Article pubs.acs.org/JAFC
Rapid Authentication of Coffee Blends and Quantification of 16‑O‑Methylcafestol in Roasted Coffee Beans by Nuclear Magnetic Resonance Elisabetta Schievano,*,† Claudia Finotello,† Elisabetta De Angelis,‡,§ Stefano Mammi,† and Luciano Navarini‡,§ †
Department of Chemical Sciences, University of Padova, via Marzolo 1, 35131 Padova, Italy Aromalab, illycaffè S.p.A., AREA Science Park, Padriciano 99, 34149 Trieste, Italy § illycaffè S.p.A., via Flavia 110, 34127 Trieste, Italy ‡
ABSTRACT: Roasted coffee is subject to commercial frauds, because the high-quality Coffea arabica species, described as “100% Arabica” or “Highland coffee”, is often mixed with the less expensive Coffea canephora var. Robusta. The quantification of 16-Omethylcafestol (16-OMC) is useful to monitor the authenticity of the products as well as the Robusta content in blends. The German standard method DIN 10779 is used in the determination of 16-OMC in roasted coffee beans to detect C. canephora in blends, but it is laborious and time-consuming. Here, we introduce a new method that provides a quantitative determination of esterified 16-OMC directly in coffee extracts by means of high-resolution proton nuclear magnetic resonance spectroscopy. Limit of detection and limit of quantitation were 5 and 20 mg/kg, respectively, which are adequate to detect the presence of Robusta at percentages lower than 0.9%. The proposed method is much faster, more sensitive, and much more reproducible than the DIN standard method. KEYWORDS: 16-O-methylcafestol, Cof fea canephora var. Robusta, qHNMR, food control, coffee blends
■
INTRODUCTION Coffee is one of the most popular beverages, consumed in different ways all over the world. Only two coffee species are commercially exploited, Coffea arabica and Coffea canephora var. Robusta, while Coffea liberica Bull ex Hiern contributes less than 1% of marketed coffee. Their differences in agronomical performance and plant metabolism influence the organoleptic properties and the price of the resulting beverage, with a definite preference of the consumer for C. arabica (Arabica), although it is more expensive. The lower cost of Robusta beans opens the possibility of commercial frauds aimed at affecting the authenticity of 100% Arabica blends, and it renders the detection and quantification of such admixtures in commercial samples a necessary analytical tool to protect consumers. The morphological characteristics of the green beans of the two species are often sufficient to discriminate between them; however, visual inspection is inadequate in the case of roasted or ground beans, and a method to determine the presence of Robusta in the powder is necessary. To this end, organoleptic properties may be insufficient and subjective, so that differences in chemical composition must be exploited. To date, the only official method available is the German standard method DIN 10779;1 it is used to quantify the amount of 16-Omethylcafestol (16-OMC) in roasted beans, for which it was originally developed, but also in green coffee beans and coffee brews.2−4 This method is based on the observation that 16OMC is present exclusively in Robusta, whereas other, more abundant diterpenes, such as cafestol and kahweol, cannot be used for this discrimination; cafestol is present in both species and kahweol is present in higher amounts but not exclusively in Arabica. The high-performance liquid chromatography © XXXX American Chemical Society
(HPLC) separation of 16-OMC using the DIN 10779 methodology is quite good, but the procedure is toilsome and time- and solvent-consuming because it involves a 5 h long lipid extraction using methyl tert-butyl ether (MTBE), a saponification step, a second extraction with MTBE, a further step in dichloromethane, and a final HPLC run of 45 min. Moreover, it requires an external standard calibration, and the recovery of the method was never determined. For these reasons, other methods are being sought that would be easier, faster, and accurate. The compositional differences between Arabica and Robusta are such that many compounds or classes of compounds can be selected to provide information on the composition of a blend. For example, studies are available in the literature suggesting the content or profile of fatty acids,5 tocopherols,6 or sterols7−9 as discriminant parameters. Also, the ratio between different diterpenes has been suggested as a discriminating criterion.10,11 Unfortunately, all of these methods require extraction and separation procedures similar or more complicated than those used in the DIN method. The use of near-infrared (NIR)12 or Raman13−15 spectra has been suggested, which offer the enormous advantage of allowing the investigation of whole or ground beans directly or with little manipulation. Nevertheless, the NIR method proposed entails a laborious calibration procedure, and the Raman methods suffer from other drawbacks, such as light instability of the molecules used for Received: October 16, 2014 Revised: November 27, 2014 Accepted: November 28, 2014
A
dx.doi.org/10.1021/jf505013d | J. Agric. Food Chem. XXXX, XXX, XXX−XXX
Journal of Agricultural and Food Chemistry
Article
The ACD software (ACD labs 12.0) was used to process the spectra. Fourier transformation was performed after exponential line broadenings of 0.2−0.4 Hz. Integrations were manually obtained after careful manual phase and baseline correction. Quantification. To determine the total concentration of the 16OMC esters, a scale factor of 0.108 was used, corresponding to the ratio between the volume of the coaxial insert (Vc) and the volume occupied by the sample in the outer part of the NMR tube (Vt), calculated as reported by Rastrelli et al.17 The absolute concentration of 16-OMC esters (Ca) was determined by comparing the integral value of the area of the methyl protons at position 21 (Ia) and that of the methyl protons of DMF at 3.02 ppm (Is), using the relationship
quantification and limits of detection of Robusta that may be too high. Also, a DNA-based food authenticity assay was recently proposed16 that is able to detect 5% Robusta in Arabica coffee. The content of 16-OMC remains a valuable option for detection of Robusta additions to Arabica blends, but a fast, reliable, and sensitive method is needed. The aim of the present work is to report an analytical method based on nuclear magnetic resonance (NMR) spectroscopy that could be used as an alternative to the DIN standard and that could guarantee the authenticity of 100% Arabica blends by detecting possible commercial frauds in a rapid and dependable way. NMR spectra can be acquired in a very short time; the technique is inherently reproducible and provides quantitative and structural information. These features combined with the minimal sample preparation required and the absence of any derivatization step make NMR a promising choice for our purposes.
■
Ca (mol/L) = (Ia /Is)(Vc/Vt)Cs where Cs is the concentration of the external standard. These methyl protons fall in a free region in all of the analyzed samples. The following formula gives the concentration of the free 16-OMC in mg/kg:
C16‐OMC (mg/kg) = Ca MW16‐OMCVextr (mL)/mcoffee (kg)
MATERIALS AND METHODS
where MW16‑OMC is the molecular weight of 16-OMC, Vextr is the CDCl3 volume used in the extraction, and mcoffee is the weight of the coffee powder. The longitudinal relaxation time T1 of all signals of interest was determined in the matrices, in the standard solutions, and in chloroform using the inversion−recovery sequence. To produce quantitative data, the relaxation delay was at least 5 times the longest measured T1, corresponding to one of the methyl signals of DMF in D2O, which determined a relaxation delay of 43 s. Method Validation. To verify the linearity of the proposed method with respect to the concentration of 16-OMC, 12 coffee blends with increasing content of Robusta were prepared. For each mixture, the CDCl3 extraction was performed in triplicate and a 1H NMR spectrum of each sample was acquired using identical experimental conditions. The integrals of the non-overlapping signals of esterified 16-OMC (i.e., protons 17 and 21) were plotted versus the nominal concentration of 16-OMC. To calculate the limit of detection (LoD) and limit of quantification (LoQ), a signal-to-noise ratio of 3 and 10 was used, respectively. These limits were determined by plotting the S/N value versus the 16-OMC amount using a concentration range of 0−450 mg/kg. The coffee powder/solvent ratio chosen was 0.3 g/mL (see the Method Development section), and the spectra were acquired with 64 scans for a total acquisition time of about 48 min. Extraction Efficiency. To assess the extraction efficiency of the proposed method, three different samples were extracted twice. Specifically, after the first extraction, the residual solvent remaining on the ground beans was evaporated and the solid residue was subjected to a second extraction procedure. The amount of esterified 16-OMC recovered in the second extraction was calculated taking into account the amount of esterified 16-OMC present in the solvent that was not recovered during the first filtration. The following equations were used:
Chemicals and Materials. Deuterated chloroform (CDCl3, 99.96% D, stabilized with silver, Sigma-Aldrich), deuterated water (≥99.96% D, Eurisotop), N,N-dimethylformamide [DMF, ≥99.99% (GC), Fluka-Sigma-Aldrich], 16-OMC (phyproof References Substances, PhytoLab), 5 mm precision glass NMR tubes (535-pp, Wilmad), and coaxial insert (wgs-5bl, Wilmad) were used. Coffee Samples. C. canephora from Ivory Coast, India, and Uganda, hereafter named R1, R2, and R3, respectively, were roasted to a medium roasting degree in a lab roaster (Probat). The C. arabica sample came from Brazil. Several commercial roasted coffee samples (in whole beans) were purchased in local markets. These samples were named CB0, ..., CBn. Sample Treatment. Coffee powder was obtained by grinding the roasted coffee beans in liquid nitrogen. Extraction of ca. 0.45 g of powder (accurately weighed) was carried out by vortexing for 15 min with 1.5 mL of CDCl3. Samples were then quickly filtered through a cotton wool filter directly into the NMR tube. All of the operations were conducted at about 4 °C to minimize solvent evaporation. A 1H NMR spectrum was acquired immediately after filtration. Preparation of the External Standard Solution. To quantify absolute concentrations, a coaxial insert filled with a DMF solution in D2O was placed inside the NMR tube. This standard was chosen because of its long-term stability that allows for a large set of samples to be measured and quantified with respect to the same reference solution. The solution was prepared by accurately weighing DMF and the D2O solvent for a final concentration of the standard in the capillary tube of 5.99 mM to provide a S/N ratio comparable to that of the test substance. NMR Spectroscopy. The assignment of 1H and 13C of the standard 16-OMC was achieved using one-dimensional (1D) and twodimensional (2D) spectra. The complete assignment of the esterified form of 16-OMC was also achieved using 2D NMR experiments performed directly on the extracts and by comparison to the NMR data of the standard 16-OMC. A Bruker (Rheinstetten, Germany) Avance DMX600 spectrometer operating at 599.90 MHz for 1H and equipped with a 5 mm TXI xyz-triple gradient probe was used. The 1D spectra were acquired in deuterated chloroform using a common onepulse sequence with a spectral width of 6000 Hz, with 32 768 data points; the receiver gain ranged from 64 to 128; and scans varied from 32 to 64 depending upon the 16-OMC concentration. Two-dimensional total correlation spectroscopy (TOCSY), heteronuclear multiple-quantum coherence (HMQC), heteronuclear multiple-bond coherence (HMBC), and nuclear Overhauser effect spectroscopy (NOESY) spectra were recorded at 298 K using standard Bruker pulse programs. Chemical shifts are reported in δ values using the residual signal of CHCl3 set to 7.27 and 77.00 ppm as reference for 1 H and 13C, respectively.
1 add 1 n16 ‐OMC = V1 C16‐OMC
2 add 2 add 1 n16 − V1rec)C16 ‐OMC = V 2 C16‐OMC − (V1 ‐OMC
where n116‑OMC and n216‑OMC are the moles of 16-OMC recovered in the add first and second extraction, respectively, Vadd 1 and V2 are the volumes of CDCl3 used in the first and second extraction, respectively, C116‑OMC and C216‑OMC are the concentrations of 16-OMC determined in the first and second extraction, respectively, and Vrec 1 is the volume recovered in the first extraction. The value of n216‑OMC is taken as a measure of the efficiency of the first extraction.
■
RESULTS AND DISCUSSION Analysis of Spectra. The CDCl3 extraction of coffee powder provides a mixture of the main components of coffee
B
dx.doi.org/10.1021/jf505013d | J. Agric. Food Chem. XXXX, XXX, XXX−XXX
Journal of Agricultural and Food Chemistry
Article
Figure 1. (a) 1H NMR of samples extracted from 100% Arabica (bottom) and 100% Robusta (top) roasted coffee powder. (b) Expanded region containing the olefinic proton resonances of cafestol and kaweol. The olefinic protons of 16-OMC and those of cafestol are overlapped. (c) NMR profile of the fatty acid proton region.
The free form of 16-OMC is generally present in trace amounts, visible in some samples, as shown in the inset of Figure 3. As a first consideration, a visual inspection of the spectrum allows for an immediate distinction of a 100% Arabica coffee from a mixture with Robusta. Specifically, the protons 21 of esterified 16-OMC immediately reveal the presence of Robusta, while the signals attributed to kahweol (18, 19, 1, and 2 of Figure 1b) that almost disappear in the Robusta sample reveal the presence of Arabica. Method Development. We optimized a very fast and simple procedure that requires only a few steps using a deuterated solvent directly. The selection of CDCl3 as the solvent for extraction was based on the observation that the diterpenes are very soluble in this solvent and that it is not expensive. Also, the residual signal of the protonated form does not overlap with important peaks to integrate. The coffee powder/solvent ratio was optimized to obtain an easily filterable solution and a well-resolved spectrum. The best solid/liquid ratio was found to be ≤0.3 g/mL.
oil, i.e., fatty acids, caffeine, diterpenes (principally present as esters of fatty acids18), and other less concentrated species (see Figure 1). Identification of the 1H and 13C signals of the marker (i.e., esterified 16-OMC) was carried out using the assignment of free 16-OMC, which was compared to that reported in the literature19 as an aid, and it is reported in Table 1. The chemical structures of esterified 16-OMC, cafestol, and kahweol are shown in Figure 2. In Figure 3a, the diagnostic region (protons 17 and 21) of 16-OMC in a Robusta extract is shown. Protons 21 in esterified 16-OMC resonate as a singlet at 3.17 ppm, slightly shifted with respect to the free form (3.18 ppm; see Figure 3c). Protons 17 form a second-order system at 3.78 ppm in the free form, while, in the ester, two doublets are obtained at 4.28 and 4.45 ppm, respectively, with a scalar coupling constant JAM = 12.8 Hz. Only the doublet at 4.45 ppm is well-resolved in the spectrum, while the other doublet is partially overlapping the signals of the glycerol moiety of triglycerides. The same region in a 1H spectrum of Arabica is shown in Figure 3b. As expected, the signals of protons 17 and 21 from the 16-OMC unit are absent. C
dx.doi.org/10.1021/jf505013d | J. Agric. Food Chem. XXXX, XXX, XXX−XXX
Journal of Agricultural and Food Chemistry
Article
Table 1. 1H and 13C Resonances of Esterified 16-OMC in the Chloroform Extract of Robusta Roasted Coffee Beansa
obtained for the methyl singlet at 3.17 ppm, this signal was used for quantification of 16-OMC. The values of the relative standard deviation (RSD) ranged from 2 to 5% and confirm the suitability of the proposed method. By plotting the S/N versus nominal concentration (R2 = 0.9906, data not shown), the sensitivity of the method was assessed in terms of LoD and LoQ, which were 5 and 20 mg/ kg, respectively. These LoD and LoQ values are satisfactory; by considering the reported range of the 16-OMC content found in Robusta green beans (0.8−2.5 g/kg of dry weight, with both the median and mean values equal to approximately 1.7 g/kg of dry weight3), this method allows for the detection of the presence of Robusta at percentages lower than 0.9% and down to 0.2%. This level is lower than that reported by the DIN method (about 2%). Extraction Efficiency. For all three of the samples 2 investigated, the corrected amount of marker (n16‑OMC ) recovered from the second extraction was lower than 2% with respect to the first extraction. This indicates that the first extraction is practically quantitative. Interestingly, by grinding at room temperature with a coffee mill, according to the DIN recommendations, a lower extraction efficiency was obtained for all examined samples (73−82%, data not shown). Therefore, grinding at a low temperature seems to be crucial. Comparison to the DIN Method. The method presented here is able to detect and to quantify 16-OMC in roasted coffee samples without complex sample manipulation and in a few minutes. Exploiting the power of NMR to determine the presence and amount of a substance in a complex mixture, this method avoids long Soxhlet extractions, a saponification step, and chromatographic procedures to isolate and separate the analyte of interest. In addition to easy detection, the method that we propose permits the quantification of 16-OMC and, in principle, determines the content of Robusta ingredients in roasted coffee blends. To evaluate the accuracy of the proposed method, the entire procedure was repeated twice on the same sample and the quantities of 16-OMC found were compared to that determined by the DIN method. In Table 2, a comparison between the 16-OMC content determined by the two methods is reported. The agreement between the two different determinations is satisfactory, although the values obtained by the DIN method are often lower than those determined by NMR. An underestimation of the 16-OMC content by the DIN method is possibly related to the complex manipulation necessary to perform the analysis, while the much simpler sample preparation of our procedure avoids the danger of loss of analyte. Therefore, the small amounts of sample required, the rapidity, precision, and accuracy, and the low detection limit of the analysis make the NMR method promising for screening purposes. The wide range of the 16-OMC concentration in different Robusta samples poses a serious obstacle for the quantification of Robusta percentage in roasted coffee blends. This problem is common to every method that exploits 16-OMC and makes the determination of this molecular marker more appropriate for authentication of 100% Arabica blends. This point, however, could be the subject of further investigation. In this regard, a large database might help in better defining the real range of the 16-OMC concentration in different Robusta samples and in a possible correlation with geographical origin and agronomical practices.
esterified 16-OMC (esterified cafestol) δ1Hb 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 CO fatty acid
1.25, 2.07 2.63
δ 13Cb 33.8 21.4 149 120.1 43.1 22.7 41
2.29 1.81, 1.85 1.57−1.67 1.19 1.57−1.67 1.57−1.67 2.26 (2.10) 1.48, 2.00 1.57, 1.67 (1.54, 1.58) 4.24, 4.44 (4.27−4.29) 6.24 7.24 0.83 3.17 (-)
53.1 38.5 19.8 22.9 45.6 (46.8) 38.34 55.3 (53.0) 84.58 (79.9) 62.72 (68.37) 107.83 140.52 14.16 49.84 (-) 173.72
a Only the cafestol assignments that differ from those of 16-OMC are reported in parentheses. bδ is in parts per million. NMR solvent was CDCl3.
Figure 2. Chemical structure and atom numbering of esterified diterpenes: 16-OMC, cafestol, and kahweol.
Quantification of 16-OMC. 16-OMC was quantified by integrating the resolved peaks at 3.17 ppm (protons 21) and the signal of one of the protons in position 17 (4.45 ppm) of the analyte compared to the integral of the methyl signal of DMF at 3.02 ppm, which did not present any overlap in any of the samples studied. The proximity of the methyl 21 peaks of the free and esterified forms allowed us to integrate them together; therefore, the value of 16-OMC that is obtained is always referring to the total amount. Linearity, Precision, and Sensitivity of the Analytical Procedure. Good linearity was achieved for both signals, as indicated by the equation shown in Figure 4 for concentrations ranging from 5 to at least 1500 mg/kg, with a satisfactory determination coefficient R2. Because a higher R2 value was D
dx.doi.org/10.1021/jf505013d | J. Agric. Food Chem. XXXX, XXX, XXX−XXX
Journal of Agricultural and Food Chemistry
Article
Figure 3. Expanded region of the 1H NMR spectrum showing the diagnostic resonances of protons 17 and 21: (a) 100% Robusta sample, (b) 100% Arabica sample, and (c) 16-OMC standard. In the insets on the top of the figure, the signals of 17 and 21 protons of 16-OMC ester are enlarged. They are clearly visible in the Robusta sample (top), but they are absent from the Arabica sample. The arrow shows the weak methyl signals in position 21 of 16-OMC in its free form.
Table 2. 16-OMC Content in Commercial Roasted Coffee Samples 16-OMC (mg/kg) blend b
R1 R2b R3b CB0 CB1 CB2 CB3 CB4 CB5 CB6 CB7 CB8 CB9 CB10
Figure 4. Plot of the integrals of non-overlapping signals of esterified 16-OMC (i.e., protons 17 and 21) versus nominal 16-OMC content of 12 coffee blends with increasing content of Robusta. For each mixture, the CDCl3 extraction was performed in triplicate.
100% 100% 100% 100% 100% A/R ndc nd nd 100% A/R nd nd nd
R R R R R
R
NMRa
DINa
1841 (36) 1770 (43) 1490 (40) 1442 (44) 1739 (14) 1160 (24) 619 (2) 284 (2) 736 (31) 1750 (7) 1537 (14) 1348 (61) 32 (2) 591(11)
1980 (100) 1630 (240) 1008 (200) 1428 (18) 1450 (18) 1193 (51) 434 (9) 130 (40) 945 (100) 1279 (230) 1135 (14) 1305 (80)
