
 - 105 -

N CO ti) M -4 CO N N M 01 0 00
0 CO %O O O O O O '0 u') N M M dt
N r-1 . -1 . -1 . -I r-1 r-1 . -1 . -1 "-ý r1 . --1

a
Cd . -1 di 00 er CO N O CO N CO N
N 00 00 N 00 00 c CO CO ON CO
.
N . O . O . O. O. O. N
.
4
. O. O. O. O
A

. -1 N M Q 'O r1 %O [N N N r-
10 Oý 0 CO ON 00 it) ýO 0 CV -4 . d{
r-I . . . . . . . .

O O 1; O. O. O. O O "-1 "-i .-ý "-i

U
V 0 IN N CO
-44 1 O 0 CO 'D
r-1 N M in '0 I'O [*- -+ 0 N CO 00 00
O O O O O
ö ö O
O

ed
a-)
O
H

ä
O\ CO 1I) [N cM %0 O CO . i+ ý "-1 0
ý+ CV N N N N to M M N N
r-s
. . . .
U O. O O O O O. O. O. O. O. O. OA

a Zý
ctt
E
ýi ö, CO
O
N
O
CO
O O
CO
O
ýt+
O
U)
O
to
O
tf)
O
N
O O O
" N
G O. O. O. O. O w
O. O. O. O. O. O.
C)
a)
U)
Q
a)

ä N %0 Cl ON 0 % In 00 00 O1 O' 00 +
cd N N N d+ lw CO CO M M Lf) Lf)
0 9-1
m
0 0 0 0 0 0 0 0 0 0 0 0
4J
cd
0
0

C)
it) ý 1
Oý to N O O\ t1) N O
1 1 i 1 1 1 1 t 1 1
cd
E U)

A
O
O
C7 15 00 0
. -1 t/,
C14 C?)
Cd Cd C
4) 00 4. ý

N
G) O O
Cd G4' . d' `a
. -i N CO d4 1f) \O . -1 N CO tn
A
H
 -106 -

C) 00 mi %
p .
C O O O O O

w O
00 o% + O 8-4
0-4 V-4
b0 ö ö W-1 V-4

O O O O O O

p - to 0 110
IF! 000 r

-C4 -4 0 r- cq cr)
%o
Cd oo 0 N 10
Ö . -i* chi)
O O O r+ - -

a
U
Cf)
.-H
v ö ö ö ö ö
`d ö ö ö ö ö ö
H
O W
0
U
a) Lti'
19
4
cd
I rI. co Zo
H
+ý-r O O ZO E
O C)
G) O O r-i rl rl N
G) w
A 0 0)

W
G Q
(t3 0)
a> a)

U) cd +
cd
0 N
U 00
O `dl O

pi O O O "-ý .--1 -
,Ö .
0
1-4
a)
0
a

w LOj N-1
0 .1 Q1 If) N 0
1 1 1 1
rA
aý
0
U

C67
U a)

N
N
W
E
. -1 CV M ýM If) 'D
CD
r-I
A

H
 -107-

M N 0 co Cq

M O VVB ti
4 4
w Co . -i "-4
0 1.4
.. 4
4-1 -
cd
d+ O . -, oo . -4
N O O'. 00 00 00
[z O O O O

Ö 00 M M M
Q', 000 N

O "--; . -i .4 . -i

Qi

U
U t4lý 0% 00 00 0-4 m

a r- o r-4
1.4 ' -4 114

Cd
H

0 C+ N
0 0 -4
CO
Co
11;
l
C C C; C C; C a
cý
E
C)
H

A
Q)
N ö ö ö ö ö
A w C C ö C C
U)
C13

+
Ca
a)

ä Ö O M 00 -4 0
Cd di '.O 00
0 C C C C C
U V C

cd

.,. N
N O
.P -t O\ U')
w 4-J t t t t t
0

cd
C)
0
U O
O

bý0
0 ... 4) C)
V-4
M
N
r .4 -. 1 N M 1144 LO '.O

Eý
 - 108 -

probably cochromatograph. In view of the high HPLC DCQA values, particularly

IVC 182, these substances may well cochromatograph with the DCQA.

By -12 weeks IVC 182 has lost 72% of its initial HPLC DCQA and the
PV has fallen by some 45%. Thereafter MV, PV, TBA BDQA, HPLC CQA and
HPLC DCQA show progressive increases to full maturity. Both measures of
FQA are low and steady. The -12 week dip is not observed in IVC 503. Instead

there is a steady rise in MV, PV, estimated FQA, TBA BDQA and HPLC CQA

and a steady decline in HPLC DCQA. In neither case do the Analytical Ratios

achieve the provisional norms.

Indenie

This cultivar, which resembles an-Arabica, is characterised by very

low TBA BDQA at -15 weeks (Table 22). This is probably consistent with the
low total HPLC CGA values despite the very high values for Ratios 2 and 3 (see

Table 23). This observation in part can be accounted for by experimental errors

associated with the unusually dilute extracts. Ratio 2 at -12 weeks suggests the

presence of some amount of a cochromatographing interfering substance(s) but

otherwise the Analytical Ratios approach the provisional norms set for mature
beans.

Estimated and HPLC DCQA increase progressively to -2 weeks and

then decline at full maturity.

Kent'

For two cultivars the beans were supplied at immature, under ripe

and mature stages, and for the third cultivar they were supplied only as
immature and mature (Table 24). Processing was typical of commercial practice -
for most samples dry processing was used.

Judging by the Analytical Ratios all three cultivars have typical

composition at maturity (Table 25). The SL 34 and FM were almost identical
in composition at maturity but different from SL 28 which has slightly more
HPLC CQA and less HPLC DCQA (Table 25).

There is no common pattern during maturation, although SL 34 and

FM show progressive increases for most analyses. In contrast, SL 28 showed
 - 109 -

ä
00 N N \O 10 ON N
00

Ö Ö Ö O O O
O O
Cd
Ei

00
° ö ö ö ö ö 0 0
0
w ö ö ö ö ö ö ö
ö ö

00 00 00 00 ° N N GO
au ö ö ö ö C C C; C C
"ý o
0 41,
d rn rn 0 rn C rn rn
0 O" CO
Ö Ö Ö C C C
U °
-i r-i
H .Ö

V
9) ä
P4 Ö 0 Ö Cl) Ö CO Cl) N
00
O O O O
C C C O O
Cd O O C;
aý
A
U)
cý
m
cd
aý
C)
E!
I

ý4 ý 00 00 cr% 00 cr, oo Oo 00 C7%

0) `4 ö ö ö ö ö ö ö ö ö
Av
4.1
cd
E
w

co
C14 rn ö 00 rn
0 C)
a) Goo Goo
r? Omi
Ö Ö Ö Ö Ö Ö
Io "ý 0 0 ri
°

U I)
Cd
q)

O i
0

bý0
v
co l al Ej
ý44 1 Cd
C N
ýd+
N
Cl)
0 cd 40

a)
14 a Cd 04
.0
E-ý
 - 110 -

a decline for most analyses. More frequent sampling at more precisely

defined stages of maturity might have provided data that were more amenable
to discussion.

A wet processed sample of mature SL 34 yielded a bean of lower dry

matter content, with lower content of phenolic compounds compared to the
corresponding dry processed sample. This suggests some loss of CGA by
leaching.

Reflectance spectra for all immature and the SL 34 under ripe samples

showed the presence of chlorophyll in the silverskin.

During the wet processing of Arabica coffee beans following the

removal of the pulp, the beans are left in water at ambient temperatures. In

this state soluble components diffuse out of the bean into the surrounding water
(Wootton, 1971). This effect can contribute to the loss of CGA from the bean.
Chlorogenic acid and other phenols are located throughout the whole bean but at

a greater concentration near the surface (Amorim -et al, 1977). Since CGA are

relatively soluble in water (Hamidi and Wanner, 1964; Van der Stegen and Van

Duijn, 1980), it is possible that some CGA diffused out during wet processing,

thus resulting in a lower CGA content for beans processed in this manner.

Assuming that the CGA content in these samples was identical to start
'
with, there would be a significant loss of dry matter and some CGA loss not

simply by leaching but also by hydrolysis of coffee components. Dicaffeoylquinic

be hydrolysed to give caffeic acid, which undoubtedly would be

acid could
dissolved in the processing water and leached. It was impossible to detect any

substantial peak of caffeic acid in the HPLC chromatogram of samples that were

wet processed, compared to those that were dry processed.

Colombia

Dry processed samples were supplied in the cherry at various stages

of maturity defined as green, yellow, semi ripe and mature. The last three

were also supplied in a wet processed form.

There are three striking features about these beans (Table 26 and Table

27): (1) the most immature sample is characterised by a low TBA BD QA

 - 111 -

N LO 144 --ý O-ý

O - N
-ý
Co

O
. -4

A
CSý 'O O' ýO N . -1 O'.
.-"1 If) . -1 N
u CO N N [N N ýM Co

c
1.0
c; C; O O O O

cö

N cr%
%0
0 10 r-4 N
Co - M
r-4
N r1 . -1
ri . -i . -1 .4

0
14

U
C) co LO N N N N -4
a '. O Co tf) N
U13 10
ö ö ö ö ö ö ö ö ö

A cd
H
Co N
0 ä --1 i-1 N N N4 0
co . N V-1 . "
. " '

CA
U
ö ö ö ö O O O
0 c;

O
ý-1 ý-I 00 N if)
Qi O - -
. . r-
°, ' ö ö ö ö ö ö ö ö
.ö
w
O

O N N1 N
M %0
U N 00
N CO
r-1
Co CO CO .- CO
CO
"v O O O O O O O O O
U U
444

3
a)
a ß. 1 1 a)
4 0)
04 C13 cu
`y N E a
LO 'CS d
N
A w n ý 5 n
Q
(
(n 1-4 w '.
cd
 - 112 -

b
v
gip %0 N
n4 m -4 °i `e
' N ö
4-1
U)

aý \o
O rn
O Ln
9
o
N Co 'o N
. -4 9-4 'I
v GI O O O O O O O

CY% to 00 tf)
CD 1- ui 0 L- 10 c; %
ö Ö Ö Ö Ö Ö
r-i

U) dq N
ö '-. 1 '0 N Cl) "-4
N 0% CO d O O' cN

cd
rZ CD c; c;
U 0
H
a) b
ý Ö c4 10 P-f Ö
0 Ö Ö O O
H E Ö Ö Ö Ö Ö
O O O
O
2
0 U)

O
'd
0
cd

I ^
N o
oC 0o Ö oCo 00 O
0 0 0 ri 0 0 '-+

C)

ctS
00 O N CO . -1 N
O
Q) n N O> Co . -4 CO G'. O

10 O O O O O
O
., r
H
Q
cd

a a
,
wO
cii

Q
O .A
U O
O
r-I C) C)
0
b`0 m U)
C
v C) ai
U U
H
OC i-1
a N
N 0 . -1 N M d+ Cl d1
Q
ý--4

H
 - 113 -

- N -
- i d+ - -M
M . . .
M

I
"-1 . --ý . --1 . -1 ri . -1

r,
A N
O
[-
N
cr%
O ö
0 C)N
so

V C Ö O O
ý-+ O O

ct3

N N *10 m %0 cq
v M M M

0
00 ce) Lo -A

E-+ U ö ö C ö

ýº N N t- C7%
U Ö r-4 0-4 - -
U ö ö ö ö ö ö ö

ö ö ö ö ö ö ö
aý ö ö ö ö ö ö ö
aý
A

ril

m ° ° mt
cd
v ö ö ö ö ö ö ö
E
0
q0 10
U 1
Q
i 0
vi a) v
c) a, aý a
-.
cd . M
G 9
cd
aý Q
o 'U 4-1 .0 4-1
F-4 F-4
A
Ö

Cý7 ho
0 ...

t'- a4 N M
N .-I N M "Im
0 cd
tn
A
Ey
 - 114 -

content. Ratio 2 and 3 suggest the presence of a cochromatographing quinic

acid free substance that is periodate-inflating; (2) the mature samples do not

achieve the provisional norms set out in Chapter 2; (3) when the mature beans

were roasted, brewed and tasted by Illycaffe, it was found that the beans were

of low beverage quality. Illycaffe suggested this may reflect the beans having
been transported in the cherry. Amorim et al, (1977) commented that such a
practice adversely affected cup quality.

The dry processed samples showed a progressive increase in HPLC

CQA and HPLC DCQA. In contrast, PV, MV, TBA BDQA and HPLC FQA showed

an overall rise with'a temporary decline at the semi ripe stage.

Comparison of the dry and wet processed samples showed that at

maturity all analytical values and average bean weight were lower for the wet
processed samples.

Reflectance spectra for the most immature sample indicated the

presence of chlorophyll in the silverskin, but this had disappeared by the yellow
"The presence of chlorophyll is indicated by the dip at 650-700 nm.
stage.

Summa

The C. Arabica var. Marogogipe and IV C Indenie were characterised

by very low TBA BDQA, total HPLC CGA, PV and MV in the most immature

samples. (These values rose progressively. ) Even allowing for experimental

errors, the values for Ratio 3 were substantially higher than the norms.

The C. Arabica cultivars from Kenya and those from Colombia were

quite different. The most immature samples had higher compositional values

and near normal values for Ratio 3. This observation could indicate that these

samples were more mature than the least mature Indenie and Kew Arabicas.

In contrast, the most immature IVC Robusta samples had a higher PV.
MV, TBA BDQA and total HPLC CGA.

For both Arabicas and Robustas the Analytical Ratios suggest the

presence of periodate-inflating cochromatographing substance(s) that do not

yield quinic acid. Apparently the Robustas have a much higher content than the

Arabicas, ' but this is the normal pattern for phenols in these species. It is not
 - 115 -

possible to identify these substances. However, sugar esters would probably

interfere as described above and are possible biosynthetic intermediates to

CGA. It should be noted that these substances are also detected by the

molybdate reagent suggesting the presence of a caffeic acid residue. Griffiths

(1982) who examined the same samples of beans from Kew, suspected that a

caffeoylglucose was present and that the level rose from approximately 0.5mg/

100 beans at -19 weeks to a maximum of 11mg/100 beans at -2, and then declined
to approximately 6mg/100 beans at full maturity. This substance cannot

account for all the interference reported here, since it did not cochromatograph,
and the quantity is far too small.

Chlorogenic acid quinones, if present, would not have been detected

by HPLC. Caffeoyl shikimates would have yielded shikimic acid on saponification,

and this would have been detected as quinic acid. Accordingly, neither could
these substances have been responsible, and for the present the precise nature

of these substances must remain unknown.

The CGA content generally rose in both Arabicas and Robustas as the
bean increased in development until full maturity.
 - 116 -

CHAPTER FIVE

CHARACTERISATION OF PECULIARLY COLOURED COFFEE BEANS

 - 117 -

I Introduction

The sorting of green coffee beans by colour was described In Chapter

1. This is an important step in producing green coffees of consistently high

quality. Such an operation can only be efficient if the relevant pigments have
been characterised.

The substances responsible for the green bean endosperm pigments

are not known, but several theories exist:

(1) Green and blue-green pigments - oxidation of magnesium chorogenate at

alkaline pH values. Northmore (1967).

(2) Green and blue-green pigments -oxidation of kahweol esters Gibson (1971).

(3) Yellow - brown - black pigments - enzymic oxidation of CGA. Luh and
Phithakpol (1972).

Gibson has reported also that the silverskin of some coffees may

contain chlorophyll (1971).

This investigation centres upon the peculiarly coloured beans.

Illycaffe (personal communication), using a Perkin Elmer Model 554

reflectance spectrophotometer as an analytical tool had observed that certain

beans with peculiar green silverskins gave a spectrum with strong absorbance
in the 650-700nm region. This suggested the presence of a chlorophyll or
similar pigment. Preliminary tests showed that:

(a) these beans turned brown in acid, consistent with the acid catalysed

conversion of chlorophyll to phaeophytin (see Fig. 26, P. 119).

(b) these beans turned blue-green when heated in dilute copper sulphate,
consistent with the formation of a copper-chlorophyll;
(c) if a typical green coffee bean was coated with commercial chlorophyll then
its reflectance spectrum became identical to the peculiar beans.

These results were accepted as prima facie evidence for the presence
.
of a chlorophyll in the peculiar beans. However, chlorophyll is not commonly
found in plant tissues that are not exposed to light, and since the nature of green
coffee pigments is controversial and such pigments are commonly linked with
 - 118 -

quality, it was considered desirable to confirm these preliminary observations.

II Extraction and Characterisation of Chlorophyll

Chlorophyll is the most widely distributed natural pigment and occurs

in the leaves and other parts of almost all plants (Humphrey, 1980). All

chlorophylls are tetrapyrrolic pigments. All contain magnesium. All exhibit

pronounced absorbance peaks in the blue-green and red regions of the visible
spectrum. All exhibit similar, though not identical, solubility in various

solvents.

Chlorophyll in living plant tissue is confined to specialised chloroplast

cells in which it is present in a colloidal suspension, and spectrophotometric

studies indicate that a major proportion is an associated form, possible with

proteins and carbohydrate (Aronoff, ' 1966).

Robinson (1980), reviewing such pigments, reported that by careful

spectroscopy it has been possible to distinguish as many as six different forms

of chlorophyll a in vivo on the basis of absorbance maxima ranging from 663

to 700nm. Chlorophyll b in vivo has an absorbance maximum at 650nm with a

small band at 640nm.

The structural relationships of the chlorophylls are summarised in

Fig. 26. Protochlorophylls are of interest since they have been found in seed
coats of Cucurbitaceae (Jones, 1966) and are identical to bacterial chlorophylls

which may be biosynthesised in the dark (Jones 1965).

Chlorophylls are labile. Holden (1965) reported that they may be

destroyed by lipoxygenase to yield colourless oxidation products. She later
(1976) suggested that extracted chlorophyll may be stored safely in the

extraction solvent but other workers have suggested the use of liquid nitrogen
at -20 to -30°C (Wickliff and Aronoff (1962).

Humphrey (1980) described some of the more important characteristics

of chlorophylls. These include (1) the ease with which magnesium is lost by

the action of dilute acids or replaced by other divalent metals; (2) the ease with

which the phytyl ester is hydrolysed by dilute alkali or transesterified by the

lower alcohols. This is illustrated in Fig. 26

 - 119 -

Chlorophyll a.

-Phytol -2H (C-76,8b)

-Mg

Phaeophytin Chlorophyillide a. Protochlorophyll

-Phytol -2H (C-76,8b)

Phaeophorbide a Protochlorophyllide a
(vinyl Mg phaeoporphyrin a5)

-carbomethoxy +H20
of C-6e hydrolytic
cleavage between
C-6d, e

Pyrophaeophorbide a Chlorin
ebmonomethyl ester

Figure 26: Nomenclature of the Chlorophyll and

some of its breakdown products.
(Aronoff, 1966)
 -120-

Acetone is a solvent commonly employed to release chlorophyll from

protein (Holden, 1976) because:

(a) 807 or higher concentration lipoxygenase is inhibited;

(b) transesterification does not occur.

Accordingly, the acetone extraction method recently developed by Humphrey

(1980) was chosen for this investigation:

1. Extraction of dried plant material with 80-90% acetone

(10mL x3 times)
2. Petroleum ether (40mL is added. )
3. Portions of water are added.

4. The water portion is removed with acetone. When this is complete,

petroleum ether solvent is partitioned with 80% methanol.

5. Petroleum ether portion is washed several times to remove traces

of acetone and methanol.

At this stage chlorophyll separates out as a fine suspension. This is

filtered off, and the precipitate is dissolved in diethyl ether. Reprecipitation

might be necessary with petroleum ether.

However, a trial showed that the chlorophyll was lost during the
cleaning up procedure and the method was simplified to an extraction of

chlorophyll with 807 acetone.

Method

Samples of ten whole beans were extracted twice with 5OmL portions of
80% acetone. The extracts were immediately pooled, concentrated under

vacuum and the absorbance spectra re corded from 350 to 700nm. The four

types of coffee beans used were normal beans (A), beans with the dark green
silverskin (B), light green (C), and 'foxy' red silverskin (D). For comparison

a standard was prepared using chlorophyll extracted from fresh grass (E).

The extracts were examined by TLC. Aliquots (2mL) were concentrated

and then streaked on a 500.µn layer of silica gel (Merck, Germany) prepared
using a Shandon TLC spreader. The plates were air dried at room temperature
 - 121 -

Hj
i
cn
w
b
r
cý c
tz
to
0

0
ni

C)
tti

tu
ru

0
121

t3l

m
zt: d
p

Q1
CD o'°
n

Oc2ý

fD
cD

r"

r"
 - 122 -

C
m
r"
L7'
rr
CD
CA
b
fD
A
I.
Pi
I.
'b 0
cý r, h
fD
00
N
w
ä
C)
öö 11
0 0
a 0
CD

pM
7r
r.

v
O
C

w
W
b
O
f
t

Fd
r"

fD O
W tiO
O

a
co
n

0
H
0
b

v
C
0
 - 123 -

and activated at 110°C for two hours. The plates were developed in hexane/

acetone (60/40) using a vapour-saturated tank in the dark.

Results and Discussion

Two peaks are distinct in the spectra, one in the region 420 to 440nm
and the second at 664 to 666nm (Figs. 27, '28 and 30). These peaks are
typical of chlorophyll (Robinson, 1980), thus suggesting that for samples B and
C the pigments extracted included chlorophyll and/or related compounds.
These peaks were not detected in the spectrum obtained for good quality beans

or beans with foxy'red silverskin (Fig. 29).

TLC: Fig. 31 shows the chromatograms. No pigments were separated

from the good quality beans.

Chromatograms B to E (Fig. 31) were compared to those of Gibson

(1971) who used the same system.

It was shown that coffee with dark green and light green silverskins

gave almost identical chromatograms. The beans with a 'foxy' red silverskin

gave only two components. Identifications are based primarily upon the data of
Gibson (1971) and cannot be taken as unequivocal. However, it would appear

that both green silverskin samples contain chlorophyll and/or related pigments,

whereas 'foxy' red silverskin coffee is quite different. It had been thought that

the pigment of 'foxy' coffees might be phaeophytin', a brown haem produced from

chlorophyll when the magnesium is replaced by hydrogen. The separation of

extracts in Fig. does not support this; one band from 'foxy' coffee appears
to be lutein (yellow-orange) and the other may well be another carotenoid.

One assumes that the pigments extracted from the silverskin have been

synthesised within the coffee cherry in the absence of light. Accordingly they

may resemble the protochlorophylls that Jones (1965, '1966) found in the seed

coat of Cucurbita pepo. However, Boardmann (1966) reported that a proto-

chlorophyll can be converted to chlorophyll by light and heat. The rate and

extent of conversion is temperature dependent, reaching a maximum at 400 C.

Such conversions may occur during sun drying or mechanical drying of coffee,

yielding a more complex mixture of pigments.

 - 124 -

Figure 31: Thin Layer Chromatographic Separation of

B Dark Green "silverskin' coffee beans;
C Light Green "silverskin' coffee beans ;
D Red foxy "silverskin" coffee beans ;
E Synthetic chlorophyll (STD)

11 00000

10
- "- "- " - -"- " -"
" " " " " " """ " "" 9 """""
Y " " y f y yf y f .Y 8 Yyyyy

V v v v v v 7 v v v v v 7 vv V-7 v7
i 6

--- ý5
4
........... 3

1A fffVfv
1

BCDE

Key: 1. Neoxanthin 7. Chlorophyll a

2. Violaxanthin 8. Unknown
3. Unknown 9. Phaeophytin
4. Unknown 10. Cryptoxanthin
5. Lutein 11. B= Carotene
6. Chlorophyll b 1A. Chlorin
 -125-

4 However, since this examination of coffee silverskin pigments,

Pierpoint (1982) has refocussed attention on allagochromes. These water-

soluble pigments are now known to be formed from CQA-quinone-protein or

amino acid interactions.

The colour produced depends upon the pH value, being blue-green at

pH 10.5 and red at pH 5.0. According to Pierpoint the blue pigment has an
absorbance maximum at 680nm, but Haberman (1972) stated that this peak is

variable and falls in the range 640-680nm. As stated earlier, chlorophylls

have maxima in the range 640-700nm and the Illycaffe spectra show a peak in
the 650-700nm range.

Retrospectively one cannot be certain that these silverskin pigments

did not include an allagochrome, especially as Garrick and Habermann (1962)
have reported such a pigment in the leaves of Coffea Arabica. However, ' it is

unlikely that allagochromes would show the same chromatographic behaviour as

the chlorophylls, especially since they differ in their polarity. Nor is there any

evidence of allagochromes having maxima near 420nm.

Conclusions

Chlorophylls or chlorophyll derivatives were found in the dark green

and light green silverskins.

III Chlorogenic Acid Analysis of Peculiarly Coloured Beans

Six batches of peculiarly coloured beans were supplied. Four were

batches of commercial green coffee beans supposedly of the same maturity, but

different degrees of spoilage or discoloration. The other two batches were of

beans that discoloured during processing while preparing the immature samples

previously discussed (Chapter 4).

(1) Kenya Arabica beans supplied by Illycaffe which had been sorted by two

methods: (a) by eye, yielding five samples of decreasing acceptability and

increasing discolouration;
(b) by machine, yielding selected and rejected fractions.
 - 126 -

0 00 N C7% 0 ý 00 N
"1 N M 0
ec
ýaQ ö ö C C C ö ö ö
w "
ý4 N V-4 cq 00 C) N N
Co U N N N
U N r-1
U) b O O O O O O O O
U
Qi

4 CO
<4 U b
ý N
'O
'. O
0
NO 1
Cl)
%0 mil 00 CO
r -f
IN
U
.. 4 Ö O O O
N '5 O O O O O
N =
a
0 M 00
.< ..I ti) co
O
m
ON
"D
0
N
M O
O "--4
O E-4 4 O "-ý O
U
'0
C)
Cf) ON 00 C14 CY) 00
-4
C)
O O O O O O O O

O
v
'o C)
In Cd N
00
00 C N
cu 1.0 114
CO CO
Co >, "ý ö C ö ö C ö ö ö
o
U

'U
4) C)
Fa
`d CO 0 i 4 ' N
ö 00 "-ON
O 00 co 3 t f) cý
O "Pd ö ö ö C ö ö ö ö
U

C)4
al CY)
P0 - . -1 r-1 r-f rl r1 "--1 . -f
w cl) Q) O
0 w 0 0 O O O O O
V

cd U
C) W
O U
U O C. 4-1
U, E O N N
co Cl) IN r-1
O
N N N ' N N 03
0 >E o 4 N
to cn
0 v P-4

u Cd Q)
Q)
00 v Ü
N «i
"ý U
C) U
1-4
fa W
. z '- .ý v 1-. .ý '0 00
C13 .--ý N M v to .ý
H
 - 127 -

0-
es 6

ABCDE

Figure 32: Photograph of Discoloured Coffee Beans

 - 128 -

(2) A batch of Kenya Arabica sorted by eye in our laboratory,

similar to 1(a).

(3) A batch of Santos Arabica sorted by standard commercial

equipment at Illycaffe. The rejected fraction further sorted by eye to provide
seven discoloured samples. The UV visible reflectance spectrum was recorded
for each bean. Beans with similar spectra were grouped together in an attempt
to produce homogeneous samples.

(4) A batch of Santos green coffee beans sorted by the standard

commercial machinery to remove defective beans. The defective beans were

re-sorted, using a similar machine under computer control. The greater

discrimination generated six groups of defective beans; each bean in these

groups was examined on the Perkin Elmer Model 554 spectrophotometer, and

any beans not typical of the group were discarded. Thus these groups were homo-

geneous.

(5) A batch of immature Arabicas from Colombia which discoloured

during dry processing.

(6) A batch of mature Arabicas from Kenya which discoloured during

dry processing.

Methods

The six batches were extracted and analysed with periodate, molybdate

and thiobarbituric acid (TBA) '

reagents, as previously described.

Batches la, lb, 4, 'S and 6 were also examined by HPLC.

The discoloured beans from Kenya (Batch 1) are illustrated in Figure 32.

Results and Discussion

Batch 1: Discoloured Kenya Arabica from Illycaffe

Batch 2: Discoloured Kenya Arabica sorted by eye in our laboratory

The data are presented on Table 28. The higher moisture content of
Sample 5 (Batch 1 and 2) is noteworthy. The presence of water could suggest
 - 129 -

° 0 ö 0 ö
...,
4.1 CY2
-
r(
«i
N

cd
Qi

E ö ö 00 c4 -1
a
ö ö ö ö ö

G) tt) CO IN N o.
-4 O O O O O
.d ö ö ö ö ö

000 000 00
Ö
O O O O
CD a te,
ö ö ö a 00
0 CD 0
U

+-+ 00 N IN '0 O
'8 O O O - N
G Ö O Ö Ö Ö
.p

O
4.1

'O U
to CD
cd 10 00 N
ý, C. cc 114
P-4
4 . . .

ctt 0 > O O O O (D
O
C4
"Vy

W ril U
a C v. O
O ,ö öN.
O
° `e r1.4 0ö 00 ö ö ö
U G)
Q

CY) c4
4 11
4-4 O
O Ul ö ö ö ö ö
t I ca
aý3
ö,
N

00 N N N e
Qi cý
'g °
tn 0 in 'O % 'D N
co

rn
CV CC)
C)
A --4 m ii in
csS
H
 -130-

that there had been a favourable environment for enzymatic activity during

storage. However, ' the higher moisture content could have resulted from

enzymic changes such as the oxidation of caffcoylquinic acid to the

corresponding quinone (Fig. 21, p. 84).

All samples show variable bean weight, but there is no doubt that

sample 5 (Batch 1) is significantly lighter. This could suggest that these beans

were harvested at an immature stage. Alternatively there would have been

loss of dry matter during storage.

The MV, PV and TBA BDQA value decreased as the degree of

discoloration increased. Estimated FQA and DCQA values increased at the

same time.

The selected. bean sample had a higher PV, MV and TBA BDQA value

than the rejected bean 'sample. Estimated FQA and DCQA are higher in the

rejects than in the selected samples. These results are consistent with those
for 1a.

The colorimetric data for Batch 2 (Table 29) is essentially the same.

The HPLC data (Batch la and lb, Table 30) was in agreement with that

from colorimetric analysis. There was a progressive fall in CQA and DCQA

and a rise in FQA (Batch la). In Batch lb the selected samples had a higher

CQA and lower FQA and DCQA than the rejected sample.

Unexpectedly the best sample from Batch 1 had abnormal values for

Ratios 1 and 2, and the worst sample had apparently normal values for all three

Analytical Ratios (Table 30). For Batch 2 only Ratio 3 could be calculated.

Again there was no significant change in this value from the best to the worst

sample. Two tentative conclusions followed:

(a) possibly the best sample had deteriorated and was not of

particularly good quality;

(b) the provisional norms adopted in Chapter 2 are not appropriate
because they are not sufficiently discriminating to reject

obviously discoloured beans.

 - 131 -

.1 Cf) N 00 Ö
N
cM )
CV) C? O
M

Cd
a
0
Öi Q Os N 0 0
v c'; 0 Co 00 0o
t' O - O O O O O O
j,
cd
Qi

0 O5 LI) N N N %0 CO
r1 "-i 0 . -4 N N ý-4 . -I . --I

Qi

a
U
a N N V-4 % ON cq N 00
te' %0 Cl) N '0
O O O O O O O O
ad

r-
ä COr-t :° co r-f Ln °-
0
c; c; ö ö ö ö - ö
ö
a

U)
C14 " N If) t- r- r- co

b ö ö ö ö ö ö ö ö
0
0
U

.A-

O
U ý ö ö ö ö ö ö ö ö
cHtt
En a
O O
Ö O
U
UCD c
CD 04
w
o9 O
cd C)

ö moo a HG)
ö ,
M v G)
Q
c) n
A 4 N Ln E
r1 co ýr+

Eý
 - 132 -

4 Ö Ö 00 CN c4
O O cr) O

"-1 Cl) Cl) N 00 M e

Ä N N O "-i N V)
Ö O O O
v O O O O
-4

O Cl) CD ` clq ö Ö
.O
-4 H '. 1
i-I ' 4
0 b w ö ö ö ö 0 ö ö 0
U

'O U 0 0 Ö N
C) 00 00 i 000 N 1t) Cl)

I Ö Ö O O O O
O O

G)
4)
'C
9
4.1
0
0 4 P-4 P-4 ö O\ IM 00
U) O O O
csS
U)

CA a ö Co ö ö ö ö ° ö
G)
N ö c ö c c ci ö ö
0 W
U
v 00 rn c C) Co ý i
C) cc c 000 cc vi
ö ö ö ö ö ö ö ö
C7

%0 tn CD u3 lief e \o L*-
CY% CY\ C) a- 00 00 \o e
0 a ö
LO 0 ö ö ö ö

cd
U

O
N
U ö ö ö ö ö ö ö ö
C)
W
W
O

[A lý 'i N %M Co %D If) N Cl
O
cd '0 N N
4) 00 `n .ö ýc
O
C] M :5u
6v

N
- Cn rý
cr) o "ý
C)
6 Ä
.
t3
cn cCU O4 U W C7 x .-,
F
 - 133 -

Table 32: Type and Taste Assessment of Rejects from the

Computer-linked Standard Commercial Sorting Equipment

Sample Numbers Type and Taste of the Coffee Beans

Sample 3 Whitish beans: balanced, slightly woody taste, little

body.

Sample- 4 Partially foxy beans with silverskin: good aroma,

considerable body.

Sample 5 Deep-green beans with silverskin: bitter, astringent

taste (slight stinker).

Sample 6 Dark-green beans with silverskin: bitter, astringent

taste (strong stinker).

Sample 7 Dark-green/brown beans with silverskin: bitter,

astringent taste covered by very strong stinker.

Sample 8 Black beans with silverskin: limited sample - not

tasted.
 - 134 -

Batch 3: Samples of Green Coffee Beans sorted Electronically

It had been intended that these samples should be homogeneous with

reference to their reflectance spectra. In practice this was not achieved, due

in part to the limited discrimination of the sorting machine.

The colorimetric analysis data of Santos 571 (Green Arabica Coffee)

are presented in Table 31.

The results showed that the discoloured samples fall into three groups:
(i) A/B, C and D/E;
(ii) F and G;

(iii) H and I.

Within the groups the samples are similar, but PV, MV and TBA BD QA decline

progressively from group (i) to group (iii). Estimated FQA and estimated

DCQA are little different in groups (i) and (ii), but higher in most discoloured

group. Group (iii) was distinctly smaller and had a slightly higher moisture

content. Ratio 3 was able to discriminate only the darkest beans (Sample I) and

not sample H from group (iii). This suggests that Ratio 3 lacks discrimination.

Batch 4: Santos Green Coffee Beans sorted by the Standard Commercial

and Computer-

Standard commercial sorting yielded a rejected sample having lower

PV, ° MV, TBA BDQA and HPLC CQA, I but higher FQA and DCQA by both methods

of analysis (Table 33 and 34).

Computer-linked sorting of the rejects provided six samples. Table 32

contains descriptions of the colour and liquoring characteristics of these, and

UV-visible reflectance spectra are provided in Appendix I. For any one sample

obtained by computer controlled sorting, there is very little bean-to-bean

variation. The occasional spurious bean was found and removed from the sample

to prevent any undue influence on the analytical data.

On the basis of these descriptions these six samples fall into three

groups.

Sample 4, the partially foxy beans, with a carotenoid-pigmented silver-

 - 135-

CD \O %0 CD c. e4 °
2C Ne u3 M 414 "

ýA

CO vom' ° co r4 ° 0 ý
C)
t3 w ö ö ö ö ö ö ö ö

'O N ý CO
CO
0 u3
ý 0 0 0 0 0
0 O O 0

01 0ý O% 00 00 00 %0

Ö Ö Ö O
[--ý 0 O O

`C7
CV O d+ N CO 0
%, '0 CO
O . r-1 "-4 r-1 O O O
'-4
Ici ö ö ö ö ö ö ö ö
a) a

::I

a)
4J O N ö 0
0MO rn e
? c> r 0Mp
ö Ö Ö Ö Ö
c; "-i

'o
r
cd
'L7 0
H 1o ö CO ö 0 CO n
Cd O\ CO
b o 0 Ö
G ce . -i . -i
cd
U)
E
0
w
4.1
l) ho %0 ;° i
-
»i
ö ö ö c c ö ö ö
c) U

0
U H
O
O 00 0. o in "-+ Co CO
aý IU ö CO ö o: oý CO or

aý
ýdi
U)
cd
'L1 'U
9)
0
W

0 ý w y C)
Id Cl "ý U
aý ýn U . i H 4-4 Ll U G) U
CD .

0
to
cu ý. CA 8 H (1) u rd " = b b
r 4-4 E 41 Ei Zý4 8
m
u2 U cn vl ° vn U co a A
Cr, a , Co
Q)
v r- 00
A (n "--1 N M d+ In
ýýi
E-+
 - 136 -

%0 C7% C7% tr) n in -11 N

M O CY) LI) M 1O)
cl

O
4-1
cd
P4
P-4
v n cq 00 00 rlý C14 CY)
N (ON O 00 ON Ö CIM
O O ý-i O r+ O "-+ "-+

..r4
0
W N N N 0 00 N 00
10 '-I
N CV N
a) . . . . . . .
x

U
r- cq

Ü
%0 %0
N O N N ° r- CY )

ö ö ö ö ö ö ö ö
'p
r
cd H
cd
U

E
E
O
U ý Cq N N
d
V-I 00
.- I N N-1 r-I
N
bH O O O
O O O O O
cd
"O
cu
4.1
u2

E
0
a
COö ö ö ö ö ö Cl)
G". w ö ö ö ö ö ö ö ö
cd
C)
CA
C)
G)

O (7%
N Lf) 00
c
U q cM 1 W " d+ N

Q)

4-4
0
P, 'a
rb
r..4
Cd
S O0 4
öýA .N
`r ý ý
ßi0 .Q
14
40, u (U
-4 Q)
b 'L7 OO ºý O O O
NA
Um d ö Ä ý b) ý'o
ö o
,, Ü
,eA b0
cn cn
j s
c
Ä Ä w

E
11, m

Cl)
C)
E9
zO 4-4 -1 C,
9
,ä e*; w in o N 06 1
 - 137 -

skin have good liquoring characteristics supported by Analytical Ratios which

are just within the normal ranges provisionally proposed.

Samples 7 and 8, the dark green-brown and black beans, have poor
liquoring characteristics. These samples had extreme values for Ratios 2

and 3, and sample 8 had the significantly lower average bean weight (Table 34).
Periodate Value, MV, TBA BDQA and HPLC CQA typical of the blackest

samples from previous batches (batches -1 and 2).

Samples 3,5 and 6 are low quality beans of types which are either not
present in, or not resolved from, previous batches. Sample 3 was pale,
bleached, ' and samples 5 and 6 had spectral properties characteristic
possibly
of chlorophyll-pigmented silverskins. The values for Ratio 1 were within the

provisional norms, although samples 3 and 5 are borderline. The values for

Ratios 2 and 3 are extreme.

Batch 5: Immature Arabicas from Colombia discoloured during Dry

Processing

Batch 6: Mature Arabicas from Kenya discoloured during Dry Processin

In both cases the discoloured beans had lower MV, PV, BDQA and

HPLC CQA, and in the case of batch 6, lower average bean weight (Tables 35

and 36). The samples in batch 5 had identical average bean weights. This

tends to suggest that discoloration does not cause a particular weight loss, and

that the lower average bean weight associated with the darker beans in batch 5

reflects the relative immaturity of these beans.

The Analytical Ratio (Table 36) for batch 5 were extreme, but this was

not surprising since these are immature beans and must be treated as a special

case.

The seco nd sample from batch 6 had Analytical Ratios within the

norms. With progressive discoloration Ratio 1 rose to an extreme value and

then fell to an extreme value, presumably passing through the normal range.
Ratio 3 rose to an extreme value and then fell towards the norm, whereas Ratio

2 rose progressively and did not fall back.

Ratios 1 and 3 diverge from the norm as discoloration begins, but

 -138-

º3
N tr)
d
I--1
to LI)
0
Ä
'p O Ö
ö ö ö

O
Q %O
o
%0
O
CO ö ö
(l O O ö ö ö

Ö ö ON
tO)
O Ö Ö
C; ö ö
a m
rn N
N r-4
O O ö ö ö

10
O
ý ö ö ö 0 ö
0 o ö ö ö
0
m
Q)

Q
4-1
O
'0Cd 000 CO
O 13) N O ON
0 ý :
A
ON 0

p O O ö ö ö
b
N

cl
Ö Q) 00 4 Cl) Cl)
a) N --1 O\ O\

b O O O O O
(n
cd
U)
41
Cd ý
a) Cd M Cl) Lf) Cl)
Hp . -. O O 1-4 V-1
O Ö Ö
'a >v O O O
a)
O
O
U HC)
Co cd
U LO tl) N
Cl)
«3
W OO 0
O cd
H U tp, Cd
En
9.1 cd
a) cd
a)
O A
9 U) Ö U)
U 10
0O C) Q) N
Ülýbvo ý4
: ? äi Cýl Q)
%-D a a A°
N Q) Ü a
M O
Cam) H
O J G HV
cu Ü as Ö 0) Ö
U .'
.0
cd
i. +
cd
q
11 4 a A
H
 - 139 -

Co - N CN
N I'O
O M CO
4-1
cti
r4
ce 00 M N 0 N
O 00
N n M
O O "-+

O M eM oo O
e}+ 00 N t! ) O

4 4

U oo u3 r-1 M
N LI) "--1
,,° --+ o
ö O O O
F
a

U 0
ä ö ö
N
ý-4
U Ö
O O
A
.Ä

H to Cl) Ö
O ö . -+ .i
a) a ö
'O ö ö O O O

cd
U)
cd
N

a) N Cf)
C) 0 - ö M
10 a O O O
O M
0
O Cd
O C4 m
U
U) Cd Cd
U U)
U
O cd
U
U) 8 'G 10
O 0
a+
Cd
m 'a
10
U) Ö i
vi
ý 0
a)ý, Q.C. EnP-4
M rn 'CJ -.4 (U P-4
Uo
CD O
Q)r. Q)_ yi P L) >'
0
i
0
ääi
0`
Üý 0
1-4 1-4 U
LO
cc Q) ;: 5 00 äo ' 0 Cd
"ö 'o ä äa
m Cd
Q.
) Ui U
- N -I N Cl)
Cd cß
Eý 94
 - 140 -

because these ratios approach the norm once more at extreme discoloration,

neither can be considered particularly discriminating. Accordingly Ratio 2

would seem more appropriate.

Summary

Integrating the results from the six batches one finds that there are

two types of peculiarly coloured beans: (a) those that fall in the sequence

yellow-brown-black, and (b) those with silverskins which do not, e. g. pales,

greens or foxy beans.

Type (a), which accounts for the majority of samples used in this
investigation were characterised by progressive decline in PV, MV, TBA
.a
BDQA, HPLC CQA and usually lower average bean weight and higher moisture

content.

In many cases (e. g. Batches 1,2 and 4) the most discoloured beans

had a higher estimated FQA content. This could reflect the presence of

quinones, since these are intermediates in enzymic browning that in model

systems were determined as estimated FQA (Chapter 3.). Amorim et al (1977)

had associated CGA quinone with low quality coffee. Increased discoloration

with decreased CQA content also suggests an enzymic browning mechanism

could be responsible for the discoloration. Model system studies (Chapter 3)

showed similar changes in PV, MV and HPLC CQA but not in TBA BDQA.

The Data from batches 5 and 6 suggested that the discoloured beans

might have been immature. Immature beans would have a higher moisture

content and higher physiological activity, and thus be more prone to mechanical
damage and subsequent enzymic and chemical changes.

Only in batches 4,5 6

and do the Analytical Ratios succeed in
discriminating the peculiar samples. Ratio 1 was least useful in this respect

and this is consistent with the rise and fall observed in batch 6 as discoloration

progressed. Ratios 2 and 3 were significantly higher than the provisional norms
in all reject or peculiarly low in batches 4 to 6. (The
coloured quality samples
partially foxy sample fell within the norm and had acceptable liquoring

properties. ) Rises in Ratios 2 and 3 could be explained by loss of TBA BDQA.

 -141-

Loss by hydrolysis is possible but there was never a marked rise in free QA

that might be expected as a consequence. Alternatively this could be explained

by the production of a compound(s) which does not contain QA but which co-

chromatographed with the major CGA and which reacted with periodate.

The unsatisfactory performance of the Analytical Ratios in batches 1

to 3 could be due to the relatively poor sorting and thus non-homogeneous nature
of the peculiarly coloured beans, since Ratios 2 and 3 perform much more
satisfactorily in the better defined samples of batches 4 to 6.

Further improvement in discrimination could occur if the width of the

norms could be reduced. This may be possible and would in turn be favoured
by the use of samples of guaranteed homogeneity.

The origin(s)-of peculiarly coloured samples of type b is unknown.

There does not seem to be an obvious reason for some coffee beans to synthesize

chrlorophyll in the silverskin. Nevertheless this appears to be a characteristic

of the most immature sample of beans from Kenya and Colombia (see Chapter 4,

Appendix D& E) which normally retain their silverskin during processing.

However the beans in samples 5 and 6 (batch 4) are too large to be more than

slightly immature and possibly the fundamental peculiarity is the failure to shed

the silverskin, rather than any immaturity. Such a synthesis is known in

other seed coats (Jones 1966) and Sheen (1973) recorded that tobacco with

chlorophyll content also has a high CQA content. Of the two current theories

regarding chlorophyll biosynthesis it is interesting that one has a common point

with CGA biosynthesis (Bogorad, 1967; Gassman et al, 1968).

One involves ö-aminolevulinic acid synthetase which catalyzes the

formation of ö -amino-levulinic acid by condensation of succinyl-CoA and glycine.

The other hypothesis suggests the formation of b;-aminolevulßnic acid through
transamination of L-alanine, L-glutamic acid and L-phenylalanine to 1, ö,-
dioxovaleric acid. When L-phenylalanine is the amino donor, phenylpyruvate

will be the by-product. If the latter metabolic pathway is functional in green

coffee beans, then the more chlorophyll synthesized in the beans, the higher

will be the phenylpyruvate production (Sheen, 1973). This may lead to an

increase in the pool size of cinnamate and consequently an accumulation of CGA.

 - 142 -

The silverskin of the foxy beans contains the same carotenoids as

the chlorophyll-pigmented silverskins. These carotenoids appear to be the

accessary pigments and thus there would appear to be

normal photosynthesis
a link between the origin of beans with foxy and green silverskins. Possibly

foxy beans can be formed from the greens by destruction of chlorophyll which

unmasks the carotenoids. Subsequent loss of the carotenoids could then

produce the pales.

Conclusions

(1) The yellow, brown and black discolourations may arise by an enzymic
browning mechanism and may be associated with small immature beans with

higher than normal moisture content.

(2) The Analytical Ratios lack discrimination when applied to

peculiarly coloured beans, but improvements may be possible.

(3) The computer-controlled sorting is capable of producing almost

homogeneous samples.
 - 143 -

CHAPTER SIX

COFFEE ASTRINGENCY
 - 144 -

I Introduction

It has been said that coffee brew occasionally may be astringent (Illy

personal communication; Clifford personal communication).

Astringency is a sensation detected in the mouth when salivary

glycoproteins are precipitated and their lubricating action destroyed (Bate-Smith,

1973). The most studied group of astringent compounds are the vegetable
tannins. Haslam (1981) considering the present knowledge on vegetable tannins

recommended the definition "water-soluble phenolic compounds having molecular

mass between 500-3000 and, besides giving the usual phenolic reactions, they
have special properties such as the ability to precipitate alkaloids, gelatin and
other proteins".

Vegetable tannins are structurally classified into two major groups:

1. Non-hydrolysable or condensed tannins,
with a flavan-3-ol repeating unit;
2. Hydrolysable tannins - these are easily
hydrolised by acids, bases and in some
cases hydrolytic enzymes yielding either:
(a) ellagic acid (ellagitannins); or
(b) gallic acid (gallotannins).
Both are usually found esterfied to glucose or a related polyol. In one case,

tara tannin from Caesalpinia spinosa quinic acid replaces the glucose.

In a preliminary investigation (Clifford, unpublished results) coffee

beans of three species were tested for the presence of traditional tannins using

the screening tests developed by Haslam (1974). These tests were negative.
It is generally accepted that the structural feature responsible for the astringent

properties is the presence of at least two separate 1,2-dihydroxyphenol residues

in a compound to produce the glycoprotein cross-linking and precipitation
(Haslam, 1974).

Only one group of known coffee constituents appears to resemble this

description. This group is the DCQA (Fig. 34). It must be noted that some

workers,. e. g. Haslam (1973), have said that CGA are not astringent, but it is

probable that he was referring to the mono-caffeoylquinic acids rather than

DCQA. The DCQA resemble the galloylquinic acid tannin (Fig. 33) found in
 -145-

OH

Figure 33:
Tara Quinic Acid Tannin (n = 0,1,2). (Haslam, E. 1981)

Hooc OH

Hö CH

Figure 34:
3,4 Dicaffeoylquinic Acid. (Waiss et al, 1964)
 - 146 -

Caesalpinia spinosa and since this compound is known to be astringent this

resemblance reinforces the slk &$, that DCQA may also be

structural

astringent.

To further clarify the taste properties of the CGA professional coffee

tasters were asked to taste:

(a) 0.1% aqueous CQA
(b) 0.1% aqueous DCQA

The general concensus of opinion was that CQA tasted acidic. In

contrast DCQA had a lingering astringent taste that was quite distinct from

CQA. It was felt that these results justified a more thorough investigation of

coffee astringency.

II Organoleptic Investigation

Sensory Assessment of Coffee Astringency

The sensory value of food or drink can be defined as the scientific

measure, analysis and interpretation of the sum of the effects of those

properties which can be perceived by the human senses (1FT, 1975). Although

sensory investigations are subjective and imperfect and efforts to replace them

with more reliable objective chemical and physical tests have been made
(Reudnitz, 1980), subjective methods are still necessary. However, if possible

subjective methods should be supported by objective.

The methods developed for sensory evaluation are of two types :

(a) analytical, and
(b) affective (Table 37; IFT, 1981).

Analytical evaluation of a product identifies differences or similarities

and quantifies these sensory characteristics. In contrast, the affective method

evaluates preference and/or acceptance and/or opinions about a product. The

affective taste panels are usually composed of a larger number of untrained

persons than are analytical panels.

For the present study an analytical Triangle Test and an affective

Rating Test were chosen.

 - 147 -

Table 37: Classification of Sensory Evaluation Methods

(Modified from IFT, 1981)

Analytical or Objective Affective or Subjective

Methods Methods

I. Discriminative: I. Preference tests:

a. Difference test. Paired-preference

Ranking
Paired-comparison Rating
Duo-trio Hedonic (verbal or
Triangle faciaD scale
Ranking
Food action scale
Rating difference/
Scalar difference from
control,

b. Sensitivity.
Threshold
Dilution

II. Descriptive:
Attribute rating
Category scaling
Ratio scaling
(magnitude estimation)
Descriptive analysis
Flavour profile analysis
Texture profile analysis
Quantitative descriptive analysis
 - 148 -

The triangle test consists of three coded samples, two identical and
one different, presented at the same time. None of the samples are identified

as the control. The judge determines which of the three samples presented
differs from the other two.

The data are analysed statistically (Amerine et al, 1965) to determine

whether a significant difference exists between treatments. The probability of

choosing the different or odd sample by chance alone is one-third (Smith, 1981).

The rating test, also known as the Hedonic Scaling or Affective

Responses, measure like and dislike on a rating scale. This method is more

appropriate at the consumer level, where a large number of people (consumers)

representing the population can assess the food or drink under a given set of

conditions.

The rating scale is simple and flexible; judges do not need previous

experience and the data can be analysed by analysis of variance (Amerine et al,
1965; IFT, 1981). Usually a nine-point structural scale is used, although
several variations of this have been reported (IFT, 1981). These include:
(1) a reduced number of ratings, although not fewer than five; (2) a greater

number of like ratings; (3) omission of the neutral rating category; (4) verbal

categories replaced by, caricatures representing degrees of pleasure and

displeasure, and (5) use of a non-structured, non-numerical line scale, replaced
by like and dislike.

Rating scales suffer from a lack of precision, high variability in

ratings from the judges, and limited application unless large populations are

used.

Method

The threshold for CQA and DCQA were defined by using ten volunteer

panelists from the Department of Biochemistry. They were asked to taste

aqueous solutions of both CQA and DCQA each at concentrations of 0.01,0.025,

0.05,0.1 and 0.2%. All panellists detected CQA at 0.05%. The threshold for

DCQA was more variable, some detecting at 0.05%, others requiring 0.1%.

These data were utilised in preparing test materials for all future investigations.
 - 149 -

Triangle Tests

The taste panel assessments were conducted in the Department of

Home Economics using their specialist taste panel facilities. The panel

members were drawn from the University community using the Home
Economics register of volunteers. First, fifteen people tasted water with

added 0.1% DCQA.

Twenty-nine people took part in the second testing experiment with 0.1 0
DCQA added to coffee brew. The taste panel were informed of the purposes of
the test and this helped to increase tasters' confidence.

The coffee was made at three grams of instant coffee (Nescafe Gold

Blend) to 300 mL of hot distilled water.

All test sampies were served at 60°C. The panellists were asked to

pick out by taste the single different sample among the three cups presented.

The samples presented were: Ist Triangle Test -

(a) distilled water containing added 0.1% DCQA;

(b) distilled water;
(c) distilled water;
2nd Triangle Test:
(1) instant coffee (Nescafe Gold Blend (1%) in distilled water containing
added 0.1% DCQA;
(2) instant coffee (1%) in distilled water;
(3) instant coffee (11) in distilled water.

Rinsing agents such as water, milk or unsalted crackers were used to remove
traces of sample.

Rating Test

The third taste assessment took the form of a rating test. This method

was used to find out the effect of CQA on the perception of DCQA. The

concentration of the acids were 0.1% CQA, 0.1% DCQA and 0.1% CQA plus
DCQA, i. e. 0. lg DCQA added to 0. lg CQA then dissolved in 100. OmL of instant

coffee.

The HPLC analysis of the instant coffee is shown in Table 38. The
 - 150 -

actual test concentrations were 1.45,1.08 and 0.08mg/mL of CQA, DCQA and
FQA respectively. The corresponding data for the original green beans were

not known.

Tests were conducted in two separate batches, each consisting of over

thirty people. In total, sixty-eight judges took part in the taste assessment.

Table 38: Concentrationof CQA, DCQA in the instant coffee brew as

measured by HPLC (mg/mL)

Coffee Brew CQA FQA DCQA

1% instant coffee brew 0.45 0.08 0.08

Coffee brew + 0.1979CQA 1.45 0.08 0.08
Coffee brew + 0.176 DCQA 0.45 0.08 1.08
Coffee brew + mixture
(CQA : DCQA, 1: 1) 1.45 0.08 1.08

At each session individual judges were placed in single booths and were

presented with the three different randomly coded samples for evaluation. All

samples were served at 60°C. Judges were asked to follow the instruction sheet
provided (Appendix F ). The scale was a seven-point scoresheet which was
clearly explained as having the stronger flavour at point 7 and the milder and

more pleasant flavour at 1,

The data were analysed by analyses of variance and difference between

the beverages were tested using a multiple-range test (Amerine et al, 1965).

Results

Triangle Tests
Distilled water with added 0.1% DCQA
Out of the fifteen people who took part in the taste assessment, only
eight selected the odd sample correctly. Using the significant table for triangular
tests (Amerine et al, 1965) these data were found not significant at 5% level

of significance. (One needs at least nine people to make the correct selection
before significance can be recorded at 5% level. )
 - 151 -

Coffee brew with added 0.1% DCQA

Twenty-nine people took part in this taste assessment. Twenty-two

panellists were able to correctly identify the odd sample. Only nineteen are
required to record 0.1% level of significance. Therefore the panellists were

able to select the odd sample and the taste of instant coffee with added 0.1970

DCQA is different to instant coffee with no addition.

When the panellists were asked to describe the taste experienced in

the odd sample, they chose unprompted descriptions like: bitter lingering

after-taste, metallic bitter taste, bitter and sour at first on top front of tongue

and bitter later down both sides of tongue.

Table 39: The Effect of CQA on the Taste of DCQA

Analysis of Variance of the Effect of CQA on DCQA

Source of Variance Degrees of Sum of Mean square Calculated

freedom squares (variance) F values

Total 192 558 2.9

Judges 67 127 1.9 2.0 p( 0.001

Products 2 317 158.5 163.8 p( 0.001

Error '124 120 1.0

I
Since, for products and judges, ' the calculated values of F (Table 39)

exceed the tabulated values the statistical analysis indicates, that differences,

significant at 0.1% level existed between mean scores of products and mean

scores of judges. Arnold and Noble (1978) reported that even when the judges

were rigorously selected and trained, the judges were significantly different.
Thus such results with an untrained panel are not unexpected. However, the

products were found to be different and therefore the judges were using the scale
correctly.

Ideally the data from the Rating Scale should have been analysed by a

non-parametric method, but this method would not show which of the beverages

was different. Therefore, a parametric method was chosen. Notwithstanding

 - 152 -

the use of a parametric method the data for the different beverages had an

equal variability (SD. 1.02,1.03 or 1.14 respectively), thus implying that

the method of analysing the data did not affect the results of the analysis.

To determine which, if any, product mean scores are significantly

different, the tabulated values of Qp for 124 degrees of freedom, arc each

multiplied by the standard error 1.0/68 = 0.015, ' to form the shortest

significant ranges, Rp. The results are summarised in Table 40.

Table 40: Multiple Range Test for the Effect of CQA on DCQA

Shortest significant ranges Comparisons

5970Level 1%oLevel
P=2 =-3 p=2 p=3
Qp 2.80 2.95 3.70 3.86 Products CQA DCQA Mixture
(1 : 1)
Rp 0.042 0.044 0.056 0.056 Mean scores 5.74 11.50 8.04

At 17 level the mean score of product CQA, DCQA and their mixtures

at 1: 1 are significantly different from one another.

In the samples evaluated, it was shown that mean score of the judges

was higher for DCQA > mixtures (1 : 1) > CQA respectively. Therefore the

presence of CQA in a mixture with DCQA mellowed the effects of DCQA.

In reply to the question: "Which of the coffee samples was found to be

objectionable? " Everyone in the panel stated instant coffee with added 0.1%
DCQA.

Discussion

The preliminary taste evaluation of DCQA indicated that it has a

peculiar taste which can be easily detected. The professional tasters
describe the taste as astringent and metallic. The consumers describe the
taste as objectionable, bitter and metallic with a lingering after taste.

Lea and Arnold (1978) defined astringency as a drying, puckering

sensation in the mouth which affected the whole of the tongue uniformly. One
 - 153 -

cannot expect someone who is not in the coffee trade or experienced in tasting
to be able to use accurately a technical word such as astringent. Whatever the

different groups called the sensation, it is obvious that the effect would

negatively influence the quality of coffee brew.

There has been little previous work upon the organoleptaic properties

of the CGA: Lee et al, (1975) reported that salts of CQA, caffeic acid and

cynarine (1.4 DCQA) have a sweet taste. Whiting and Coggins (1975) studied

the astringency content of ciders and found that a high 5-CQA content mellowed
the astringent notes associated with condensed tannins.

The observation is in agreement with the data in the present study.

III Objective Investigations

In the preceding section it has been established that DCQA have a

peculiar taste which is detectable in instant coffee brew.

It was considered desirable to confirm by an objective method that

DCQA have the properties of a typical astringent compound. Several objective

methods have been reported, based upon the ability of such compounds to

proteins, ' e. g. gelatin (Jones, 1927); casein (Handley, 1961);

precipitate
haemolysed blood (Bate-Smith, 1973), and ß glycosidase (Haslam, 1974). The

cheapest and most convenient of these appeared to be Bate-Smith's method

using haemolysed blood.

Investigations using the Haemoglobin Precipitation Method

1. The ability of tannic acid, CQA or DCQA to precipitate haemoglobin.

2. The ability of green coffee extracts to precipitate haemoglobin.

Method
Rat blood was haemolysed by diluting with distilled water at a ratio of
1: 50 immediately after collection. From this, 1. OmL was taken and added to
1. OmL of tannic acid, 5-CQA, DCQA or green bean extract at various concentrations

in water. (Propanol cannot be used because it produces turbidity. )

The test solution was added to the haemoglobin solution in a centrifuge

tube, using a vortex mixer to avoid the formation of temporary protein gradients
 -154-

during mixing (Bate-Smith, 1973). After shaking for 15 seconds, the mixture

was centrifuged for 15 minutes at 4000 r. p. m. The supernatant was decanted

and its absorbance recorded at 578nm on a Unican SP1800 spectrophotometer.

The standard curve was constructed with different concentrations of tannic acid
(Sigma Chemical Co., U. S. A. ) in water, a known astringent compound.

In each case a control was prepared by adding to 1. OmL of haemolysed

blood an appropriate volume of water instead of the test solution.

Results and Discussion

1. Tannic acid, DCQA and CQA

Preliminary investigations ninde use of propanolic '

extracts, but these
lead to a turbid supernatant. Such extracts were evaporated to dryness and the

residues were dissolved in water.

The addition of aqueous 5-CQA to haemolysed blood produced a brown

colour; aqueous DCQA left a red supernatant. These observations led to the

scanning of the supernatants between 400nm to. 700nm.

.
Haemolysed blood normally shows peaks at 540nm and 578nm and these

peaks were retained in the supernatant after addition of DCQA. In contrast,

the supernatant remaining after addition of 5-CQA showed no such peaks (Fig.

39). According to Kisugawa et al, (1981) the haemoglobin may have been

oxidised to methaemoglobin. Whatever the explanation this change in super-

natant spectrum led to a false indication of precipitation.

A plot of tannic acid concentration against the reduction in supernatant

absorbance (0 A) produced a sigmoid curve of which the precise nature

depended upon the concentration of haemoglobin and the concentration of tannic

acid.

These results fitted a logistic curve (Causton, 1977) which had a linear

form loge a- 1) b- kt.

= Loge

Where a= upper asymptote value i. e. supernatant absorbance of blank

y= sample DA
t= final concentration of test substance that would
have occurred in the reaction mixture in the absence of precipitation.
 - 155 -

FIG. 35 Spectra of the haemoglobin precipitation of 5-CQA and DCQA.

CQA supernatant DCQA supernatant

 -156-

The Intercept on the x axis gives t50%, the concentration of astrigent

corresponding to 5070precipitation, a value which previous workers have used

as an indication of relative astringency (e. g. Haslam, 1974).

The logistic curves are compared in Table 41. Dicaffeoylquinic acid

behaved similarly to tannic acid. The DCQA had approximately 207 the

astringency shown by tannic acid, thus confirming the hypothesis that DCQA

are astringent.

Table 41: Summary of Logistic Curve Data and Relative Astringency

of Tannic Acid and DCQA.

Analytical Method b k t507" Relative

and Astringency
Test Substance mg/mL

Bate-Smith (Colorimetric)
Tannic acid 107 -12.90 0.36 1
DCQA 23.1 - 1.78 1.76 0.21

Gravimetric

Tannic Acid 12.1 -19.20 0.13 1

DCQA 7.3 - 4.05 0.49 0.27

2. Green Bean extracts

Some twenty green bean extracts were examined by the method

described on p. 153. All extracts produced some precipitation, but it was not

possible to establish a consistent correlation with the DCQA content as measured

by HPLC. These data are tabulated in Appendix H.

It was thought that the failure to correlate might be due to the poor
solubility of DCQA in water.

In an attempt to overcome this problem the colorimetric method of

Bate-Smith was converted to a gravimetric method as reported by Clifford and
Ohiokpehai (see Appendix G).

Tannic acid and DCQA behaved as expected (see Table 41), but the data
 -157-

for the green bean extracts were still too

complex to give a simple
correlation. The raw data are presented in Appendix H.

It would seem that other substances in the coffee extracts are able to
interfere in the DCQA haemoglobin interaction. Among the substances known
to be present in the extract are caffeine and CQA.

Caffeine is known to form a hydrophobic complex with CQA via the

caffeic acid residue, and thus possibly with DCQA (Hormann and Viani, 1972).
It was anticipated that caffeine might interfere in the DCQA-haem interaction.
Non-astringent phenols such as CQA are relatively poor protein precipitators..
One must assume that this is because they do not cross-link, rather than
because they do not associate with the protein, since they offer a virtually
identical shape. Indeed, their smaller size might be expected to favour

association. It has been suggested (Bate-Smith, 1973) that an astringent phenol

must be present above a certain threshold level to cause precipitation, i. e. a

certain minimum number of sites must be occupied. If this is correct, then

CQA by occupying some of these sites might be expected to behave synergistically

and reduce the DCQA threshold for precipitation. However the astringent phenols
to which Bate-Smith refers do not, in general, contain carboxylic acid residues
and alternatively one can postulate antagonism using the following model:

Protein with n readily available binding sites may bind (1) n molecules
n/2
of CQA yielding n carboxylic acid groups; (2) molecules of DCQA yielding
n/2 (3) some combination yielding intermediate
carboxylic acid groups; number

of carboxylic acid groups. The pka values for 5-CQA and that of DCQA is 4.25

at 20°C (Inoue et al, 1965) indicating extensive ionisation at the interaction pH

value, (approximately pH 6.5) and one can envisage that the accumulation of
sufficient charges would significantly impede precipitation. Such a mechanism

might explain the mellowing effect of 5-CQA on astringency observed in this

investigation and by Whiting and Coggins.

Although the mechanism(s) cannot be defined, it is not difficult to sec

that CQA, caffeine and possibly other substances could interfere and thus prevent
the observation of a simple correlation.
 - 158 -

CHAPTER SEVEN

GENERAL SUMMARY OF RESULTS

 - 159 -

survey presented a summary of data upon the

The literature

production and composition of coffee and the factors known or thought to

influence beverage quality. It was clear and still is that beverage quality is a

complex function of many factors which include:

a. Inherent chemical and physical properties of the green bean;

b. variations in green bean processing;

c. variations in coffee bean sorting, roasting and brewing.

Previous investigators have drawn attention to particular green bean

constituents as determinants or indices of quality. One such index which has

received considerable attention over a number of years is the CGA content.

This has been linked with quality because of a possible association with
(1) green bean colour;
(2) mechanical and/or microbial damage;
(3) coffee aroma via volatile, particularly smoky degradation
products;
(4) coffee flavour via non-volatile products.

Accordingly the CGA were selected for further study. In Chapter 2

methods of analysis were described and their attributes and limitations

considered critically. The more promising were selected and much time was
and TBA
spent in gaining expertise with HPLC and the molybdate, periodate
In all cases the precision was determined and in the case of the TBA
reagents.

reagent significantly improved. Consideration of the differing specificities of

these chosen methods led to the postulation of three Analytical Ratios.

Pv
1. J

2. HPLC BDQA
TBA BDQA

3. Pv
TBA BDQ A

Chapter 3 included the testing of these hypotheses for good

commercial green beans and led to the further hypothesis that these ratios
 - 160 -

serve as provisional indices of the acceptable composition.

Application of these ratios to six batches of immature beans and six

batches of peculiarly coloured beans met with limited success. Very

immature samples could be discriminated, but samples harvested a few weeks

early could not. It is these nearly mature samples that arc more likely to

enter commercial batches and thus for which some discriminating method is
desirable.

Ratios 2 and 3 performed acceptably for homogeneous samples of

peculiarly coloured beans. However with commercial heterogeneous batches
there will be a decline in discriminating power as the various peculiar groups
are diluted.

Thus one must conclude that these Analytical Ratios are likely to
distinguish only very good and very bad batches and to be of little help at
intermediate quality levels unless further work permits the width of the norms
to be reduced. However, these Analytical Ratios have some merit as an

analytical tool and clearly indicated the presence of unusual phenolic compounds
in the immature beans. The presence of these substances would not necessarily
have been obvious if only one method of analysis had been employed. The

possible identities of these unusual phenolic compounds were discussed in

Chapter 4, but it was not possible to reach firm conclusions. The very
immature beans may have very high contents of phenolic compounds, e. g. 8.9%
dmb in Arabicas and 17.4% dmb in Robustas. In all six batches examined the

composition changed progressively as the beans matured and eventually

approached typical levels (IVC 503 was unusually high at maturity, 11.4% dmb).
There were no common patterns of behaviour.

Attempts to synthesise CGA-quinones were not especially successful.

It would appear that some enzymic oxidation products are detected by periodate
but not by the HPLC method used in this study.

Such products would thus be determined as estimated FQA, but not as

HPLC FQA. In general estimated FQA is larger than the value obtained by

HPLC. In part this may be a reflection of the relative precision of the two
methods, but this observation does tend to suggest that some CQA oxidation
 - 161 -

products are present in many of the samples, particularly the darker ones.
Arising incidentally from this section of the investigation was the interesting

observation that DCQA, in contrast to 5-CQA is not an in-vitro substrate for

PPO. This selectivity does not necessarily occur in-vivo. However, in

Chapter 5 it was observed for Batch 1 that although DCQA content declined

with blackening, it declined much more slowly than CQA, and in the darkest
beans accounted for approximately one-third of the CGA. This is consistent

with CQA also being the preferred substrate in-vivo. The possible organoleptic
implications of an increased proportion of DCQA were discussed in Chapter 6.

The dilution effect referred to above is not a problem with the

computer-controlled colour sorting, since each bean is examined individually.

Such a method easily discriminated all the peculiarly coloured samples used in

this study. Immature beans were detected if they retained their chlorophyll-

pigmented silverskin, although usually this was easily shed during processing

as the beans approached maturity. However it is clear that there is a group of

peculiar beans which retain their silverskin during green bean processing.

The reason for this is not known. Some of these beans, characterised by

chlorophyll-pigmentation and a higher than normal CGA content, are undesirable.

Others, e. g. the foxy, characterised by carotenoid pigmentation, are acceptable.

The CGA content of the yellow, brown and black beans is consistent

with their being discoloured by enzymic browning. These discoloured samples

had low average bean weights, suggesting that they were harvested
strongly
while immature.

In Chapter 6 and Appendix H are presented the data on astringency.

Subjective and objective methods show that DCQA have some astringent effect,

whereas CQA do not. This astringency can be detected in instant coffee brew at
1.08mg/mL if the CQA content is low. However, raising the proportion of
CQA from 1: 2.5 to 1: 1 and finally 18 :1 progressively reduces the intensity.

To avoid the unpleasant organoleptic effect of DCQA it is necessary to

this is linked
avoid green beans with high DCQA content, particularly where

with a relatively low CQA content. Both CQA and DCQA are destroyed

progressively during roasting, but CQA is destroyed more rapidly and thus the
 - 162 -

CQA : DCQA ratio may change to a less acceptable value.

The investigation reported here suggests that typically Arabicas

contain 0.8' to 1.5% dmb DCQA with a CQA : DCQA ratio of 4: 1 and Robustas
contain 1.8 to 2.2% dmb DCQA with a CQA : DCQA ratio of 3: 1.

Exceptional DCQA contents may occur in some slightly immature

beans, e. g. 1.9% dmb IVC 503 with a CQA : DCQA ratio of *: 1, and as

discussed previously some discoloured beans may have a disadvantageous

initial CQA : DCQA ratio, e. g. 1: 1, Batch 1, sample 5.

Efficient sorting can eliminate the discoloured beans, but slightly

immature beans may be more difficult. Haemoglobin precipitation techniques

were tested as a method which might detect potentially astringent beans.

Unfortunately these were unsuccessful (Chapter 6). The DCQA content was

determined by HPLC and estimated by colorimetry. The values obtained

correlated poorly and since the HPLC method was more precise and simpler
for this purpose, that would be the preferred quality control method.

Future Work

1. Me development
of a rapid method for determining the CQA:DCrA ratio
the testing of the hypothesis and
that this ratio is related to beverage
acceptability.

2. Characterisation
of unknowns on the chromatograms and development the
existing analytical of
methods for use with roasted beans.
3. The analysis of samples similar to those used in this investigation
other components that might influence for
the quality, e. g. diterpenes.
 - 163 -

FINAL CONCLUSIONS

1. The CGA can influence beverage quality via the astringent property of
the DCQA. To produce high quality beverage green beans of high DCQA

content and/or low CQA : DCQA ratio should be avoided.

2. Potential sources of such beans Indude:

(a) certain discoloured green beans.
(b) certain immature green beans.

3. Immature green beans appear to be an important source of discoloured

green beans.

4, None of the individual analytical methods nor the Analytical Ratios

tested proved to have much potential as a commercial index of quality.

5. High pressure liquid chromatography was the most precise and most
informative single analytical method, but is expensive.

6. Molybdate reagent offered the best compromise of precision, simplicity

and rapidity.

7. Periodate reagent detects some enzymic browning intermediates and

these are determined as estimated FQA. This does not significantly
detract from the value of this reagent.

8. Dicaffeoylquinic acid (DCQA) is not an in vitro substrate for PPO.

 - 164 -

BIBLIOGRAPHY
 -165, -

BIBLIOGRAPHY

Adrian J., Frangne R., Xabregas J., Corte dos Santos A. (1967) 3rd Int.
Sci. Coll. on Coffee, 427-435, Trieste, ASIC, Paris.
Adrian J., Pena J., Morals de Carvalho J., Miranda A., Xabregas J., Corte
dos Santos A. (1969) 4th Int. Sci. Coll. on Coffee, 232-242, Amsterdam,
ASIC, Paris.
Alcaide A., Devys M., Banbier M., Kaufman H. P. Sen. Gupta A. K. (1971)
Phytochem. 10,209-210.
Amerine M. A., Pangborn R. M., Roessler E. B. (1965) Fd. Sci. & Techn. -
A series of mongraphs. Principles of sensory evaluation of Food.
Chap. 10,437-493.
Amorin H. V. and Silva D. M. (1968) Nature (London), 219,381-382.

Amorin H. V., Guercio M. A., Cortez J. G., Malavolta (1973) An Esc. Super.
Agric. "Luiz de Queros" Univ. Sao Paulo, 30,281-289.

Amorin H. V. and Josephson R. V. (1975) J. Food Sci. 40 (6) p. 1179-1184.

Amorin H. V., Cruz A. R., St. Angelo A. J., Dias R. M., Melo M., Teixeira A. A.,
Gutierrez L. E., Ory R. L. (1977) 8th Int. Sci. Coil. on Coffee, 183-186,
Abidjan, 1977, ASIC, Paris.

Andersen R. A. and Vaughn T. H. (1972)Phytochem, 11,2593-2595.

Ara V. and Thaler H. (1976) Z. Lebensm. Unters-Forsch. 161,143-150.

Arcila P. J. and Valencia A. G. (1975) Cenicafe (Colombia), 26, (2) 55-71.

Arnold R. A. and Noble A. C. (1978) Am. J. Enol. Vitic. 29, (3), 150-152.

Aronoff S. (1966) In: The Chlorophylls. Vernon L. P. and Seely G. R. (eds)

3-20.

Assi, A. (1977) 8th Int. Sci. Coll. on Coffee, 359-363, Abidjan, 1977, ASIC,
Paris.

Baltes W. (1975) 7th Int. Sci. Coll. on Coffee, 91-107, Hambourg, ASIC,
Paris.

Barbirolli G. (1965) Rass Chim. 17, p 261-263.

Basu S. R. J. (1977) Indian Coffee, 41, (6), 213-217.
Bate-Smith E. C. (1973) Phytochem, 12,907-912.
Barnes H. M., Fieldman J. R. and White W. V. (1950), 72,4178-482
Berger St. and Rieker A (1974) In. The Chemistry of The Quinonoid Compounds
(1), 163-230, John Wiley, London.

Bernstein J., Cohen M. D., Leiserowitz L. (1974) In. The Chemistry of the
Quinoid Compounds, 37-110. Patai S. (ed), John Wiley, London.
 -166 -

Boardman N. K. (1966) In. The Chlorophylls. Vernon L. P. and Seely G. R. (eds),

p. 437-476, Academic Press, London.
Bogorad L. (1966) In: The Chlorophylls, 481-510, Vernon L. P. and Seely (eds)
Acad. Press, London.

Bothe and Cammenga (1980) 9th Int. Sci. Coll. on Coffee, 135-149, London,
ASIC, Paris.
British Standard Institution,
BS 57 52 : Part 1: 19179. ISO 1446 -197 8.
Butty M. (1973) Kenya Coffee 38,214-224.

Calzolari C. and Lokar L. C. 3rd Int. Sci. Coll. on Coffee, 240-247, Trieste,
ASIC, Paris. '
Campos L. S. and Rodrigues J. M. L. 5th Int. Sci. Coll. on Coffee, 91-96,
Lisbonne, ASIC, Paris.

Carisano A., Gariboldi L. (1964) J. Sci. Food Agric. 15,619-622.

Causton D. R. (1977) A Biologists Mathematics, Edward Arnold.
Centi-Grossi M., Tassi-Micco C., Silano V. (1969) Phytochem. 8,1749-1751.
Chassevent F., Vincent J.C. Hahn D, PougneaudS., Wilbaux R. (1969) 4th
Int. Sci. Coll. on Coffee, 179-185, Amsterdam, ASIC, Paris.

Chassevent F., Dalger G., Gernwig S., Vincent J. C. (1973) 6th Int. Sci.
Coll. on Coffee, 147 -154, Bogota, ASIC, Paris.

Clarke R. J. (1979) Fd Chem. 4,81-97.

Clarke R. J. (1976) Chem. and Industry 17,362-365.
Clarke R. J., Walker L. J. (1974) J. Sci. Food Agric. 25, (11), 1389-1404.

Clarke R. J., (1967) Process Biochem., 2,15-19.

Clifford M. N. (1979) Food Chem. 4, (1), 63-71.

Clifford M. N. and Stanifbrth P. S. (1977) 8th International Sci. Coll. on Coffee
109-113, Abidjan, ASIC, Paris.

Clifford M. N. and Wight J. (1976) J. Sci. Fd. Agric., 27,73-84.

Clifford M. N. (1975) Process Biochemistry 10, (2) 20-23,29 and (4) 13-16,19.
Clifford M. N. (1974) J. Chromatography 94,261-266.
Clifford M. N. and Wight J. (1973) J. Chromatography 86,222-224.
Clifford M. N. (1972) Ph. D. Thesis, University of Strathclyde, Glasgow.
Corse J., Layton L. L., Patterson D. C. (1970) J. Sci. Fd. Agric., 21, (3)
164-168.
Clifford M. N. Dept. of Biochemistry, University of Surrey, Guildford,
Personal Communication.
 - 167 -
D'ornano M., Chassevent F., Pougneaud S. (1963) 2nd Int. Sci. Coll. on
Coffee, 131-144, ASIC, Paris.

Dorrell D. G. (1976) Can. J. Plant Science, 56,901-905.

Dorrel D. G. (1974) Proceedings of the 6th Int. Sunflower Conference 22-24.
Bucharest, Romania.
Dorrel D. G. (1976) Crop Science, 16,422-424.
Dorrel D. G. (1978) Crop Science, 18,667-670.
Dentan E. (1977) 8th Int. Sci. Coll, on Coffee, 59-64, Abidjan, ASIC, Paris.
Durst H. D., Mack M. P., Wudl F. (1975) J. Org. Chem., 40, (2), 268-269.
Dyke S. F. (1965) In: The Chemistry of Vitamins, Interscience, London.

Ehlers G. M. (1980) 9th Int. Sci. Coll. on Coffee, 267-271, London, ASIC, Paris.

Feldman J. R., Ryder W. S. and Kung J. T. (1969) J. Agric. Fd. Chem. 17,
(4), 733-739.

Fiegl F. and Anger V. (1966) In: Organic Analysis, 7th English edition,
Elsevier, Amsterdam.
Fobe L. A., Nery J. P. and Tango J. S. (1967) 3rd Int. Sci. Coll. on Coffee,
389-397, Trieste, ASIC, Paris.
Folstar P., van de Plas H. C., Pilnic W., de Heus J. G. (1977) J. Agric. Food
Chem., 25, (2), 283-285.
Folstar P., Pilnik W., de Haus J. G., van de Plas H. C. (1975) 7th Int. Sci.
Coll. on Coffee, 253-258 Hambourg, ASIC, Paris.

Folstar P., Pilnik W., Heus J. G., van de Plas H. C., van dez (1975) Mitt.
Gebiete Lebanam. Hyg., 66,502-506.
Folster P., Van der Plas H. C., Pilnik W., Schols H. A., Melger P. (1979)
J. Agric. Food Chem., 27 (1), 12 15.
Forsyth W. G. C., Quesnel V. C. and Roberts J. B. (1958) J. Sci. Food Agric. 9
181-184.

Fujimaki M., Kim K., Kurata T. (1974) Agr. Biol. Chem., (Tokyo) 38,45-52.

Garrick L. S., and Haberman H. M. (1962) American J. Botany, 49,1078-1088.

Gautschi F., Winter M., Flament I., Willhalm B. and Stoll M. (1967) J. Agric.
Food Chem., 15,15-23.
Grassman M., Pluscec J., and Bogorad L. (1968) Plant Physiol. 43,1411-1414.
Gibson A. (1971) 5th Int. Sci. Coll. on Coffee, 246-258, Lisbonne, 1971,
ASIC, Paris.
 -168 -

Gibson A. (1973) 6th Int. Sci. Coll on Coffee, 319-324, Bogota, ASIC, Paris.
.
Goldoni L. (1979) Prehrambeno-tehnoloska revija PTRVB 17 (4) 185-189.

Griffiths T. I. A B. Sc. Thesis, Dept of Biochemistry, University of Surrey,

U. K. 1982.
Gutmann W., Werkhoff P., Barthels M., Vitzthum O. G., 8th Int. Sci. Coll. on
Coffee, 153-161, Abidjan, ASIC, Paris.

Haarer A. E. (1963) Coffee Growing, Oxford University Press, London.

Habermann H. M. (1,972) Bot. GAZ. 134 (3), 221-232.

Hamidi A. and Wanner H.. (1964) Planta 61,90-96.

Handley W. R. C. (1961) Plant Sci. 1,15,37-73.

Harborne J. B. (1967) -Comparative Biochemistry of the Flavonoids Ch. 3 and

4, Acad. Press, London.

Harborne J. B. (1973) Phytochemical Methods 33-43, Chapman and Hall, London.

Hartman H. T., Flocker W. J. and Kofranek A. M. (1981) In: Tropical Fruit

and Nut Crops 621-635, Prentice-Hall, N. J.

Hartmann L., Lago R. C. A., Tango J. S., Teixeira C. G. (1968) J. Amer.

Oil Chem. Soc. 45,577 -57 9.
Haslam E. (1974) Biochem. J. 139,285-288.

Haslam E. (1981) In: The Biochem of Plants, 527, Conn E. E. (ed) Academic
Press, London.

Hillis W. E. and Swain T. J. Sci. Food Agric. 1959,10,135-144.

Holden M. (1976) In: Chemistry and Biochem. of Plant Pigments, 2,2-32, Goodwin
T. W. (ed), Acad. Press, London.
Holden M. (1965) J. Sci. Food Agric., 16,312-325.
Hormann I. and Viani R. (1972) J. Fd. Sci., 37,925-927.

Humphrey A. M. (1980) Food Chemistry, 5,57-67

IFT (1981) Sensory Evaluation Guide for Testing Food and Beverage Products,
50-58.

IFT (197 5) Minutes of Sensory Evaluation Div. Business meeting at 35th Ann.
Meet. Inst. of Food Technologists, Chicago, June 10.

Ikan, R. (1969) A lab. guide, 48-55, Academic Press, London.

Inoue Y., Aoyagi S. and Nakanishi K. (1965) Chem. Pham, Bull., 13 101-104.
,
Illy E., Illycaffe. S. p. A. Trieste, Italy. Personal Communication.

Illycaffe, S. p. A. Trieste, Italy. Personal Communication.

 -169-

IUPAC, (1974) 37,285-297.

International Coffee Organisation (1962) Text of ICO Agreement (Revised 1968),
London.

Jones A. E. (1927) Analyst (London), 52,27 5-27 6.

Jones O. T. G. (1965) Biochem. J., 96,6P-7P.
Jones O. T. G. (1966) Biochem. J., 101,153-160.

Kato S., Kurata T.; Fujimaki M. (1971) Agric. Biol. Chem. (Tokyo), 35,
2106-2112.
Katz S. N. (1980) 9th Int. Sci. Coll. on Coffee, 295-302, London, ASIC, Paris.
Kisugawa K., Arai S;, and Kurechi T. (1981) Chem. Pharm. Bull., 29,
1694-1701.

Kuc J. (1964) Symposium Proceedings on Phenolics in normal and diseased

fruits and vegetables, 63-77, Runeckles V. C. (ed).
Kung J. T., Ryder W. S. and Feldman J. R. (1967) 3rd Int. Sci. Coll. on
Coffee, 223-230, Trieste, ASIC, Paris.
Kulaba G. W. (1978) M. Sc. Thesis, University of Nairobi, Kenya.

Kulaba G. W. (1979) Kenya Coffee, 44, (520), 23-34.

Kulaba G. W. and Robins P. A. (1981) Kenya J. Sci. Technol., 2,63-72.

Lattanzio V. and Morone I. (1979) Experentia 35,993-994, Birkhauser Verlag

Basel, (Schweig).
Lea A. G. H. and Arnold G. H. (1978) J. Sci. Fd. Agric., 29,478-483.
Lee Chi-H., Scarpellino R. J., Murtagh M. M. (1975) United State Patent
No. B, 332,527.

Legrand M., Fritig B., and Hirth L. (1978) Planta 144,101-108.

Luh B. S. and Phithakpol (1972) J. Food Sci. 37,264-268.
Lyons Central Laboratories Report No. D 1852,1977.

Maga J. A. (197 8) CRC Food Sci. and Nutr. 10,323-372.

Maga'J. A. (1981b) CRC Critical Revs in Fd. Sci. and Nutr. 14,295-307.
Maga J. A. (1981a) J. Agric. Food Chem. 29 (5), 295-898.
 - 170 -

Marigo G. and Boudet A. M. (1979) Z. Pflanzenphysiol. Bd., 92,33-38.

Martino V. S., Debenedetti S. L. and Coussio J. D. (1979) Phytochem., 18,
2052.
Mason H. S. (1955) Adv. Enzymol., 16,105-175.
Mason H. S. (1955) Nature, 175,771-772.
Maughan W. S., Milo S., Roarzi L. (1980) 9th Int. Sci. Coll. on Coffee,
201-210, London, ASIC, Paris.
Merrit C. Jr., Robertson D. H. and McAdoo D. J. (1969) 4th Int. Sci. Coll. on
Coffee, 144-148, Amsterdam, ASIC, Paris.
Minard R. D., Lui S-Y and Bollag J-M. (1981) J. Agric. Food Chem., 29,
250-253.
Moores R. G., McDermott D. L. and Wood T. R. (1948) Anal. Chem. 20
(7), 620-624.

Molderez M., Nagels; L., Parmentier F. (1978) Phytochem., 17,1747-1750.

Munene S. G. (1973) Kenya Coffee, 38, (452), 332-334.

Murphy J. B. and Stutte C. A. (1978) Anal. Biochem. 86,220-228.

Nagels L., van Dongen W., and Parmentier F. (1979) Arch. Intern. Physiol.
Biöchem., 87,585-591.
Nakabayashi T. (1977) Nippon Shokuhin Kogyo Gakkashi (J. Jap. Soc. Food
Sci. Techn. ) 24, (9) 474-483.
Nakabayashi T., Watanabe C. (197 5) Nippon Shokuhin Kogyo Gakkashi, 24, (3)
124-129.

Natarajan C. P., Urs M. K., Bhatia D. S. (1955) J. Indian Chem. Soc., 16, (1)
9-12.

i Navellier P. and Brunin R. (1962) Cafe, Cacao, The VI, 1,47 -54.
Navellier P. (197 0) In: Encyclopedia of Industrial Chemical Analysis,
373-447. John Miley, N. Y.
Nichiforesco E. (1970) Plantes Medicinales et Phytotherapie IV (1), 56-62.

Northmore J. M. (1965) Kenya Coffee (Nairobi), 30,373-382.

Northmore J. M. (1967) 3rd Int. Sci. Coll. Coffee, 405-414, Trieste,
on
A SIC, Paris.
Northmore J. M. (1968) Kenya Coffee 33, (392) 297 -302.
Nursten H. E. (1981) Food Chem., 6, (3) 263-277.
 -171 -

Oliveira J. C., Silva D. M., Teixeira A. A., Amorim H. V. (1977) Turrialba 27 (1)
7 5-82.

Payen M. (1846) Compt rend., 23,244-251.

Pereira A. and Periera M. M. (1971) 5th Int. Sci. Coil. on Coffee, 85-90,
Lisbonne, ASIC, Paris.
Pierpoint W. S. (1969) Biochem. J., 112,619-629.
Pierpoint W. S., Ireland R. J., Carpenter J. M. (1977) Phytochem., 16,
001-006.
Pierpoint W. S. (1982) Phytochemistry 21, (1), 91-95.
Poisson J. (1977) 8th Int. Sci. Coll. on Coffee, 33-57, Abidjan, ASIC, Paris.
Proksch P., Wisdom C. and Rodriguez E. (1.981) Z. Naturforsch., 36C,
357-360.
Purseglove J. W. (1968) Tropical Crops, Dicotyledons 2, p. 458-492.

Pypker J. and Brouwer H. (1969) 4th Int. Sci. Coll. on Coffee, 122-130,
Amsterdam, ASIC, Paris.

Quijano-Rico M. and Spettel B. (1975) 7th Int. Sci. Coll. on Coffee, 165-173,
Hambourg, ASIC, Paris.

Radtke-Granzer R. and Piringer D-G. (1981) Deutsche Lebensmittel -

Rundschau 77 (6), 203-210.

Rao N. G., Bulachandran J. M., Matarajan C. P. (1969) 4th Int. Sci. Coll.
on Coffee 55-58, Amsterdam, ASIC, Paris.
Rees D, I. and Theaker P. D. (1977) 8th Int. Sci. Coll, on Coffee, 97-84,
Abidjan, ASIC, Paris.

Reudnitz K. (1980) Dragoco Report Flavour Imformation Service, 2,46-53.

Robinson T. (1980) The Organic Constituents of Higher Plants, Ch. 4.,

54-86, Condus Press, North Amherst, Mass., U. S.
Roffi J,, Corte Dos Santos A,, Mexia J. T., Busson F., Maigrot M. (1971)
5th Int. Sci. Coll, on Coffee, 179-200, Lisbonne, ASIC, Paris.
Rolz C., Menchu J. F., Arimany E. (1969) 4th Int. Sci. Coll, on Coffee,
166-173, Amsterdam, ASIC, Paris.
Rubach K. (1969) Dissertation, Techn. Univ. Berlin.
 -172 -

Ruckenbrod H. (1955) Planta 46,19-45.

Russwurm H. (1969) 4th Int. Sci. Coll, on Coffee, 103-106, Amsterdam,

ASIC, Paris.

Sabir M. A., Sosulski F. W. and Finlayson A. J. (1974) J. Agric. Food Chem.,

22, (4), 575-578.
Sanint O. B. and Valencia G. (1972) Cenicape (Colombia), 23,59-71.
Schulz J. M. and Hermann (1980) J. Chromatogr. 195,85-94.
Schutte L. (1974) C. R. Fd. Techn., 4 457 -505.
Simatupang H. H. and Hausen B. M. (1970) J. Chromatography 52,180-183.
Sheen S. J. (1969) Phytochemistry 8,1839-1847.
Sheen S. J. (1973) Plant Physiol. 52,422-426.
Shibata S., Takito M.;. Tanaka O. (1950) J. Am. Chem. Soc., 72,2789-2790.

Sivetz M. and Des rosier N. W. (1979) Coffee Technology, 185-190, AVI Publ.
Westport, Ct., U. S.

Sivetz M. S. and Foote H. E. (1963) Coffee Processing Technology V. I. Avi

Publishing Ct., U. S.

Sivetz M. (1973) Coffee, Origin and Use, Coffee Pub. 3635 Elmwood,
Corvallis, Oregon, U. S.
Smith, G. L. (1981) J. Sci. Food Agric. 32,513-520.

Smith R. F. (1965) 2nd Int. Sci. Coll, on Coffee, 27 -38, Paris, ASIC.

Soloveva T., and Simonova V. N. (1976) lzv. Vyssh. Uchebn. Zaved. Pishch.
Tekhasl., 5,165-166.

Steiner E. R. In: W. J. Youden Statistical Techniques for Collaborative Tests,

A. O. A. C. Washington, D. C. 1967.
Stirling H. G. and Jackson N. W. (1979) Tea and Coffee Trade J. 151, (5) 20
and 46.
Stirling H. (1980) 9th Int. Sci. Coll. on Coffee, 1,189-200, London, ASIC,
Paris.

Streuli H. (1973) 6th Int. Sci. Coll. on Coffee, 61-72, Bogota, ASIC, Paris.

Sutherland G. K. and Gortner W. A. (1959) Australian J. Chem., 12,240-246.

Swain T. and Hillis W. E. (1959) J. Sci. Food Agric. 10,63-68.
Synge R. L. M. (1978) Bulletin de Liaison No 8 du Groupe polyphenols narlonne,
France Compte-Rendre des journees internationales D'Etude et de
L'Assemble Generale.
Smith R. F. (1963) Ist Int. Sci. Coll. on Coffee, 245-252, Paris, ASIC.
 - 173 -

Taylor R. J. (1975) Carbohydrates, A Unilever Educational Booklet.

Tepley L. J., Prior R. F. (1957) J. Agric. Food Chem. 5,375-377.

Tiscornia E., Centi-Grossi M., Tassi-Micco C., and Evangelist F. (1979)

Riv. Ital. Sostanze Grasse, 56, (8), 283-292.
Thaler H. and Arneth W. (1967) 3rd Int. Sci. Coll, on Coffee, 127-136, Trieste,
ASIC, Paris.

Thaler H. (1963) Cafe, Cacao, The 3,240-244. Ist Int. Sci. Coll. on Coffee, Paris.

Thaler H. and Gaigl R. (1963) Z. Lebensm-Unters-Forsch. 120,357-363.

Thaler H. (1979) Fd. Chem. 4 13-23.

Thaler H. (1975) 7th Int. Sci. Coll. on Coffee, 175-187, Hambourg, ASIC,
Paris.

Thomson R. H. (1971) In: Naturally Ocurring Quinones, 1-634, Acad. Press,

London. '

Thomson R. H. (1976) 'In: Chemistry and Biochemistry of Plant Pigments,

207-232, Goodwin T. W. (ed), Academic Press, London.

Tressel R. Silwar R. (1981) J. Agric. Food Chem., 29 (5), p. 1078-82.

Tressl R., Grunewald K. G., Kamperschroer H. and Silivar R. (1981) Prog.

Food Nutrition Sci., 5,71-79.

Tressl R. (1977) 8th Int. Sci. Coll. on Coffee, 115-120, Abidjan, ASIC, Paris.

Uritani I. (1961) In: Biochemistry of plant phenol substances 98-124, Symposium,

V. C. Runecles (eds) (Colorado State University), Fort Collins, U. S. A.

Ulbrich B., Zenk M. H. (1979) Phytochem., 18,929-933.

Underwood G. E. and Deathrage F. E. (1952) Food Res. 17 , 425-432.

Van der Stegen G. H. D. (1979) Food Chem., 4 (1) 23-29.

Van der Stegen G. and Noomen P. J. (1977) Lebensm. -Wis s. Technol., 10,
(6), 321-323.

Van der Stegen G. H. D. and Van Duijn J. (1980) 9th Int. Sci. Coll. on Coffee,
107 -112, London, ASIC, Paris.

Villar H. and Ferreira L. A. B. (1971) 5th Int. Sci. Coll. on Coffee, 120-127,
Lisbonne, ASIC, Paris.
Viani R. and Hormann I. (1975) 7th Int. Sci. Coll. on Coffee, 273-278, Hambourg,
ASIC, Paris.

Waiss A. L., Lundin R. E., Corse J. (1964) Chem. and Industry, 1984-1985.

Walker E. K. and Lee T. T. (1968) Can. J. Plant Sci. 48,381-385.

 -174 -

Walker R. L. and McCallion R. F. (1980) Phytochem. 19,373-377.

Walker W. and Weidemann H. L. (1968) Z. Ernahrungswiss, 9,123-128.

Weidemann H. L. and Mohr W. (1970) Lebens-Wiss. Technol. 3,23-32.

Weiss L. C. (1953) J. Ass. of Anal. Chem., 36,663-670.
Whiting G. C. and Coggins R. A. (1975) J. Sci. Fd. Agric., 26,1833-1838.
Wickliff J. L., Arronoff S. (1962) Pl. Physiol., Lancaster 37, p. 584-589.
Williams C. N. (1975) The Agronomy of the Major Tropical Crops, 84-96,
Kuala Lumpar, Oxford Univ. Press.

Wootton A. E. (1971) 5th Int. Sci. Coll. on Coffee, 316-324, Lisbonne, ASIC,
Paris.

Wolfram M. L., Plunkett R. A. and Laver M. L. (1960) J. Agric. Food Chem.,

8, (1), 58-65.

Wolfrom M. L. and Patin D. L. (1964) J. Agric. Fd. Chem., 12,376-377.

Walter W., Grigat H. '-E., Heukeshoven J. (1970) Natarwiss 57,246-247.

Wurziger J. and Harms U. (1969) 4th Int. Sci. Coll. on Coffee, 85-91, Lisbonne,
ASIC, Paris.
Wurziger J. and Asomoa Y. (1974) Kaffee und Tee Markt., 24, (6), 4-6.

Wurziger J. (1977) Fette, Seifen, Anstrichm., 79 (8), 334-339.

Wulf L. W. and Nagel C. W. (1976) J. Chromatography, 116,271-279.

-175 -

APPENDIX
 - 176 -

Appendix A (p. 177, to 181)

IR Spectra of:

(i) 3,5-diltertlbutylcatechol (the starting material).

(ii) 3,5-di-tert-butyl 1,2-benzoquinone (Aldrich)

(iii) Synthesised 3,5-di-tert-butyl 1,2 benzoquinone.

(iv) Technical Grade 1,4-benzoquinone (Aldrich).

(v) Nuj of in NaCl plates (medium of scan).

 m

1
p -

i! O

N yf
0
T
D
D zz
-
0 33
m tD
m
Iý
n

0i
m
m
z> 0
O 0
Z m
m
X
v

C',
; 0
_ý 0
ýý
ýý
m
 H

öý ý,
A 0
0

i
ý.
i
wýT P
,,
ý,
7
O
11
0

1
i

I,

A
O
O
iT

LA
9 N
NN
>0 O

_h
n<

io
m
ýD zm Z
C
1
O
mm
m

n
3ö
Io p
D -o

ýým

)I

4 O
3
m
c
z
Z
Dp 1
z N

a
Qý
ýQ

p
m
T
 HIE-
1
T1

lri, ift 51"., ` 3

)

x
j
Y"'ý
.
t-. 1T
.l 1h--L-I -1I
II.. +f
-I
11
ý4
'1-
ýY +r :1
C7
O
Z
m vi
3
t t

1 J. :L L. }j
N iý r Y ý r . T ý.
T "- ! -r-- i , + . t
-rTi Y'- '*fit »ý-"r -L. -+ - r ...
r +

oý
ö
0 . ý} t} 1i
t}

Zý .4

YTr 1T7.1 .-} I ,ý

+"-" ýL r+ . iy . Lý1- {'*t Y'. L
_

l+ttýrr ýiF+++ý 1, y
_

vi
fiY +i t"r`' t Tý"-: r- O
..

ij

.ýO

oi

NN

rým 41
ý r 1f
ZZ , ýý+- ýL f r 1++ 1-.

.ý . .y

t-ý r .. T.. 'o

co
tr
6
m 1 T " ýrt L

n rr ;r (
y it i:
3
25
ýt
C
- T T
+ '1+i aI I
j
i

O
ý,
0' I
1! t
4 J
. _:. f-y I 1 I A,.. ý ý. I i I . räI
' ý --" :1,j
w ;
Oa
>
m
ý D
Y.
rý °O I. 1-
tL-l. ! ?;
L T11
1
m
x -

N
m
ý
1:.
Ti ,f.
" t ' i ý -` Y' ' rI l +
+ t rl-rt-FT :T:. ý
1' t 'Ll i rX. 1: 1-.:,
.ý L

C) 1- 7

ý11tr i
rlr; t t r_
---, r,ý it f }ý. T-

I}{
T+rr- i- -i

-
 m
Am 0
0
ýD
v
n
1N

'
0
0

3ý

A
O
O

73

*
>0
o0 <0
r) m
D z Z
C
o 3
p co
m
m

n
I
10
D ýU

m2 co
z °
0- 0
zm
>m
z
-I I N
1 m

a
0
0

A
m
T
A

A
1! Oý O
 i, rt' e`t .' r'
rF -t "l. tf mi.
f
II f . -j A

ý)
ö
rý :ý
n
.T. O

Fi
1
z
sý 1. ;; 0'
CD
.., 1

,.
0 ;j i:
ýi
-c, ... , .ý
ý,
ý:,::
' a
r i` . {. ý : . ý..,,:.:
..
i 'i' ..

". ý
:. a.... 1. : `,

rI

.. ý
_.
:,..
.
..: ,
ý'J Ir T1I ýIrt"r
11!
rýi 'llrr-rý

C)
If 1r

D1;
Iýr I!, ir_ !I
Z

ih ;ý1rt
ix-I .
Iý N
\ý ! 'I' Id
Co
irr
li ý

O . 1. fr
_..

'JU
 - 182 -

Appendix B (p. 183 and 184)

NMR Spectra of:

(i) 3,5-di-tert-butyl 1,2-benzoquinone.

(ii) Synthesized 3,5-di-tert-butyl 1,2-benzoquinone.

 4

0
2

Ii
 0
Z

1
 - 185 -

Appendix C

Typical Reflectance Spectra of Green Coffee Beans.

Reflectance spectra by II Iyc, II'I'" Yxxl c1u;l IityII; I1 ('lr cif 1),.: kII .
- supplivcl k) raI;

r, -
_ll`

86

Z
7` iljt iili
I fall }1 i I-
lit

j
Ii ll ` ,ý
1 1ý -I ß {! I Iii I.
-fý\ i1,
II

l
ý--1111 j:. i;
ii I ý'}
LIýýýlIIJý
-ýi 1.1

M-
-7

I
rý fI try \ t.
+
O LL 1Ij''_ Ia1

}'_L_: ý1. t
111_Iýýý ýýI. \`I I, ": f'ýIýýý }l;
. ýý1ý, i ý *
t'i{{{}lý:

71 '1L {1ý I' t.

t fFýý tt i_j± jl- iT i
ýt ýý", ýI! 'ý'. r.
ý1

-'ýj'i.

{. tI 1, L
ll t_ 1 if11. I. G., 11{, 1+4+ii, =..
ý L1 . !. .
-7-1 L

ýý ý'
IL111111 1 '`
r1I l' ý1 ''
ll III
1. rtj r

1, .. 1 af ar. il
_ 1, t iý 1.: 1t
I
1,

1c -i
 -187 -

Appendix D (p. 188 to 192)

Reflectance Spectra of Immature Green Coffee Beans of

Colombian origin from green stage to maturity.
 Rellectance spectra or itrTnatur(ý k;rý"ý'n ('OI( l i: in tw; iii:;.

{: 1

0
1"\\\\
1\

7
 Reflectance spectra of' inmatiinY black

.p( I_ '__ 1_-ýL I .. -

"Al

+-
{r i
 Reflectance spectra of ý.ý' 11cwl )cpari.
.

it t: ,'E
fj, I
=t.
190

------------- -- f
ýýr

L_ý.
_ _ý

_... _
_1.. _..

ýi ý ...

I y. i f
--)i Wir. ` ...
::: F:
_)-. 11_"_ .: ._ ýý_

v
: L1; y. ý
ý 41

ýy r 1
yý

=1 ii c_
t: T lr.

111
-1týI ' `ý

77

ýý ý\(ý4
_1 111 J
-
`I \t frý1 ýý

If

--------- --- -
 Reflectance spectra c.L u nCel -1:ßp;,

ITF
-Olp es '-_r t-
71 lrA
f. _ .. 1 z

-47 -

0 - __:

i
-- ý\ X

ii
Reflectance spectra of nuLu u Iw uls.

4 t_.

I t
_L1i. i_. _

t Tý :a'

<r

Y1
411-

}ý
rl _
{

'1I.
 - 193 -

Appendix E (p. 194 to 201)

Reflectance Spectra of different Immature Green Coffee

Beans of Kenya origin from green stage to maturity.

 Reflectance spectra of irrrriature b( U1ti
.

77W':

."TtI:. X94 r;
1: Iii

-ý'
Ö

YjI ----ý
I ýTý w

.. -i

-t
.,
-

ý-.... -- -a-- --i_. _ :.. t. ý

ý _

.;
_. T. _=_

.
.........

i
I
-i TT-1 ..
 kýflectante spectra of under-ripe beans.

! ýý

-- ;

41 TT

%y 71-
 Reflectance spectra beans.
of mztum

i1 IT i
I` ; fir

1--;

T-` ý, f4
1 ,ý. - r<"-I

77. -ET ci
I(ý i

-_ -_- T. to
1-- --ý --

ý__ v.._ ý_ _7 rºT Ti1 it 41 ý_..: i.

.ct

I
 Reflectance
tiýxýcýtr"ýýv[' i11ffl tu ly,
ý;I( fl hýý,It
"

1- :ýiI __ IF
1 ti 197

;
Til

T T-

L.
L-D

L -1
-1 -7 ^'

(- ; 1+ -"--- a"º
cý tt Iii
mal'

_.

4
 Reflectance spectra of matum, b aliti,

"1ý ý. --Y1 rte

1__1- . --:.:

:i:..: 198
_f:..
....

E
....

O
-law
0
do

_,
_ _. ..... J
._...
O
TTT IAW v
r

IL
_ý--T

17
 Reflectance spectra cf' irnnaturc: e grctýn

Ff

k-

I
I. I.. 199

y-k

Ji

10

'1I:

I: -

T_1

4-1 14

ýý I

-ýý. ý'
L -- '. r'te'. ' . ----I_--
-: -- ý`'" - -- --ý-..
. _-i- _

Far

ýt F i'ifi+ý'ft.
'{-. _.-r-} r-+f Y'1111: )- '.. ý'.. 1I
t1IIý_
11 Gt ýt +1 1ý} 1f "-' tý :
rt }1111-, 1 -.
. a" ý.
-+ ý.
I.
{{-]j ý. 11 1,
1 r: '_- +lt-
ia. 4, ? ý7 iJ{lfT1{1
i'lljl
iýl,
11111.1
11 ýý..
.1
}_..
ýrl ýt "
{
ýt
,.
.: ýý
__
1...
I'
`_
..
t11I,. , -.. _. _I__"
'ý.. 'ý .L.,. 11.. .r.., .I
 Reflectance spectra of under-ri ixv b; ', uls.

ý.; iý ý,

ý! ?ý!
9ý ýyp"ýr
ýý1ý 200-
y'"ý
/ý

ý, '.,

^ý /, fir'

0ý\

1--
=---
-

-'fi t,. ý,-" Tom' ý-"

ý..

ý
ý_ý
-ý

..... J_

1--i -_..
.. __.. _-_... ' ....

44

-
"- ----r-'ý'-ý'ýý,

,1
. ýT

0
7- 1

1 1
ý,
ýr-+t-ý ýJ.,
ý_" ý T4 -i "-ýýI: 1I
r... ýý}Tý_rl 11_
_,
--

ý, J "I"- ý
ýýýý_ ýr ---, ýý=iii,--ý ý.rý .-ý-ý'
OA
O

3 1
Iý_-1
"- 71
_1 I-ý .ýJ1ýýI1ý1

,
-r-- {j{. Iýýi il I, I 11i'
I ti'1 rýl
.. - ---1111-1----"" 1 :; 'I I'
11 i;
,

' ý11 ý,,

ý1 ";: I 111 ýýt
" .. ýI " _-ý
.. _j... _-"
1'ý1
 Reflectance spectra of bear is.

wlý,
12
.-.... :i:.... ;
i01'_

'ß+41T-`

J1{ T / j
--- _ý-_ _".. :ý4.. -L_.
-l rr{ý'-`T 71 1/'ý 1ýT"I I'
-ý-
..: _ý. _., _

op
IMF--
 -2 02 -

Appendix F

RATING SCALE

Coffee Taste Panel Instruction Sheet

1. Please sip a mouthful of each beverage.

2. Hold it in your mouth for 30 seconds.

3. Spit out or drink the beverage if you wish.

4. Please rate the beverage you tasted on a scale provided, by

circling the appropriate number. The stronger flavour at the

higher end of the scale (5,6,7) and the milder at the lower

end of the scale (1,2,3).

Beverage A1234567

B1234567

C. 1234567

5. Which of these coffee samples did you find objectionable, if any?

 - 203 -

Appendix G (p. 203 to 209)

COFFEE ASTRINGENCY

M. N. Clifford and Omozoje Ohiokpehai

Division of Nutrition and Food Science

Department of Biochemistry
University of Surrey
Guildford, GU2 5XH

Presented at a Royal Society of Chemistry,

Special Techniques Group Meeting
held at Metal Box Ltd., Wantage,
May 12th 1982
 -204 -

COFFEE ASTRINGENCY

Introduction

During informal discussions at the Eighth International Colloquium on

the Chemistry of Coffee (1977) two experienced coffee tasters expressed the
opinion that black coffee brew occasionally had a rather unpleasant astringent
taste. This sensation was encountered with green beans that were otherwise

unremarkable. It was often accompanied by a metallic note but was quite

distinct from bitterness. The cause was not knoten
.

The Shorter Oxford Dictionary defines "astringent" as "Having the

power to draw together the organic tissues; binding, constrictive, styptic".
According to Bate-Smith (1973) the sensation occurs when an astringent

compound precipitates salivary proteins and glycoproteins thus removing their

lubricating action. Among the compounds to have such a property are the
vegetable tannins, for which Haslam (1981) recommends the definition, "water-

soluble phenolic compounds having molecular weight between 500 and 3000, and
besides giving the usual phenolic reactions they have special properties such as
the ability to precipitate alkaloids, gelatin and other proteins". On structural

features these substances may be classified in two major groups: -

1) non-hydrolysable or condensed tannins which have a flavon-3-ol
repeating unit;
2) hydrolysable tannins which are easily hydrolysed by acids, bases and

some hydrolytic enzymes yielding either: -

a) ellagic acid (ellagitannins);

or b) gallic acid (gallotannins), both normally accompanied by glucose or

related polyol. Tara Tannin from Caesalpinia spinosa yields

gallic acid and quinic acid.

Haslam (1974,1981) considers that the distinctive structural feature

responsible for the astringent properties of these substances is, "the

accumulation within a molecule of moderate size of a substantial number of

unconjugated phenolic groups, many of which are associated with an o-dihydroxy

or o-trihydroxy orientation within a phenyl ring" and that at least two such

sites are required for an astringent cross links to form. Some data, although
 - 205 -

not all support the hypothesis that binding is via hydrogen bonding and Van
Sumere (1975) has suggested that unionised hydroxyls and unionised carboxyls

may act as binding sites. Carboxyl groups are found in some ellagitannins and
the quinic acid tannins.

Haslam (1974) has examined a range of tannins for the ability to

precipitate ß"-glycosidase. He has shown that these tannins contained 0.41 to

0.56 effective sites per 100 Daltons (effective site = 0-dihydroxy or o-trihydroxy

pnehyl group). An increase in this ratio from 0.41 to 0.47 caused a marked

increase in precipitation.

These data led us to ask if coffee beans might contain a similar

compound, but no literature references could be found, (for a review see

Clifford 1975,1979) to the presence of condensed or hydrolysable tannins.

There were data indicating the presence of dicaffeoyl quinic acids (DCQA) which

have a structure resembling the galloylquinic acids of Tara Tannin. The

effective site mass ratio for the DCQA is 0.38 if only the two caffcoyl residues

considered or 0.59 if the quinic acid carboxyl is also considered. Inoue,

are

et al (1965) reported for DCQA a pKa of 4.25 (at 20°) indicating extensive

ionisation at neutral pH values and thus the carboxyl group is unlikely to be an

important hydrogen bonding site. Nevertheless it seemed possible that the

DCQA would be astringent and we summarise our results to date.

Experimental
Green coffee beans were ground and extracted as previously described
(Clifford,1979) and analysed by reversed phase high pressure liquid

chromatography using the method of Van der Stegen and Van Dui jn (1980).

Screening for "traditional" tannins was by the method of Bate-Smith

(1973). Astringency was assessed by the haemoglobin precipitation method of
Bate-Smith (1973) and by the 0-glycosidase precipitation method of Haslam (1974).

Very crude DCQA was obtained from Roth GmbH, Halsruhe and purified

by repeated recrystallisation from water until the three DCQA peaks separated

by HPLC accounted for better than 95% of the material injected.

The ß-glycosidase was obtained from Roth GmbH, Karlsruhe.

 206 -

Haemolysed rats blood was supplied by the University of Surrey

Animal Unit.

Other reagents were normal commercial products.

Results and Discussion

1. Screening for "traditional'tannins

Negative results were obtained for five samples of beans representing

the species Coffee arabica, C. canephora var robusta and C. arabusta. It was

concluded that neither condensed nor hydrolysable tannins were present In these
commercial green beans.

2. DCQA Content

Fifty six samples of commercial green coffee contained 0.8-2.47 (dmb)

DCQA. Robustas generally contain more (1.8-2.2%) than arabicas (0.8-1.57)

and the arabusta hybrid is intermediate (1.477). Two arabica Santos samples

had exceptional DCQA contents of 2.1 and 2.47. The reason(s) for this are not

clear, but in a separate investigation to be reported elsewhere, it has been

observed that beans from slightly immature fruit may contain markedly higher

levels of DCQA than the corresponding fully mature beans.

Roasting causes a progressive loss such that the content in the roasted

bean is a function of the amount originally present and the severity of the roast.
In severe roasts there may be complete destruction. A brew prepared from

typical beans subjected to a moderate roast could contain up to 0.5 mg/ml.

Haemoglobin precipitation
Preliminary investigations using tannic acid showed that: -
1) haemoglobin precipitation is a function of haemoglobin concentration

and tannic acid concentration;

2) a plot of reduction in supernatant absorbance (0 A) against tannic acid
concentration is sigmoid, the precise nature of the curve depending on
the haemoglobin-tannic acid ratio.

These data fit a logistic curve (Causton, 1977) which has a linear form

Loge (y - 1) = Loge b- kt

where a= upper asymptote value i. e. supernatant absorbance of blank,

 207

y= sample A A,

t= final concentration of test substance that would have occurred

in the reaction mixture in the absence of precipitation. Using this trans-
formation it was possible to fit the data for seven sigmoid curves to one straight

line.

The data so obtained are summarised in Table 1. It can be seen that

relative astringency is not only a function of the phenol but also of the protein

used in assessment. These methods are not entirely satisfactory for application

to unknowns. Caffeoylquinic acid interferes with Bate-Smith's method probably

by converting red hemoglobin to brown methaemoglobin (see Kisugawa, et al,
1981) and Haslam's -glycosidase method is relatively insensitive to DCQA.
We converted Bate-Smith's spectrophotometric method to gravimetry using
freeze-drying to constant weight to avoid thermal damage to the precipitate.

Application of this method to bean extracts so far has failed to establish a

simple correlation between DCQA content and the extent of precipitation possibly
because extraneous substances are entrained in the precipitate.

Preliminary organoleptic evaluation suggests that the DCQA have a

threshold in the range 0.05 to 0.10 mg/ml in water and in coffee beans. Our

investigations are continuing but our provisional conclusions arc that DCQA
has an easily detected and peculiar lingering metallic taste which can influence

the acceptability of coffee brew.

Acknowledgements

The authors wish to thank the following for supplying coffee samples :-
Fed Cafe in Colombia
IFCC in the Ivory Coast

KIRDI in Kenya

The Royal Botanic Gardens at Kew and

Illycaffe spa of Trieste who also gave financial support.
The technical assistance of Fraulein Gerti Deffur, Fraulcin Karin

Schneider and Mr. M. Ascongh is also gratefully acknowledged.

 - 208 -

References

Bate-Smith, E. C. (1973) Phytochemistry 12,907-912.

Causton, D. R. (1977) A Biologist's Methematics, Edward Arnold.

Clifford, M. N. (1975) Process Biochemistry 10, (2) 20-23,29 and (4) 13-16,19.
Clifford, M. N. (1979) Food Chem. 4 63-71.
Haslam, E. (1974) Biochem. J. 139,285-288.
Haslam, E. (1981) in Conn, E. E. The Biochemistry of Plants, p. 527,
Academic Press.

Inoue, Y., Aoyagi S. and Nakanishi, K (1965) Chem. Pharm. Bull. 13,101-104.

Kisugawa, K., Arai, S. and Kurechi, T. (1981) Chem. Pharm. Bull. 29, '1694-1701.
Van der Stegen, G. H. D. and Van Duijn, J. (1981) 9th International Scientific
Colloquium on Coffee, 1,107-122, London, 1980, Association
Scientifique Internationale du Cafe, Paris.
Van Sumere, C. F. (1975) in The Chemistry and Biochemistry of Plant Proteins,
Phytochemical Society Symposium, Series No. 11, Edited Harborne, J. B.
and Van Sumere, C. F.
 - 209-

Table 1:

Logistic Curve Data and Relative As of

Tannic Acid and DCQA

Analytical Method
t50J Relative
and Test Substance bk
mg/mL Astringency

Bate-Smith
Tannic Acid 107.7 -12.90 0.36 11

DCQA 23.1 - 1.78 1.76 0.21 -

Haslam

Tannic Acid 2.0 -10.41 0.07 1 5.1

DCQA No precipitation at 1.25 mg/mL
Gravimetric

Tannic Acid 12.1 -19.20 0.13 1 2.8

DCQA 7.3 - 4.05 0.49 0.27 -

 210
Appendix H

SUMMARY OF LOGISTIC DATA OF GREEN COFFEE BEANS

Green Coffee Extract HPLC DCQA Haemoglobin Precipitation

(mg/mL Loge (y - 1)
extract) Bate-Smith Gravimetric

Kenya Discoloured Beans a 1.39 aa 27.2

1 9.3 3.5 0.9

2 12.9 2.9 -0.09

3 13.8 2.9 -0.27
4 11.4 2.2 -0.40
5 11.8 2.0 -0.46
r=0.29 ra0.72
Astringent Coffee Beans a"1.58 a" 27.1

IA Sautes Sagille 19.7 0.0 -0.27

10 Sautes Sagille 22.0 5.1 "0.41
1C Sautes Sagille (with
green silverskin) 22.0 2.1 -0.48
2A Sautes Aldebaran 20.4 3.9 -0.32
3 Zairi Robusta sample
no. 373 37.8 0.0 0.21

4 Zairi Robusta sample

no. 374 39.5 -0.13
5 Zairi Robuste sample
no. 375 39.2 -0.0W
6 Sautes batch no. 565 24.2 2.5 -0.41
7A Robusta from Zaire 39.5 3.5 -0.2
7B Robusta from Zaire 39.5 3.1 -0.3
8 Java Robuste 34.6 4.4 -0.1
Standard Commercial and r-0.01 r" 34

Computer-linked sorting
samples a- 28.2
1 10.6 0.1
2 11.3 -0.03
3 12.9 0.2
4 13.1 0.6

5 15.0 0.7

6 15.0 0.3

7 14.0 -0.3
8 9.2 -0.8
r"0.68
 -211 -

Appendix I (p. 212 to 217)

Reflectance spectra of peculiarly coloured beans.

 Reflectance spectra of' wh ; te beaus.

ýii

..
ý_
__--4-..
--- _.::

-_: t: --
r. =ý _.
; 7ý .
ý: ý. i

. -V. .;.. ý- .---

ij

1!
11

ý'
k
- --
 Reflectance spectra of foxy beans.

213
i.,
..

F- ,">, .,,

t--
- (_ t_z

-. . -,... - ..
==fiI-.
FT_
I_
_

I rI1 STr'.
jrr1h rr -
r r"
1

Trt

HH rtt r

1- ýr
t --
_
.
ý-i
f 'ý
_ .:. _- __T _. _. J_

.... .... _.. _, __. _. r. .... ..

Ii:
7 -t-'TTY
TL

ý7 77-
 R! f 1('( kulc e sptýct, 1' t of ' (I( ýýýt>-t;rr ýE 1u
ý11 ýýin .

i ý-
. -- c
ý: E
:. "i _-.:
ý
ý i
_'
,...
i
"!

9u
O

0
_.

ý-

--- ----r - ---

... i-ý----4ýr... -

.1
O
0

-r4
}i-

ff(fL
ýr1.

L j,

11ý - 'I;
\'
yý i
 flect, ance spectx
Fý-e a of

ý Tr r
trtr'it ýFr rýýt#, ýr -t
firL Yrrt rr"r r 1,1
.r r}+If I. t ri #1 ft tt l't. I
t11.
rtýrt rt Ir r} ýrr Y"" f
r t- ißt- l =1 it ttrýl f
rrr -rtr"-t
"l_.

\{
.]i-
r=T _t -
ýfC
C7'ß" .. A
ý-tI
+Tr/Tj(tF-rI-r1-.
rT

1r
$trt

Tý
r'
jr
,T
rT rl
ýlt,

::: :TT.
}
_r

iIr. 'ýr 1 Y- tr
^T trrr r it rr
ýýý:: ;{ rýrirt tart 1.
Ö{ I tlrýrý 4lr
r1 r j' 1rTI _
rr r_ ý{
'_ rrf rrf t+
it
rrr rr
rfrrl
r_". r1iýr rrI
rr. 7
.- . _. _ 1rtr r} t _ ({?
_ pl -r\..

i 1{ rt (j jtt
r +iri {rri I rr
riF rr1{ rf{ rl# rit4[r
F+ TFrý T,FTr 11

ýt ýc
rLtr! ýT {{ ý;: ## ; jE i{## i##f rli r
hr_ i} r{#

{t #{rr{ ý..
fi ff}rrj ir1T it-t
fZ =ý =f# I t: Lf t rrri trýttl, jt t}. il/1f ^-t-
I{t rt rfr j r-
f rrr}f ; r" F\týI
irirF'
r-t}-ýr-r ýr ,'fhrrrt. i rrt -T-1 7L ý'
-1
r,
-'_I rI
r
=- Ij
ti
-Lý
ri r ýý
\
fr: 1: irt
t 11F

fT
rtt }} ! ', ' S'
77 t rI _T
11 rý- rr }_ }_-r-: i i .. (d 'O ýi. ri
ýr
.
ýrL
47 _T1-
r!
ir I1 . 'I. -r.? rr.
"ý.
t- -Z__

irr t t( I It 't lifte ir ýr

ýýý r 1; ý{1 t(Iý +;, i iii
ci , trr r} ' i, -4-. 1-
_: t--
h'?`t-ýý{-}'ý'frl ._
1r !# ý-..
f

r 7i I T- Ti

rf

}- ` rt L rT I-It{ r-'
ý1
-r}r

\. . rl r

Ilr riEý (:
r Ei} :ý "I t__, rý ý"ý1
"i-ýr, r`ý, "` -r
`I1
. ..
rrrr t; ýF "r Ff , rrrrIr
If rtrr kr .rýr : ý:
I
r if jitr rýý r_i`" f
r -7
"f
rrT r "r ýr 'rr
ii 'r ý, rrt ' ýf
 Reflectance spectra of dark i)1v)wn t)E,,in: >.

L1i

- -=-.

-L- ,

cy
rt-t . r__:
rrT

C ý__ ý r(
ii tt -(
--. ý
t f. ,. rT fr -{
r
Ire
T_...
F }r
r
.. _.
7 7.

Tir T- T
rl fit' + r_ +ý +L

7-T- r

Fhr}

Et_ : _ff

Or
ýr `t

- '1
./
. r. I[ rf
I_fr
ý---
r'r r1
"Lt`

't
:. _.. .. "- ------i
1LFI(Ctýýtýc, ýý s --Irn of I)I, LCk

ritfr jý ", T ,+rj-r

!j