DETERMINATION OF GEOGRAPHICAL ORIGIN OF GREEN AND ROASTED COFFEE BASED ON SELECTED CHEMICAL PARAMETERS

Coffee is one of the most consumed beverages in the world. Its quality depends on many factors, such as, country of origin, altitude, climate, post-harvesting processing and others. This paper is focused on the possibility to determinate origin of American, African, and Asian coffees based on chemical properties of the final beverage, such as total antioxidant capacity (TAC) measured using DPPH radical, content of chlorogenic acids and caffeine determined by HPLC-DAD. Samples of green and roasted coffee (roasting level medium dark Full City ++) were used. In green samples the highest values of TAC and caffeine were measured in American samples (averagely 93.014 % inhibition of DPPH and 0.854 g.100 g-1 of caffeine respectively), the highest content of chlorogenic acids showed samples from Africa (averagely 5,037 g.100 g-1). In samples of roasted coffees values of TAC decreased by 7, 47 % in Africa samples, by 18,12 % in American, and 13,73 % in samples from Asia. Roasted African coffees showed on average 1.035 g.100 g-1 of caffeine, the highest average was measured in American samples (1.201 g.100 g-1), and lowest Asian samples (1.089 g.100 g-1). Lowest content of CGAs was obtained from African samples (0.595 g.100 g-1), and the higher from American (0.596 g.100 g-1) and African samples (0.6345 g.100 g-1). ANOVA single factor showed significant differences between green samples regarding the TAC and caffeine content. However, content of chlorogenic acids did not show any difference (p-value=0,6809) regarding the geographical origin. Same results were obtained comparing roasted samples.

This research was aimed to evaluate whether chemical parameters such as total antioxidant capacity, caffeine and chlorogenic acids content can be used for the determination of coffee origin.

Extract preparation
Green coffee samples were obtained from Barzzuz s.r.o., a company that is focused on importing and processing of coffee from various well-known producers. Our monitored group of 15 samples were divided into three subgroups based on the declared country of origin. Group one contained five samples from America, group two five samples from Asia, and group three similarly contained five samples from Africa. Subsequently, all of our analyses were done for both green and roasted samples. Samples were roasted at the medium roasting level, Full City ++, using traditional technology in a drum roasting machine with gas heating (Toper TKSMX 10, Turkey; a batch of 1 sample of green coffee was 500 g) under air or nitrogen at 232 °C. The roasting time was in the range between 500 and 840 s. After roasting, coffee beans were cooled in the same atmosphere as was used during roasting. Green coffee beans are characterized by their structure, so we froze them at -80 °C for 30 minutes before the analysis due to easier and better homogenization. We weighed 10 g of roasted coffee using Kern PLJ 600-3-CM (Germany). The coffee sample was then homogenized using an IKA A10 basic (IKA Works, Wilmington, USA) for 30 s, to the required grinding thickness. The homogenized samples were then sieved through a sieve with a mesh diameter of 1 mm. Particle size analysis showed that a sample particle size of 0.23 to 0.5 mm is optimal for both roasted and green coffee use, as a size is commonly used to prepare espresso. For final extract preparation, 7 g of homogenized sample was weighted using Kern 120-5DM (Germany) in glass beakers with a volume of 150 mL and added 120 mL of demineralized water (95 °C), extraction time was 5 min with occasional stirring. Final extracts were filtered through Sartorius filter paper (type Grade-390-quantitative, Germany) and the obtained stock solution was used to determine the TAC using DPPH method according Brandt-Williams, caffeine and chlorogenic acid content by HPLC.

Determination of Total Antioxidant Capacity (TAC) using DPPH radical method according Brandt-Williams
The total antioxidant capacity (TAC) was determined as free radical scavenging assay using DPPH (2,2-diphenyl1-picrylhydrazyl) according to Brand-Williams et al. method with minor modifications Bobková et al. (2020). The amount of 0.025 mg of DPPH was weighed properly, then dissolved with methanol and the volumetric flask with the stock solution was filled up to 100 mL. The stock solution was diluted with methanol (1:9) to obtain absorbance of approximately 0.7. The diluted 3.9 mL DPPH solution was put into the glass cuvettes and the initial DPPH absorbance (A0) was measured at a wavelength of 515.6 nm. Then 100 μL of sample extract was pipetted into cuvette and the mixture was stirred with a glass stick. Absorbance (At) was measured after 10 minutes at 515.6 nm (T80 UV/VIS Spectrometer). The scavenging capacity was calculated using the following equation: . 100

Preparation of coffee extracts for HPLC analysis
Prepared coffee extracts for HPLC analysis were additionally filtered using 25 mm PVDF syringe filters, pore size 0.45 µmm (Agilent Technologies, Germany -Agilent 1260 Infinity equipped with DAD detector (1260 DAD VL+), and subsequently placed in HPLC vials.

HPLC determination of caffeine and chlorogenic acids
Agilent Infinity 1260 (Agilent Technologies GmbH, Waldbronn, Germany) was used for the analysis. The separation itself was performed on a C-18 Poroshell 120 column (150 mm x 3 mm x 2.7 μm) (Agilent Technologies, Waldbronn, Germany). The mobile phases were as follows: acetonitrile (A) and 0.1% H3PO4 in ddH2O (v/v) (B), the separation being carried out by gradient elution according to the following program: 0-1 min. isocratic elution (20% A + 80% B), 1-5 min. linear gradient elution (25% A + 75% B), 5-15 min. (30% A + 70% B), 15-25 min. (40% A + 60% B). The equilibration time was 3 min. before the next injection. Flow was mobile phase was 1 mL.min -1 , the injection was 10 μL and the separation temperature was set at 30 °C, the samples were stored at 4 °C prior the analysis. The detection wavelengths were set as follows: 276 nm (caffeine) and 320 nm (chlorogenic acids), with a total of data collected in the wavelength range 210-400 nm. According to the UV spectrum and retention times content of chlorogenic acids is expressed as sum of abovementioned acids.

Statistical Analysis
For the summarizing and describing of our results descriptive statistic was used. Parameters such as minimum, maximum, arithmetic means, and standard deviation were used for interpreting the results. For the determination of significant differences between selected geographical origin, regarding the TAC, caffeine and content of chlorogenic acids, ANOVA single factor was used separately for both green and roasted coffee. This statistical analysis was performed using Microsoft Office Excel 2016 for Windows. Moreover, for an appropriate graphical representation of results, Quadratic Discrimination Analysis (QDA) was performed using XLSTAT Excel for Windows.

RESULTS AND DISCUSSION
To evaluate the total antioxidant capacity of green and roasted coffees, we, similarly to Patay et al. (2016), used the method by Brand-Williams (1995) using the DPPH radical, which consists in the reaction of antioxidants with the stable radical 2,2-diphenyl-1-picrylhydrazyl (DPPH) in methanol dilution. It is well known that the antioxidant activity of green coffee is related to the content of chlorogenic, ferulic, caffeic and n-coumaric acids contained in it, and their content depends mainly on the type of coffee (C. arabica, C canephora) and also of origin (Stalmach et al., 2006). Among geographical group of green samples from America, Africa and Asia, the highest values were measured in samples from America, ranged from 91.89 % to 93.96% inhibition of DPPH. Lower values were obtained from coffee samples from Africa, ranging from 81.62 % to 91.52%. The lowest values were measured in Asian coffees, and ranged from 83.81% to 91.63%. Moreover, using ANOVA single factor, and by further comparing the values of total antioxidant capacity of green coffee samples, we proved statistically significant differences between places of coffee origin. Based on results showed in Table 1, is clear that there are statistically significant differences between samples (marked as upper index a and b in Table 1) from Africa and America and America and Asia regarding the TAC. Nevertheless, there is no significant difference between green samples from Asia and Africa regarding the total antioxidant capacity (Table 1). Chlorogenic acids are the main polyphenolic compound with antioxidant properties (Monteiro, 2005; Mullen, 2013), According to Doo and Maskarinec (2014), chlorogenic acids are found in amounts from 70 to 350 mg per cup (approx. 140 mL) and can be the cause of the beneficial effects of coffee. Green coffee bean can contain up to 5 chlorogenic acid derivatives, caffeoylquinic acid (CQA), dicafeoylquinic acid (diCQA), feruoylquinic acid (FQA), pcoumaroylquinic acid (CoQA) and caffeeoferluoylquinic acid (CFQA). Narita and Inouye (2011) reported that green coffee beans of Coffea arabica contain approximately from 3.5 to 7.5% of chlorogenic acids and their derivatives. Farah and Lima (2019) added that beverages prepared from various cultivars of Coffea arabica can contain 4-6 g CGA. 100 g -1 . Which is in accordance with our measured values. Highest values of chlorogenic acids were measured in samples from Africa, ranged from 3.89 to 5.79 of g CGA. 100 g -1 , on the other hand, the lowest were found in samples from Asia (from 4.507 to 4.913 g CGA. 100 g -1 ). However, based on our results is clear that there is no significant difference between samples from Asia, America and Asia regarding the content of chlorogenic acids. Another measured parameter in green samples was caffeine content. According to Belitz et al (2009), caffeine (1,3,7-trimethylxanthine) is due to its physiological effect the best-known nitrogen compound of coffee. The average caffeine content in beans of Coffea arabica is 0.8 -1.4%, which is roughly 1.0 -1.5 g.100g -1 , depending of Arabica variety. Highest amount of caffeine was recorded in samples from America (0,761 to 0,890 g.100 g -1 ). In African coffee, values ranged from 0,7617 to 0,8923 g.100 g -1 of caffeine. On the other hand, Asian coffees showed averagely lowest values, ranging from 0,7264 to 0,7969 g.100 g -1 . Based on results shown in Table 1, we observed a significant difference between samples from Africa and America regarding the caffeine content, however, we are not able to differentiate between samples from Asia and Africa just based on this parameter. In samples of roasted coffees values of TAC decreased by 7, 47 % in Africa samples, by 18,12 % in American, and 13,73 % in samples from Asia. By comparing values of total antioxidant capacity of roasted coffee samples we proved statistically significant differences between places of coffee origin, comparing samples from Africa and Asia, and the biggest difference of variances is clearly visible between these two groups. On the contrary though, by comparing Africa vs America and America vs Asia, we can see differences in their variance. However, statistical analysis shows that those differences are not significant, but p-value of both mutual comparisons are approaching the set α of 0,05 of used test. These results are a hint that TAC content can be used as a tool to separate coffee samples by their geographical origin (Table 2). ANOVA test of chlorogenic acids showed that we accept null hypothesis, there is no difference among groups of samples. Average values of CGA is relatively even among all groups with very small standard deviation of whole groups, and very small variance within and among geographical groups. Similarly, as in green samples, this parameter did not show a significant difference among groups (Table 2).
In all our samples of coffee roasted at roasting level Full City ++ (232 °C), we found a slight increase in the level of extractable caffeine compared to green coffee samples. These findings are in accordance with Farah (2012) so that the caffeine content in roasted coffees can ranged from 1.1 to 1.3 g.100 g -1 , and the author also noted a slight increase in the caffeine content compared to green samples. Possible, although minimal changes in caffeine content is justified by the fact that caffeine does not change significantly during roasting, but a slight decrease in its content may be due to sublimation (overheating) and vice versa, its increase may be observed due to decline of other components. Content of caffeine seems to be the best factor for separating groups based on geographical origin in samples of roasted coffee given p-value among groups is 0.0083 which is highly significant. Based on comparations between pairs we are able to separate Africa and America samples, America and Asia samples. However, Africa and Asia samples showed p-value over 0.05, so we, again, accepting null hypothesis (Table 2). For multi-component analysis QDA was used. Based on Figure 1it is obvious that selected parameters (variables) are almost equally suitable and affect the possible determination of coffee origin. Wilks' Lambda test (Rao's approximation) showed significant differences among groups (p-value=0.026). Confusion matrix for the cross-validation results showed that we were able to determine the origin of green American coffee with accuracy of 100 % based on total antioxidant activity, caffeine and chlorogenic acids content, nevertheless, regarding the other two geographical groups, more parameters, or samples respectively, would be needed to achieve higher percentage accuracy.  Figure 2 is obvious that TAC and caffeine content appears to be more suitable parameters and for determination of geographical origin of roasted coffee more than content of chlorogenic acids. This may be caused by the fact that CGAs are thermally unstable, therefore are degraded during roasting to produce products that contribute to the formation of coffee aroma, the final taste of the drink and according to Jeszka-Skowron et al. (2016) their levels in the coffee beverage also depend on the degree of roasting. Coffee roasted to dark roasting level can containing less than 10% of their original content. On the contrary, caffeine is more stable and even though some of antioxidants are degraded during roasting, this decrease is compensated by formatting of products of Maillard reaction, which show antioxidant properties too. Wilks' Lambda test (Rao's approximation) showed significant differences among groups (p-value=0.002).

CONCLUSION
This research suggested that certain chemical parameters may be used for the determination of geographical origin of both green and roasted coffee samples. Using Multi-component QDA analysis was observed that in green coffee samples TAC, and both caffeine and chlorogenic acids contend almost equally affected the determination among selected groups, on the other hand regarding the roasted samples, content of chlorogenic acids may be a less effective parameter. It is essential to add that even though we have been able to determine significant difference between American, African and Asian coffees, further research that would cover more samples or chemical parameters such as total polyphenol content or content of certain volatile substances, respectively, is needed to achieve higher accuracy in determination of geographical origin of coffee.