Article. Discrimination of Commercial Roasted and Ground Coffees According to Chemical Composition. Romilaine M. N. de Souza and Marta T.

Transcription

1 Article J. Brz. Chem. Soc., Vol. 23, No. 7, , Printed in Brzil Sociedde Brsileir de Químic $ A Discrimintion of Commercil Rosted nd Ground Coffees According to Chemicl Composition Romiline M. N. de Souz nd Mrt T. Benssi* Deprtmento de Ciênci e Tecnologi de Alimentos, Universidde Estdul de Londrin, CP 6001, Londrin-PR, Brzil Cfés torrdos e moídos comerciis (38 mostrs) e cfés com espécie conhecid (rábic e robust) form crcterizdos por nálise de componentes principis, utilizndo-se como vriáveis, ácido nicotínico, trigonelin, ácido 5-o-cfeoilquínico (5-CQA), cfeín, cveol e cfestol, potencilmente indicdors ds espécies. O estudo objetivou vlir importânci desses prâmetros n discriminção entre cfés. Ácido nicotínico foi eficiente n crcterizção d torr. Trigonelin e 5-CQA destcrm-se pel vribilidde, tnto entre s mostrs purs, como nos cfés comerciis. O mior potencil de discriminção entre s espécies foi tribuído os prâmetros termoestáveis (cfeín, cveol e cfestol). No gerl, miores teores de cfeín e menores de diterpenos (cveol e cfestol) form relciondos proporções de robust mis elevds no produto, consttndo-se um tendênci de diminuição n rzão cveol/cfestol e umento n rzão cfeín/cveol. Esses dois novos prâmetros (rzões cveol/cfestol e cfeín/cveol) form propostos como ferrments pr vlição d dição de robust em cfés comerciis. Rosted nd ground 38 commercil coffees nd coffees of known species (rbic, robust) were chrcterized by principl component nlysis using s vribles nicotinic cid, trigonelline, 5-o-cffeoylquinic cid (5-CQA), cffeine, khweol nd cfestol, which re potentilly indictive of species. The objective of the study ws to ssess the relevnce of such prmeters in coffee discrimintion. Nicotinic cid llowed the chrcteriztion of rosting degree. Trigonelline nd 5-CQA presented vribility mong rbic nd robust coffees s well s mong commercil ones. Thermostble prmeters (cffeine, khweol nd cfestol) hd high discrimintive potentil between the species. In generl, high levels of cffeine nd low levels of diterpenes (khweol nd cfestol) were relted with higher proportions of robust in the products, which were observed by the decresing khweol/cfestol rtio nd incresing cffeine/khweol rtio. The use of these new prmeters (khweol/cfestol nd cffeine/khweol rtios) ws suggested s tools for ssessing the ddition of robust in commercil coffees. Keywords: rbic, robust, pc, khweol/cfestol, cffeine/khweol Introduction Coffee is one of the min food commodities in the worldwide economy, nd it is the most consumed beverge fter te. 1 Brzil is the world leder in the production nd exporttion of coffee 2 becuse Brzil is hevy producer of both Coffe rbic (rbic) nd Coffe cnephor (robust) species. 3 Arbic nd robust coffees differ considerbly in price, sensoril qulity nd cceptnce. 2,4 In generl, rbic coffee hs higher concentrtion of crbohydrtes, lipids nd trigonelline in the green nd rosted products, providing superior sensoril cup qulity, *e-mil: mrttb@uel.br nd robust coffee hs high levels of cffeine. 5-8 Conversely, robust coffee is widely ccepted in mrket due to its low cost, nd it is considerbly used in blends with rbic. 2 In Brzil, more thn 80% of the commercil rosted coffee re blends of the C. rbic nd C. cnephor species, which offer low cost product nd beverge djusted to the preference or hbits of consumers. 2,4 The Brzilin legisltion provides criteri for coffee qulity clssifiction with indicted proportions of robust used in the blends. 9 The green bens of C. rbic nd C. cnephor hve different colors, shpes nd sizes. After rosting nd grinding, however, such species cnnot be visully distinguished, which is why it is necessry to use chemicl prmeters for their discrimintion. As they belong to

2 1348 Discrimintion of Commercil Rosted nd Ground Coffees According to Chemicl Composition J. Brz. Chem. Soc. the sme genus, rbic nd robust coffees hve little differences in their compositions. There is no consensus on which compounds should be employed for detecting the ddition of the lower vlue robust coffee to rbic coffee in commercil blends. 10 Severl components could be used s indictors or discrimintors for the rbic nd robust species. However, their concentrtions generlly depend not only on the species but lso on the geogrphic origin, presence of defective bens nd rosting degree. This vribility nd the lrge mount of compounds studied require the use of multivrite sttisticl techniques, such s principl component nlysis (PCA), to ssess the dt. Chemicl constituents proposed s possible discrimintors include wter-soluble (cffeine, trigonelline, nicotinic nd chlorogenic cids) nd ft-soluble compounds (khweol nd cfestol). 7,15,16 Cffeine is stble to the rosting process. The highest cffeine contents hve been reported for robust coffee nd coffees with smll proportions of defective grins. 6,11,13,14,17-21 The rosting process promotes the degrdtion of trigonelline nd chlorogenic cid (CGA), 19,22 cusing vrition in their concentrtions depending on the presence nd on the type of defective bens present. 20,23 Higher levels of trigonelline nd nicotinic cid (product of the trigonelline degrdtion) hve been reported for rosted rbic coffee, nd higher levels of CGAs hve been reported for rosted robust coffee. 11,14,17,18,24 Cfestol is present in both species. However, khweol hs been reported to be specific to rbic coffees. 8,14,16,25 Furthermore, diterpenes present reltive stbility to het tretment, nd their levels re only slightly ffected by the presence of defective grins. 16 The quntittive determintion of such compounds in commercil blends llowed us to ssess the importnce nd vibility of these compounds s tool for identifying species in different blends of rosted coffee. 11,14,16,26,27 Hence, the im of this work ws to chrcterize rosted nd ground commercil coffees (38 smples) nd coffees of known species (rbic nd robust) through multivrite nlysis using compositionl vribles (nicotinic cid, trigonelline, 5-o-cffeoylquinic cid (5-CQA), cffeine, khweol nd cfestol) nd ssessing the importnce of such prmeters in the discrimintion of coffee products. Experimentl Commercil rosted nd ground coffees (38 smples) cquired from the locl mrket or supplied by industries were studied. Twenty-three brnds from 13 industries ssocited to the Brzilin Coffee Industry Assocition (ABIC) were ssessed. All smples were sifted through 0.84 mm mesh for stndrdiztion. The smples were commercilized under different denomintions s follows: trdicionl (trditionl), forte (strong), extr forte (extr strong), gourmet, premium, rlto (high ltitude), exportção (export) nd espresso (express). Despite the denomintion, ll coffees presented lightness between 19 nd 21 indicting medium/drk rost degree. Bsed on the lbeling informtion, only five of the coffees (less thn 1%) contined only rbic coffee (gourmet), nd the other smples hd the possibility of n ddition of robust up to 30% ccording to the Brzilin legisltion. 9 Pure smples of rbic (A) nd robust (R) coffees were used s references for the comprison with the commercil products. They were provided by the Cfé Iguçu Compny (Cornélio Procópio-PR, Brzil) nd Instituto Agronômico do Prná (IAPAR, Londrin-PR, Brzil). The coffee bens hd distinct geogrphic origins nd different qulity levels, which resulted in high heterogeneity mong the smples. The R1 nd R2 robust coffees were from Rondôni nd Espírito Snto Sttes, respectively, nd R1 hd higher number of defective bens thn R2. R3 ws clssified s type 4 coffee ccording to officil dt. 28 The A1 rbic coffee from Mins Geris Stte hd less defects thn A2, which ws from Prná Stte. The following three rbic coffees of known vrieties were lso studied: A3 (IAPAR-59), A4 (Ctuí) nd A5 (Mundo Novo). The coffees were processed to chieve rosting degree similr to commercil smples (Tble 4). The smples were pcked in plstic bgs nd were stored t 10 ºC until nlyses (color, moisture, wter-soluble compounds nd ft-soluble compounds), which were performed in duplicte. Chromtogrphic nlysis Potssium hydroxide (Vetec, Rio de Jneiro, Brzil), tert-butyl methyl ether (Acrós Orgnics, NJ, USA) of nlyticl grde, cetonitrile (J. T. Bker, USA) nd cetic cid (J. T. Bker, USA) of chromtogrphic grde were used s solvents for the extrction nd preprtion of the mobile phse. The wter ws obtined from Milli-Q system (Millipore, USA), nd the mobile phses were filtered (0.45 mm; Millipore). Anlyticl grde stndrds of cffeine (1,3,7-trimethylxnthine; Acrós Orgnics, NJ, USA), nicotinic cid (n-methylnicotinmide; Vetec, Rio de Jneiro, Brzil), chlorogenic cid (1,3,4,5-tetrhydroxycyclohexne-crboxylic cid; 5-CQA) nd trigonelline (1-methylpyridine-3-crboxylte monohydrte; Sigm, Steinhelm, Germny) were

3 Vol. 23, No. 7, 2012 Souz nd Benssi 1349 employed. Khweol nd cfestol stndrds from Axxor (Sn Diego, USA) were kept in freezer t 18 ºC. The wter-soluble compounds were determined by extrcting the smples ( g) with n cetonitrile:wter solution (5:95; v/v) t 80 ºC for 10 min. The smples were then filtered (7 mm) into ml volumetric flsk. An liquot (5.0 ml) of this solution ws trnsferred to volumetric flsk (25.0 ml). For the nlysis, Spherisorb ODS-1 column (250 mm 4.6 mm; 5 µm) (Wters, Milford, USA) nd grdient of 5% cetic cid nd cetonitrile were employed. The detection ws crried out t 260 nm (nicotinic cid nd trigonelline), 272 nm (cffeine) nd 320 nm (5-CQA). 26 For the ft-soluble compounds, the smples ( g) were sponified with 2.0 ml of 2.5 mol L 1 potssium hydroxide (in 96% ethnol) t 80 ºC for 1 h. After dding 2.0 ml of distilled wter, the unsponifible frction ws extrcted with 2.0 ml of tert-butyl methyl ether. This frction ws gitted nd centrifuged for 3 min t 3,000 rpm, nd the orgnic phse ws collected. The extrction procedure ws repeted three times. To clen up, 2.0 ml of distilled wter ws dded to the extrct. After homogenizing nd discrding the queous phse, the extrct ws left t 70 ºC until complete dryness. The ether extrct ws resuspended in 4.0 ml of mobile phse (55% cetonitrile in wter). The Spherisorb ODS-1 column with the sme specifictions s used before nd n isocrtic elution with n cetonitrile:wter solution (55:45; v/v) were employed. The detection ws crried out t 230 nm nd 290 nm for cfestol nd khweol, respectively. 27 Before the injection, the smples were filtered (0.45 mm; Millipore, USA). The identifiction of compounds ws bsed on the retention times, UV-Vis (DAD) spectrum nd spiking. The quntifiction ws crried out through externl stndrdiztion with t lest five concentrtions in triplicte. For the clcultion of the concentrtions in dry bse, the moisture (g per100 g) ws determined (105 ºC for 7 min) using infrred equipment (OHAUS-MB200, USA). Sttisticl nlysis The results were nlyzed by principl component nlysis (PCA) using Sttistic 7.1 softwre. 29 Results nd Discussion Tbles 1 nd 2 show the contents of wter-soluble (nicotinic cid, trigonelline, 5-CQA nd cffeine) nd ft soluble (khweol nd cfestol) compounds in the robust nd rbic coffees. The sme prmeters were individully described for ech commercil smple in preliminry study 30 nd re summrized in Tble 3. The robust smples (R1, R2 nd R3) hd the highest contents of cffeine (pproximtely 2.00 g per 100 g) nd lowest levels of cfestol (between 0.19 nd 0.24 g per 100 g). Moreover, khweol ws not detected in the robust smples (Tble 1). The rbic coffees (A1 to A5) hd the highest levels of khweol (0.82 g per 100 g) nd the lowest concentrtions of cffeine (between 1.10 nd 1.29 g per 100 g) (Tble 2). The cffeine contents were similr to the contents observed for the gourmet coffees (between 0.99 nd 1.29 g per 100 g) 30 nd lmost hlf of the cffeine content found in the robust smples (Tble 1). The behvior observed for the rbic nd robust smples ws in ccordnce with the literture, which describes higher levels of trigonelline nd nicotinic cid for rbic coffee. 14,17 In green nd light rosting coffee, 5-CQA is predominnt in robust. 11,14,18,24 However, s the compound hs higher susceptibility to the het process in Tble 1. Levels of wter-soluble nd ft-soluble compounds (g per 100 g of smple in dry bse) in ddition to khweol/cfestol nd cffeine/khweol rtios found in pure smples of robust coffee Prmeters R1 R2 R3 Men ± SD (CV / %) Nicotinic cid 0.02 ± ± ± ± 0.00 (17) Trigonelline 0.26 ± ± ± ± 0.14 (63) 5-CQA 0.24 ± ± ± ± 0.15 (72) Cffeine 2.01 ± ± ± ± 0.03 (2) Khweol n.d. n.d. n.d. n.d. Cfestol 0.24 ± ± ± ± 0.03 (13) Khweol/Cfestol Cffeine/Khweol Men of two repetitions ± stndrd devition (SD); CV: coefficient of vrition; n.d.: compound not detected under the conditions of nlysis; (-): not clculted.

4 1350 Discrimintion of Commercil Rosted nd Ground Coffees According to Chemicl Composition J. Brz. Chem. Soc. Tble 2. Levels of wter-soluble nd ft-soluble compounds (g per 100 g of smple in dry bse) in ddition to the khweol/cfestol nd cffeine/khweol rtios found in pure smples of rbic coffee Prmeters A1 A2 A3 A4 A5 Men ± SD (CV / %) Nicotinic cid 0.03 ± ± ± ± ± ± 0.01 (20) Trigonelline 0.40 ± ± ± ± ± ± 0.20 (42) 5-CQA 0.24 ± ± ± ± ± ± 0.13 (46) Cffeine 1.29 ± ± ± ± ± ± 0.09 (7) Khweol 0.80 ± ± ± ± ± ± 0.10 (13) Cfestol 0.42 ± ± ± ± ± ± 0.08 (21) Khweol/Cfestol 1.90 ± ± ± ± ± ± 0.73 (31) Cffeine/Khweol 1.61 ± ± ± ± ± ± 0.27 (18%) Men of two repetitions ± stndrd devition (SD); CV: coefficient of vrition. Tble 3. Levels of wter-soluble nd ft-soluble compounds (g 100 g 1 of smple in dry bse) in ddition to khweol/cfestol nd cffeine/khweol rtios found in commercil rosted nd ground coffees Prmeters Men ± SD (CV / %) Rnge of concentrtion Nicotinic cid 0.03 ± 0.00 (15) Trigonelline 0.49 ± 0.17 (34) CQA 0.43 ± 0.22 (50) Cffeine 1.47 ± 0.25 (17) Khweol 0.47 ± 0.20 (43) Cfestol 0.42 ± 0.06 (15) Khweol/Cfestol 1.07 ± 0.38 (35) Cffeine/Khweol 4.19 ± 3.29 (79) Men ± stndrd devition (SD) of 38 smples; CV: coefficient of vrition; individul vlues for ech product previously described by De Souz et l. 30 the robust mtrix, higher contents of 5-CQA is observed in drk rost rbic coffee. 14,31 Robust coffee hs higher levels of cffeine, lower levels of cfestol nd n bsence of khweol when compred to rbic coffee. 6,8,11,13,14,16-19,25 Considering tht ech brnd cn dopt different blend, the differences mong the rw mteril, especilly the species employed (rbic nd robust), in ddition to the differences of the rosting degrees nd defects, generted vribility between the smples for ll compounds studied. The prmeters tht hd less vrition mong the products were cffeine, nicotinic cid nd cfestol with coefficient of vrition (CV) less thn 20%. Trigonelline, khweol nd 5-CQA were highly vried mong the products (CV from 34 to 50%) (Tble 3). Trigonelline nd 5-CQA, however, hd wide vribility even mong the pure species: CV from 63 to 72% for robust (Tble 1) nd from 42 to 46% for rbic (Tble 2). Thus, it ws not possible to ttribute the vribility of these compounds only to the vrition between the species becuse even slight chnge in the rosting process could result in n expressive vrition of some of the compounds. Principl component nlysis (PCA) ws employed to ssess the discrimintion cpbility of the following vribles: nicotinic cid, trigonelline, 5-CQA, cffeine, khweol nd cfestol (Figure 1). The first two principl components (PC) ccounted for 71% of the vrince observed (Figure 1). Equtions 1 nd 2 were obtined by the lodings of the PCs, nd they generted the coordintes of the principl component grphic (Figure 1b). PC1 = nicotinic trigonelline chlorogenic cffeine khweol cfestol (1) PC2 = nicotinic trigonelline chlorogenic cffeine khweol cfestol (2) The min prmeters of PC1 were s follows: cffeine (negtive correltion), nd cfestol, khweol, 5-CQA nd trigonelline (positive correltion) (Figure 1). Nicotinic cid (positive correltion) ws the prmeter of more relevnce in PC2. Trigonelline nd 5-CQA (negtive correltion) lso contributed to PC2. Thus, there ws evidence tht PC1 ws ssocited with the rw mteril employed, becuse PC1 ws correlted with compounds tht presented lrge vribility mong the species (Tbles 1 nd 2). Considering the correltion of PC2 with compounds formed (nicotinic cid) or degrded (5-CQA nd trigonelline) in the rosting process, PC2 contributed to the rosting chrcteriztion. In generl, the coffees were within medium/drk rosting rnge, which ws indicted by the vlues of

5 Vol. 23, No. 7, 2012 Souz nd Benssi 1351 Figure 1. Projection of the vribles () nd smples (b) in the principl components (PC1 nd PC2). R1, R2 nd R3 (robust); A1, A2, A3, A4 nd A5 (rbic); numbers 1 to 38 (commercil smples). color (L* nd H*) nd weight loss percentge estimted ccording to Dis 14 (Tble 4). tht re configured on the top of the grphic in Figure 1b hd higher levels of nicotinic cid nd lower concentrtions of 5-CQA nd trigonelline s compred to the other coffees, indicting more ggressive rosting Tble 4. Rosting ssessment of commercil, rbic nd robust coffees expressed s weight loss percentge (WL%) nd prmeters of color (lightness, L*, nd hue, H*) b Prmeters Commercil Arbic Robust L* ± ± ± 3.79 H* ± ± ± 1.89 WL% ± ± ± 1.90 Estimted ccording to the following eqution: WL% = SUM (nicotinic cid + trigonelline) RATIO (5-CQA/cffeine) 0.13 L* 0.06 H* (Dis 14 ); b men ± stndrd devition of 38 commercil coffees, 5 rbic coffees nd 3 robust coffees. process. For exmple, there were estimted weight losses of 17 nd 18% for smples 32 nd 16, respectively, ccording to Tble S1 (Supplementry Informtion (SI))., such s coffees 20 nd 38, which re locted on the bottom of the grphic in Figure 1b, hd 14% weight loss indicting light rosting degree (Tble S1 in the SI section). Considering PC1, smples shown on the right side of the grphic in Figure 1b hd higher levels of diterpenes (khweol nd cfestol), 5-CQA nd trigonelline nd lower concentrtion of cffeine s compred to the other coffees. The stndrd rbic coffees (A1, A2, A3, A4 nd A5) nd the gourmet smples (3, 5, 15, 21 nd 35) were plced in this region, which ws chrcteristic of rbic coffee, nd these coffee smples hd different behvior from the other smples. 30 In generl, the pure smples were seprted into the following two groups: (i) robust coffee nd (ii) rbic coffee. The blended smples were well distributed in the intermedite spce where the upper right region ws reserved for smples close to rbic coffee nd the lower left region for those close to robust coffee. The robust coffees were better grouped nd seprted from the other smples. The rbic coffees hd vried chrcteristics, which indicted tht differences not ssocited with the species (vriety, soil fctors, climte fctors, defects nd rosting process) influenced the smples composition, hindering more ccurte species chrcteriztion. Tking into ccount tht the rosting process ws poorly differentited (Tble 4) nd ssessing the projection of the vribles, smples configurtion (Figure 1) nd the dt shown in Tbles 1, 2 nd 3, it ws concluded tht the commercil coffee with high proportion of C. rbic species should hve high levels of diterpenes (notbly khweol), trigonelline nd 5-CQA. Therefore, the ddition of robust coffee to the product should decrese the levels of such compounds nd rise the level of cffeine. The highest concentrtions of diterpenes, trigonelline nd 5-CQA nd the lowest concentrtion of cffeine found in the gourmet smples indicted higher probbility for the presence of pure rbic coffee. The low levels of such compounds, which hve lso been observed in some trditionl smples, 30 could be ttributed to different proportions of robust coffee in the products. Considering studies tht hve pplied multivrite techniques to discriminte species, some were limited to the ssessment of green coffees 32 or to pure species. 33 Working with rosted coffees, some uthors hve chrcterized cffeine, trigonelline, 5-CQA 11,13,14 nd khweol 11,14 s vribles responsible for the seprtion of species nd blends. According to Csl et l., 13 nicotinic cid does

6 1352 Discrimintion of Commercil Rosted nd Ground Coffees According to Chemicl Composition J. Brz. Chem. Soc. not hve significnt importnce in the distinction between the species. Dis 14 hs reported tht nicotinic cid nd trigonelline cn be used for the discrimintion of species for smples with light nd drk rosting degrees but not with smples of medium rosting degree. In generl, higher levels of cffeine nd lower levels of khweol, trigonelline nd 5-CQA were relted to elevted proportions of robust. Importntly, these studies were crried out with coffees of known origin, nd comprison mong the stndrd nd commercil smples ws not found in the literture. The use of models to estimte the percentge of robust coffee (R%) dded to rbic coffee hs been reported in the literture. Dis 14 proposed the use of multiliner models for wter-soluble compounds nd color prmeters (R% = H* trigonelline cffeine), nd Cmpnh 16 proposed the use of models for khweol nd cfestol in drk rosting coffee (R% = khweol cfestol). These models were pplied to the studied commercilized smples (trditionl nd gourmet) for ssessing the probble proportion of robust coffee. In order to observe the efficiency of the models, they were lso pplied to rbic nd robust coffees. The best results were observed for smples of 100% robust coffee, which hd men estimtive of robust coffee between 97 nd 113%. A huge vribility of results nd low concordnce mong the models were shown for rbic coffees, which presented since negtive vlues ( 20%) in the estimtive bsed on diterpene content until positive vlues (16%) in the model bsed on the wter-soluble compounds (Tble 5). Although considering the possibility of underestimtion nd overestimtion of the vlues nd the error in ech modeling, the gourmet coffees, which should be comprised of only rbic species, 9 hd low percentges of robust coffee in both models (7 nd 8%, Tble 5). Concerning the trditionl coffees, both models predicted high men percentge of robust (50%), which ws bove the limit llowed by the legisltion (30%). 9 Depending on the composition of the smples, the models were more or less consistent. Regrding coffee smples 23 nd 27, which were plced close to the robust species in the PCA grphic (Figure 1b), higher percentges of robust coffee were estimted by both models (between 82 nd 98%). These two smples hd low levels of khweol (0.19 nd 0.10 g per 100 g, respectively) nd high concentrtions of cffeine (1.79 nd 2.02 g per 100 g, respectively), effectively showing similr behvior to robust coffees (Tble 1) nd no similrity with profile of rbic coffee (Tble 2). On the other hnd, for coffee smples 29, 30 nd 31, which were plced t the center of the PCA grphic (Figure 1b), low vlues in the Dis 14 model (R% from 15 to 11%) nd high vlues in the Cmpnh 16 model (R% from 67 to 80%) were observed. Such divergence occurred becuse these smples hd both low levels of cffeine (between 1.05 nd 1.19 g per 100 g) nd khweol (between 0.13 nd 0.27 g per 100 g). 30 These smples did not present typicl profile of either robust or rbic coffee (Tbles 1 nd 2), indicting possible ddition of n dulternt different of robust coffee. Thus, models tht seprtely study the wter-soluble nd ft-soluble vribles my not be efficient to determine the percentges of robust coffee in commercil smples becuse of the possible ddition of other dulternts to the coffee blends. In the present study, trigonelline nd 5-CQA were relevnt prmeters in PC1, but they hd the highest vribility in commercil (Tble 3), pure rbic nd pure robust smples (Tbles 1 nd 2). These prmeters my be correlted with the vrition mong the species. However, they my lso be ssocited with other chnges, such s the rosting, processing nd the presence of defective bens. Thus, prmeters less sensitive to rosting (cffeine, khweol nd cfestol) presented the highest potentil of discrimintion mong the smples. Considering tht wter-soluble (cffeine) nd ft soluble compounds (khweol nd cfestol) re mong the most importnt vribles in species discrimintion nd tht the models previously mentioned 14,16 were restricted in the estimtion of robust in commercil coffees due to the seprte use of the wter-soluble nd ft-soluble vribles, we proposed the use of combined prmeters. The use of reltions between such ft-soluble nd wter- Tble 5. Assessment of the percentge of robust coffee (R%) in commercil (trditionl nd gourmet), rbic nd robust coffees by the mthemticl models proposed by Dis 14 nd Cmpnh 16 Models Trditionl Gourmet Arbic Robust Dis 14 b 49 ± 27 ( 15-93) 8 ± 14 ( 2-32) 16 ± 20 ( 7-43) 113 ± 5 ( ) Cmpnh 16 c 50 ± 25 ( 4-98) 7 ± 15 ( 7-32) 20 ± 22 ( 47-5) 97 ± 3 (95-100) Men ± stndrd devition of 33 trditionl coffees, 5 gourmet coffees, 5 rbic coffees nd 3 robust coffees; b estimted ccording to the following eqution: R% = H* trigonelline cffeine; c estimted ccording to the following eqution: R% = khweol cfestol.

7 Vol. 23, No. 7, 2012 Souz nd Benssi 1353 soluble vribles (khweol/cfestol nd cffeine/khweol), proposed by the first time in the literture, would llow more brod detection of differences. In this study, the concentrtions rnged from 1.10 to 1.29 g per 100 g for cffeine, from 0.66 to 0.93 g per 100 g for khweol nd from 0.28 to 0.48 g per 100 g for cfestol feturing tipicl rbic coffee. It corresponded to khweol/cfestol rtio (k/c) vrition from 1.73 to 3.40 nd cffeine/khweol rtio (cf/k) vrition from 1.18 to 1.91 (Tble 2). Bsed on these dt nd previous reports 14,16,34 regrding the vribility in coffee composition, k/c rtio bove 1.00 is indictive of rbic coffee. Regrding the cf/k rtio, vlues between 1.00 nd 3.00 re indictive of rbic coffee, nd cf/k vlues greter thn 4.00 re indictive of robust coffee. Thus, the ddition of robust coffee should decrese the k/c rtio nd increse the cf/k rtio. In ddition to robust coffee, the ddition of other dulternts (e.g., mize or husk) or even the presence of defects would result in decrese in the correltion between cffeine nd khweol. However, the correltion between khweol nd cfestol should remin unchnged becuse diterpenes re exclusive to coffee nd do not present vrition with defect levels. 16 A lrge vribility for the rtios ws observed in the commercil products. The k/c rtios were between 0.27 nd 1.82 (verge of 1.07 ± 0.38; CV of 35%), nd the cf/k rtios were between 1.34 nd (verge of 4.19 ± 3.29; CV of 79%) (Tble 3). The gourmet smples were similr to C. rbic regrding both rtios (k/c rtio rnged from 1.21 to 1.74; nd cf/k rtio rnged from 1.34 nd 2.25). Concerning the trditionl coffees, 45% of smples hd k/c rtio below 1.00 nd cf/k rtio close to or bove 4.00 indicting the probbility of significnt ddition of robust coffee to the product (Tble S2 in the SI section). 23 nd 27 (k/c rtios of 0.54 nd 0.27, respectively; nd cf/k rtios of 9.37 nd 20.07, respectively), which were well-described in the models (Tble 5), in ddition to smples 29, 30 nd 31 (k/c rtios rnged from 0.52 to 0.80; nd cf/k rtios rnged from 3.83 to 9.26), which hve not been djusted, my be ssocited with robust coffee ddition by pplying the proposed rtios (Tble S2 in the SI section). Conclusions The ssessment using PCA showed the relevnce of the composition vribles in the identifiction of coffee species in commercil products. The prmeters nicotinic cid, trigonelline, 5-CQA, cffeine, khweol nd cfestol were useful to chrcterize coffee species. Cffeine nd diterpenes presented more potentil for discriminte the species due to their reltive het stbility, while the other prmeters presented interction with the rosting process. In generl, higher levels of cffeine nd lower levels of khweol nd cfestol were relted with higher proportions of robust coffee in the product. The ddition of robust coffee to rbic coffee could be observed by the decrese in khweol/cfestol rtio nd increse in cffeine/cfestol rtio. Thus, we suggest the use of these two new prmeters (khweol/cfestol nd cffeine/khweol rtios) s tools for ssessing the ddition of robust coffee to commercil products. Acknowledgments The uthors would like to cknowledge CNPq (Conselho Ncionl de Desenvolvimento Científico e Tecnológico), Fundção Arucári/Prná, FAEP/UEL (Fundo de Apoio o Ensino, Pesquis e Extensão) nd CBP&D/Cfé (Consórcio Brsileiro de Pesquis e Desenvolvimento do Cfé) for scholrships nd finncil support. Supplementry Informtion The estimted dt for weight loss (WL%) in commercil (1 to 38), rbic (A1 to A5) nd robust (R1 to R3) coffees nlyzed in the present study re shown in Tble S1. The khweol/cfestol nd cffeine/ khweol rtios for ech commercil smple re shown in Tble S2. Both of these tbles re vilble free of chrge t s PDF file. References 1. ccessed in My 2012; Meztu, L.; Andrez, S.; Ibáñez, C.; Peñ, M. P.; Bello, J.; Cid, C.; J. Agric. Food Chem. 2001, 49, ccessed in My ccessed in My ccessed in My 2012; ccessed in My De Mri, C. A. B.; Moreir, R. F. A.; Quim. Nov 2004, 27, 586; Frnç, A. S.; Oliveir, L. S.; Borges, M. L.; Vitorino, M. D.; Rev. Brs. Armzen. 2001, Especil Cfé, Ky, C. L.; Lourn, J.; Dussert, S.; Guyot, B.; Hmon, S.; Noirot, M.; Food Chem. 2001, 75, Lgo, R. C. A.; Boletim do CEPPA 2001, 19, Speer, K.; Kölling-Speer, I.; Brz. J. Plnt Physiol. 2006, 18, consultrlegislco.do?operco=visulizr& id =18524 ccessed in My 2012.

8 1354 Discrimintion of Commercil Rosted nd Ground Coffees According to Chemicl Composition J. Brz. Chem. Soc. 10. Alves, R. C.; Csl, S.; Alves, M. R.; Oliveir, M. B.; Food Chem. 2009, 114, 295; González, A. G.; Pblos, F.; Mrtín, M. J.; León-Cmcho, M.; Vldenebro, M. S.; Food Chem. 2001, 73, 93; Amorim, A. C. L.; Hovell, A. M. C.; Pinto, A. C.; Eberlin, M. N.; Arrud, N. P.; Pereir, E. J.; Bizzo, H. R.; Cthrino, R. R.; Moris Filho, Z. B.; Rezende, C. M.; J. Brz. Chem. Soc. 2009, 20, Alves, S. T.; document/?code=vtls ccessed in My Bicchi, C. P.; Binello, A. E.; Pellegrino, G. M.; Vinni, A. C.; J. Agric. Food Chem. 1995, 43, 1549; Cmp, C.; Bllester, J. F.; Doulbeu, S.; Dussert, S.; Hmon, S.; Nirot, M.; Food Chem. 2004, 88, 39; Mrtín, M. J.; Pblos, F.; González, A. G.; Tlnt 1998, 46, Csl, S.; Oliveir, M. B. P. P.; Alves, M. R.; Ferreir, M. A.; J. Agric. Food Chem. 2000, 48, Dis, R. C. E.; document/?code=vtls ccessed in My Alves, S. T.; Scholz, M. B. S.; Benssi, M. T.; Abstrcts of the 8º Encontro Regionl Sul de Ciênci e Tecnologi de Alimentos, Curitib, Brzil, 2003; Speer, K.; Tewis, R.; Montg, A.; Abstrcts of the 14º Interntionl Conference on Coffee Science - ASIC, Sn Frncisco, USA, Cmpnh, F. G.; document/?code=vtls ccessed in My 2012; Cmpnh, F. G.; Dis, R. C. E.; Benssi, M. T.; Coffee Sci. 2010, 5, Csl, S.; Oliveir, M. B.; Ferreir, M. A.; Food Chem. 2000, 68, Dgli, M.; Cuzzoni, M. T.; Decno, C.; J. Agric. Food Chem. 1994, 42, Moreir, R. F. A.; Trugo, L. C.; De Mri, C. A. B.; Quim. Nov 2000, 23, Rmlkshmi, K.; Kubr, I. R.; Ro, L. J. M.; J. Food Sci. 2007, 72, S Vsconcelos, A. L. S.; Frnç, A. S.; Glóri, M. B. A.; Mendonç, J. C. F.; Food Chem. 2007, 101, De Mri, C. A. B.; Moreir, R. F. A.; Trugo, L. C.; Quim. Nov 1999, 22, Frh, A.; Donngelo, C. M.; Brz. J. Plnt Physiol. 2006, 18, 23; Frnç, A. S.; Mendonç, J. C. F.; Oliveir, S. D.; LWT - Food Sci. Technol. 2005, 38, Perrone, D.; Frh, A.; Donngelo, C. M.; De Pulis, T.; Mrtin, P. R.; Food Chem. 2008, 106, Kurzrock, T.; Speer, K.; Food Rev. Int. 2001, 17, 433; Rubyiz, A. B.; Meurens, M.; J. Agric. Food Chem. 2005, 53, Alves, S. T.; Dis, R. C. E.; Benssi, M. T.; Scholz, M. B. S.; Quím. Nov 2006, 29, Dis, R. C. E.; Scholz, M. B. S.; Benssi, M. T.; Abstrcts of the 2 nd Congresso Brsileiro de Ciênci e Tecnologi de Alimentos, Curitib, Brzil, contribuintes/cordos/2001/3/ pdf ccessed in My SttSoft 7.1; Sttistic for Windows: Computer Progrm Mnul; Softwre Inc.: Tuls, De Souz, R. M. N.; Cnuto, G. A. B.; Dis, R. C. E.; Benssi, M. T.; Quim. Nov 2010, 33, Trugo, L. C.; Mcre, R.; Food Chem. 1984, 15, Mendonç, J. C. F.; Frnç, A. S.; Oliveir, L. S.; Nunes, M.; Food Chem. 2008, 111, Pizrro, C.; Estebn-Díez, I.; González-Sáiz, J. M.; Anl. Chim. Act 2007, 585, Kitzberger-Scomn, C. S.; Scholz, M. B. S.; Pereir, L. F. P.; Vieir, L. G. E.; Ser, T.; Silv, J. B. G. D.; Benssi, M. T.; Abstrcts of the 23 rd Interntionl Conference on Coffee Science- ASIC, Bli, Indonesi, Submitted: April 25, 2011 Published online: June 28, 2012

9 Supplementry Informtion J. Brz. Chem. Soc., Vol. 23, No. 7, S1-S2, Printed in Brzil Sociedde Brsileir de Químic $ SI Discrimintion of Commercil Rosted nd Ground Coffees According to Chemicl Composition Romiline M. N. de Souz nd Mrt T. Benssi* Deprtmento de Ciênci e Tecnologi de Alimentos, Universidde Estdul de Londrin, CP 6001, Londrin-PR, Brzil Tble S1. Estimted dt for weight loss (WL%) in commercil (1 to 38), rbic (A1 to A5) nd robust (R1 to R3) coffees WL% Numbers/letters Code b Numbers/letters Code b WL% 1 AEF LT BP MT BG NF CT OEF CG PF DT QT DEF QEF DA REF DEX SEF ET SEF clss EF TF EEF UG FT VT FEF VEF FG WES GT A HT A HEF A HP A IEF A IG R JEF R KT R Estimted ccording to the following eqution: WL% = SUM (nicotinic cid + trigonelline) RATIO (5-CQA/cffeine) 0.13 L* 0.06 H* (Dis 14 ); b Coding for ech product previously pplied by De Souz et l. 30 *e-mil: mrttb@uel.br

10 S2 Discrimintion of Commercil Rosted nd Ground Coffees According to Chemicl Composition J. Brz. Chem. Soc. Tble S2. Khweol/cfestol nd cffeine/khweol rtios for commercil coffee smples Khweol/cfestol rtio Cffeine/khweol rtio