- Email: [email protected]
Analysis of the content of the diterpenes cafestol and kahweol in coffee brews
Pergamola
Food and Chemical Toxicology 35 (1997) 547-554
Analysis of the Content of the Diterpenes Cafestol and Kahweol in Coffee Brews G . G R O S S , E. J A C C A U D a n d A. C. H U G G E T T * Department of Quality and Safety Assurance, Nestl6 Research Centre, Vers-chez-les-Blanc, PC Box 44, CH-1000 Lausanne 26, Switzerland (Accepted 30 July 1996)
Abstract--The diterpenes cafestol and kahweol have been implicated as the components in boiled coffee responsible for its hypercholesterolaemic effects. These particular coffee constituents have also been shown to possess anticarcinogenic effects. A simple and sensitive reverse-phase HPLC method using solid-phase extraction has been developed for the analysis of cafestol and kahweol in coffee brews. This method was used to confirm that the method of coffee brewing is a major determinant of the cup content and hence level of consumption of these diterpenes. Scandinavian-style boiled coffee and Turkishstyle coffee contained the highest amounts, equivalent to 7.2 and 5.3 mg cafestol per cup and 7.2 and 5.4 mg kahweol per cup, respectively. In contrast, instant and drip-filtered coffee brews contained negligible amounts of these diterpenes, and espresso coffee contained intermediate amounts, about 1 mg cafestol and 1 mg kahweol per cup. These findings provide an explanation for the hypercholesterolaemic effect previously observed for boiled coffee and Turkish-style coffee, and the lack of effect of instant or drip-filtered coffee brews. This methodology will be of value in more correctly assessing the human exposure to these diterpenes through the consumption of coffee, and hence the potential physiological effects of different brews. © 1997 Elsevier Science Ltd
AbbreviatioJts: C = cafestol; K = kahweol; C + K = cafestol and kahweol; C + K/P = cafestol and kahweol palmitates.
INTRODUCTION Ever since the introduction of coffee into Europe in the 17th century ~:here has been interest concerning its possible beneficial or negative effects on human health. Recently, attention has focused on the biological effects of the major diterpenes present in coffee--cafestol and kahweol (Di Simplicio et al., 1989; Lam et al., 1982; Mensink et al., 1995; Miller et al., 1991; Wattenberg et al., 1985; Weusten-Van der Wouw et al., 1994). These components, which appear to be rela~fively specific to coffee, are found in both Arabica and Robusta (predominantly cafestol) varieties (Lercker et al., 1995) and comprise up to about 10-15% of the lipidic fraction of roasted coffee beans (Ratnayake et al., 1993). Although the total diterpene content of various brewed coffees has been reported previously (Ratnayake et al., 1993), there has been only one report on the specific content of cafestol and kahweol in different brews (Urgert et al., 1995b). The consumption of 'boiled' coffee, a brew peculiar to Scandinavian countries, has been shown to be associated with elevated levels of serum cholesterol (Pietinen et al., 1990; Stensvold et al., 1989; *Author for correspondence.
Thelle et al., 1983 and 1987). Epidemiological and controlled clinical studies have indicated that this hypercholesterolaemic effect is dependent on the method of preparation of the coffee brew (Arc et al., 1987; Bak and Grobbee, 1989; Theile and Van der Stegen, 1990). For example, in contrast to 'boiled' coffee, the consumption of instant or filtered coffee has no significant effect on serum cholesterol levels (Burr et al., 1989, Van Dusseldorp et al., 1991), whereas Turkish or Middle Easternstyle coffees appear to increase serum cholesterol (Kark et al., 1985). A recent series of clinical trials have confirmed that the hypercholesterolaemic agents are present in the lipid fraction of 'boiled' coffee and have identified the major causative agents as the diterpenes cafestol and kahweol which are present mainly as fatty acid esters (WeustenVan der Wouw et al., 1994). These investigations demonstrated a dose-response effect of cafestol and kahweol on increasing serum cholesterol levels. Feeding studies in rodents using coffee oils and purified coffee diterpenes have confirmed the hypercholesterolaemic effects of these compounds (Huggett et al., 1993; Ratnayake et al., 1995; Sanders and Sandaradura, 1992). However, the sensitivity of experimental animals to the hypercholesterolaemic effects of coffee diterpenes appears to be modulated
0278-6915/97/$17.00 + 0.00 © 1997 Elsevier Science Ltd. All rights reserved. Printed in Great Britain PII S0278-69 i 5(96)00123-8
548
G. Gross et al.
by the diet composition (Ratnayake et al., 1995), which may explain why some investigators were unable to demonstrate a cholesterol-raising effect of boiled coffee (Mensink et al., 1992). Coffee has been shown to contain cancer chemopreventive agents (Wattenberg, 1983). Aeschbacher and Jaccaud (1990) demonstrated that oral administration of coffee to mice decreased nitrosoureamediated D N A damage. When added to the diet of experimental animals, coffee inhibited the development of 7,12-dimethylbenz[a]anthracene (DMBA)induced oral carcinomas and mammary tumours (Miller et al., 1988 and 1993; Wattenberg, 1983). Further studies demonstrated that a (50:50) mixture of cafestol and kahweol produced a similar inhibition of buccal pouch tumours (Miller et al., 1991). The mechanisms responsible for the anticarcinogenic effects of cafestol and kahweol have not been elucidated; however, these compounds have been shown to be potent inducers of glutathione Stransferases (Di Simplicio et al., 1989; Huggett and Schilter, 1995; Lam et al., 1982). The enhancement of these activities may play a role in the chemopreventive action of the diterpenes by catalysing the detoxification of reactive carcinogenic electrophiles. Information on the content of cafestol and kahweol in coffee brews allows a prediction of the likely effects of different coffee brews on serum cholesterol levels and hence their potential role in cardiovascular disease, and also provides a basis for determining the relevance of these biologically active constituents with regard to the reported chemoprotective effects of coffee (Baron et al., 1994). Although methods for the analysis of total coffee diterpenes have been reported, these are of inadequate sensitivity or specificity for application to the measurement of individual diterpenes in coffee brews (Ratnayake et al., 1993). We report the development of a simple and sensitive reverse-phase HPLC method using solid-phase extraction procedures that has been applied to the analysis of cafestol and kahweol in various coffee brews. This methodology complements the gas chromatographic method for these components reported recently by Urgert et al. (1995b) and permits a corroboration of the finding that the method of preparation of the brew is a critical determining factor in determining the daily intake of these diterpenes from coffee consumption.
MATERIALS AND M E T H O D S
Materials and reagents
All solvents used were of analytical or HPLC grade. If not otherwise stated, the water used was 18 MfLcm quality obtained from a Milli-Q water purification system (Millipore, Bedford, MA, USA). A mixture of 17-O-palmitoyl-cafestol (52.5%, w/w) a n d 17-O-palmitoyl-kahweol (47.5%, w/w), hereafter referred to as C + K/P, prepared according to the method of Bertholet (1987), was used as refer-
ence standard. Its purity was checked by saponification and quantification of liberated diterpene alcohols, cafestol (C) and kahweol (K), by HPLC using the method described below. Major by-products in C + K/P were identified as isokahweolene (0.3%, w/w) and cafestolene (0.2%, w/w). Free cafestol and kahweol reference standard mixture [54.5% (w/w) cafestol; 45.5% (w/w) kahweol], hereafter referred to as C + K, was prepared from coffee oil as previously reported (Bertholet, 1987). C + K purity was confirmed by mass spectroscopy as well as by proton and carbon-13 nuclear magnetic resonance spectroscopy. Stock solutions of C + K/P in hexane (2.5mg/ml), and C + K in ethanol (0.5, 2.5 and 5 mg/ml) were prepared. Extrelut extraction cartridges (3 ml) were from Merck Ltd (Darmstadt, Germany). Bond-Elut PRS (propyl sulfonic acid; 100rag) and C18 (100rag) cartridges and cartridge coupling adapters were from Analytichem International (ICT AG, Basle, Switzerland). Purified extracts were stored in 1.1 ml Chromacol gold microvials (Infochroma AG, Zug, Switzerland). Solid-phase extraction cartridge conditioning
C18 cartridges were rinsed with 3 ml 2-propanolethyl acetate (1:1). PRS cartridges were rinsed with, in order, 3 ml 1 y hydrochloric acid solution, 20 ml water and 3 ml 2-propanol-ethyl acetate (1:1). One C18 was coupled to a PRS cartridge using an adapter piece. This assembly is hereafter referred to as 'tandem'. HPLC
Chromatographic analysis was performed using a Merck L-6200 pump, a Gilson 231-401 automated sample processor and injector, and a Hewlett Packard model 1040A/mark II multi-channel ultraviolet detector driven by a HP9000 series 300 workstation (Chemstation). A 250 x 4.6 mm Merck Superspher Lichrocart 100 RP-18 column (3/~m particle size) was used to separate the diterpenes. The mobile phase consisted of water (solvent A) and methanol (solvent B). Separations were carried out at ambient temperature using linear gradient elution conditions (0 rain: 30% A, 70% B; 20 rain. 5% A, 95% B) and a flow rate of 1 ml/min. C and K were detected at 230 nm and 290 nm, respectively. Peaks were quantified by area measurement and identified by retention times and by matching on-line recorded spectra with library ultraviolet spectra. Injections were performed in the partial loop filling mode using a custom program for the Gilson '231-401 injector. In general, a 20 ml portion of purified extract or a 5 pl standard preparation w a s injected into a 50 pl loop, followed by dilutant liquid (20/A) to compensate for the injector port dead volume. The HPLC instrument was calibrated on a weekly basis by injecting each C + K stock solution.
Cafestol and kahweol in coffee brews
Preparation of coffee brews All coffee brews were prepared using a commercial Arabica blend of roasted coffee purchased in Switzerland, except for single-serving size espresso brews which were purchased as prepacked capsules containing different blends, and instant coffee, which was prepared from two different commercial soluble coffee blends. For preparation of 'boiled coffee', coffee beans were coarsely ground (800900 pm), whereas for other coffee brews the coffee beans were ground to the same medium-fine particle size (500-550 pm). Tap water was used to prepare all coffee brews. Brewing methods, brew strengths and cup sizes were standardized according to IARC (1991). Boiled coffee. Was prepared in the traditional manner by boiling 9 g coarsely ground roasted coffee with 150 ml boiling water for 10 min. The mixture was hot-filtered through a metal screen (mesh size 0.5 mm) and the filtrate used for analysis. Turkish coffee. Was prepared by bringing to a gentle boil a mixture of 5 g roast and ground coffee, 10 g saccharose and 60 ml cold water. Boiling was interrupted when a foam developed on the surface of the brew. The brew was left to settle and the supernatant was taken for analysis. Mocha (moka, Neapolitan) coffee. Was prepared using a commercial Italian coffee maker. The water reservoir, filter cup and coffee recipient were separated and the filter cup filled with 20 g roasted coffee powder. Cold water (300 ml) was added to the bottom reservoir, and the filter cup was inserted. The coffee maker was reassembled and then heated on a hotplate until the bottom water reservoir was empty. The resulting coffee brew was mixed prior to analysis. Standard espresso coffee. Was prepared using a mechanical-type Turmix EX10 coffee machine (Turmix Ltd, Jona, Switzerland). 20 g of roast and ground coffee was used to draw a 130 ml volume. Single-serving size espresso. Was prepared using a Turmix CII01 coffee machine (Turmix Ltd, Jona, Switzerland). One commercially available disposable capsule containing 5 g roasted coffee was used to draw a cup of about 50 ml. Five different commercial prepacked coffee blends were examined. Drip-filtered coffee. Was prepared by adding 13 g roast and ground coffee into a Tchibo No. 4 type paper filter (Tchibo GmbH, Hamburg, Germany) placed on an aplgropriate funnel. Boiling water (200 ml) was added and the filtered brewed coffee was collected. Instant coffee. Was prepared by mixing 170 ml boiling water with 2 g instant coffee granules per cup. Two different commercial coffee blends were examined. Extraction procedure. The brewed coffees were allowed to cool to room temperature and adjusted to pH 9-10 using 4 N sodium hydroxide solution. Aliquots of 2.5 ml were transferred into plastic test tubes (Falcon 2059, Becton Dickinson, N J, USA)
549
and spiked with suitable quantities of C + K/P standard solution using a graduated microlitre syringe (Hamilton 1705/1710, Hamilton Ltd, Bonaduz, Switzerland). The resulting mixtures were vortex mixed and applied to an Extrelut-3 columns placed on a Supelco Visiprep ® vacuum manifold. The application of a slight vacuum accelerated the adsorption of the liquid phase onto the column material. The Extrelut columns were then placed on a rack and fine luer lock needles (25G x 5/8 inch) were applied to the tips of the columns to serve as flow restrictors. Each sample tube was rinsed with 2.5 ml hexane-dichloromethane (95:5, v/v) which was then poured into the respective Extrelut column. After standing for approximately 15-30 minutes, the Extrelut columns were eluted under gravity flow by the addition of 12.5ml hexane-dichloromethane (95:5) to the column reservoir. Approximately 11 ml eluate per column was collected into 25 ml pearshaped flasks. The eluates were evaporated to dryness at ambient temperature under reduced pressure. The resulting residues were saponified by the addition of 2 ml methanolic 0.3 N potassium hydroxide solution. Following incubation with agitation for 60 min at room temperature using a model KL2 agitator (Universal Kleinstschtittler, Edmund BiJhler GmbH, Tiibingen, Germany), 800/~1 2-propanol-ethyl acetate (1:1, v/v) and 800/~1 0.5 M aqueous sodium dihydrogen phosphate were added. After thorough mixing, this preparation was transferred into the top reservoir of previously conditioned C18/PRS solid-phase extraction cartridge tandems and slowly eluted through both cartridges by applying gentle air pressure with a plastic syringe attached to the top of the tandem. The eluate was collected in 16 x 76 mm poly-allomer test-tubes (Beckman Instruments, Palo Alto, CA, USA). The flask containing the incubation mixtures was rinsed with 1 ml 2-propanol-ethyl acetate (1:1, v/v), and the washings were passed through the tandem and combined with the previous eluate. The combined eluate was concentrated using a model SVCI00H Speed-Vac vacuum centrifuge with a RT100A cryotrap (Savant Instruments, Farmingdale, NY, USA). The residue was reconstituted in 400/~1 methanol with vortex-mixing and then filtered through a 0.2pro zero dead volume Nalgene filter tip (Nalgene Company, Rochester, NY, USA) into an HPLC microvial. Samples were stored at +4°C prior to HPLC analysis.
Quantification Quantitative analysis was performed using external standard calibrations. The method of standard additions was employed to check C and K extraction efficiencies and the results were corrected accordingly. In general, one preliminary determination was carried out to approximate the C and K levels in the examined product. This was then followed by triplicate determinations using one unspiked sample and two parallel samples spiked
G. Gross et al.
550
with the C + K/P mixture prior to the extraction procedure so as to double and triple the estimated levels of C and K. The extraction efficiencies were calculated for both analyses as the slope of the linear regression line using the parameters added (x) v. measured (y) analyte concentration. If not otherwise stated, the results are expressed as concentrations of free diterpene alcohols. RESULTS
Optimization of analytical methodology Diterpene alcohols were analysed by HPLC following extraction of the corresponding esters from the coffee brews, saponification, and clean-up of the saponified mixture using solid-phase extraction technology. The method used is schematized in Fig. 1. Initially, water/hexane partitioning of the coffee brew was evaluated as a means to pre-concentrate the diterpene esters. However, this procedure was abandoned owing to technical difficulties (i.e. formation of stable emulsions and multiple liquid phases). The use of diatomaceous earth (Extrelut ®) to immobilize the aqueous liquid phase permitted the efficient extraction of diterpene esters with hexane while avoiding the formation of emulsions. The saponification of C and K proceeded smoothly at room temperature and was complete within 1 hr. Longer hydrolysis times diminished the yields, while much shorter times resulted in incomplete saponification (data not shown). The crude hydrolysate was purified using Coffee brew pH 9 Partition on Extrelut-3 Extraction with hexane- DCM, 95:5 (IncreasedC + K palmitateextractibility) Drying/evaporation (Concentration) Saponification (KOH/MeOH) Neutralization (NaH2PO4.2H20) SPE:
C18
l
PRS
Residual linidsremoval Extraction withAcOEt: 2-propanol, 1:1 (IncreasedC + K extractibility) Salts removal Drying/evaporation
Reconstitution(MeOH) HPLC analysis Fig. 1. Outline of the procedure employed for the solidphase extraction of diterpenes from coffee brews: DCM = dichloromethane; SPE = solid-phase extraction; AcOEt = ethyl acetate; MeOH = methanol.
C18 silica to remove residual lipids, followed by a propyl sulfonic acid cation exchanger to reduce the salt content of the extracts. These two processes were conveniently combined in a one-step procedure by the use of coupled cartridges. Following reconstitution, the resulting extract could then be directly injected into the liquid chromatograph. The accuracy of the quantification of C and K levels depends critically on the extraction efficiency of the corresponding esters from crude coffee, the yield of the saponification, and the recovery of free C and K during the final filtration. In order to optimize these parameters, we used the method of standard additions. Employing the protocol described, including the use of triplicate extractions, a set of three data pairs [spiked concentration (x)/determined concentration (y)] was created for each analyte and the extraction efficiency expressed as the slope of the respective linear regression line (r 2 = 0.999 and r 2 = 0.997 for C and K, respectively). The mean extraction efficiencies were 97 + 19% (n = 16) for C and 96_+ 19% (n = 16) for K. The specificity of C and K determinations was provided by the selective detection of both compounds at their respective wavelength of maximal absorption, as well as the match of on-peak recorded ultraviolet spectra with references. The limit of quantification was estimated as approximately 0.05 mg/litre for both C and K, sufficiently sensitive for the analysis of all of the coffee brews investigated in the present study. The chromatograms obtained from analyses of an espresso blend soluble coffee are shown in Fig. 2. In our laboratory the method showed good repeatability as demonstrated by the coefficient of variation of 8.6% (C) and 8.7% (K) obtained following triplicate determinations of the same sample (containing 17.3 mg C/litre and 16.3 mg K/litre, respectively).
Quantification of cafestol and kahweol m various coffee brews C and K were analysed in seven different coffee brews prepared using a variety of methods. The results are summarized in Table 1. Interestingly, Turkish and not boiled coffee contained the highest concentration of C and K. However, when the results were expressed on the basis of cup size, it was apparent that boiled coffee (14.4mg/cup) provides more of these diterpenes than Turkish-style coffee (10.7 mg/cup). The amount of C and K measured in mocha (Neopolitan)-style coffee was also relatively high, about 75% of the concentrations found in boiled coffee and about 32% when expressed as a ratio of cup size (4.6 rag/cup). The C and K concentrations found in standard espresso brewed coffee were almost three times less than observed for boiled coffee, despite a brewing strength almost 2.5-fold greater than that used for boiled coffee. When expressed as the amount per cup, the C + K level
Cafestol and kahweol in coffee brews
551
230 nm !
230 nm
~
Reference spectrum ._= E eq
290 nm I
c)
I~" Kahweol jK~~ference spectrum
250 30O 35O Waye!ength(nm) C ]B A 290 nm
0
2
4
6
8
IO
12
14
16
18
Time (min) Fig. 2. Chromatographic profiles of espresso blend instant coffee recorded at 230 nm (top) and 290 nm (bottom): A = unspiked sample containing 44 pg C/g and 40 #g K/g (values not corrected for incomplete extraction efficiency); B and C = profiles of the same sample spiked with 209/~gC/g and 171 pg K/g, and 418 #g C/g and 342 ttg K/g, respectively, prior to analysis; mAU = milli Absorbance Units.
of espresso (2 mg/cup) was over seven times lower than for boiled coffee. The C + K concentration of espresso coffee brews prepared from single-serving size prepacked roast and ground coffees was also considerably lower than for boiled coffee. Furthermore, the amount of C and K in these espresso brews prepared from capsules (Table 1, Blends A-E, 0.15-0.35mg C + K / c u p ) ranged from about five to 14 times lower than for the espressos prepared using the traditional method. The blend of coffee employed in the capsules appeared to influence the diterpene content of the brew to some exterLt. Instant coffees contained very low concentrations of C and K (0.2-0.5 rag/cup), providing between about 25 and 70 times less C + K per cup than boiled coffee. Drip-filtered cof-
fee contained the least C + K, equivalent to a total of 0.04 mg per cup. DISCUSSION A simple and sensitive reverse-phase HPLC method using solid-phase extraction has been developed for the analysis of C and K in coffee brews. This method benefits from a straightforward extraction procedure that overcomes the problems of emulsion formation and the presence of multiple liquid phases encountered when using direct solvent extractions of coffee. The method of standard additions was selected in place of an internal standard for quantification, since an appropriate internal standard was not available. An internal standard
552
G. Gross et al. Table 1. Analysis of diterpene alcohols in coffee brews Diterpene alcohols Brew strength (mg/ml)
Cafestol
Kahweol
Coffee brew
IARC
Used
mg/litre+ SD
mg/cup*
mg/g
mg/litre + SD
mg/cup*
mg/g
Turkish coffee Boiled coffee Mocha coffee
83 50-75
83 60 67
88.7 ± 4.0 48.3 + 3.8 37.5 + 1.3
5.3 7.2 2.3
1.07 0.81 0.56
89.9 + 4.1 48 + 2.5 38.5 + 0.9
5.4 7.2 2.3
1.08 0.8 0.57
Espresso coffee Exp. 1 100-300 Exp. 2 100-300
154 154
17.3 ± 1.5 16.7+0.3
1.0 1.0
0.11 0.11
16.3 + 1.4 17.1 +0.5
1.0 1.0
0.11 0.11
Single-servingsize espresso Blend A Blend B Blend C Blend D Blend E Filtered coffee 25-75
96 106 114 125 106 65
3.4 + 0.3 2.1 _+0.3 2.2 + 0.2 1.6 + 0.7 1.2 + 0.3 0.12 + 0.02
0.17 0.10 0.11 0.08 0.06 0.02
0.04 0.02 0.02 0.01 0.01 0.002
3.5 + 0.3 2.9 + 0.2 2.6 ± 0.3 1.8 + 0.8 1.7 + 0.3 0.14 + 0.03
0.18 0.15 0.13 0.09 0.09 0.02
0.04 0.03 0.02 0.01 0.02 0.002
Instant coffee Standard blend Espresso blend
12 12
1.9+0.05 0.7 + 0.07
0.3 0.1
0.16 0.06
1.9+0.01 0.7 + 0.10
0.3 0.1
0.16 0.06
10 16 10-16
*Coffee cup sizes are as follows: 150 ml for boiled, filtered and instant coffees; 60 ml for Turkish, mocha and espresso coffees; and 50 ml for the single serving size espresso coffees.
selected purely on the basis of its chromatographic behaviour (Urgert et al., 1995b) suffers from an inability to take into account the variability that may occur in saponification or extraction efficiencies. Turkish-style coffee, and not boiled coffee, contained the highest concentration of C and K. A similar result was previously reported by Ratnayake et al. (1993) and this may be a consequence of the higher amount of fines present in Turkish-style coffee, compared with 'boiled coffee'. Thus the method by which 'boiled coffee' is decanted may have a large influence on its diterpene content. Furthermore, increasing the time of boiling did not increase the C and K contents of the 'boiled coffee' (data not shown). Similar findings have been reported by Urgert et al. (1995b), who observed extremely large variations in field samples of these coffee brews and attributed this variation to differences in brewing strength and the amount of bean particles decanted with the brew. They also reported that field samples of Scandinavian boiled and Turkish/Greek coffees contained between 1 and 10mg cafestol per cup (equivalent to about 220 mg/cup) of C + K. The values that we obtained for a standardized boiled coffee (14.4 mg/cup) and Turkish coffee (10.7 mg/cup) prepared under laboratory conditions from the same roast and ground coffee beans according to IARC (1991) fall within this range. The C + K content for mocha (Neopolitan)-style coffee (4.6mg/cup) is a little higher than that previously reported by Urgert et al. (1995b) for a laboratory-prepared brew (2.5mg/ cup); nevertheless, in both studies this brewing method results in a coffee providing intermediate amounts of C and K. It is interesting to note that it was recently reported that consumption of three cups per day of Italian brewed coffee (moka,
mocha) for 5 wk in a double-blind crossover clinical trial had no significant effect on serum cholesterol levels (Sanguigni et al., 1995). The C and K concentrations found in standard espresso brewed coffee (2 mg/cup) is consistent with the field samples (about 0-6 mg/cup) reported elsewhere (Urgert et al., 1995b). However, this finding is in contrast to that reported by Ratnayake et al. (1993) which indicated that espresso coffee contained twice the level of total diterpenes compared to 'boiled' and Turkish coffees. The reason for this discrepancy is unclear, but it is likely that diterpene levels may be influenced by steam pressure, contact time of steam with grounds, and the efficacy of the metal screen strainer in retaining fine coffee particles, thereby preventing their contribution to the diterpene content of the brewed coffee. The C + K contents of coffee brews prepared from single-serving size prepacked roast and ground coffees (0.150.35 mg/cup) were considerably lower than those of traditional espresso coffee. The reason for this difference is not obviously apparent but it seems likely that the prepacked capsules act as a filter, thus allowing fewer fine particles of coffee grounds to pass into the brew. This explanation is consistent with a recent report indicating that spent fine coffee grounds raise serum cholesterol levels when administered to human volunteers, indicating that the fines are an important source of diterpenes (Urgert et al., 1995a). Both drip-filtered and instant coffees contained very low concentrations of C + K, a finding which is in good agreement with that reported by Urgert et al. (1995b) and also with the low levels of total lipids and total diterpenes found by Ratnayake et al. (1993) for these types of coffee. The low levels of diterpene esters in drip-filtered coffee prepared
Cafestol and kahweol in coffee brews
1.5 --
= 1.0 O
g
~ 0.5 -
e-
0.0
'
Turkish
s' r' so,I I
iI
i
i
Ii
513
J
~
I
I
i
i
I
100
I
[
150
I
k
i
[
200
C + K (mg/day) Fig. 3. Predicted increase in serum total cholesterol levels following the consunaption of five cups/day of different coffee brews for 4 wk. The dose-response curve is based upon data points (O) taken from studies in human volunteers previously repo:rted by Weusten-Van der Wouw et al. (1994). The estimated C + K intakes are based upon the consumption of five cups of coffee brews of the following sizes: boiled, filtered, instant--150ml; Turkish, mocha, espresso--60 ml; prepacked espresso capsules--50 ml. The predicted increase in serum cholesterol is less than 0.1 mmol/litre for coffee brews (single-serving size espresso, instant, filtered) not shown in the figure. using a paper filter can be explained by both the removal of fines and the strong lipid-binding effect of filter paper (Ahola et al., 1991). For instant coffee, a large proportion of the diterpenes is lost during industrial processing. The potential cholesterol-raising effects of the different coffee brews can be estimated on the basis of the results of clinical studies in which 'boiled coffee', coffee oils or purified diterpenes were administered to volunteers, as reported by Weusten-Van der W o u w et al. (1994). A review of these studies demonstrates a dose-dependent effect of the daily ingestion of preparations containing C and K for 4 wk on the increase in serum cholesterol levels (Fig. 3). If it is assumed that five cups of coffee are consumed daily, oll the basis of the results reported in the present investigation, the expected increase in cholesterol levels would range from less than 0.05mmol/litre for filtered coffee to about 0.85 mmol/litre for boiled coffee. It must be emphasized that this extrapolation is based on assumptions and uncertainities concerning the diterpene contents of some of the test materials used in the clinical studies. Nevertheless, it is interesting to note that Pietinen et al. (1990) reported that Finnish men and women who regularly drank seven to nine cups of boiled coffee per day had serum cholesterol levels 0.6 mmol/litre higher than those of consumers of filtered coffee. Furthermore, a cross-sectional study in Norwegians found that subjects habitually consuming five or more cups of boiled coffee per day had mean serum cholesterol levels 0.3 mmol/ litre higher than those of matched filter coffee drinkers. (Weusten-Van der W o u w et al., 1994). In conclusion, the use of solid-phase extraction techniques eliminates the problems of multiple phase separation that are often encountered when
553
liquid-liquid extraction techniques are used with coffee brews. The results for C and K in different coffee brews are consistent with those previously reported and with the results predicted from an evaluation of clinical and epidemiological studies comparing the hypercholesterolaemic effects of different brews. Finally, our fndings indicate that the levels of C and K found in traditional espresso coffees are lower than previously suspected and predict that instant and filtered coffees would not have any effect on serum cholesterol levels. This new methodology should help to assess more correctly the exposure to these compounds through the consumption of coffee and to predict more accurately the physiological effects of different coffee brews. REFERENCES
Aeschbacher H. U. and Jaccaud E. (1990) Inhibition by coffee of nitrosourea mediated DNA damage in mice. Food and Chemical Toxicology 28, 633-637. Ahola I., Jauhiainen M. and Aro A. (1991) The hypercholesterolaemic factor of boiled coffee is retained by a paper filter. Journal of Internal Medicine 230, 293-297. Aro A., Tuomilehto J., Kostiainen E., Uusitalo U. and Pietinen P. (1987) Boiled coffee increases serum low density lipoprotein concentration. Metabolism 36, 10271030. Bak A. A. A. and Grobbee D. E. (1989) The effect on serum cholesterol levels of coffee brewed by filtering or boiling. New England Journal of Medicine 321, 14321437. Baron J. A., de Verdier M. G. and Ekbom A. (1994) Coffee, tea, tobacco, and cancer of the large bowel. Cancer Epidemiology, Biomarkers and Prevention 3, 565570. Bertholet R. (1987) Preparation of cafestol. United States Patent 4,692,534. Burr M. L., Gallacher J. E., Butland B. K., Bolton C. H. and Downs L. G. (1989) Coffee, blood pressure and plasma lipids: a randomized controlled trial. European Journal of Clinical Nutrition 43, 477-483. Di Simplicio P., Jensson H. and Mannervik B. (1989) Effects of inducers of drug metabolism on basic hepatic forms of mouse glutathione transferase. Biochemical Journal 263, 679-685. Huggett A. C., Marchesini M. and Perrin I. (1993) Effects of coffee oil on plasma cholesterol in the rat. In 15th International Scient!fic Colloquium on Coffee. pp. 497503. Association Scientifique International du Caf& Paris. Huggett A. C. and Schilter B. (1995) The chemoprotective effects of cafestol and kahweol: effects on xenobiotic metabolising enzymes. Journal of Cellular Biochemistry 19A, 190. IARC (1991) IARC Monographs on the Evaluation of the Carcinogenic Risks to Humans. Vol. 51. Coffee, Tea, Mat~, Methylxanthines and Methylglyoxal. pp. 41-206.
International Agency for Research on Cancer, Lyon. Kark J. D., Friedlander Y., Kaufmann N. A. and Stein Y. (1985) Coffee, tea and plasma cholesterol: the Jerusalem Lipid Research Clinic prevalence study. British Medical Journal 291, 699-704. Lam L. T. K., Sparnins V. L. and Wattenberg L. W. (1982) Isolation and identification of kahweol palmitate and cafestol palmitate as active constituents of green coffee beans that enhance glutathione S-transferase activity in the mouse. Cancer Research 42, 1193-1198. Lercker G., Frega N., Bocci F. and Rodriguez-Estrada M. T. (1995) High resolution gas chromatographic determination of diterpenic alcohols and sterols in coffee lipids. Chromatographia 41, 29-33.
554
G. Gross et al.
Mensink R. P., Lebbink W. J., Lobbezoo 1. E., WeustenVan der Wouw M. P. M. E., Zock P. L. and Katan M. B. (1995) Diterpene composition of oils from Arabica and Robusta coffee beans and their effects on serum lipids in man. Journal of Internal Medicine 237, 543-550. Mensink R. P., Zock P. L., Katan M. B. and Beynen A. C. (1992) Boiled coffee does not increase serum cholesterol in gerbils and hamsters. Zeitschrift far Ernahrungswissenschaft 31, 82-85. Miller E. G., Formby W. A., Rivera-Hidalgo F. and Wright J. M. (1988) Inhibition of hamster buccal pouch carcinogenesis by green coffee beans. Oral Surgery 65, 745-749. Miller E. G., Gonzales-Sanders A. P., Couvillon A. M., Binnie W. H., Sunahara G. I. and Bertholet R. (1993) Inhibition of oral carcinogenesis by roasted beans and roasted coffee bean fractions. In 15th International Scientific Colloquium on Coffee. pp. 420-425. Association Scientifique International du Caf6, Paris. Miller E. G., McWhorter K., Rivera-Hidalgo F., Wright J. M., Hirsbrunner P. and Sunahara G. I. (1991) Kahweol and cafestol: inhibitors of hamster buccal pouch carcinogenesis. Nutrition and Cancer 15, 41-46. Pietinen P., Aro A., Tuomitehto J., Uusitalo U. and Korhonen H. (1990) Consumption of boiled coffee is correlated with serum cholesterol in Finland. International Journal of Epidemiology 19, 586-590. Ratnayake W. M. N., Hollywood R., O'Grady E. and Stavric B. (1993) Lipid content and composition of coffee brews prepared by different methods. Food and Chemical Toxicology 31, 263-269. Ratnayake W. M. N., Pelletier G., Hollywood R., Malcolm S. and Stavric B. (1995) Investigation of the coffee lipids on serum cholesterol in hamsters. Food and Chemical Toxicology 33, 195-201. Sanders T. A. B. and Sandaradura S. (1992) The cholesterol-raising effect of coffee in the Syrian hamster. British Journal of Nutrition 68, 431434. Sanguigni V., Gallu M. P., Rul~ni M. P. and Strano A. (1995) Effects of coffee on serum cholesterol and lipoproteins: the Italian brewing method. European Journal of Epidemiology 11, 75-78.
Stensvold I., Tverdal A. and Foss O. P. (1989) The effect of coffee on blood lipids and blood pressure. Results from a Norwegian cross-sectional study, men and women, 40-42 years. Journal of Clinical Epidemiology 42, 877-884. Thelle D. S.~ Arnesen E. and Forde O. H. (1983) The Tromso heart study: does coffee raise serum cholesterol? New England Journal of Medicine 308, 1454-1457. Thelle D. S., Heyden S. and Fodor J. G. (1987) Coffee and cholesterol in epidemiological and experimental studies. Atherosclerosis 67, 97 103. Thelle D.S. and Van der Stegen G. (1990) In Coffee and
Coronary Heart Disease, Special Emphasis on the CoffeeBlood Lipids Relationship. Edited by D.S. Thelle and G. Van der Stegen. pp. 9-15. Gotab, Stockholm. Urgert R., Schulz A. G. M. and Katan M. B. (1995a) Effects of cafestol and kahweol from coffee grounds on serum lipids and serum liver enzymes in humans. American Journal of Clinical Nutrition 61, 149-154. Urgert R., van der Weg G., Kosmeijer-Schuil T. G., van de Bovenkamp P., Hovenier R. and Katan M. B. (1995b) Levels of the cholesterol-elevating diterpenes cafestol and kahweol in various coffee brews. Journal of Agricultural and Food Chemistry 43, 2167-2172. Van Dusseldorp M., Katan M. B., Van Vliet T., Demacker P. N. M. and Stalenhoef A. F. H. (1991) Cholesterol-raising factor from boiled coffee does not pass a paper filter. Arteriosclerosis and Thrombosis 11, 586 593. Wattenberg L. W. (1983) Inhibition of neoplasia by minor dietary constituents. Cancer Research 43, 2448 2453. Wattenberg L. W., Hanley A. B., Barany G., Sparnins V. L., Lam L. K. T. and Fenwick G. R. (1985) Inhibition of carcinogenesis by some minor dietary constituents. International Symposium. Princess Takamatsu Cancer Research Fund 16, 193-203. Weusten-Van der Wouw M. P. M. E., Katan M. B., Viani R., Huggett A. C., Liardon R., Lund-Larsen P. G., Thelle D. S., Ahola I., Aro A., Meyboom S. and Beynen A. C. (1994) Identity of the cholesterol-raising factor from boiled coffee and its effect on liver function enzymes. Journal of Lipid Research 35, 721-733.
Food and Chemical Toxicology 35 (1997) 547-554
Analysis of the Content of the Diterpenes Cafestol and Kahweol in Coffee Brews G . G R O S S , E. J A C C A U D a n d A. C. H U G G E T T * Department of Quality and Safety Assurance, Nestl6 Research Centre, Vers-chez-les-Blanc, PC Box 44, CH-1000 Lausanne 26, Switzerland (Accepted 30 July 1996)
Abstract--The diterpenes cafestol and kahweol have been implicated as the components in boiled coffee responsible for its hypercholesterolaemic effects. These particular coffee constituents have also been shown to possess anticarcinogenic effects. A simple and sensitive reverse-phase HPLC method using solid-phase extraction has been developed for the analysis of cafestol and kahweol in coffee brews. This method was used to confirm that the method of coffee brewing is a major determinant of the cup content and hence level of consumption of these diterpenes. Scandinavian-style boiled coffee and Turkishstyle coffee contained the highest amounts, equivalent to 7.2 and 5.3 mg cafestol per cup and 7.2 and 5.4 mg kahweol per cup, respectively. In contrast, instant and drip-filtered coffee brews contained negligible amounts of these diterpenes, and espresso coffee contained intermediate amounts, about 1 mg cafestol and 1 mg kahweol per cup. These findings provide an explanation for the hypercholesterolaemic effect previously observed for boiled coffee and Turkish-style coffee, and the lack of effect of instant or drip-filtered coffee brews. This methodology will be of value in more correctly assessing the human exposure to these diterpenes through the consumption of coffee, and hence the potential physiological effects of different brews. © 1997 Elsevier Science Ltd
AbbreviatioJts: C = cafestol; K = kahweol; C + K = cafestol and kahweol; C + K/P = cafestol and kahweol palmitates.
INTRODUCTION Ever since the introduction of coffee into Europe in the 17th century ~:here has been interest concerning its possible beneficial or negative effects on human health. Recently, attention has focused on the biological effects of the major diterpenes present in coffee--cafestol and kahweol (Di Simplicio et al., 1989; Lam et al., 1982; Mensink et al., 1995; Miller et al., 1991; Wattenberg et al., 1985; Weusten-Van der Wouw et al., 1994). These components, which appear to be rela~fively specific to coffee, are found in both Arabica and Robusta (predominantly cafestol) varieties (Lercker et al., 1995) and comprise up to about 10-15% of the lipidic fraction of roasted coffee beans (Ratnayake et al., 1993). Although the total diterpene content of various brewed coffees has been reported previously (Ratnayake et al., 1993), there has been only one report on the specific content of cafestol and kahweol in different brews (Urgert et al., 1995b). The consumption of 'boiled' coffee, a brew peculiar to Scandinavian countries, has been shown to be associated with elevated levels of serum cholesterol (Pietinen et al., 1990; Stensvold et al., 1989; *Author for correspondence.
Thelle et al., 1983 and 1987). Epidemiological and controlled clinical studies have indicated that this hypercholesterolaemic effect is dependent on the method of preparation of the coffee brew (Arc et al., 1987; Bak and Grobbee, 1989; Theile and Van der Stegen, 1990). For example, in contrast to 'boiled' coffee, the consumption of instant or filtered coffee has no significant effect on serum cholesterol levels (Burr et al., 1989, Van Dusseldorp et al., 1991), whereas Turkish or Middle Easternstyle coffees appear to increase serum cholesterol (Kark et al., 1985). A recent series of clinical trials have confirmed that the hypercholesterolaemic agents are present in the lipid fraction of 'boiled' coffee and have identified the major causative agents as the diterpenes cafestol and kahweol which are present mainly as fatty acid esters (WeustenVan der Wouw et al., 1994). These investigations demonstrated a dose-response effect of cafestol and kahweol on increasing serum cholesterol levels. Feeding studies in rodents using coffee oils and purified coffee diterpenes have confirmed the hypercholesterolaemic effects of these compounds (Huggett et al., 1993; Ratnayake et al., 1995; Sanders and Sandaradura, 1992). However, the sensitivity of experimental animals to the hypercholesterolaemic effects of coffee diterpenes appears to be modulated
0278-6915/97/$17.00 + 0.00 © 1997 Elsevier Science Ltd. All rights reserved. Printed in Great Britain PII S0278-69 i 5(96)00123-8
548
G. Gross et al.
by the diet composition (Ratnayake et al., 1995), which may explain why some investigators were unable to demonstrate a cholesterol-raising effect of boiled coffee (Mensink et al., 1992). Coffee has been shown to contain cancer chemopreventive agents (Wattenberg, 1983). Aeschbacher and Jaccaud (1990) demonstrated that oral administration of coffee to mice decreased nitrosoureamediated D N A damage. When added to the diet of experimental animals, coffee inhibited the development of 7,12-dimethylbenz[a]anthracene (DMBA)induced oral carcinomas and mammary tumours (Miller et al., 1988 and 1993; Wattenberg, 1983). Further studies demonstrated that a (50:50) mixture of cafestol and kahweol produced a similar inhibition of buccal pouch tumours (Miller et al., 1991). The mechanisms responsible for the anticarcinogenic effects of cafestol and kahweol have not been elucidated; however, these compounds have been shown to be potent inducers of glutathione Stransferases (Di Simplicio et al., 1989; Huggett and Schilter, 1995; Lam et al., 1982). The enhancement of these activities may play a role in the chemopreventive action of the diterpenes by catalysing the detoxification of reactive carcinogenic electrophiles. Information on the content of cafestol and kahweol in coffee brews allows a prediction of the likely effects of different coffee brews on serum cholesterol levels and hence their potential role in cardiovascular disease, and also provides a basis for determining the relevance of these biologically active constituents with regard to the reported chemoprotective effects of coffee (Baron et al., 1994). Although methods for the analysis of total coffee diterpenes have been reported, these are of inadequate sensitivity or specificity for application to the measurement of individual diterpenes in coffee brews (Ratnayake et al., 1993). We report the development of a simple and sensitive reverse-phase HPLC method using solid-phase extraction procedures that has been applied to the analysis of cafestol and kahweol in various coffee brews. This methodology complements the gas chromatographic method for these components reported recently by Urgert et al. (1995b) and permits a corroboration of the finding that the method of preparation of the brew is a critical determining factor in determining the daily intake of these diterpenes from coffee consumption.
MATERIALS AND M E T H O D S
Materials and reagents
All solvents used were of analytical or HPLC grade. If not otherwise stated, the water used was 18 MfLcm quality obtained from a Milli-Q water purification system (Millipore, Bedford, MA, USA). A mixture of 17-O-palmitoyl-cafestol (52.5%, w/w) a n d 17-O-palmitoyl-kahweol (47.5%, w/w), hereafter referred to as C + K/P, prepared according to the method of Bertholet (1987), was used as refer-
ence standard. Its purity was checked by saponification and quantification of liberated diterpene alcohols, cafestol (C) and kahweol (K), by HPLC using the method described below. Major by-products in C + K/P were identified as isokahweolene (0.3%, w/w) and cafestolene (0.2%, w/w). Free cafestol and kahweol reference standard mixture [54.5% (w/w) cafestol; 45.5% (w/w) kahweol], hereafter referred to as C + K, was prepared from coffee oil as previously reported (Bertholet, 1987). C + K purity was confirmed by mass spectroscopy as well as by proton and carbon-13 nuclear magnetic resonance spectroscopy. Stock solutions of C + K/P in hexane (2.5mg/ml), and C + K in ethanol (0.5, 2.5 and 5 mg/ml) were prepared. Extrelut extraction cartridges (3 ml) were from Merck Ltd (Darmstadt, Germany). Bond-Elut PRS (propyl sulfonic acid; 100rag) and C18 (100rag) cartridges and cartridge coupling adapters were from Analytichem International (ICT AG, Basle, Switzerland). Purified extracts were stored in 1.1 ml Chromacol gold microvials (Infochroma AG, Zug, Switzerland). Solid-phase extraction cartridge conditioning
C18 cartridges were rinsed with 3 ml 2-propanolethyl acetate (1:1). PRS cartridges were rinsed with, in order, 3 ml 1 y hydrochloric acid solution, 20 ml water and 3 ml 2-propanol-ethyl acetate (1:1). One C18 was coupled to a PRS cartridge using an adapter piece. This assembly is hereafter referred to as 'tandem'. HPLC
Chromatographic analysis was performed using a Merck L-6200 pump, a Gilson 231-401 automated sample processor and injector, and a Hewlett Packard model 1040A/mark II multi-channel ultraviolet detector driven by a HP9000 series 300 workstation (Chemstation). A 250 x 4.6 mm Merck Superspher Lichrocart 100 RP-18 column (3/~m particle size) was used to separate the diterpenes. The mobile phase consisted of water (solvent A) and methanol (solvent B). Separations were carried out at ambient temperature using linear gradient elution conditions (0 rain: 30% A, 70% B; 20 rain. 5% A, 95% B) and a flow rate of 1 ml/min. C and K were detected at 230 nm and 290 nm, respectively. Peaks were quantified by area measurement and identified by retention times and by matching on-line recorded spectra with library ultraviolet spectra. Injections were performed in the partial loop filling mode using a custom program for the Gilson '231-401 injector. In general, a 20 ml portion of purified extract or a 5 pl standard preparation w a s injected into a 50 pl loop, followed by dilutant liquid (20/A) to compensate for the injector port dead volume. The HPLC instrument was calibrated on a weekly basis by injecting each C + K stock solution.
Cafestol and kahweol in coffee brews
Preparation of coffee brews All coffee brews were prepared using a commercial Arabica blend of roasted coffee purchased in Switzerland, except for single-serving size espresso brews which were purchased as prepacked capsules containing different blends, and instant coffee, which was prepared from two different commercial soluble coffee blends. For preparation of 'boiled coffee', coffee beans were coarsely ground (800900 pm), whereas for other coffee brews the coffee beans were ground to the same medium-fine particle size (500-550 pm). Tap water was used to prepare all coffee brews. Brewing methods, brew strengths and cup sizes were standardized according to IARC (1991). Boiled coffee. Was prepared in the traditional manner by boiling 9 g coarsely ground roasted coffee with 150 ml boiling water for 10 min. The mixture was hot-filtered through a metal screen (mesh size 0.5 mm) and the filtrate used for analysis. Turkish coffee. Was prepared by bringing to a gentle boil a mixture of 5 g roast and ground coffee, 10 g saccharose and 60 ml cold water. Boiling was interrupted when a foam developed on the surface of the brew. The brew was left to settle and the supernatant was taken for analysis. Mocha (moka, Neapolitan) coffee. Was prepared using a commercial Italian coffee maker. The water reservoir, filter cup and coffee recipient were separated and the filter cup filled with 20 g roasted coffee powder. Cold water (300 ml) was added to the bottom reservoir, and the filter cup was inserted. The coffee maker was reassembled and then heated on a hotplate until the bottom water reservoir was empty. The resulting coffee brew was mixed prior to analysis. Standard espresso coffee. Was prepared using a mechanical-type Turmix EX10 coffee machine (Turmix Ltd, Jona, Switzerland). 20 g of roast and ground coffee was used to draw a 130 ml volume. Single-serving size espresso. Was prepared using a Turmix CII01 coffee machine (Turmix Ltd, Jona, Switzerland). One commercially available disposable capsule containing 5 g roasted coffee was used to draw a cup of about 50 ml. Five different commercial prepacked coffee blends were examined. Drip-filtered coffee. Was prepared by adding 13 g roast and ground coffee into a Tchibo No. 4 type paper filter (Tchibo GmbH, Hamburg, Germany) placed on an aplgropriate funnel. Boiling water (200 ml) was added and the filtered brewed coffee was collected. Instant coffee. Was prepared by mixing 170 ml boiling water with 2 g instant coffee granules per cup. Two different commercial coffee blends were examined. Extraction procedure. The brewed coffees were allowed to cool to room temperature and adjusted to pH 9-10 using 4 N sodium hydroxide solution. Aliquots of 2.5 ml were transferred into plastic test tubes (Falcon 2059, Becton Dickinson, N J, USA)
549
and spiked with suitable quantities of C + K/P standard solution using a graduated microlitre syringe (Hamilton 1705/1710, Hamilton Ltd, Bonaduz, Switzerland). The resulting mixtures were vortex mixed and applied to an Extrelut-3 columns placed on a Supelco Visiprep ® vacuum manifold. The application of a slight vacuum accelerated the adsorption of the liquid phase onto the column material. The Extrelut columns were then placed on a rack and fine luer lock needles (25G x 5/8 inch) were applied to the tips of the columns to serve as flow restrictors. Each sample tube was rinsed with 2.5 ml hexane-dichloromethane (95:5, v/v) which was then poured into the respective Extrelut column. After standing for approximately 15-30 minutes, the Extrelut columns were eluted under gravity flow by the addition of 12.5ml hexane-dichloromethane (95:5) to the column reservoir. Approximately 11 ml eluate per column was collected into 25 ml pearshaped flasks. The eluates were evaporated to dryness at ambient temperature under reduced pressure. The resulting residues were saponified by the addition of 2 ml methanolic 0.3 N potassium hydroxide solution. Following incubation with agitation for 60 min at room temperature using a model KL2 agitator (Universal Kleinstschtittler, Edmund BiJhler GmbH, Tiibingen, Germany), 800/~1 2-propanol-ethyl acetate (1:1, v/v) and 800/~1 0.5 M aqueous sodium dihydrogen phosphate were added. After thorough mixing, this preparation was transferred into the top reservoir of previously conditioned C18/PRS solid-phase extraction cartridge tandems and slowly eluted through both cartridges by applying gentle air pressure with a plastic syringe attached to the top of the tandem. The eluate was collected in 16 x 76 mm poly-allomer test-tubes (Beckman Instruments, Palo Alto, CA, USA). The flask containing the incubation mixtures was rinsed with 1 ml 2-propanol-ethyl acetate (1:1, v/v), and the washings were passed through the tandem and combined with the previous eluate. The combined eluate was concentrated using a model SVCI00H Speed-Vac vacuum centrifuge with a RT100A cryotrap (Savant Instruments, Farmingdale, NY, USA). The residue was reconstituted in 400/~1 methanol with vortex-mixing and then filtered through a 0.2pro zero dead volume Nalgene filter tip (Nalgene Company, Rochester, NY, USA) into an HPLC microvial. Samples were stored at +4°C prior to HPLC analysis.
Quantification Quantitative analysis was performed using external standard calibrations. The method of standard additions was employed to check C and K extraction efficiencies and the results were corrected accordingly. In general, one preliminary determination was carried out to approximate the C and K levels in the examined product. This was then followed by triplicate determinations using one unspiked sample and two parallel samples spiked
G. Gross et al.
550
with the C + K/P mixture prior to the extraction procedure so as to double and triple the estimated levels of C and K. The extraction efficiencies were calculated for both analyses as the slope of the linear regression line using the parameters added (x) v. measured (y) analyte concentration. If not otherwise stated, the results are expressed as concentrations of free diterpene alcohols. RESULTS
Optimization of analytical methodology Diterpene alcohols were analysed by HPLC following extraction of the corresponding esters from the coffee brews, saponification, and clean-up of the saponified mixture using solid-phase extraction technology. The method used is schematized in Fig. 1. Initially, water/hexane partitioning of the coffee brew was evaluated as a means to pre-concentrate the diterpene esters. However, this procedure was abandoned owing to technical difficulties (i.e. formation of stable emulsions and multiple liquid phases). The use of diatomaceous earth (Extrelut ®) to immobilize the aqueous liquid phase permitted the efficient extraction of diterpene esters with hexane while avoiding the formation of emulsions. The saponification of C and K proceeded smoothly at room temperature and was complete within 1 hr. Longer hydrolysis times diminished the yields, while much shorter times resulted in incomplete saponification (data not shown). The crude hydrolysate was purified using Coffee brew pH 9 Partition on Extrelut-3 Extraction with hexane- DCM, 95:5 (IncreasedC + K palmitateextractibility) Drying/evaporation (Concentration) Saponification (KOH/MeOH) Neutralization (NaH2PO4.2H20) SPE:
C18
l
PRS
Residual linidsremoval Extraction withAcOEt: 2-propanol, 1:1 (IncreasedC + K extractibility) Salts removal Drying/evaporation
Reconstitution(MeOH) HPLC analysis Fig. 1. Outline of the procedure employed for the solidphase extraction of diterpenes from coffee brews: DCM = dichloromethane; SPE = solid-phase extraction; AcOEt = ethyl acetate; MeOH = methanol.
C18 silica to remove residual lipids, followed by a propyl sulfonic acid cation exchanger to reduce the salt content of the extracts. These two processes were conveniently combined in a one-step procedure by the use of coupled cartridges. Following reconstitution, the resulting extract could then be directly injected into the liquid chromatograph. The accuracy of the quantification of C and K levels depends critically on the extraction efficiency of the corresponding esters from crude coffee, the yield of the saponification, and the recovery of free C and K during the final filtration. In order to optimize these parameters, we used the method of standard additions. Employing the protocol described, including the use of triplicate extractions, a set of three data pairs [spiked concentration (x)/determined concentration (y)] was created for each analyte and the extraction efficiency expressed as the slope of the respective linear regression line (r 2 = 0.999 and r 2 = 0.997 for C and K, respectively). The mean extraction efficiencies were 97 + 19% (n = 16) for C and 96_+ 19% (n = 16) for K. The specificity of C and K determinations was provided by the selective detection of both compounds at their respective wavelength of maximal absorption, as well as the match of on-peak recorded ultraviolet spectra with references. The limit of quantification was estimated as approximately 0.05 mg/litre for both C and K, sufficiently sensitive for the analysis of all of the coffee brews investigated in the present study. The chromatograms obtained from analyses of an espresso blend soluble coffee are shown in Fig. 2. In our laboratory the method showed good repeatability as demonstrated by the coefficient of variation of 8.6% (C) and 8.7% (K) obtained following triplicate determinations of the same sample (containing 17.3 mg C/litre and 16.3 mg K/litre, respectively).
Quantification of cafestol and kahweol m various coffee brews C and K were analysed in seven different coffee brews prepared using a variety of methods. The results are summarized in Table 1. Interestingly, Turkish and not boiled coffee contained the highest concentration of C and K. However, when the results were expressed on the basis of cup size, it was apparent that boiled coffee (14.4mg/cup) provides more of these diterpenes than Turkish-style coffee (10.7 mg/cup). The amount of C and K measured in mocha (Neopolitan)-style coffee was also relatively high, about 75% of the concentrations found in boiled coffee and about 32% when expressed as a ratio of cup size (4.6 rag/cup). The C and K concentrations found in standard espresso brewed coffee were almost three times less than observed for boiled coffee, despite a brewing strength almost 2.5-fold greater than that used for boiled coffee. When expressed as the amount per cup, the C + K level
Cafestol and kahweol in coffee brews
551
230 nm !
230 nm
~
Reference spectrum ._= E eq
290 nm I
c)
I~" Kahweol jK~~ference spectrum
250 30O 35O Waye!ength(nm) C ]B A 290 nm
0
2
4
6
8
IO
12
14
16
18
Time (min) Fig. 2. Chromatographic profiles of espresso blend instant coffee recorded at 230 nm (top) and 290 nm (bottom): A = unspiked sample containing 44 pg C/g and 40 #g K/g (values not corrected for incomplete extraction efficiency); B and C = profiles of the same sample spiked with 209/~gC/g and 171 pg K/g, and 418 #g C/g and 342 ttg K/g, respectively, prior to analysis; mAU = milli Absorbance Units.
of espresso (2 mg/cup) was over seven times lower than for boiled coffee. The C + K concentration of espresso coffee brews prepared from single-serving size prepacked roast and ground coffees was also considerably lower than for boiled coffee. Furthermore, the amount of C and K in these espresso brews prepared from capsules (Table 1, Blends A-E, 0.15-0.35mg C + K / c u p ) ranged from about five to 14 times lower than for the espressos prepared using the traditional method. The blend of coffee employed in the capsules appeared to influence the diterpene content of the brew to some exterLt. Instant coffees contained very low concentrations of C and K (0.2-0.5 rag/cup), providing between about 25 and 70 times less C + K per cup than boiled coffee. Drip-filtered cof-
fee contained the least C + K, equivalent to a total of 0.04 mg per cup. DISCUSSION A simple and sensitive reverse-phase HPLC method using solid-phase extraction has been developed for the analysis of C and K in coffee brews. This method benefits from a straightforward extraction procedure that overcomes the problems of emulsion formation and the presence of multiple liquid phases encountered when using direct solvent extractions of coffee. The method of standard additions was selected in place of an internal standard for quantification, since an appropriate internal standard was not available. An internal standard
552
G. Gross et al. Table 1. Analysis of diterpene alcohols in coffee brews Diterpene alcohols Brew strength (mg/ml)
Cafestol
Kahweol
Coffee brew
IARC
Used
mg/litre+ SD
mg/cup*
mg/g
mg/litre + SD
mg/cup*
mg/g
Turkish coffee Boiled coffee Mocha coffee
83 50-75
83 60 67
88.7 ± 4.0 48.3 + 3.8 37.5 + 1.3
5.3 7.2 2.3
1.07 0.81 0.56
89.9 + 4.1 48 + 2.5 38.5 + 0.9
5.4 7.2 2.3
1.08 0.8 0.57
Espresso coffee Exp. 1 100-300 Exp. 2 100-300
154 154
17.3 ± 1.5 16.7+0.3
1.0 1.0
0.11 0.11
16.3 + 1.4 17.1 +0.5
1.0 1.0
0.11 0.11
Single-servingsize espresso Blend A Blend B Blend C Blend D Blend E Filtered coffee 25-75
96 106 114 125 106 65
3.4 + 0.3 2.1 _+0.3 2.2 + 0.2 1.6 + 0.7 1.2 + 0.3 0.12 + 0.02
0.17 0.10 0.11 0.08 0.06 0.02
0.04 0.02 0.02 0.01 0.01 0.002
3.5 + 0.3 2.9 + 0.2 2.6 ± 0.3 1.8 + 0.8 1.7 + 0.3 0.14 + 0.03
0.18 0.15 0.13 0.09 0.09 0.02
0.04 0.03 0.02 0.01 0.02 0.002
Instant coffee Standard blend Espresso blend
12 12
1.9+0.05 0.7 + 0.07
0.3 0.1
0.16 0.06
1.9+0.01 0.7 + 0.10
0.3 0.1
0.16 0.06
10 16 10-16
*Coffee cup sizes are as follows: 150 ml for boiled, filtered and instant coffees; 60 ml for Turkish, mocha and espresso coffees; and 50 ml for the single serving size espresso coffees.
selected purely on the basis of its chromatographic behaviour (Urgert et al., 1995b) suffers from an inability to take into account the variability that may occur in saponification or extraction efficiencies. Turkish-style coffee, and not boiled coffee, contained the highest concentration of C and K. A similar result was previously reported by Ratnayake et al. (1993) and this may be a consequence of the higher amount of fines present in Turkish-style coffee, compared with 'boiled coffee'. Thus the method by which 'boiled coffee' is decanted may have a large influence on its diterpene content. Furthermore, increasing the time of boiling did not increase the C and K contents of the 'boiled coffee' (data not shown). Similar findings have been reported by Urgert et al. (1995b), who observed extremely large variations in field samples of these coffee brews and attributed this variation to differences in brewing strength and the amount of bean particles decanted with the brew. They also reported that field samples of Scandinavian boiled and Turkish/Greek coffees contained between 1 and 10mg cafestol per cup (equivalent to about 220 mg/cup) of C + K. The values that we obtained for a standardized boiled coffee (14.4 mg/cup) and Turkish coffee (10.7 mg/cup) prepared under laboratory conditions from the same roast and ground coffee beans according to IARC (1991) fall within this range. The C + K content for mocha (Neopolitan)-style coffee (4.6mg/cup) is a little higher than that previously reported by Urgert et al. (1995b) for a laboratory-prepared brew (2.5mg/ cup); nevertheless, in both studies this brewing method results in a coffee providing intermediate amounts of C and K. It is interesting to note that it was recently reported that consumption of three cups per day of Italian brewed coffee (moka,
mocha) for 5 wk in a double-blind crossover clinical trial had no significant effect on serum cholesterol levels (Sanguigni et al., 1995). The C and K concentrations found in standard espresso brewed coffee (2 mg/cup) is consistent with the field samples (about 0-6 mg/cup) reported elsewhere (Urgert et al., 1995b). However, this finding is in contrast to that reported by Ratnayake et al. (1993) which indicated that espresso coffee contained twice the level of total diterpenes compared to 'boiled' and Turkish coffees. The reason for this discrepancy is unclear, but it is likely that diterpene levels may be influenced by steam pressure, contact time of steam with grounds, and the efficacy of the metal screen strainer in retaining fine coffee particles, thereby preventing their contribution to the diterpene content of the brewed coffee. The C + K contents of coffee brews prepared from single-serving size prepacked roast and ground coffees (0.150.35 mg/cup) were considerably lower than those of traditional espresso coffee. The reason for this difference is not obviously apparent but it seems likely that the prepacked capsules act as a filter, thus allowing fewer fine particles of coffee grounds to pass into the brew. This explanation is consistent with a recent report indicating that spent fine coffee grounds raise serum cholesterol levels when administered to human volunteers, indicating that the fines are an important source of diterpenes (Urgert et al., 1995a). Both drip-filtered and instant coffees contained very low concentrations of C + K, a finding which is in good agreement with that reported by Urgert et al. (1995b) and also with the low levels of total lipids and total diterpenes found by Ratnayake et al. (1993) for these types of coffee. The low levels of diterpene esters in drip-filtered coffee prepared
Cafestol and kahweol in coffee brews
1.5 --
= 1.0 O
g
~ 0.5 -
e-
0.0
'
Turkish
s' r' so,I I
iI
i
i
Ii
513
J
~
I
I
i
i
I
100
I
[
150
I
k
i
[
200
C + K (mg/day) Fig. 3. Predicted increase in serum total cholesterol levels following the consunaption of five cups/day of different coffee brews for 4 wk. The dose-response curve is based upon data points (O) taken from studies in human volunteers previously repo:rted by Weusten-Van der Wouw et al. (1994). The estimated C + K intakes are based upon the consumption of five cups of coffee brews of the following sizes: boiled, filtered, instant--150ml; Turkish, mocha, espresso--60 ml; prepacked espresso capsules--50 ml. The predicted increase in serum cholesterol is less than 0.1 mmol/litre for coffee brews (single-serving size espresso, instant, filtered) not shown in the figure. using a paper filter can be explained by both the removal of fines and the strong lipid-binding effect of filter paper (Ahola et al., 1991). For instant coffee, a large proportion of the diterpenes is lost during industrial processing. The potential cholesterol-raising effects of the different coffee brews can be estimated on the basis of the results of clinical studies in which 'boiled coffee', coffee oils or purified diterpenes were administered to volunteers, as reported by Weusten-Van der W o u w et al. (1994). A review of these studies demonstrates a dose-dependent effect of the daily ingestion of preparations containing C and K for 4 wk on the increase in serum cholesterol levels (Fig. 3). If it is assumed that five cups of coffee are consumed daily, oll the basis of the results reported in the present investigation, the expected increase in cholesterol levels would range from less than 0.05mmol/litre for filtered coffee to about 0.85 mmol/litre for boiled coffee. It must be emphasized that this extrapolation is based on assumptions and uncertainities concerning the diterpene contents of some of the test materials used in the clinical studies. Nevertheless, it is interesting to note that Pietinen et al. (1990) reported that Finnish men and women who regularly drank seven to nine cups of boiled coffee per day had serum cholesterol levels 0.6 mmol/litre higher than those of consumers of filtered coffee. Furthermore, a cross-sectional study in Norwegians found that subjects habitually consuming five or more cups of boiled coffee per day had mean serum cholesterol levels 0.3 mmol/ litre higher than those of matched filter coffee drinkers. (Weusten-Van der W o u w et al., 1994). In conclusion, the use of solid-phase extraction techniques eliminates the problems of multiple phase separation that are often encountered when
553
liquid-liquid extraction techniques are used with coffee brews. The results for C and K in different coffee brews are consistent with those previously reported and with the results predicted from an evaluation of clinical and epidemiological studies comparing the hypercholesterolaemic effects of different brews. Finally, our fndings indicate that the levels of C and K found in traditional espresso coffees are lower than previously suspected and predict that instant and filtered coffees would not have any effect on serum cholesterol levels. This new methodology should help to assess more correctly the exposure to these compounds through the consumption of coffee and to predict more accurately the physiological effects of different coffee brews. REFERENCES
Aeschbacher H. U. and Jaccaud E. (1990) Inhibition by coffee of nitrosourea mediated DNA damage in mice. Food and Chemical Toxicology 28, 633-637. Ahola I., Jauhiainen M. and Aro A. (1991) The hypercholesterolaemic factor of boiled coffee is retained by a paper filter. Journal of Internal Medicine 230, 293-297. Aro A., Tuomilehto J., Kostiainen E., Uusitalo U. and Pietinen P. (1987) Boiled coffee increases serum low density lipoprotein concentration. Metabolism 36, 10271030. Bak A. A. A. and Grobbee D. E. (1989) The effect on serum cholesterol levels of coffee brewed by filtering or boiling. New England Journal of Medicine 321, 14321437. Baron J. A., de Verdier M. G. and Ekbom A. (1994) Coffee, tea, tobacco, and cancer of the large bowel. Cancer Epidemiology, Biomarkers and Prevention 3, 565570. Bertholet R. (1987) Preparation of cafestol. United States Patent 4,692,534. Burr M. L., Gallacher J. E., Butland B. K., Bolton C. H. and Downs L. G. (1989) Coffee, blood pressure and plasma lipids: a randomized controlled trial. European Journal of Clinical Nutrition 43, 477-483. Di Simplicio P., Jensson H. and Mannervik B. (1989) Effects of inducers of drug metabolism on basic hepatic forms of mouse glutathione transferase. Biochemical Journal 263, 679-685. Huggett A. C., Marchesini M. and Perrin I. (1993) Effects of coffee oil on plasma cholesterol in the rat. In 15th International Scient!fic Colloquium on Coffee. pp. 497503. Association Scientifique International du Caf& Paris. Huggett A. C. and Schilter B. (1995) The chemoprotective effects of cafestol and kahweol: effects on xenobiotic metabolising enzymes. Journal of Cellular Biochemistry 19A, 190. IARC (1991) IARC Monographs on the Evaluation of the Carcinogenic Risks to Humans. Vol. 51. Coffee, Tea, Mat~, Methylxanthines and Methylglyoxal. pp. 41-206.
International Agency for Research on Cancer, Lyon. Kark J. D., Friedlander Y., Kaufmann N. A. and Stein Y. (1985) Coffee, tea and plasma cholesterol: the Jerusalem Lipid Research Clinic prevalence study. British Medical Journal 291, 699-704. Lam L. T. K., Sparnins V. L. and Wattenberg L. W. (1982) Isolation and identification of kahweol palmitate and cafestol palmitate as active constituents of green coffee beans that enhance glutathione S-transferase activity in the mouse. Cancer Research 42, 1193-1198. Lercker G., Frega N., Bocci F. and Rodriguez-Estrada M. T. (1995) High resolution gas chromatographic determination of diterpenic alcohols and sterols in coffee lipids. Chromatographia 41, 29-33.
554
G. Gross et al.
Mensink R. P., Lebbink W. J., Lobbezoo 1. E., WeustenVan der Wouw M. P. M. E., Zock P. L. and Katan M. B. (1995) Diterpene composition of oils from Arabica and Robusta coffee beans and their effects on serum lipids in man. Journal of Internal Medicine 237, 543-550. Mensink R. P., Zock P. L., Katan M. B. and Beynen A. C. (1992) Boiled coffee does not increase serum cholesterol in gerbils and hamsters. Zeitschrift far Ernahrungswissenschaft 31, 82-85. Miller E. G., Formby W. A., Rivera-Hidalgo F. and Wright J. M. (1988) Inhibition of hamster buccal pouch carcinogenesis by green coffee beans. Oral Surgery 65, 745-749. Miller E. G., Gonzales-Sanders A. P., Couvillon A. M., Binnie W. H., Sunahara G. I. and Bertholet R. (1993) Inhibition of oral carcinogenesis by roasted beans and roasted coffee bean fractions. In 15th International Scientific Colloquium on Coffee. pp. 420-425. Association Scientifique International du Caf6, Paris. Miller E. G., McWhorter K., Rivera-Hidalgo F., Wright J. M., Hirsbrunner P. and Sunahara G. I. (1991) Kahweol and cafestol: inhibitors of hamster buccal pouch carcinogenesis. Nutrition and Cancer 15, 41-46. Pietinen P., Aro A., Tuomitehto J., Uusitalo U. and Korhonen H. (1990) Consumption of boiled coffee is correlated with serum cholesterol in Finland. International Journal of Epidemiology 19, 586-590. Ratnayake W. M. N., Hollywood R., O'Grady E. and Stavric B. (1993) Lipid content and composition of coffee brews prepared by different methods. Food and Chemical Toxicology 31, 263-269. Ratnayake W. M. N., Pelletier G., Hollywood R., Malcolm S. and Stavric B. (1995) Investigation of the coffee lipids on serum cholesterol in hamsters. Food and Chemical Toxicology 33, 195-201. Sanders T. A. B. and Sandaradura S. (1992) The cholesterol-raising effect of coffee in the Syrian hamster. British Journal of Nutrition 68, 431434. Sanguigni V., Gallu M. P., Rul~ni M. P. and Strano A. (1995) Effects of coffee on serum cholesterol and lipoproteins: the Italian brewing method. European Journal of Epidemiology 11, 75-78.
Stensvold I., Tverdal A. and Foss O. P. (1989) The effect of coffee on blood lipids and blood pressure. Results from a Norwegian cross-sectional study, men and women, 40-42 years. Journal of Clinical Epidemiology 42, 877-884. Thelle D. S.~ Arnesen E. and Forde O. H. (1983) The Tromso heart study: does coffee raise serum cholesterol? New England Journal of Medicine 308, 1454-1457. Thelle D. S., Heyden S. and Fodor J. G. (1987) Coffee and cholesterol in epidemiological and experimental studies. Atherosclerosis 67, 97 103. Thelle D.S. and Van der Stegen G. (1990) In Coffee and
Coronary Heart Disease, Special Emphasis on the CoffeeBlood Lipids Relationship. Edited by D.S. Thelle and G. Van der Stegen. pp. 9-15. Gotab, Stockholm. Urgert R., Schulz A. G. M. and Katan M. B. (1995a) Effects of cafestol and kahweol from coffee grounds on serum lipids and serum liver enzymes in humans. American Journal of Clinical Nutrition 61, 149-154. Urgert R., van der Weg G., Kosmeijer-Schuil T. G., van de Bovenkamp P., Hovenier R. and Katan M. B. (1995b) Levels of the cholesterol-elevating diterpenes cafestol and kahweol in various coffee brews. Journal of Agricultural and Food Chemistry 43, 2167-2172. Van Dusseldorp M., Katan M. B., Van Vliet T., Demacker P. N. M. and Stalenhoef A. F. H. (1991) Cholesterol-raising factor from boiled coffee does not pass a paper filter. Arteriosclerosis and Thrombosis 11, 586 593. Wattenberg L. W. (1983) Inhibition of neoplasia by minor dietary constituents. Cancer Research 43, 2448 2453. Wattenberg L. W., Hanley A. B., Barany G., Sparnins V. L., Lam L. K. T. and Fenwick G. R. (1985) Inhibition of carcinogenesis by some minor dietary constituents. International Symposium. Princess Takamatsu Cancer Research Fund 16, 193-203. Weusten-Van der Wouw M. P. M. E., Katan M. B., Viani R., Huggett A. C., Liardon R., Lund-Larsen P. G., Thelle D. S., Ahola I., Aro A., Meyboom S. and Beynen A. C. (1994) Identity of the cholesterol-raising factor from boiled coffee and its effect on liver function enzymes. Journal of Lipid Research 35, 721-733.