Article pubs.acs.org/JAFC
Identification of 3‑Methylbutanoyl Glycosides in Green Coffea arabica Beans as Causative Determinants for the Quality of Coffee Flavors Keiko Iwasa,*,† Daiki Setoyama,‡ Hiroaki Shimizu,† Harumichi Seta,† Yoshinori Fujimura,‡ Daisuke Miura,‡ Hiroyuki Wariishi,‡,§,∥ Chifumi Nagai,⊥ and Koichi Nakahara*,† †
Research Division, Suntory Global Innovation Center Limited (Suntory SIC), 5-2-5 Yamazaki, Shimamoto-cho, Mishima-gun, Osaka 618-0001, Japan ‡ Innovation Center for Medical Redox Navigation, Kyushu University, 3-1-1 Maidashi, Higashi-ku, Fukuoka 812-8582, Japan § Faculty of Arts and Science, Kyushu University, 744 Motooka, Nishi-ku, Fukuoka 819-0395, Japan ∥ Graduate School of Bioresource and Bioenvironmental Sciences, Kyushu University, 6-10-1 Hakozaki, Higashi-ku, Fukuoka 512-8581, Japan ⊥ Hawaii Agriculture Research Center, P.O. Box 100, Kunia, Hawaii 96759, United States S Supporting Information *
ABSTRACT: The quality of coffee green beans is generally evaluated by the sensory cupping test, rather than by chemical compound-based criteria. In this study, we examined the relationship between metabolites and cupping scores for 36 varieties of beans, using a nontargeted LC−MS-based metabolic profiling technique. The cupping score was precisely predicted with the metabolic information measured using LC−MS. Two markers that strongly correlated with high cupping scores were determined to be isomers of 3-methylbutanoyl disaccharides (3MDs; 0.01−0.035 g/kg of beans) by spectroscopic analyses after purification, and one of them was a novel structure. Further, both the 3MDs were determined to be precursors of 3-methylbutanoic acid that enhance the quality of coffee. The applicability of 3MDs as universal quality indicators was validated with another sample set. It was concluded that 3MDs are the causative metabolites determining beverage quality and can be utilized for green bean selection and as key compounds for improving the beverage quality. KEYWORDS: Cof fea arabica green coffee bean, metabolic profiling, cupping score, 3-methylbutanoyl disaccharide, LC−MS, NMR, pyrolysis, GC−MS
1. INTRODUCTION Coffee is one of the most important commodities traded worldwide. Coffee green beans are produced in more than 70 countries.1 However, there is no universal standard to assess the quality of coffee green beans in the coffee market. Each producing country has its own grading criteria for export and trade purposes.2 In general, the green beans are graded on the basis of the number of defective beans (Brazil, Jamaica, Ethiopia, and others), the screen size of the beans (Brazil, Colombia, Tanzania, and others), and the altitude of the cultivation area (Guatemala, El Salvador, Costa Rica, and others).2 Although the flavor of the coffee beverage is strongly affected by the species (or cultivars) and maturities, the green beans are not graded by these criteria, because it is difficult to determine these attributes based on the appearance of the green beans. Alternatively, the sensory cupping test of the resultant beverage is also utilized in evaluating the quality of the green beans. However, the cupping test requires experienced assessors with trained techniques and sensitive surrounding conditions for evaluation. From a chemical viewpoint, beverage quality is known to be determined by flavor and aroma compounds, which are generated from the metabolites in green beans during roasting. Therefore, the metabolite properties are the causative © XXXX American Chemical Society
determinants of beverage quality, which makes it crucial for evaluating the quality of green beans based on the metabolite properties. In this study, we focused on the metabolites to determine the beverage quality. A chemical-compound-based standard could act as an objective and systematic assessment criterion for both producing and consuming countries in the trading market. So far, many approaches have been adopted for metabolite analyses in green coffee beans. Most of these studies have aimed to distinguish (or classify) the origins3−6 between arabica (Cof fea arabica L.) and robusta (Cof fea canephora P.)7−10 or between defective beans and nondefective beans10,11 for compound-targeted measurements or nontargeted measurements coupled with multivariate statistical analyses. Only a few studies have reported on the correlation between metabolites and beverage quality12 or on the correlation between metabolites and maturity levels,13−18 which influence the quality. However, these studies were performed for specific targeted compounds (the major constituents in green beans), such as caffeine,13,14 chlorogenic acid,14−16 and trigonelline.13,14 Received: November 9, 2014 Accepted: February 1, 2015
A
DOI: 10.1021/jf5054047 J. Agric. Food Chem. XXXX, XXX, XXX−XXX
Article
Journal of Agricultural and Food Chemistry
studied. The ground powder (30 g) of each variety of roasted beans was extracted with hot water (300 mL, 97 °C). The coffee brew resulting from the coffee beans that were of good quality was used as the basic brew for studying the influence of 0.1 μ/L of flavor compound additives, whereas the brew prepared from the coffee beans of excellent quality was used for calibration reference. The panel assessed and scaled these brews with an SCAA cupping score. Only the aftertaste grades are provided in this article. The intensity values are assigned as line-scale values from 6 to 10 points in quarter point increments. 2.4. Metabolite Extraction and LC−MS Analysis for Metabolic Profiling. The green beans (10 g) frozen in liquid nitrogen were crushed with a small metal device by shaking using a Multibeads Shocker (Biomedical Science, Nagasaki, Japan) at 1100 rpm for 7 min, twice. The crushed powder (50 mg) was extracted with 1.5 mL of 70% methanol/H2O for 1 h at 80 °C. After centrifugation at 10000 rpm for 10 min at 10 °C, the supernatants were diluted four times with 20% acetonitrile and filtered using a Ultrafree-MC filter units (0.22 μm, Polyvinylidene Fluoride (PVDF), Millipore, Billerica, MA) for use in the LC−MS analysis. The LC−MS measurement of coffee metabolites was performed as described in our previous study,20 on a Luna-C18 column (250 × 1.0 mm, 5-μm particle size, Phenomenex, Torrance, CA) using a LC− MS−ion trap (IT)−time-of-flight (TOF) instrument (Shimadzu, Kyoto, Japan). 2.5. Data Processing, Multivariate Statistical Analysis, and Correlation Plots. The data processing (peak picking and alignment) and normalization by the sum of total ion counts of each sample for the ion signals in positive mode were performed as described in our previous report.20 The ion signals were subsequently filtered according to two criteria: (1) at least one group of a total of 36 categorical groups (36 varieties, n = 3) should exhibit
