Microwave-Assisted Extraction of Phenolic Compounds from Spent Coffee Grounds. Process Optimization Applying Design of Experiments

In this study, sustainable technology microwave-assisted extraction (MAE) in association with green solvents was applied to recover phenolic compounds from spent coffee grounds (SCGs). A design of experiments (DOE) was used for process optimization. Initially, a 24−1 two level Fractional Factorial Design was used and ratios “solvent to solute” and “ethanol to water” were identified as the significant experimental factors. Consequently, Central Composite Design (CCD) was applied to analyze the effects of the significant variables on the response yield, total polyphenols content (TPC), and antioxidant activity (AA) by the DPPH assay method, and quadratic surfaces to optimize those responses were generated. The values of the significant factors of 16.7 (solvent/solute) and 68.9% (ethanol/water) were optimized simultaneously the yield (%) at 6.98 ± 0.27, TPC (mg GAE/g) at 117.7 ± 6.1, and AA (µmol TE/g) at 143.8 ± 8.6 and were in excellent agreement with those predicted from the CCD model. The variations of the compositions of the lipids, caffeine, pentacyclic diterpenes, and FAME as a function of the dominant factor % ethanol in the solvent mixture were analyzed by applying NMR and GC-FID, and the results obtained confirmed their determinative significance.


Introduction
Coffee is one of the most used plants on Earth; the world coffee production in 2019 to 2020 was estimated at 169.34 million 60-kg bags [1]. Taking into consideration that, from each kg of coffee, 0.91 kg of solid waste is produced, the storage of those huge quantities of spent coffee grains, SCGs, is a critical environmental problem that should be resolved by applying innovative approaches based on the principles of circular economy. Within those concepts, the design and implementation of one feed, multiproduct biorefineries focused on the sustainable use and valorization of SCG waste biomass are a promising alternative. SCG biorefinery should integrate predominantly mild, advanced processes and techniques targeting the clean, safe production of high-value bioactives with enhanced properties and a wide spectrum of applications in the pharmaceutical, nutraceuticals, and cosmetic industries, as well as in the development of functional foods [2][3][4].
From this perspective, among the available techniques for biological and biomass matrix valorization, microwave-assisted extraction (MAE) has proven to be one of the most promising. It is considered an advanced and viable alternative to conventional and other green extraction techniques (e.g., supercritical and subcritical fluid extraction, pressurized liquid extractions, and accelerated fluid and ultrasound-assisted extraction) that has important advantages, among which are controllable and effective heating, faster

Fractional Factorial Design (FFD)
In the present study, the effects of four independent parameters (factors) were investigated: microwave power, Power (Watt); ratio between the volumes of the solvent and the solute (SCGs), Ratio Solv/sol (mL/g); composition of the solvent, namely ethanol-to-water ratio, Ratio Eth/water (%Ethanol), and time of microwave irradiation, Time (min). The levels choice was based on the knowledge acquired previously. For a 2 4−1 (two-level) FFD with four factors, 16 experimental runs were required. It should be noted that, in all the experiments performed, the maximum percentage of water was 50%, since the increase of its content in the final (extract plus solvent) solutions increased considerably the amount of energy required to obtain the final dry extract.
The maximum and minimum of the different factors and their respective coded values generated by the program Design-Expert 11 (DS-11) are shown in Table 1. Ratio (Solv/sol) mL/g −1 ≡ 5.00 +1 ≡ 20.00 X 3 Ratio (Eth/water) % Ethanol −1 ≡ 50.00 +1 ≡ 99.00 X 4 Time min −1 ≡ 3.00 +1 ≡ 6.00 The responses characterizing the quality of the extracts in the three response functions are: the yield (%), the total phenolic content, TPC, represented in mg of gallic acid equivalents (GAE) by g of extract (mg GAE/g), and AA as the DPPH assay in micromole of trolox equivalents (TE) by g of extract (µmol TE/g). The results of the FFD design tests are displayed in Table 2.

Statistical Analysis of the FDD Experimental Data
The experimental design applied allowed the identification of the significant experimental factors influencing the yield, TPC, and AA of the extracts recovered. Statistical testing of the model was performed in the form of an analysis of variance (ANOVA) for each response. The yield, TPC, and AA F-values-65.59, 146.57, and 14.40, respectively-implied that the model was significant. This was corroborated by the values of p < 0.0001 and p < 0.0006, indicating that there was only a probability of 0.01% for the two first responses and 0.06% for the AA that a higher F-value could occur due to noise. p-Values less than 0.05 indicate that the individual model terms are significant.
These results identified the independent variables X 2 -Ratio (Solv/sol)-and X 3 -Ratio (Eth/water)-as the most significant factors that influence the experimental extraction yield, TPC, and AA, with X 3 being the dominant one between the two. Thus, for the yield, the contribution percentages of the ratios Solv/sol and Eth/water were 13.52 and 81.44%, respectively, and the combined contribution to the overall model of these two variables was 95%. The same trend was observed concerning TPC and AA, for which the combined contributions were 97.2 and 79.4%, respectively.
The ANOVA results on the FFD model selected are shown in Table 3, showing the most significant terms. The time (independent variable X 4 ), as well as the interaction between X 1 X 2 (Power * Ratio (Solv/sol)) and X 1 X 4 (Power*Time), have some significance on AA but with lower key values for the overall model. p-Values greater than 0.1000 indicate that the model terms are not significant. Based on the ANOVA analysis and fitting the factors with the responses using the least squares method, the extract yield of the SCGs (Yield), as a function of the independent variables X 1 -X 4 , is obtained by Equation (1) The ANOVA attested that a lack of fit (with a p-value < 0.0001) was not significant for all response surface models at a 95% confidence level, which means that all models represented the data satisfactorily. Simultaneously, the R 2 , adjusted R 2 (R 2 adj ), predicted R 2 (R 2 Pred ), coefficient of variation (CV), and adequate precision (Ad Precision) were calculated to check the model adequacy and are presented in Table 4. The values of the determination coefficient (R 2 ) calculated (98.29, 99.23, and 92.65%) infer that the accuracy and general predictive ability of the quadratic polynomial regression models represented by Equations (1)-(3) are very good, since it is accepted that, for a good fitting model, R 2 should not be less than 80% [19,20].
Furthermore, the value of the adjusted R 2 should be close to R 2 adj , suggesting that a high degree of correlation between the observed and predicted values exists. In addition, R 2 adj values should also be in a reasonable agreement with the R 2 Pred (differences less than 0.2), showing that nonsignificant terms have not been included in the model.
The coefficient of variation, CV, is a measure expressing the standard deviation as a percentage of the mean. Small values of the CV represent better reproducibility. Overall, a CV higher than 10% signifies that the variation in the mean value is high and does not satisfactorily develop an adequate response model [21].
In our case, the CVs of the yield and TPC are reasonable, while, for the AA, the CV value is the only parameter that does not satisfy the below 10% criterion. Finally, adequate precision (Ad Precision) measures the signal-to-noise ratio, and a value greater than 4 is appropriate, which is verified in all the cases examined (Table 5). The maximum value of a respective response can be obtained from Equations (1)-(3), respectively. Additionally, if the values corresponding to the simultaneous maximum of the three independent variables-yield, TPC, and AA-are required, those can be obtained by solving the system of Equations (1)-(3). However, as shown, the most important factor for these variations is the ethanol/water ratio, followed by the solvent/solute ratio.
To locate the ideal region of the design space for the chosen experimental factors and intended response, a FFD analysis should be used to assess the main effects and the essential interactions between them. Although Equations (1)-(3) can estimate the fitted values at the corner points of the purpose design, quadratic terms in the model to shape the curvature across the whole response surface are required. That can be achieved by applying a response surface design with axial points generated by a central composite design (CCD).

Central Composite Design (CCD)
The results of the FFD design applied at stage one indicated that the influence of power and time on the MAE was not substantial. Hence, in the CCD, the values of those independent parameters were fixed at 90 W and 4.5 min, respectively.
The other two factors-Ratio Solv/sol (10-20 mL/g) and Ratio Eth/water (50-70% Ethanol)-were used to generate the CCD test design. The ANOVA results on the CCD models selected are shown in Table 5.
It was demonstrated that, for the responses yield and AA, the best fit equation was quadratic, while, for the TPC, a linear one. The F-values showed that the models for the yield, TPC, and AA were significant, and there was only a 0.03, 0.01, and 0.07% chance, respectively, that an F-value that high could occur due to noise. The lack of fit F-values to the three responses was nonsignificant relative to the pure error. The latter was required to have a good model fit and determine the adequacy of the model.
The final models for the responses (yield, TPC, and AA) in terms of actual factors are represented by Equations (4) The ANOVA results on the CCD models selected are shown in Table 6. All terms (independent variables) were significant, since the p-values were less than 0.05 (at the 5% probability level), and only values < 0.100 were insignificant. The ANOVA analysis demonstrated that the models were adequate and reproducible and the results were reliable. Thus, it was ascertained that the regression models obtained were suitable for the determination of significant extraction parameters values that optimized the responses-yield, TPC, and AA-of the SCG extracts recovered.
The response surface analysis was designed centered on the three-dimensional model polynomial function defined, which established the effect of the significant independent variables chosen on each observed response. The contour response plots allowed visual identification of the optimal levels of each factor and a choice of the most suitable values of the different response factors.
It can be deduced from Figure 1 that, when the ratio ethanol/water is kept at a lower level, the yield increases with the growth of the ratio solvent/solute (SCGs). The extraction yield, in terms of a combination of binary factors, can change from 6.58 to 7.64%. a Sums of squares. b Mean square.
The response surface analysis was designed centered on the three-dim model polynomial function defined, which established the effect of the sig independent variables chosen on each observed response. The contour respon allowed visual identification of the optimal levels of each factor and a choice of t suitable values of the different response factors.
It can be deduced from Figure 1 that, when the ratio ethanol/water is kept at level, the yield increases with the growth of the ratio solvent/solute (SCGs). The ex yield, in terms of a combination of binary factors, can change from 6.58 to 7.64%. Figure 2a depicts the linear effect of the ratio ethanol/water in the solvent mi the TPC. It was demonstrated that, within the entire range of the ratio Solv increase of the water percentage in the mixture raised the TPC content. This complete agreement with the data of Table 5, where the parameter ratio Eth/wat p-value < 0.0001. On the other hand, the p-value of the ratio Solv/sol was not sig since it was much higher than 0.0001.    Table 5, where the parameter ratio Eth/water had a p-value < 0.0001. On the other hand, the p-value of the ratio Solv/sol was not significant, since it was much higher than 0.0001. Figure 2b shows the quadratic effects of the solvent/solute and ethanol/water ratios on the AA. The maximum was achieved at the lower values of the independent parameters, a behavior pattern completely different from that depicted in Figure 2b.

Optimization of the MAE Process
The target of the response surface methodology is process optimization. Thus, we aimed to determine the values of the two significant independent variables that led to the optimum responses. It should be noted, however, that this was not a trivial task, since their behaviors were divergent, and there was not a single unique point that could be achieved applying Equations (4)- (6). Instead, adjusting a maximum or a minimum to the response, the optimal conditions could be defined randomly, and a few optimal points could be located for them.
To realize that, the DS11 software was applied to determine the best conditions to optimize the three responses. Subsequently, those values were verified by performing three experimental tests to compare them with the optimized ones. The results obtained are  Table 7, and they show that the models for predicting the optimum response values were adequate, since there was a very good agreement with the experimental values.  Figure 2b shows the quadratic effects of the solvent/solute and ethanol/water ratios on the AA. The maximum was achieved at the lower values of the independent parameters, a behavior pattern completely different from that depicted in Figure 2b.

Optimization of the MAE Process
The target of the response surface methodology is process optimization. Thus, we aimed to determine the values of the two significant independent variables that led to the optimum responses. It should be noted, however, that this was not a trivial task, since their behaviors were divergent, and there was not a single unique point that could be achieved applying Equations (4)- (6). Instead, adjusting a maximum or a minimum to the response, the optimal conditions could be defined randomly, and a few optimal points could be located for them.
To realize that, the DS11 software was applied to determine the best conditions to optimize the three responses. Subsequently, those values were verified by performing three experimental tests to compare them with the optimized ones. The results obtained are presented in Table 7, and they show that the models for predicting the optimum response values were adequate, since there was a very good agreement with the experimental values. Table 7. Predicted and experimental values of the responses were obtained at the optimum conditions of the independent variables. The experimental data are given as the mean ± SD (n = 3).

Influence of Ethanol:Water Ratio on the Extracts Composition-1 H NMR Analysis
In addition to examining the influence of Ratio (Solv/sol) and Ratio (Eth/water) on the yield, TPC, and AA of the MAE extracts recovered, the second objective of our work was to study the impact of Ratio (Eth/water), identified as the most significant experimental parameter, on the extract's principal compound compositions. To realize that, an analysis of the extracts obtained in the overall range of the ethanol:water mixture composition employed was carried out. The analyses were performed by applying 1 H-NMR, which has proven to be a fast and useful tool for composition identification and quantification. The  Table 7. Predicted and experimental values of the responses were obtained at the optimum conditions of the independent variables. The experimental data are given as the mean ± SD (n = 3).

Influence of Ethanol: Water Ratio on the Extracts Composition-1 H NMR Analysis
In addition to examining the influence of Ratio (Solv/sol) and Ratio (Eth/water) on the yield, TPC, and AA of the MAE extracts recovered, the second objective of our work was to study the impact of Ratio (Eth/water), identified as the most significant experimental parameter, on the extract's principal compound compositions. To realize that, an analysis of the extracts obtained in the overall range of the ethanol:water mixture composition employed was carried out. The analyses were performed by applying 1 H-NMR, which has proven to be a fast and useful tool for composition identification and quantification. The influence of Ratio (Eth/water) on the compositions of lipids and diterpenes in the extracts is presented in Tables 8 and 9, respectively.   Table 8 shows that the % of molar fractions of TAGs decrease smoothly from 96.1 to 85.35%, with the decrease of the percentage of ethanol down to 50%.
The 1,2 DAGs composition, however, exhibited an opposite trend-their % molar fractions increased sharply from 1.96 to 9.04% in the range 99-60% ethanol, reached their maximum at 50% ethanol, and then decreased to 6.53% at 45.9% ethanol.
There was not a clear trend in the change of the composition of the mono-(MUFA) and di-unsaturated fatty acids (DUFA) with the change of the ethanol:water ratio. The only exception was the extract of Run 4, recovered by 99% of ethanol, for which the highest concentration of DUFA (40.4) and the lowest of MUFA (15.5) were recorded. Table 9 shows the influence of Ratio (Eth/water) on the compositions of the diterpenes. The reduction of ethanol % down to 60% positively affected the recovery of the diterpenes. Their % of molar fractions increased, with the highest increment registered in the range of 70-60%, the most pronounced being for cafestol. Interestingly, the diterpene % molar fractions diminished at lower than 60% ethanol in the solvent, and the most significant drop-from 11.12 to 3.17%-was observed for 16-O-Methyl-Cafestol for a relatively small increment from 50 to 45.9% in the ethanol composition. The present results are in good agreement with a previous work, where supercritical CO 2 was used to obtain the oil.

Influence of Ethanol: Water Ratio on the Eextract Compositions-Fatty Acid Methyl Esters (FAMEs) Analysis
The influence of the ethanol:water ratio on the extracts' fatty acid compositions was also studied, and the results obtained are presented in Table 10.
It should be noted, that, unlike the lipids and diterpenes, the fatty acid compositions did not exhibit any clear trends regarding the ethanol:water ratio influence. Still, the highest composition of DUFA (41.8) was registered at the highest value of the ethanol:water ratio, while, for MUFA and SFA, the opposite trend was observed: the highest quantities were registered at the lowest value of that ratio.
As could be expected, the polyunsaturated/saturated ratio UI increased with the increase of the ethanol percentage, and the highest UI was calculated for Run 4, due to the highest content of linoleic acid registered in the extract. The values obtained agreed with the previous results [4,22,23], where the differences in the origin and variety of the coffee and, consequently, the resulting SCGs used could explain some small variations.

Discussion
There have been few studies published in the literature that have examined alternative techniques for SCG valorization and the recovery of extracts rich in bioactives with pronounced AA and high TPC. The data reported, however, are very scattered and depend on multiple factors, such as the types of solvents, extraction temperatures, solid-liquid ratios, coffee blends and SCGs, and/or different preparation processes, among others, which makes comparisons of the results challenging.
Generally, Soxhlet n-hexane extraction is used as the reference method (see, for example, the study of Acevedo et al. [24], where the values of TPC = 273.34 mg GAE/g and DPPH = 82.65 µmol TE/g, respectively, were reported). However, nowadays, green technologies and more environmentally friendly solvents are preferred.
Higher contents of bioactive compounds were registered in SCG extracts of 100% Arabica coffee blends by changing the temperature and the solvent extraction volume at a constant time [25]. The TPC and AA by DPPH of the extracts were 61.49 mg GAE/g and 324.51 µmol TE/g, respectively.
A more comprehensive comparison of our results can be performed with the works of References [12,15], discussed briefly above, who studied the ethanol influence on MAE of SCGs, utilizing RSM. In both works, the % of ethanol in the solvent mixture, however, was different. Ranic et al. [15] applied a fixed percentage of ethanol (20%). The yields achieved were low, but the extracts TPC and AA of the DPPH assay were considerable. On the other hand, Pavlović et al. [12] used mixtures with 20-80% of ethanol. Values of the TPC between 175.08 and 398.95-mg GAE/g, as well as very good DDPH activity (%), were reported.
In view of the above, the most appropriate comparison of our TPC and AA values by DPPH was with those of Pavlović [12].
There were some differences observed in the values of TPC and AA reported by both studies. This was not unexpected, taking into consideration the nature of the SCGs used-filter coffee vs. espresso. Still, our conclusion that the decrease of the % of ethanol in the solvent mixture diminishes the yield of the MAE process while increasing the TPC or DPPH values of the extracts recovered (Tables 3 and 5) was confirmed and substantiated.

Raw Material
The SCGs, a blended mixture of Arabica and Robusta species, obtained from an espresso machine of a Bulgarian coffee shop were oven-dried to a constant mass at 378 K and stored frozen in a refrigerator at 255 K. The moisture content (4.0 ± 0.3%) was measured with a thermogravimetric balance, Kern MRS 120-3 (KERN & Sohn GmbH, D-72336 Balingen, Germany), and the average particle diameter, dp (0.273 ± 0.023) mm, was calculated as described before [4].

Microwave Extractions
The microwave-assisted extractions of the SCGs samples were performed using a CEM Discover SP microwave reactor (2.45 GHz, 300 W) (CEM Corporation, Matthews, NC, USA), equipped with a noncontact infrared temperature sensor. The temperature was controlled by the variable microwave irradiations, and the samples were cooled by the nitrogen current at the end of extraction. The microwave radiation ranged from 60 to 120 W and was adjusted to stabilize the temperatures at 75 • C for different extraction times.
For extraction, 1 g of SCGs were placed in a 35-mL pressure vessel containing 15 mL of solvent (a mixture of water-ethanol at different ethanol concentrations (v/v)). The suspension was irradiated for different periods and at different microwave irradiation powers with stirring. The influence of the solvent mixtures, microwave irradiation powers (60 W and 120 W), irradiation times at the defined temperature (3-6 min), and solvent-to-SCGs ratios on the extraction yield during the extraction process were studied.
The resulting suspension (solvent + extract) was filtered, and the solvent evaporated in a rotary evaporator, Büchi, model R-205 (BUCHI AG, Flawil, Switzerland). The global yield was calculated from the ratio between the mass of extract and the mass of raw material. The extract was stored at −18 • C until use and analyzed for the TPC and DPPH assay.

Antioxidant Activity of DPPH Assay
In our study, the method applied to determine the AA of the extracts was the radical scavenging activity by DPPH. It should be noted that, because the DPPH assay method is usually employed to determine the AA of a sample, a straightforward and direct comparison of the results reported by different authors might be challenging due to dissimilarities in the conditions (reactions and solvents) applied [26].
The DPPH assay of SCG extracts was determined according to the method in the microplates system described previously [27], with modifications. Considering the solvents used in the extractions and the extract solubility, a mixture of ethanol/water (50:50, v/v) was used. Thirty microliters of the SCG extracts dissolved in the solvent mixture were set in a microplate (Nunc) with 270 µL of DPPH solution (100 mM) in the same mixture solvents. The solutions were kept at 298 K in the dark, and the absorbance was measured at 517 nm after 40 min in a microplate reader (BioTek Synergy 2, Winooski, VT, USA) in triplicate.
A calibration curve for trolox was used to express the DPPH assay of the extracts as µmol of trolox equivalents by g of extract (µmol TE/g). The inhibition capacity was determined applying Equation (7): where A S , A b , and A c are the measured absorbance of the sample, the blank with pure solvent, and the control with the solution of DPPH, respectively. All measurements were done in triplicate. The analysis was done by linear regression and using ANOVA [28], presented in Equation (8), where C S is the concentration of the sample in µg/mL, with an R 2 = 0.998. The inhibition concentration to trolox at 50% was IC 50 = 46.60 ± 1.19 µg/mL. The sensitivity was determined according to the limit of detection (LOD = 2.35 µg/mL) and limit of quantification (LOQ = 7.83 µg/mL) [29].

Total Polyphenol Content
The total polyphenol content of the SCG extracts was determined quantitatively using the Folin-Ciocalteu reagent in a microplates system with modifications [27]. In a microplate (Nunc), 30 µL of the extracts (previously dissolved in ethanol/water, 50/50 (v/v)) were mixed with 150 µL of Folin-Ciocalteu reagent, prepared from a 2-N solution, and diluted in distilled water (1:10, v/v). After standing for 4 min, 120 µL of a solution of Na 2 CO 3 (75 g/L) were added, the mixture was left at 40 • C for 30 min, and the absorbance was measured at 765 nm in a microplate reader (BioTek Synergy 2, Winooski, VT, USA) in triplicate. The TPC was expressed as mg of gallic acid equivalents (GAE) by g of extract (mg GAE/g), determined from a calibration curve.
The calibration curve of gallic acid was built following the method described above to the samples in triplicate, and the results are represented by Equation (9), with an R 2 = 0.991. The sensitivity was determined according to the limit of detection (LOD = 2.92 µg) and limit of quantification (LOQ = 9.75 µg).

Analysis of the Extracts by 1 H NMR
The analysis was performed according to the method described before [4], with the necessary adaptations. Briefly, samples of 0.015-0.025 g of extracts dissolved in 500 µL 75-100-mM solutions of CDCl 3 were used for recording the proton NMR spectra. The chemical shifts (δ) for the different components were assigned based on the values reported in the literature for TAGs and 1,2-DAGs [30], caffeine [31], cafestol, 16-O-methylcafestol [32], and kahweol [33].

Analysis of the Fatty Acid Methyl Esters (FAMEs)
A gas chromatographic method was developed to characterize the fatty acid ester profiles of the SCG extracts achieved. The assessments were done regarding the parameters in Annex I to Commission Regulation (EEC) No. 2568/91(1), CELEX_01991R2568 published 12 April 2016, with the required adjustments. The transesterification of the extracts into fatty acid methyl esters (FAMEs) was performed on a methanol solution of KOH (2M).

Design of Experiments
Design of experiments is a fundamental tool for research and development, which provides a way of analyzing areas where the basic knowledge is minimal and needs to be constructed methodically and successfully.
We used a 2 4−1 two-level Fractional Factorial Design with four factors on stage one. In a two-level (2-level) FFD, each experimental factor has only two levels, and the experimental runs include all combinations of these factors. Although 2-level factorial designs are unable to fully explore a wide region in the factor space, they provide useful information for relatively few runs per factor.
Once the most significant independent parameters (factors) identified (ethanol:water ratio and solid/solute ratio) in a stage two Central Composite Design (CCD) was used to examine the effects of those variables concerning their responses and quadratic surfaces to optimize the values with a minimum number of experiments generated [34,35].
An empirical model that correlated the response to the independent parameters using a polynomial equation, like the one given by Equation (10), was applied: a ij X i X j (10) where Y is the predicted response, a 0 -the constant coefficient, a i -the linear coefficients, a ii -the quadratic coefficients, a ij -the interaction coefficients, and X i and X j are the coded values of the independent parameters in the experience.

Conclusions
In this work, two DOE methods-a FFD followed by a CCD-were successfully implemented for the first time in the optimization of MAE of SCGs in terms of the extraction yield, TPC, and AA as the DDPP assay. Furthermore, the composition and fatty acid profile of the extracts recovered were analyzed by NMR and GC-FID.
It was demonstrated that the influence of the factor of the principal significance-Ratio (Eth/water)-on the extraction process responses of yield, TPC, and AA was substantial. While low ratio values (decrease of ethanol percent in the solvent mixture) decreased considerably the yield, it positively affected the TPC and AA of the extracts obtained. The significant impact of the variations of the % of ethanol in the solvent mixture on the compositions of the lipids, caffeine, pentacyclic diterpenes of the kaurene family, and FAME was also confirmed by the NMR and GC-FID analyses of the extracts recovered.