STUDY OF FUNCTIONAL FOODS AND DEVELOPMENT OF NUTRACEUTICALS

STUDY OF FUNCTIONAL FOODS AND DEVELOPMENT OF NUTRACEUTICALS: ANALYTICAL AND APPLICATIVE ASPECTS Classification of Italian propolis by HR-NMR: influence of different production techniques (G. Papotti, D. Bertelli, M. Plessi) INTRODUCTION Propolis is a product based on resins collected from plant exudates by bees and contains more than 160 constituents. The propolis is used by the bees to defend them from the intruders, to prevent decomposition of animals killed after invading the colony, to thermically isolate the beehive, to cracks or openings. The propolis is characterized by a mean content of 50% balsams and resins, 30% waxes, 10% essential and aromatic oils, 5% pollen and 5% various other substances, including organic debris. It usually contains a variety of chemical compounds, such as polyphenols (flavonoids, phenolic acids and their esters), terpenoids, steroids and amino acids. Flavonoids are thought to be responsible for many of its biological and pharmacological activities including anticancer, anti-inflammatory, antimicrobial and antioxidant effects. It is generally accepted and demonstrated that in temperate zones, including Europe, Asia and North America, the bud exudates of Populus species are the main source of propolis, indeed this is the main propolis type also in Italy. 1 The chemical profile of poplar propolis can be characterized by its content of bioactive components. A typical poplar propolis is so charachterized: balsam, minimum 45%; total phenolics, minimum 21%; total flavones and flavonols, minimum 4%; total flavanones and dihydroflavonols, minimum 4%. 2 Smell, color, constitution, and chemical composition of propolis vary not only according to the different botanical sources from which it is collected, but also to the geographical origin and the climatic conditions. Besides also the method of harvest can have some influence on the chemical composition and on the properties of propolis. AIM OF THE STUDY The present study was conducted to evaluate if the HR-NMR used as chemical fingerprint analysis coupled with multivariate statistical methods is able to classify propolis according to the method of harvest and to gain extra knowledge concerning the fingerprint of propolis. Sixty raw propolis collected with different methods were extracted with ethanol and, before performing the NMR experiments, were analyzed for quantification of total phenolics, total flavones and flavonols, total flavanones and dihydroflavonols and balsams, to confirm the poplar source. EXPERIMENTAL RESULTS AND DISCUSSION All the samples were provided by several local beekeepers and by CRA-API (Bologna, Italy) in the summer of 2007 and were As shown in Figure 3, the results are in large agreement Balsams obtained from colonies located in a restricted area of the Emilia Romagna in which it is known that bees produce propolis mainly from poplar. 60 samples collected with different methods were supplied in all: 17 obtained by scraping, which consists with the data reported in literature 2 and it is therefore (B)a TFD in scraping the whole inner surface of the beehive; 26 obtained by wooden wedges, which is obtained through scraping several wooden wedges interposed between the super and the cover of the hive; 17 obtained by plastic nets, in which the beekeeper TFF scrapes the nets interposed between the super and the cover of the hive. confirmed the poplar source of the considered samples. (A)a The (B)a (A)a TP test shows statistical significant differences between the samples (P<0. 05) for TP, TFF and TFD. The post hoc test demonstrates the difference (B)a Propolis Extraction. Ethanol was used since it is the solvent for the most common commercial products obtained by propolis. For each ethanolic extract, 1 g propolis exactly weighed was chopped in to small pieces and extracted with 10 ml of solvent under continuous stirring at room temperature (twice after 24 hours). Each extract was filtered in a volumetric flask and the volume made up to 25 m. L. Phenolic compounds and balsams contents determination. Total phenolics (TP), total flavones and flavonols (TFF) and total flavanones and dihydroflavonols (TFD) contents were estimated using the procedures reported in literature. 2 The Balsam content was evaluated using the method described by Popova et al. 2 Two m. L of each extract were evaporated to dryness under vacuum until constant weight, percentages of balsam were calculated as the ethanol soluble fraction. ANOVA (A)a between scraping and wedges and plastic nets for TP, (A)a TFF and TFD. Considering these results, the NMR a. Results of the HSD test: the same letter in the same column indicates no significant differences (P<0. 05). experiments were performed. Figure 3. Phenolic compounds and balsams mean content 1 H NMR spectra (0 -13. 0 ppm). All the 1 H NMR spectra were compared using the Spectra Positional Matching (Amix 3. 7. 10), to calculate similarities. A match factor of 95. 22% was obtained, which confirm the botanical source of the 1 H NMR 1 H-13 C HMBC Temperature 300 K Data points 16 K 3 K in F 2 (1 H) and 100 in F 1 (13 C) Acquisition time 1. 3665 s 0. 2747 s Delay time 1. 0 s 0. 5 s Numbers of scans 256 32 Dummy scans / 16 Spectral width 5995. 204 Hz 5592. 841 Hz in F 2 (1 H) and 20124. 465 Hz in F 1 (13 C) Total acquisition time 10 min 19 s 82 min and 11 s Table 1. Acquisition parameters of 1 H and 1 H-13 C NMR experiments Sample Preparation for Spectroscopic Analysis. samples. The model obtained (25 variables) by the GDA To prepare NMR samples, 1 m. L of each extract was transferred (Figure 4) is only marginally able to group samples. The to a NMR tube and evaporated to dryness at room temperature first DF explains 68. 2% of the total variance and the under nitrogen. The residue was dissolved in 0. 5 m. L of DMSO- leave-one out crossvalidation shows a predictive capacity of d 6 and TMS was added to calibrate the 1 H-NMR spectra. Each 83. 3%. Considering these results, the next step was to propolis sample was immediately analyzed and used for both 1 D analyze only the spectral region which includes the signals and 2 D NMR experiments. of the main propolis compounds. The NMR experiments were performed at 300 K. The 1 H NMR and 1 H-13 C HMBC acquisition parameters were reported in Table 1 H 1. Tipical 1 H-NMR and 1 H-13 C HMBC spectra of a propolis group the propolis samples in three evident clusters. The first 1 H-NMR. The 1 H DF explains 79. 4% of the total variance and the leave-one NMR out crossvalidation shows a predictive capacity of 96. 7%. spectra were collected in a data set consisting of 16 K Considering the good results obtained, it was verified if spectral variables and 60 samples. The number of further splitting up of the spectrum provided good results. data points was reduced obtaining a data set with 2159 variables and 60 samples. Three other data sets have been prepared, the first one referred to the region between 4. 50 -13. 00 ppm, the second one which contains the signals of aromatic compounds, between 4. 50 -8. 25 ppm (mid-frequency region) and the third one which contains the signals between 8. 25 -13. 00 ppm (high-frequency region) related to the carbonylic and carboxylic protons. The 1 H NMR spectra were also integrated based on the integration pattern obtained by the software Amix 3. 7. 10 (Bruker Biospin GMBH, Rheinstetten, Germany). The integration was performed considering the presence of the signals in the spectra obtained from all the samples. Thirty-seven signals were integrated in all. 1 H-13 C HMBC. To apply chemometric analysis to HMBC spectra, the integration was performed. Volume integration of spectral correlations was calculated by the spectrometer software; all spectra were processed using the same map of regions of interest. 161 regions of interest were defined and used for chemometric analyses. Besides a data set of 112 variables and 60 samples was obtained considering only the spectral region between 4. 50 and 13. 00 ppm Figure 2. 1 H-13 C HMBC spectrum of a propolis extract. NMR spectra (4. 5 -13. 0 ppm). Figure 5 shows the results of the GDA. In this case the model (23 variables) was able to extract are shown in Figures 1 and 2. Spectral Calculation. Figure 1. 1 H-NMR spectrum of a propolis extract. Figure 4. Score plot of the first two canonical functions for the 1 H NMR spectra (spectral region between 0 -13. 0 ppm): group centroids (●). in F 2 (1 H). Statistical Analysis. For the phenolics and balsams contents, the ANOVA was used to evaluate the statistical significance of measured differences between the propolis extracts. The Spectra Positional Matching was performed using Amix 3. 7. 10. To evaluate the most important discriminant variables, the Tukey “Honest Significant Difference test’’ (HSD) (P<0. 05) was used as post hoc test. Before the statistical analyses all data were normalized. Factor analysis (FA) and general discriminant analysis (GDA) were used to achieve a reliable classification of the different propolis samples according to their NMR fingerprint. To perform GDA, a reduction of variables obtained considering only the signals which presented a factorial weight during FA>[0. 8] was necessary. The data set of 37 variables and 60 samples obtained by applying the integration pattern was statistically analyzed by ANOVA, therefore only the stastically significant variables (P<0. 05) were considered and used for the GDA analysis. All statistical calculations were performed using Statistica 6. 1 and SPSS 13. 0 for Windows. References 1. Bankova, V. Chemical diversity of propolis and the problem of standardization. J. Ethnopharmacol. 2005, 100, 114 -117. 2. Popova, M. P. ; Bankova, V. S: ; Bogdanov, S. ; Tsvetkova, I. ; Naydenski, C. ; Marcazzan, G. L. ; Sabatini, A. G. Apidologie 2007, 38, 306 -311. Therefore, the mid-frequency region and the high-frequency Figure 5. Score plot of the first two canonical functions for 1 H NMR spectra (spectral region between 4. 5 -13. 0 ppm): group centroids (●). 1 H NMR spectra (mid and high-frequency regions). The first model (mid-frequency region, 4. 50 -8. 25 ppm) (20 variables) explains 73. 9% of the total variance (DF 1), whereas the second model (high-frequency region, 8. 25 -13. 00 ppm) (22 variables) explains 77. 2% of the total variance (DF 1). The leave-one out crossvalidation show a predictive capacity of 78. 3% for the first model and 63. 3% for the second one. Considering the results obtained, it is possible to affirm that the signals of characterizing compounds for propolis classification according to harvesting methods are present both in the mid and high-frequency regions. In fact, a lot of the bioactive compounds of propolis (flavonoids, phenolic acids and their esters, terpenoids and steroids) provide signals in both the spectral regions. region were statistically analyzed. 1 H NMR spectra analyzed by the integration pattern. After applying the ANOVA test, a data set with 29 variables was obtained. The first DF explains 76. 3% of the total variance, besides the model show a predictive capacity of 61. 7%. The analysis of the propolis extracts provides spectra which are too complicated to be correctly integrated and also the phasing process introduces a source of variability during the spectral calculation. All this explains the importance of directly applying the chemical fingerprint analysis to the spectral signals avoiding other processing procedures. HMBC spectra. After applying the FA, data sets with 96 variables for the complete spectral region and 61 variables for the spectral region between 4. 50 and 13. 00 ppm in F 2 (1 H) were obtained. Figure 6 shows the results of the GDA. With respect to the model obtained by the first data set (complete spectral region), the first DF explains 82. 1% of the total variance and the model show a predictive capacity of 91. 7%. Considering the model obtained by the second data set (spectral region between 4. 50 and 13. 00 ppm in F 2 (1 H)), the first DF explains 63. 6% of the total variance and the model show a predictive capacity of 81. 7%. 1 H-13 C HMBC technique provides useful information from the complete spectral region and furnishes well-defined signals also in the low frequency region (0 -4. 5 ppm), which on the contrary appears not good resolved in the 1 H NMR spectra owing to the presence of confused and overlapped signals. A) 1 H-13 C B) Figure 6. Score plot of the first two canonical functions for 1 H 13 C HMBC spectra: A) complete spectral region; B) spectral region between 4. 5 -13. 00 ppm in F 2 (1 H): group centroids (●). Considering these results, it can be concluded that the use of HR-NMR coupled with an appropriate data processing procedure and multivariate statistical methods enabled to develope sufficiently effective and appropriate models in classifying propolis according to the harvesting methods. It is interesting to note that the best results were obtained using the 1 H NMR which is the simplest and fastest technique.