Metabolomics fingerprint of Philippine coffee by SPME-GC-MS for geographical and varietal classification

25 Volatile metabolites of Philippine Arabica and Robusta coffee beans in the both forms standard (not-26


51
Coffee aroma is the result of the multiplicity of volatile compounds present in roasted coffee beans 52 (Coffea spp.). The complex balance of the most important volatile compounds in coffee has a relative 53 contribution to its overall aroma quality (Bernard, Roberts, & Kraehenbuehl, 2005). So far, more than 54 eight hundred volatile compounds belonging to a wide range of chemical classes have been identified 55 in roasted coffee (Mayer & Grosch, 2001;Rocha, Maetzu, Barros, Cid & Coimbra, 2003), including 56 aliphatic volatile metabolites (carbonyl-containing compounds, sulfur-containing compounds), alicyclic 57 compounds (including several ketones), benzenoids (phenols); heterocyclic compounds (furans, 58 hydrofurans, pyrroles, pyridines, quinolines, pyrazines, quinoxalines, indoles, thiophens, thiophenones, 59 thiazoles, oxazoles) (Clarke, 1986). 60 Nowadays, coffee drinking is the best social lubricant and people are becoming more discriminating in 61 their preference for coffee. The aroma of coffee is one of the most important consumer's preference 62 vectors due to its contribution to the palatability and appreciation of overall coffee quality. This has 63 recently given rise to a fast growing demand for specialty coffee or commonly referred to gourmet or 64 premium coffee produced from special geographic microclimates beans with unique flavor profiles and 10 min) were assessed based on the highest number of peaks and highest peak areas. All the 142 optimization analysis was carried out on the same sample of Cordillera Arabica coffee. The roasted coffee beans (1.0 g) were placed in a 20-mL crimped-top-sealed vial. Each vial was heated 147 at 70 °C for 10 min to reach sample headspace equilibrium. The volatile compounds were extracted 148 using a 50/30 µm divinylbenzene-carboxen-polydimethylsiloxane (DVB/CAR/PDMS) fiber (Supelco,149 Merck KGaA, Bellefonte, PA, USA). The fiber was inserted into the vial and exposed to the headspace 150 above the coffee sample for 20 min at 70 °C. After the extraction, the fiber was thermally desorbed into 151 the GC injection port for 5 min. Each coffee sample was analyzed thrice.

152
[Arabica (2 standard + 2 civet) + Robusta (2 standard + 2 civet)] x 3 = 24 samples (total) The analysis was performed using a gas chromatograph Hewlett-Packard (HP) 6890 series instrument for which the same identification was matched across several samples and for which a similar mass 173 spectra spectrum was observed, were identified. In cases in which unacceptable confident matches 174 were found through the libraries, the compounds were individually checked and in cases where the 175 compounds showed the same retention time, molecular ion, base ion, and fragmentation patterns in all 176 samples were taken into account and labeled as 'unknown 1-8' accordingly. The absence of said 177 compounds was verified in blank injections. Whenever it was possible, the identification of volatiles 178 was also verified based on the presence in the literature. A semi-quantitation was carried out by 179 considering the average values of the absolute peak areas.  The DVB/CAR/PDMS fiber was chosen for HS-SPME due to its high affinity towards a pool of 192 analytes characterized by a wide-range of polarity, including aromatic heterocycles, benzenoids, 193 aliphatic and alicyclic hydrocarbons. In addition, this fiber has already been successfully applied in 194 previous studies (Bicchi et al., 2002;Mondello et al., 2004;Ryan, Shellie, Tranchida, Casilli, 195 Mondello, & Marriott, 2004;Mondello et al. 2005;Toci & Farah, 2008;Franca, Oliveira, Oliveira, 196 Agresti, & Augusti, 2009).

197
Increasing the sample weight from 0.5 g to 1.0 g, the intensity peaks of most compounds substantially 198 improved. However, 1.5 g of sample did not yield a further increase in the response. This is probably 199 due to a decrease of phase ratio "β" (headspace to sample ratio), and in the retention capacity of the 200 fiber (Kolb & Ettre, 2006). For this reason, 1.0 g was used as a standard sample weight.

201
Headspace generation was held at 70 °C for 10 min and the extraction temperature was varied from 202 8 203 temperature of 60 °C generated lower peak areas for most of the semi-volatile compounds. Conversely, 204 the highest extraction temperature of 80 °C resulted in an increase of peak areas of the high boiling 205 compounds, but caused the reduction of the areas of the compounds with a high vapor pressure. This 206 was due to a displacement effect that occurred onto the fiber to the detriment of substances with a high 207 vapor pressure. Extraction temperature of 70 °C was therefore deemed the best condition to achieve the 208 maximum extraction efficiency of volatile metabolites and used for the standard protocol. purging and cleaning of SPME fiber. The fiber was desorbed in the GC injection port for 5 and 10 min 223 and subjected again to desorption in a subsequent blank run. No peaks appeared during the latter run in 224 both cases, thus indicating that 5 min was a suitable time to prevent carry-over effects. The list of volatile metabolites extracted and identified is shown in Table 1. IUPAC names are 229 indicated together with the main synonyms. The latter are used throughout the article as they are most 230 commonly used in the literature.

231
Arabica and Robusta coffees showed a high number of volatile metabolites belonging to a wide variety 232 of chemical classes, notably aromatic heterocycles (furans, pyranes, pyrazines, pyridines, pyrroles), 233 aliphatic and alicyclic hydrocarbons, phenols, aldehydes, ketones, alcohols, esters, lactones, and fatty acids. Forty-seven volatile metabolites were considered in total, 27 of which were confirmed using pure 235 reference standards, while other 12 were tentatively identified based on MS-libraries matching. Eight 236 peaks were included in the list as unknown compounds, since their presence was verified in most of the 237 samples.
238 Figure 2 presents the volatiles composition of the complete samples set. The volatile that showed by far 239 the highest concentrations was furfuryl alcohol, followed by furfuryl acetate, 5-methylfurfural, and 3-240 acetylanisole. Furfuryl alcohol has a very mild, slightly caramel-like, warm-oily smell and is well 241 correlated with the undesirable burnt and bitter note of dark-roasted coffees (Flament, 2002).

268
AR samples showed a lack of volatile substances compared not only to their corresponding civet 269 samples but also to all other samples, being its average volatiles sum from one third to one fifth lower.

270
The cause of it is unknown and might be due to the specific lot of sample. This behavior has drastically 271 affected a correct comparison of this sample within the characterization of all other Philippine coffees.  Table 1) 316 were all distributed in the negative quadrants of the PC1, except dodecane. This result confirmed that 317 using this data set PCA could separate the samples on the PC2 more than on PC1. Indeed, due to their 318 general scarcity of volatile substances, AR coffees were completely separated from all the other 319 samples. For this reason, different PCAs were run in order to reduce this effect. In particular, civet 320 coffees were subjected alone to a PCA (Fig. 4a), while Arabica (CA and MA) standard coffee samples 321 were compared individually with Asipulo Robusta (Fig. 4b) and Kalinga Robusta (Fig. 4c) standard 322 coffees in two different PCA analysis.

323
The PCA score plot of all civet coffees successfully discriminated Arabica civet (CC and MC) from regions of production. Likewise, a clear discrimination between Arabica and Robusta samples on PC1