J Sci Food Agric 1998, 76, 617È627 Analysis of Sterols: a Novel Approach for Detecting Juices of Pineapple, Passionfruit, Orange and Grapefruit in Compounded Beverages* Lay-Keow Ng” and Michel Hupe Research & Development Division, Laboratory & ScientiÐc Services Directorate, Revenue Canada, 79 Bentley Avenue, Ottawa, Ontario, Canada, K1A OL5 (Received 10 June 1996 ; revised version received 31 July 1997 ; accepted 23 September 1997) Abstract : Direct GC/MS analysis of the hexane extracts of fruit juices provides an efficient means for demonstrating that very di†erent sterol patterns exist in the juices of pineapple, passionfruit and the two citrus fruits, orange and grapefruit. Ergostanol and stigmastanol were found to be the sterol markers for pineapple juice, while passionfruit juice was characterised by the presence of an unidentiÐed but unique sterol referred to as compound C. Juices of orange and grapefruit yielded very similar sterol proÐles. They were readily distinguished from pineapple and passionfruit juices by a higher stigmasterol/campesterol ratio. Valencene/nootkatone response ratio in the hexane extracts was employed to aid in the di†erentiation of the two citrus juices. Matrix e†ects on the determination of sterol and sesquiterpenoid distributions were found to be insigniÐcant. Although natural variation and absolute uniqueness of the sterol proÐle for each of the four fruit juices were not established due to the relatively small number of fruit samples examined, the results of several compounded beverages clearly point to the potential usefulness of sterol proÐles for detecting juices of orange, grapefruit, pineapple and passionfruit in mixed drinks. ( 1998 SCI. J Sci Food Agric 76, 617È627 (1998) Key words : juice ; sterols ; sesquiterpenoids ; detection ; compounded beverages ; gas chromatography/mass spectrometry INTRODUCTION materials such as water, sugars, organic acids, Ñavourings, colorants, preservatives, other juices, etc, are added so that the composition of the original juice is altered. Therefore, for the purpose of determining the presence of a juice in beverages, the markers for the juice should be so chosen such that they remain unique to the target juice within the matrix of the compounded beverages. Sterols, in free or conjugated forms, have been identiÐed in parts of various citrus fruits, including grapefruit peel (Williams et al 1967), orange and tangor juice sacs (Nagy and Nordby 1971), and peel of Rangpur lime (Yokoyama and White 1968). Quantitative composition of free sterols has been used to characterise di†erent varieties of orange and tangor. Rangpur lime was found to be the only citrus fruit containing ergosterol. Free sterols speciÐc to strawberry fruit (Couture et al 1989) and leaves of pineapple (Pakrachi et al 1975) have also It is often necessary to conÐrm the declared presence of a particular juice in compounded beverages for the purpose of imposing Customs or Excise duty. One plausible approach would be to detect characteristic components, or markers, of the juice in the beverages. Fruit juices are complex mixtures of sugars, organic acids, volatile Ñavours, fatty acids, sterols, amino acids, Ñavonoids, pigments, etc in water. Each juice is characterised by a composition which is a unique combination of these ingredients. In a formulated juice drink, * Most of the Ðndings from this study were presented at the 21st Annual Conference of the Federation of Analytical Chemistry and Spectroscopy Societies, 2È7 October 1994, St Louis, MO, USA. ” To whom correspondence should be addressed. 617 ( 1998 SCI. J Sci Food Agric 0022È5142/98/$17.50. Printed in Great Britain L -K Ng, M Hupe 618 been reported. These Ðndings point to the potential of using free sterols as fruit indices. Unlike volatile Ñavours, sugars and acids, sterols are not added ingredients used in beverage formulation since they are not Ñavouring agents and are rather expensive chemicals. Sterols are, therefore, not likely to be masked by added substances in the juice drinks. They are also not easily removed from the juice matrices during concentration, which is important given that most commercial juice drinks are prepared from juice concentrates. Based on these considerations, sterols could constitute a useful group of chemical markers for the identiÐcation of juices in compounded beverages. Numerous reports on the detection of fruit juices in beverages have resulted from e†orts to combat juice adulteration where a juice has been diluted by a cheaper juice. Various chemical indices for fruit have been used. For example, adulteration of orange juice by grapefruit juice was conÐrmed by the detection of the Ñavonoid naringin (Greiner and Wallrauch 1984). The high concentration of proline in apple juice was taken as an indication of the addition of pear juice (Wallrauch and Faethe 1988). Sterols have not been reported to be employed as index components of fruit juices. Most of the fruit sterols investigated were derived from juice sacs or other parts of the plants. Few studies on the composition of free sterols in fruit juices have been undertaken except for orange juice (Stack et al 1986). This method consisted of gas chromatography (GC) of a fraction eluted from a C-18 disposable minicolumn. The sample preparation procedure involved tedious conditioning of the column, solvent elution of sterols, evaporation of the solvent and derivatisation. GC analysis of sterols from plant materials is difficult, due to structural similarity of the compounds which leads to overlapping of peaks. Gas chromatography/ mass spectrometry (GC/MS) is a more powerful technique for the analysis of co-eluting free sterols. Components can be extracted from the total ion chromatogram (TIC) using speciÐc ions for qualitative and quantitative analysis. They can be identiÐed by retention times (RT), as well as by their distinct mass spectral patterns. In this study, the potential usefulness of sterols as diagnostic components for identifying juices of pineapple, passionfruit, orange and grapefruit has been investigated. The analysis was accomplished by a simple procedure in which sterols and other components were extracted into hexane by a one-step liquidÈliquid extraction, followed by GC/MS analysis. MATERIALS AND METHODS Samples Three oranges from California, Florida and Uruguay, two pineapples from Costa Rica, one pink-Ñeshed and two white-Ñeshed grapefruits from Florida, one yellowskinned and two purple-skinned passionfruits, and different brands of commercial beverages composed of single or mixed juices were purchased from local stores or supplied by the Excise Department of Thailand. Solvents and standards b-Caryophyllene and valencene were obtained from Fluka Chem Corp (Ronkonkoma, NY, USA) and Janssen Chimica (Beerse, Belgium), respectively. Cholesterol [99%, stigmasterol 96%, stigmastanol 98% and b-sitosterol 98É3%, were supplied by Sigma (St Louis, IL, USA). Campesterol 99% was purchased from Applied Science Laboratories Inc (State College, PA, USA). Specially denatured alcohol, formula 3A and hexane were obtained from JT Baker (Toronto, Ontario, Canada). Extraction procedure Eight millilitres of denatured alcohol formula 3A, 10 ml of a single-strength juice sample and 1É5 ml of hexane were pipetted into a centrifuge tube. The tube was screw-capped and vortexed for 2 min. After centrifuging the mixture at 8000 rpm for 5 min, 1 ll of the hexane layer (top) was injected in splitless mode into the GC/MS system. For compounded beverages, the same procedure was used except that the hexane extracts were concentrated to 0É25 ml at room temperature under a slow stream of nitrogen before 1 ll was injected into the gas chromatograph. Gas chromatography/mass spectrometry (GC/MS) The extracts were analysed using a HewlettÈPackard gas chromatograph 5890 series II equipped with a 5970 mass selective detector, and a 5990A MS Chemstation (HP-UX) (HewlettÈPackard, North Hollywood, CA, USA). The column used was a 12É5 m DB5-MS (5% diphenyl and 95% dimethyl polysiloxane) fused silica (J & W ScientiÐc, Falsom, USA), 0É33 lm Ðlm thickness, 0É20 mm id. The carrier gas was helium with a Ñow rate of 1É0 ml min~1 at 50¡C. The split/splitless injection port and the interface were at 250¡C and 310¡C, respectively. The oven temperature was programmed as follows : 50¡C for 1 min, then ramped at 10¡C min~1 to 300¡C and held for 40 min. All mass spectra were acquired in the electron impact (EI) mode at 70 eV. The mass spectrometer was scanned in the range of 39È 500 Da at a rate of 1É24 scans s~1, starting 1 min after injection of the sample. The spectrometer was tuned using perÑuorotributylamine (PFTBA) as the calibrant and the autotune procedure incorporated in the Chemstation software package. Analysis of sterols : detecting fruit juices in compounded beverages Analysis of overlapping peaks The purity of TIC peaks was determined by examining the mass spectrum at di†erent time slices. Co-elution of components was indicated by appreciable changes of the mass spectrum as the TIC peak was scanned. A mass spectrum of each component was taken at a time slice where overlapping was absent, or by subtracting the residual spectrum contributed by the co-eluting component. Computerised library matching, using the NBS library, was carried out on the mass spectrum to identify the component. The ions identiÐed as being characteristic of a component were used to generate extracted ion chromatograms (EIC). The ion peaks in these EIC, within the time window where the component eluted, should give the same maxima since they all belong to the same component. The retention time (RT) of the component was determined from the maxima of the extracted ion peaks. Principal component analysis Data based on the presence or absence of compound C, and area ratios of the extracted ion peaks of selected components other than compound C/isofucosterol (see Table 5 below) were subjected to principal component analysis. Area ratios for which the denominator component, or both the numerator and denominator components, were not detected were assigned a value of [ 1, and zeros were used to indicate that only the numerator component was not detected. Compound C was assigned a value of 1 when detected and zero when 619 not detected. The data were Ðrst autoscaled to zero mean and unit variance before carrying out principal component analysis. The commercial software package, Ein*Sight (InfoMetrix, Seattle, WA, USA) was used for calculating principal components. RESULTS AND DISCUSSION All juice samples were subjected to a simple liquidÈ liquid extraction procedure to separate sterols from the matrices of the beverages. Hexane was used as the extraction solvent. Generally, liquidÈliquid extraction of fruit juices by an organic solvent is inefficient due to the formation of an emulsion. In our procedure, ethanol was added to the juice to precipitate pectins and other emulsifying agents. The emulsion was destabilised, and complete separation of the organic and aqueous phases was achieved. The extracts were analysed directly by capillary GC/MS without further treatment. In addition to sterols, fatty acids were also extracted. Volatile Ñavours were only detected in the two citrus juices. As shown in Fig 1, which is a representative TIC of a hexane extract, volatile Ñavours eluted in the Ðrst 14 min, followed by fatty acids, with the semivolatile sterols appearing in the range of 26È29 min. Several unidentiÐed high boiling components were detected after the sterols. Free sterols were present in all of the samples of single juices of orange, grapefruit, pineapple and passionfruit. The common plant sterols, b-sitosterol, campesterol and stigmasterol, were among the major Fig 1. Total ion chromatogram of the hexane extract of orange juice. L -K Ng, M Hupe 620 components. Some of the compounds detected in the sterol fraction were not identiÐed, they may be sterols, sterol derivatives, triterpene alcohols, etc. For simplicity, they are referred to as sterols in the following discussion. The sterols, along with several terpenic compounds, were characterised by RT and speciÐc ions as shown in Table 1. Cholesterol, campesterol, stigmasterol, b-sitosterol and stigmastanol were identiÐed by computerised library matching and by comparison with authentic chemicals. Components which partially overlapped with other peaks, such as 24-ethylidene cholesterol (also known as isofucosterol), stigmastanol, ergostanol, compounds A, B and C, etc, were analysed as described in the Experimental section. Single juices For each juice, samples of freshly squeezed juice were analysed along with several commercial samples to determine the nature and the relative abundances of the sterol components, except for passionfruit juice for which samples of the commercial counterpart were not available. The same components were found in the fresh juices and the commercial counterparts, and their sterol patterns were comparable as discussed later. Three fresh and Ðve commercial juices of orange were analysed. In all of these samples, the major sterols found were, in order of decreasing prevalence, bsitosterol [ campesterol [ stigmasterol, isofucosterol TABLE 1 Chromatographic and mass spectral characteristics of diagnostic components RT a (min) Component (tentative identity) MW Base peak m/z Characteristic ions m/z (relative intensity, %) 11É27 Valencene 204 161b 14É61 Nootkatone 218 147b 26É43 Campesterol 400 400b 26É52 Compound C 424 165 26É54 Ergostanol 402 233b 26É65 Stigmasterol 412 412b 27É24 b-Sitosterol 414 414b 27É31 Isofucosterol 412 314b 27É32 Stigmastanol 416 233b 28É21 Compound B (stigmast-4-en-3-one) 412 124b 28É28 Compound A (citrostadienol) 426 285b 91, 93 (65) 105, 107 (60) 133, 204 (50) 147, 189 (37) 79, 121 (84) 91, 133 (65) 105, 175, 203 (56) 218 (21) 315 (60) 382 (50) 367, 289 (50) 145, 213, 255 (30) 424b (99) 203, 205 (20) 215 (80) 234, 402 (70) 397 (30) 255, 271, 300 (50) 351 (33) 329, 396 (50) 303, 381 (40) 281 (34) 229, 299 (26) 416 (70) 215 (65) 234, 401 (55) 412 (80) 229 (60) 288, 289 (50) 370 (30) 328 (40) 411, 313 (10) 426 (5) a Retention times may change over time, they are listed only to show the relative order of elution. b Ions used to extract the corresponding components from the total ion chromatograms. Analysis of sterols : detecting fruit juices in compounded beverages and compound A [ cholesterol and compound B. All of the identiÐed components have been reported to be present in the juice sacs of orange fruit (Nagy and Nordby 1971). These components, with the exception of isofucosterol, have also been quantiÐed in orange juice (Stack et al 1986). A typical TIC of sterols in orange juice is illustrated in the top trace of Fig 2. Isofucosterol appears as a shoulder on the predominant b-sitosterol peak, and compound A overlaps with compound B. Using ions m/z 414, 314, 124 and 285, b-sitosterol, isofucosterol, compounds B and A, respectively, were extracted from the TIC. The EIC are presented in Figs 2(B)È(E), which show that the RT of isofucosterol was signiÐcantly di†erent from that of b-sitosterol, and the co-eluting components, compounds A and B, yielded slightly, yet noticeably, di†erent RT. The mass spectra of compounds A and B are shown in Figs 3 and 4, respectively. These chemicals were ten- Fig 2. Selective detection of coeluting sterols of orange juice. 1, Cholesterol ; 2, campesterol ; 3, stigmasterol ; 4, b-sitosterol ; 5, isofucosterol ; 6, compound B ; and 7, compound A. The individual plots are (A) the total ion chromatogram and (BÈE) the extracted ion chromatograms generated from the theoretical mass of the respective quantitation ion with a limited mass window of 1 Da. 621 tatively identiÐed by library search as (3b, 4a, 5a, 24Z)4-methylstigmasta-7,24(28)-dien-3-ol, commonly known as citrostadienol, and stigmast-4-en-3-one, respectively. The probability based matches (PBM) (McLa†erty et al 1974) with the library mass spectra were 93 and 91%, respectively. Their identity could not, however, be ascertained because of the unavailability of authentic specimens for direct comparison. These compounds could be among the unidentiÐed sterols in orange pulp reported in a previous study (Tushishvili et al 1982). The relative abundances of several sterols were also used to characterise orange juice. The results obtained for the commercial juice samples fall within, or close to, the range for the fresh juices (Table 2). Three fresh juices, prepared from one pink-Ñeshed and two white-Ñeshed grapefruits, and two commercial grapefruit juices were analysed. They all yielded a sterol proÐle similar to that reported for grapefruit peel (Williams et al 1967). The proÐle is also very similar to that of orange juice. The sterol proÐles of Ðve commercial and two fresh juice samples of pineapple can be readily distinguished from those of the citrus juices. They were all characterised by the presence of signiÐcant amounts of two saturated sterols, namely, ergostanol and stigmastanol. Stigmastanol and ergostanol have been reported to be present in the leaves of the pineapple plant (Pakrachi et al 1975). Their mass spectra displayed the same base peak at m/z 233. Integration of the ion peak at m/z 233 showed that they were present in almost equal amounts. All sterols found in orange juice were also present in pineapple juice. However, the stigmasterol/campesterol, isofucosterol/campesterol, cholesterol/campesterol, compound A/campesterol and compound A/compound B ratios in pineapple juice are smaller than those found in orange and grapefruit juices (Table 2). There are many known species of passionfruit. The best known are the purple-skinned PassiÑora edulis Sims and its closely related yellow-skinned mutant PassiÑora edulis f Ñavicarpa Degener. The sterols detected are b-sitosterol, compound A, compound C, stigmasterol, isofucosterol and cholesterol. The unknown component, compound C, is likely to be a methyl sterol. Its unique mass spectrum is presented in Fig. 5. The relative abundance of compound C to isofucosterol was much higher in the yellow-skinned variety. The ion ratios of stigmasterol/campesterol, isofucosterol/campesterol and cholesterol/campesterol were much smaller than those of the citrus juices, but were of comparable magnitude to those found for pineapple juice (Table 2). The results for the samples of single juices show that there are sterol components speciÐc to juices of passionfruit and pineapple. They are present in easily detectable quantities and would be good candidates for serving as indicators of individual juices in blended beverages. Thus, compound C was chosen as a marker for L -K Ng, M Hupe 622 Fig 3. Mass spectrum of compound A. passionfruit juice, and the saturated sterols, ergostanol and stigmastanol, were employed as index compounds for pineapple juice. Individual juices were further characterised by quantitative measurements based on response ratios of selected components shown in Table 2. Although speciÐc sterol markers could not be identiÐed for the two citrus juices, they were characterised by relatively high stigmasterol/campesterol and isofucosterol/campesterol ratios. It is noted that di†erentiation of orange and grapefruit juices cannot be accomplished by the use of sterol patterns alone. Volatile Ñavours, detected only in the two citrus juices, may be used to serve this purpose. Volatile Ñavours are not considered as good index compounds of juices in cases where the beverages are prepared from juice concentrates. In the concentration process, the majority of the volatiles are removed, and the recovered essence may not be added back to the Fig 4. Mass spectrum of compound B. Analysis of sterols : detecting fruit juices in compounded beverages 623 TABLE 2 Composition of diagnostic components in single juicesa Orange Stigmasterol/campesterol Isofucosterol/campesterol Cholesterol/campesterol Compound A/campesterol Compound C/isofucosterol Ergostanol/stigmastanol Valencene/nootkatone Compound A/compound B Grapefruit Pineapple Fresh passionfruit Fresh (3)b Comm (5) Fresh (3) Comm (2) Fresh (2) Comm (5) Purple (2) Y ellow (1) 0É35È0É41 1É02È2É02 0É09È0É19 0É45È0É86 Èc È 29É1È45É6 2É2È99É1 0É36È0É43 0É32È1É03 0É08È0É13 0É19È0É41 È È 19É6È44É1 2É0È5É2 0É22È0É30 1É16É2É01 0É06È0É09 0É04È0É43 È È 0É06È0É16 2É8È50É7 0É26È0É35 0É84È0É93 0É07È0É09 0É04È0É46 È È 0É24È0É29 9É9È20É7 0É06È0É07 0É11È0É14 0É004È0É011 0É05È0É06 È 0É75È0É96 È 0É05È0É06 0É05È0É12 0É06È0É21 0É004È0É032 0É03È0É08 È 0É73È1É2 È 0É03È0É08 0É05È0É06 0É04È0É10 0É001È0É004 0É11È0É18 0É2È1É3 È È 3É4È8É0 0É04 0É16 0É04 0É51 50É3 È È 9É4 a Ratios of components were determined from the relative areas of the component peaks extracted from the total ion chromatograms using speciÐc ions shown in Table 1. b Numbers in parentheses are number of samples. c È indicates a ratio in which the numerator, the denominator, or both the numerator and the denominator components were not detected (area \ 3000). concentrates. However, the semivolatile Ñavours, such as valencene and nootkatone, are less likely to be lost during the concentration process due to their relatively high boiling points. Valencene and nootkatone are characteristic components of juices of orange and grapefruit (Shaw 1991), respectively. They were selected to serve as secondary markers for the respective juices. However, nootkatone is also found in orange juice in small amount ; similarly valencene is present as a trace component in grapefruit juice. To increase the speciÐcity of these markers, the response ratio of valencene/ nootkatone was used to di†erentiate the citrus juices. As shown in Table 2, the valencene/nootkatone ratio was at least ten times higher in orange juice than in grapefruit juice. Matrix e†ects on sterol and sesquiterpenoid distributions ProÐle analysis based on response ratios of components could be useful for the detection of individual juices in compounded beverages if the ratios are consistent in the single juices and in the compounded beverages. Matrix e†ects on the ratios of selected components of each fruit juice were studied in the following matrices : Fig 5. Mass spectrum of compound C. L -K Ng, M Hupe 624 (1) (2) (3) a 25% juice medium having the same ¡Brix as the single juice, prepared by diluting the fresh juice with water and then making up to the ¡Brix by addition of sucrose ; a diluted juice of 5 ¡Brix, prepared by diluting the fresh juice with water until the desired ¡Brix was obtained ; and a compounded juice of 20 ¡Brix, prepared by adding sucrose to the fresh juice until the desired ¡Brix was obtained. compound C/isofucosterol ratios in all mixtures containing passionfruit juice were smaller than that found for the single juice. This is because isofucosterol was found in all four juices and compound C was detected only in passionfruit juice. In general, it is expected that in a mixed juice, a response ratio of two components, which are both present in some or all of the individual juices in the mixture, should yield a value which lies between the highest and the lowest values of the same component ratio of those individual juices. For example, in the mixtures containing both grapefruit and orange juices such as GP-OJ, OJ-GP-PN, OJ-GP-PF and OJ-GP-PN-PF, the valencene/nootkatone ratios were found to be between 0É15 and 35. Other component ratios, listed in Table 4, also indicate that matrix e†ects were insigniÐcant. As shown in Table 3, across the di†erent matrices in which the same juice was present, all of the response ratios varied with relative standard deviations (RSD) ranging 4È16%, indicating that they were essentially insensitive to changes in ¡Brix or juice concentration. The only exception was the ratio of compound A/ compound B, which changed by as much as 46È75% RSD, possibly due to di†erences in the chemical nature of the two components. Therefore, this component ratio will not be used to identify the individual juices in mixtures. The proÐles of mixed juices were also investigated. The samples were prepared by mixing two, three or all four fruit juices. In the absence of matrix e†ects, the ratio of components speciÐc to a particular juice, such as the ergostanol/stigmastanol ratio for pineapple juice, should remain the same in any mixtures containing that juice as in the single juice. As shown in Table 4, the ergostanol/stigmastanol ratios of all mixtures containing pineapple juice, ie GP-PN, OJ-PN, PN-PF, OJ-GPPN, GP-PN-PF, OJ-PN-PF and OJ-GP-PN-PF, were found to be between 0É87 and 0É96, which was very close to that (0É96) of the pineapple juice used to prepare the mixed juices, indicating that matrix e†ects were not signiÐcant. Similarly, for mixed juices containing either orange or grapefruit juice, the valencene/ nootkatone ratios were the same as that of the corresponding citrus juice present in the mixture, ie D35 and D0É15, respectively, for the orange and the grapefruit juices used in this study. On the other hand, Compound beverages The results shown in Tables 3 and 4 demonstrate the consistency of the sterol and the sesquiterpenoid distributions of the individual juices in various matrices. In this study, these proÐles were used to conÐrm the alleged presence of the juices in various commercial juice beverages. The results are presented in Table 5. The data on single juices should not be taken as deÐnitive standards of the component distributions due to the small number of samples investigated ; they are presented only to serve as a guide to the characteristic composition of the individual juices. Samples of orangeade, OJ-A1 and OJ-A2, which contain 25% and 40% of orange juice, respectively, with added water, sugars, acids and Ñavors, were found to yield sterol proÐles in accordance with orange juices. The citrus components in OJ-A1 and OJ-A2 were further conÐrmed to be derived from orange juices by the measurements of the valencene/nootkatone ratio. Three commercial samples of passionfruit drinks analysed in this study, PF-A1, PF-A2 and PF-A3, contained 25% juice with added sugar and water. Com- TABLE 3a Matrix e†ects on response ratios of selected components Orange Stigmasterol/campesterol Isofucosterol/campesterol Cholesterol/campesterol Compound A/campesterol Compound C/isofucosterol Ergostanol/stigmastanol Valencene/nootkatone Compound A/compound B Grapefruit Pineapple Passionfruit SJ M1 M2 M3 SJ M1 M2 M3 SJ M1 M2 M3 SJ M1 M2 M3 0É37 1É03 0É11 0É45 Èb È 34É56 2É20 0É38 1É03 0É11 0É46 È È 32É29 1É28 0É35 0É95 0É12 0É43 È È 34É76 0É71 0É36 0É90 0É12 0É41 È È 30É66 0É52 0É22 1É16 0É07 0É25 È È 0É16 17É45 0É22 1É17 0É07 0É28 È È 0É16 11É30 0É20 1É06 0É09 0É31 È È 0É18 2É87 0É21 1É18 0É09 0É31 È È 0É18 4É30 0É06 0É14 0É005 0É06 È 0É96 È 0É37 0É07 0É13 0É003 0É06 È 0É86 È 0É28 0É08 0É18 0É002 0É08 È 1É21 È 0É10 0É07 0É14 0É008 0É06 È 0É93 È 0É14 0É05 0É04 0É003 0É11 1É26 È È 8É01 0É06 0É04 0É004 0É12 1É41 È È 7É64 0É06 0É04 0É005 0É15 1É60 È È 3É23 0É05 0É04 0É004 0É14 1É31 È È 3É54 a SJ, single juice ; ¡Brix, 10É8 (orange) ; 10É4 (grapefruit) ; 14É3 (pineapple) ; and 15É1 (passionfruit) ; M1, 25% juice having the same ¡Brix as the corresponding single juice ; M2, diluted juice having ¡Brix 5 ; M3, compounded juice having ¡Brix 20. b È, see footnotes of Table 2. Analysis of sterols : detecting fruit juices in compounded beverages 625 TABLE 4a Comparison of found and expected response ratios of diagnostic components in mixed juices GP-PF Stigmasterol/campesterol Isofucosterol/campesterol Cholesterol/campesterol Compound A/campesterol Compound C/isofucosterol Ergostanol/stigmastanol Valencene/nootkatone GP-PN Expected Found Expected Found Expected Found Expected 0É07 0É17 0É01 0É15 0É26 Èb 0É15 0É05È0É22 0É04È1É16 0É001È0É068 0É11È0É25 0È1É26 È 0É16 0É13 0É55 0É06 0É15 È 0É95 0É14 0É06È0É22 0É14È1É16 0É005È0É068 0É06È0É25 È 0É96 0É16 0É30 1É08 0É08 0É35 È È 3É7 0É22È0É38 1É03È1É16 0É06È0É09 0É25È0É45 È È 0É16È34É6 0É09 0É14 0É02 0É17 0É27 È 31É3 0É05È0É38 0É04È1É03 0É001È0É094 0É11È0É45 0È1É26 È 34É6 PN-PF OJ-GP-PN GP-PN-PF Found Expected Found Expected Found Expected Found Expected 0É18 0É44 0É05 0É20 È 0É92 34É5 0É07È0É38 0É13È1É03 0.005È0É094 0É06È0É45 È 0É96 34É6 0É06 0É05 0É004 0É12 0É77 0É96 È 0É05È0É06 0É04È0É14 0É001È0É004 0É06È0É11 0È1É26 0É96 È 0É19 0É63 0É06 0É21 È 0É91 4É7 0É06È0É38 0É14È1É16 0É004È0É094 0É06È0É45 È 0É96 0É16È34É6 0É07 0É16 0É01 0É14 0É21 0É91 0É16 0É05È0É22 0É04È1É16 0É001È0É063 0É06È0É25 0È1É26 0É96 0É16 OJ-GP-PF Stigmasterol/campesterol Isofucosterol/campesterol Cholesterol/campesterol Compound A/campesterol Compound C/isofucosterol Ergostanol/stigmastanol Valencene/nootkatone OJ-PF Found OJ-PN Stigmasterol/campesterol Isofucosterol/campesterol Cholesterol/campesterol Compound A/campesterol Compound C/isofucosterol Ergostanol/stigmastanol Valencene/nootkatone GP-OJ OJ-PN-PF OJ-GP-PN-PF Found Expected Found Expected Found Expected 0É10 0É25 0É02 0É19 0É16 È 4É20 0É05È0É38 0É04È1É16 0É001È0É094 0É11È0É45 0È1É26 È 0É16È34É6 0É09 0É15 0É02 0É16 0É26 0É88 32É2 0É05È0É38 0É04È1É03 0É001È0É094 0É06È0É45 0È1É26 0É96 34É6 0É10 0É23 0É02 0É17 0É15 0É87 4É3 0É05È0É38 0É04È1É16 0É001È0É094 0É06È0É45 0È1É26 0É96 0É16È34É6 a OJ, orange juice ; GP, grapefruit juice ; PN, pineapple juice ; and PF, passionfruit. These were the same single juices used in Table 3. Each mixed juice sample was composed of equal volumes of the constituent juices. b È, see footnote c of Table 2. pound C was found to be present in all three samples, indicating the presence of passionfruit juice. Relative abundances of the various components were in comparable magnitudes to the fresh passionfruit juices. The isofucosterol/campesterol ratios in the beverages were, however, slightly higher than those of the three fresh juice samples analysed in this study. In BEV1, a commercial beverage comprised of pineapple and grapefruit juices in unknown proportion, the markers for pineapple, ergostanol and stigmastanol, were detected in a TABLE 5a Composition of diagnostic components in single juices and compounded beverages Compounded beverages Stigmasterol/campesterol Isofucosterol/campesterol Cholesterol/campesterol Compound A/campesterol Compound C/isofucosterol Ergostanol/stigmastanol Valencene/nootkatone Single juicesb OJ-A1 OJ-A2 PF-A1 PF-A2 PF-A3 BEV 1 BEV 2 BEV 3 Orange Grapefruit Pineapple Passionfruit 0É39 0É84 0É07 0É53 Èc È 45É2 0É36 0É54 0É06 0É20 È È 19É2 0É07 0É29 0É05 0É17 4É4 È È 0É08 0É22 0É06 0É25 5É2 È È 0É09 0É29 0É04 0É02 3É9 È È 0É14 0É26 0É04 0É16 È 0É82 0É18 0É18 0É06 0É03 0É15 1É1 1É1 25 0É08 0É23 0É03 0É11 0É54 0É93 È 0É35È0É43 0É32È2É02 0É08È0É19 0É19È0É86 È È 19É6È45É6 0É22È0É35 0É84È2É01 0É06È0É09 0É04È0É46 È È 0É06È0É29 0É05È0É12 0É06È0É21 0É004È0É01 0É03È0É08 È 0É73È1É2 È 0É04È0É06 0É04È0É16 0É001È0É04 0É11È0É51 0É2È50É3 È È a OJ-A1 and OJ-A2 orangeades ; PF-A1, PF-A2 and PF-A3, passionfruit drinks ; BEV1, a beverage composed of pineapple and grapefruit juices ; BEV2, a beverage composed of pineapple, orange and passionfruit juices ; and BEV3, a beverage composed of pineapple and passionfruit juices. b Ranges of response ratios are the combined results of the fresh and commercial juices shown in Table 2. c È, see footnote c of Table 2. L -K Ng, M Hupe 626 proportion consistent with a pineapple juice. The presence of grapefruit juice was indicated by the signiÐcant amount of nootkatone and the low valencene/ nootkatone ratio. Other component ratios assumed values which were intermediate between those of pineapple and grapefruit juices. The results conÐrmed that the sample contained juices of pineapple and grapefruit. The sterol proÐle of sample BEV2, which was a composite of juices of pineapple, orange and passionfruit, was similar to that of BEV1. This beverage was determined to contain orange, rather than grapefruit juice, based on the relative abundances of the sesquiterpenoid components. The presence of passionfruit juice was indicated by the detection of compound C. In the case of sample BEV3, a mixture of 90% pineapple and 10% passionfruit juices, the marker compounds for pineapple and passionfruit juices were detected. Quantitative aspects of the sterolic proÐle were also consistent with the alleged composition of the beverage. Several commercial beverages composed of juices of pineapple or passionfruit blended with other juices not investigated in this study were also analysed, with a view to determining the presence of markers. Stigmastanol and ergostanol were detected at approximately equal amounts in a sample comprised of pineapple and apple juices, as well as in a beverage containing 10% pineapple juice compounded with banana puree and guava juice. Similarly, in commercial drinks composed of 10% passionfruit juice in banana puree and sugar solution, and 20% passionfruit admixed with mango juice and added sugar and water, the presence of passionfruit juice was indicated by the positive detection of compound C. Principal component analysis Principal component analysis (PCA) provides a simple graphical picture of the variability in a set of data. A sample having n-variables can be represented by a point in n-dimensional space. Principal component analysis is a technique of reducing the dimensionality of the data by constructing a series of new data axes, called principal components, which are linear combinations of the variables in the data set. The mutually orthogonal principal components are given a prioritisation based on the amount of variance in the data for which they can account. In e†ect, a projection of the data onto the new axes allows one to visualise graphically the relationship of the samples on a two-dimensional plot if the variance described by the Ðrst two principal components is sufficient to account for most of the variance of the data. A data set based on the data of the selected sterol and terpenoid components of the juice samples was subjected to PCA. It was established and pre-processed before PCA as described in the Experimental section. Figure 6 presents PCA results and shows the pattern of association among the samples. In the two-dimensional plot, samples of juices of pineapple, passionfruit, orange Fig 6. Plot of the Ðrst two principal components of juices of pineapple, passionfruit, orange and grapefruit, as well as compounded beverages. Total variance \ 82%. 1, Orange juice ; 2, grapefruit juice ; 3, pineapple juice ; 4, passionfruit juice ; OJ-A, orangeade ; PF-A, passionfruit drink ; BEV1, composed of pineapple and grapefruit juices ; BEV2, composed of pineapple, orange and passionfruit juices ; and BEV3, composed of pineapple and passionfruit juices. Analysis of sterols : detecting fruit juices in compounded beverages and grapefruit form well-separated groups. The closeness of the orange and grapefruit clusters indicates that the two citrus juices are very similar in composition. The two orangeades fall within the cluster of orange juice as expected since no other juices, which would alter the ratios of the diagnostic components, have been added. The same observation was made with the three passionfruit drinks. Samples of mixed juices, on the other hand, fall between the clusters of their constituent juices. The PCA results clearly illustrate the discriminatory power of assays of sterols and sesquiterpenoids. CONCLUSION The results of this study indicate that analysis of sterol patterns is a useful approach for conÐrming the presence of juices of pineapple, passionfruit, orange and grapefruit in compounded beverages. Sterols such as ergostanol, stigmastanol, and compounds A and C, which have not been reported previously to be present in the juices, were found to be diagnostic components for the identiÐcation of these juices. The relative abundances of various selected sterols, as well as valencene and nootkatone, are of particular value for this purpose. In principle, the scope of the applicability of this approach could be extended to the detection of adulteration of fruit juices. ACKNOWLEDGEMENTS The authors thank the Excise Department of Thailand for supplying some of the juice samples. Thanks are also 627 due to Dr Jan Kovar for valuable discussions throughout this study. Particular thanks to Dr Andre H Lawrence for helpful suggestions during the preparation of this paper. REFERENCES Couture R, Arul J, Gosselin C, Willemot C 1989 The sterols of strawberry fruit. Phytochemistry 28 1276È1277. Greiner G, Wallrauch S 1984 Naringin as evidence for blending grapefruit juice with orange and tangerine juices. Fluss Obst 12 626È628. McLa†erty F W, Hertal R H, Villwock R D 1974 Computer identiÐcation of mass spectra. VI. Probability based matching of mass spectra. Rapid identiÐcation of speciÐc compounds in mixtures. Organic Mass Spectrometer 9 690È702. Nagy S, Nordby H E 1971 Distribution of free and conjugated sterols in orange and tangor juice sacs. L ipids 6 826È830. Pakrachi S C, Achari B, Majumdar P C 1975 Indian medicinal plants XXXII. Constituents of Ananas comosus leaves. Ind J Chem 13 755È756. Shaw P E 1991 Fruits II. In : V olatile Compounds in Foods and Beverages, ed Maarse H. Marcell Dekker Inc, New York, USA, pp 305È324. Stack J B, Joe F L Jr, Cunning D G, Fazio T, Roach J A G 1986 Gas chromatographic determination of fatty acids and sterols in orange juice. J AOAC 69 551È556. Tushishvili L Sh, Durmishidze S V, Sulaberidze K V 1982 Sterols in the Ñesh of grapefruit, orange and mandarin. Khimiya Prirodnykh Soedinenii 4 478È480. Wallrauch S, Faethe W 1988 Amino acids : criteria for the evaluation of fruit juices. In : Adulteration of Fruit Juice Beverages, eds Nagy S, Attaway J A & Rhodes M E. Marcel Dekker Inc, New York, USA, pp 21È47. Williams B L, Goad L J, Goodwin T W 1967 Sterols of grapefruit peel. Phytochemistry 6 1137È1145. Yokoyama H, White M J 1968 Occurrence of ergosterol in citrus. Phytochemistry 7 493È494.