Selection of Discriminant Markers for Authentication of Asian Palm

Jul 27, 2013 - *(E.F.) Mailing address: Department of Biotechnology Graduate School of Engineering, Osaka University, 2-1 Yamadaoka, Suita, Osaka ...
1 downloads 0 Views 3MB Size
Subscriber access provided by FLORIDA INTL UNIV

Article

Selection of Discriminant Marker for Authentication of Asian Palm Civet Coffee (Kopi Luwak): A Metabolomics Approach Udi Jumhawan, Sastia Prama Putri, Yusianto Yusianto, Erly Marwanni, Takeshi Bamba, and Eiichiro Fukusaki J. Agric. Food Chem., Just Accepted Manuscript • DOI: 10.1021/jf401819s • Publication Date (Web): 27 Jul 2013 Downloaded from http://pubs.acs.org on August 6, 2013

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Journal of Agricultural and Food Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 35

Journal of Agricultural and Food Chemistry

1

Selection of Discriminant Markers for Authentication of Asian Palm

2

Civet Coffee (Kopi Luwak): A Metabolomics Approach

3 4

Udi Jumhawan1, Sastia Prama Putri1, Yusianto2, Erly Marwani3, Takeshi

5

Bamba1, Eiichiro Fukusaki*,1

6 7

1

8

University, Osaka, Japan

9

2

Indonesian Coffee and Cocoa Research Institute, East Java, Indonesia

10

3

Bandung Institute of Technology, West Java, Indonesia

Department of Biotechnology, Graduate School of Engineering, Osaka

11 12 13 14

*Address for correspondence:

15

Prof. Dr. Eiichiro Fukusaki

16

Department of Biotechnology

17

Graduate School of Engineering

18

Osaka University

19

2-1 Yamadaoka, Suita, Osaka, 565-0871, Japan

20

Tel: +81-6-6879-7424

21

Fax: +81-6-6879-7424

22

E-mail: [email protected]

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 2 of 35

23

Abstract

24

Kopi Luwak, an exotic Indonesian coffee, is made from coffee berries that

25

have been eaten by the Asian palm civet (Paradoxurus hermaphroditus).

26

Despite being known as world’s most expensive coffee, there is no reliable,

27

standardized

28

multimarker profiling was employed to explore significant metabolites as

29

discriminant markers for authentication. Extracts of 21 coffee beans (Coffea

30

arabica and Coffea canephora) from three cultivation areas were analyzed

31

and subjected to multivariate analyses, principal component analysis, and

32

orthogonal projection to latent structures-discriminant analysis. Citric acid,

33

malic acid, and the inositol-pyroglutamic acid ratio were selected for further

34

verification by evaluating their differentiating abilities against various

35

commercial coffee products. The markers demonstrated potential application

36

in the differentiation of original, fake Kopi Luwak, regular coffee, and coffee

37

blend samples with 50 wt% Kopi Luwak content. This is the first report to

38

address the selection and successful validation of discriminant markers for the

39

authentication of Kopi Luwak.

method

for

determining

its

authenticity.

GC/MS-based

40 41

Keywords: Kopi Luwak, Asian palm civet, GC/MS, authentication, discriminant

42

markers

ACS Paragon Plus Environment

Page 3 of 35

Journal of Agricultural and Food Chemistry

43

Introduction

44

Coffee is one of the most popular beverages in the world. Kopi Luwak is

45

considered to be the world’s most expensive coffee, with a price tag of 300–

46

400 USD per kg.1 Kopi Luwak, the Indonesian words for coffee and civet,

47

respectively, is made from coffee berries that have been eaten by the Asian

48

palm civet (Paradoxurus hermaphroditus), a small mammal native to Southern

49

and Northern Asia.2 The civet climbs coffee trees and instinctively selects

50

coffee cherries. During digestion, the coffee pericarp is completely digested

51

and the beans are excreted. The intact beans are then collected, cleaned,

52

wet-fermented, sun-dried, and further processed by roasting. Kopi Luwak’s

53

high selling price is mainly attributed to its exotic and unexpected production

54

process.3

55

Despite its profitable prospects, there is no reliable, standardized method for

56

determining the authenticity of Kopi Luwak. Moreover, there is limited

57

scientific information on this exotic coffee. Recently, coffee adulterated to

58

resemble Kopi Luwak was reported in the coffee market.4 This poses serious

59

concern among consumers over the authenticity and quality of the products

60

currently available in the market. Discrimination between Kopi Luwak and

61

regular coffee has been achieved using electronic nose data.3 However, the

62

selection of a discriminant marker for authentication was not addressed. The

63

methods currently employed by Kopi Luwak producers is sensory analysis

64

include visual and olfactory testing, both of which are inadequate. For

65

example, visual examination is only possible for green coffee beans prior to

66

roasting, and very few trained experts can perform the highly subjective

67

sensory analysis to discriminate Kopi Luwak.

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 4 of 35

68

Information flows in metabolic pathways are highly dynamic and represent the

69

current biological states of individual cells. Hence, the metabolome has been

70

considered as the best descriptor of physiological phenomena.5 Metabolomics

71

techniques can be powerful tools to elucidate variations in phenotypes

72

imposed by perturbations such as gene modification, environmental factors,

73

or physical stress. The “black box” process during animal digestion can be

74

translated as physical and enzymatic consequences to the coffee bean, which

75

presents a smoother surface and color changes after digestion.3 Thus, a

76

metabolomics technique was chosen to screen and select discriminant

77

markers for the authenticity assessment of Kopi Luwak. Metabolomics

78

techniques have been effectively applied to distinguish the phytochemical

79

compositions of agricultural products of different origins,6 varieties,6,7 and

80

cultivars8 for quality control and breeding.

81

In

82

spectrometry (GC-Q/MS)-based multimarker profiling was employed to

83

identify discriminant markers for the differentiation of Kopi Luwak and regular

84

coffees. A combination of gas chromatography and mass spectrometry

85

(GC/MS) provides high sensitivity, reproducibility, and the quantitation of a

86

large number of metabolites with a single-step extraction.9,10 Sample

87

classification by means of chemometrics was performed using principal

88

component analysis (PCA). Subsequently, orthogonal projection to latent

89

structures combined with discriminant analysis (OPLS-DA)11 and significance

90

analysis of microarrays/metabolites (SAM)12 were used to isolate statistically

91

significant compounds as discriminant marker candidates. The applicability of

our

study,

gas

chromatography

coupled

with

ACS Paragon Plus Environment

quadrupole

mass

Page 5 of 35

Journal of Agricultural and Food Chemistry

92

these candidates as discriminant markers was verified to differentiate various

93

commercial coffee products.

94 95

Materials and Methods

96

Materials

97

Samples were divided into experimental and validation coffee groups.

98

Experimental coffees were utilized to construct the discrimination model and

99

to select significant compounds. Verification of the applicability of the

100

discriminant markers was carried out using the validation coffee set. In this

101

report, coffee that had been digested by the animal is referred to as Kopi

102

Luwak, and the other beans as regular coffee. Kopi Luwak and regular coffee

103

samples of two species, Coffea arabica (Arabica) and Coffea canephora

104

(Robusta), were used. Coffee samples were obtained from 21 sampling points

105

in three cultivation areas in Indonesia (Java, Sumatra, and Bali). Samples

106

were roasted in a Probat-Werke von Gimborn Maschinenfabrik GmbH model

107

BRZ 2 (Probat, Rhein, Germany) at 205°C for 10 min to obtain a medium

108

degree of roasting, and then were air-cooled for 5 min. Coffee beans were

109

ground and stored in sealed 50 mL BD Falcon tubes at −30°C with light-

110

shielding prior to analysis. The sample descriptions are shown in Tables S1

111

and S2 (Supporting Information). Wet-fermentation was applied to both the

112

Kopi Luwak and regular coffees. For regular coffees, conventional protocols

113

were applied after harvesting, including de-pulping, wet fermentation, and

114

drying.

115

The validation set consisted of authentic Kopi Luwak, commercial Kopi

116

Luwak, commercial regular coffee, fake coffee, and coffee blend. Authentic

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 6 of 35

117

Kopi Luwak was produced via controlled processing to ensure the quality of

118

the beans pre- and post-digestion. The remaining samples were purchased

119

commercially. The coffee blend was utilized to examine the feasibility of the

120

method for differentiating mixed and pure coffees.

121 122

Reagents

123

Methanol, chloroform, distilled water, ribitol, and pyridine were purchased

124

from Wako Pure Chemical Industries, Ltd. (Osaka, Japan). Methoxyamine

125

hydrochloride and alkane standard solution were purchased from Sigma

126

Aldrich

127

(trimethylsilyl)trifluoroacetamide (MSTFA) was purchased from GL Science,

128

Inc. (Tokyo, Japan). The six authentic standards of the discriminant markers,

129

their providers, and purities were as follows: citric acid (Nacalai-Tesque,

130

Kyoto, Japan, 99.5%), malic acid (Nacalai-Tesque, 99%), pyroglutamic acid

131

(ICN Biomedicals, Ohio, USA, 99.5%), caffeine (Sigma Aldrich, 98.5%),

132

inositol (Wako, 99%), and glycolic acid (Sigma Aldrich, 99%).

(Milwaukee,

Wisconsin,

USA).

N-Methyl-N-

133 134

Extraction

135

Coffee beans were put into a grinding mill container, cooled for 3 min on water

136

ice cubes, and then ground with a Retsch ball mill (20 Hz, 3 min). Coffee bean

137

powder (15 mg) was transferred into a 2 mL Eppendorf tube. In addition to

138

pure samples, a 50:50 (wt%) blend of Kopi Luwak and regular coffee was

139

used. One milliliter of a single-phase extraction solvent consisting of 2.5/1/1

140

(v/v/v) methanol, distilled water, and chloroform, respectively, was added to

141

extract a wide range of metabolites. A non-specific extraction procedure was

ACS Paragon Plus Environment

Page 7 of 35

Journal of Agricultural and Food Chemistry

142

applied to avoid limiting the target analysis to specific compounds and to

143

comprehensively screen the components of Kopi Luwak. As an internal

144

standard, ribitol (60 µL, diluted with deionized water to 0.2 mg/mL) was

145

utilized. The mixture was shaken for 1 min and then centrifuged at 4°C and

146

16000 g for 3 min. The supernatant (900 µL) was transferred into a 1.5 mL

147

Eppendorf tube and diluted with 400 µL Milli-Q water (Wako). The mixture

148

was then vortexed and centrifuged for 3 min. A 400 µL portion of the aqueous

149

phase was transferred into a fresh 1.5 mL Eppendorf tube with a screw cap.

150

The solvent was removed by vacuum centrifugation for 2 h, followed by

151

freeze-drying overnight. All samples were analyzed in triplicate (n = 3).

152 153

Derivatization for GC/MS analysis

154

Oximation and trimethylsilylation were used for derivatization. Methoxyamine

155

hydrochloride (100 µL, 20 mg/mL in pyridine) was added to the dried extract

156

as the first derivatization agent. The mixture was incubated at 30°C for 90

157

min. After addition of the second derivatization agent, N-methyl-N-

158

(trimethylsilyl)trifluoroacetamide (MSTFA, 50 µL), the mixture was incubated

159

at 37°C for 30 min.

160 161

GC/MS Analysis

162

GC-Q/MS analysis was performed on a GCMS-QP 2010 Ultra (Shimadzu,

163

Kyoto, Japan) equipped with a CP-SIL 8 CB low bleed column (0.25 mm x 30

164

m, 0.25 µm, Varian Inc., Palo Alto, California, USA) and an AOC-20i/s

165

(Shimadzu) as an autosampler. The mass spectrometer was tuned and

166

calibrated prior to analysis. The derivatized sample (1 µL) was injected in split

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

167

mode, 25/1 (v/v), with an injection temperature of 230°C. The carrier gas (He)

168

flow was 1.12 mL/min with a linear velocity of 39 cm/s. The column

169

temperature was held at 80°C for 2 min, increased by 15°C/min to 330°C, and

170

then held for 6 min. The transfer line and ion source temperatures were 250

171

and 200°C, respectively. Ions were generated by electron ionization (EI) at

172

0.93 kV. Spectra were recorded at 10000 u/s over the mass range 85−500

173

m/z. A standard alkane mixture (C8–C40) was injected at the beginning and

174

end of the analysis for tentative identification.

175 176

Identification and quantitation of marker candidates

177

The discriminant marker candidates were identified and quantitated against

178

six authentic standards (malic acid, citric acid, glycolic acid, pyroglutamic acid,

179

caffeine, and inositol) at various concentrations. The final concentrations of

180

the authentic standards were adjusted to 1, 10, 50, 100, 250, 500, 750, 1000,

181

1500, and 2000 µM with the extraction solvent to construct a calibration curve.

182

For extraction, the authentic standards were processed identically to the

183

coffee bean samples. The standards were co-injected during sample analysis.

184

Two blank solutions were prepared by adding only extraction solvent and

185

distilled water, respectively. The limits of detection (LOD) and quantitation

186

(LOQ) were determined via known protocols.13,14 The construction of the

187

standard curve and quantitation were conducted using GC/MS Solution

188

software (Shimadzu). No authentic standards were detected in either of the

189

blank samples.

190 191

Data Processing

ACS Paragon Plus Environment

Page 8 of 35

Page 9 of 35

Journal of Agricultural and Food Chemistry

192

Chromatographic data were converted into ANDI files (Analytical Data

193

Interchange Protocol, *.cdf) using the GC/MS Solution software package

194

(Shimadzu). Peak detection, baseline correction, and peak alignment of

195

retention times were performed on the ANDI files using the freely available

196

software package, MetAlign.15 Spectra were normalized manually by adjusting

197

the peak intensity against the ribitol internal standard.

198

Retention indexes of the eluted compounds were calculated based on the

199

standard alkane mixture. By comparing the retention indexes and unique

200

mass spectra with our in-house reference library, tentative identifications were

201

obtained. To simplify and accelerate the tentative identifications of

202

compounds that were registered in the in-house library database, AIoutput2

203

(version 1.29) annotation software, developed in the authors’ laboratory, was

204

utilized.16 For comparison with the National Institute of Standards and

205

Technology (NIST) library, retention times were used instead.

206 207

Multivariate data analysis

208

PCA and OPLS-DA were performed. OPLS-DA with an S alphabet-like plot, or

209

S-plot, was chosen to isolate and select statistically significant and potentially

210

biochemically interesting compounds. The S-plot provides covariance and

211

correlation between metabolites and the modeled class to allow easier

212

visualization. The variables that changed significantly are plotted at the top

213

and bottom of the S-plot, and those that do not significantly contribute are

214

plotted in the middle.11 A sevenfold cross validation was carried out to assess

215

the accuracy of the discrimination model in practice. The goodness-of-fit (R2)

216

and predictability (Q2) parameters were then determined. Analysis was

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

217

performed with commercial software, SIMCA-P+ version 12 (Umetrics, Umeå,

218

Sweden). Data were Pareto-scaled to reduce the effect of noise in the

219

chromatograms.

220

To confirm the selection of significant compounds by OPLS-DA, data were

221

also subjected to significance analysis of microarrays/metabolites (SAM)

222

using MetaboAnalyst 2.0. At first, SAM was projected to microarrays analysis

223

to assign significance genes on the basis of changes in their expression.12

224

Recently, MetaboAnalyst 2.0, a web server for metabolomics analysis, was

225

developed for the application of SAM to metabolomics data.17,18

226

The “relative differences,” d(i), i.e., the differences in intensity for each peak,

227

were calculated by a formula described elsewhere.12 To generate the control

228

set, data were analyzed by n-balanced permutations. The relative differences

229

of the permutated data, dp(i), were determined. Next, the “expected relative

230

differences”, dE(i), were calculated as the average of dp(i) over n-balanced

231

permutations. Scatter plots of the expected relative differences (dE(i)) and

232

relative differences (d(i)) were constructed. Insignificant compounds were

233

identified as those which satisfied d(i)

234

assigned as a certain distance displaced from d(i) = dE(i); compounds at

235

distances larger than the threshold value were considered significant. The

236

higher the threshold value, the lower the false discovery rate (FDR, the

237

percentage of falsely significant compounds) and the number of significant

238

compounds that can be obtained. Univariate statistics, performed by the freely

239

available software, R project,19 was applied to compare the means of selected

240

significant compounds using the Student’s t-test. Box plots were also

dE(i). A threshold value (∆) was

ACS Paragon Plus Environment

Page 10 of 35

Page 11 of 35

Journal of Agricultural and Food Chemistry

241

constructed to display the differentiation among coffee samples. Differences

242

were considered significant when p < 0.05.

243 244

Results and Discussion

245

GC/MS-based metabolite profiling of Kopi Luwak

246

GC-Q/MS analysis was performed on aqueous extracts of coffee beans to

247

investigate the differences in their metabolite profiles and select discriminant

248

markers for robust authentication. In addition, this research focused on

249

increasing the scientific information about Kopi Luwak. A quadrupole mass

250

spectrometer was selected because of its availability as the most widely used

251

mass analyzer. However, a conventional Q/MS can be operated only at a

252

slow scan rate.20 With processor improvements and high-speed data

253

processing, newly developed GC-Q/MS instruments provide increased

254

sensitivity at high scan speeds of up to 10.000 u/s.21

255

Because of their broad cultivation areas and commercial profitability, Coffea

256

arabica and Coffea canephora, which represent 65% and 35% of the total

257

annual coffee trade, respectively, were utilized for metabolomics analysis.22 A

258

total of 182 peaks from 21 coffee beans were extracted using MetAlign.

259

Moreover, 26 compounds were tentatively identified by comparison with our

260

in-house library (by retention index) and the NIST library (by retention time);

261

six of these were identified by co-injection with an authentic standard.

262

Tentatively identified components consisted of organic acids, sugars, amino

263

acids, and other compounds (Table S3). Previously reported coffee bean

264

constituents, including chlorogenic, quinic, succinic, citric, and malic acids;

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

265

caffeine, one of the compounds supplying bitter taste in coffee; and sucrose,

266

the most abundant simple carbohydrate, were identified.23–27

267

In recent research, unsupervised analysis, PCA, has been employed for data

268

exploration and to visualize information based on sample variance.28,29 A PCA

269

score plot derived from the 21 coffee beans differentiated two data groups

270

based on their species, Arabica and Robusta (Figure 1), and resulted in a

271

goodness-of-fit parameter (R2) of 0.844. Caffeine and quinic acid were

272

significant for the Robusta coffee data sets, whereas the Arabica data set was

273

mainly supported by various organic acids such as malic, chlorogenic, citric,

274

and succinic acids. The data differentiation was explained by 42.9% of

275

variance along PC1. The results indicated that genetic diversity more strongly

276

influenced the data separation than animal perturbation.

277

Because of the large variance among coffee species, sample differentiation

278

based on the type of coffee, Kopi Luwak or regular coffee, could not be

279

observed. Additional analyses were carried out independently for each coffee

280

species originating from the same cultivation area. The PCA score plot

281

revealed data separation based on the type of coffee, in which Kopi Luwak

282

and regular coffee could be clearly separated (Figure S1). For the Arabica

283

coffee data set, the separation was explained by 45.5% and 27.7% variances

284

in PC1 and PC2, respectively. By PC2, Kopi Luwak was closely clustered in

285

the same region, whereas regular coffees tended to separate based on their

286

cultivation areas. In the loading plot, malic and glycolic acids contributed

287

highly to the Kopi Luwak data (data not shown). Thereby, coffee beans may

288

possess similar profiles after animal digestion. Differences in cultivation areas

289

were considered to have the least significance for data separation. In Robusta

ACS Paragon Plus Environment

Page 12 of 35

Page 13 of 35

Journal of Agricultural and Food Chemistry

290

coffee, a clear separation between Kopi Luwak and regular coffee was

291

observed, which was explained by 79.1% variance of PC1. Significant

292

compounds for separation, including inositol and pyroglutamic acid for Kopi

293

Luwak and quinic acid for regular coffee, were observed.

294 295

Discriminant analysis to select candidates for discriminant markers

296

An overview of all data samples was provided by the unsupervised analysis,

297

PCA. However, detailed information regarding compounds contributing to the

298

data differentiation between Kopi Luwak and regular coffee remained unclear.

299

Therefore, coffee bean data sets were subjected to supervised discriminant

300

analysis (OPLS-DA). For analyses having two or more classes, OPLS-DA is

301

the most suitable platform for isolating and selecting differentiation markers.

302

Compounds with reliable, high contributions to the model may possess

303

potentially biochemically interesting characteristics; thus, they can be selected

304

as biomarker candidates.11 All OPLS-DA models exhibited R2 and Q2 values

305

greater than 0.8, which would be categorized as excellent.30 In addition, all

306

models were in the range of validity after permutation tests using 200

307

variables (data not shown). The model was considered valid after permutation

308

for those that met the following criteria: R2Y-intercepts and Q2-intercepts

309

which did not exceed 0.3–0.4 and 0.05, respectively.31

310

Potential candidates for discriminant markers can be selected via S-plots by

311

setting the cut-off for covariance, p[1], and the correlation value, p[corr], to >

312

|0.2|. S-plots of the coffee data sets are shown in Figures 2B and D. In

313

addition to cut-off values, candidates for discriminant markers were selected

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

314

by variable importance in projection values (VIP). Large VIP values (> 1) are

315

more relevant for model construction.

316

The OPLS-DA score plot of Arabica coffee data sets is shown in Figure 2A.

317

Discrimination between Kopi Luwak and regular coffee was obtained. The

318

model was evaluated with R2 and Q2 values of 0.965 and 0.892, respectively.

319

Interestingly, compounds that were uncorrelated with Kopi Luwak were quinic

320

acid, caffeine, and caffeic acid. These compounds have been reported as

321

contributors of bitterness as well as acidity in coffee.23–26 In contrast,

322

compounds that were predictive to Kopi Luwak, i.e., over the cut-off value,

323

included citric, malic, and glycolic acids. The OPLS-DA score plot of the

324

Robusta coffee data sets (Figure 2C) was explained by R2 and Q2 values of

325

0.957 and 0.818, respectively. Caffeine, one of the bitter principles in coffee,

326

was found to be significantly correlated with Robusta Kopi Luwak data sets.

327

Robusta coffee has been reported to contain higher amounts of caffeine than

328

Arabica. Thus, it tends to be bitter and flavorless, whereas Arabica coffee is

329

considered to be milder, contain more aromatic compounds, and is more

330

appreciated by the consumer.32

331

We employed significance analysis of microarrays/metabolites (SAM) to

332

select significant compounds as discriminant markers, as a comparison to

333

OPLS-DA. In general, we used the same data sets, applying different

334

multivariate data analysis. By assigning the lowest possible FDR value, a total

335

of 12 compounds was considered as significant in the Arabica coffee data set

336

(Figure S2). Of these, citric acid, glycolic acid, malic acid, quinic acid, and

337

other unidentified peaks were included. In the Robusta coffee data set, nine

338

significant compounds were found (Figure S2); inositol, caffeine, pyroglutamic

ACS Paragon Plus Environment

Page 14 of 35

Page 15 of 35

Journal of Agricultural and Food Chemistry

339

acid, and six unidentified peaks exhibited high significance by SAM.

340

Candidates for discriminant markers for the authentication assessment of

341

Arabica and Robusta coffees are listed in Table 1. The selected marker

342

candidates met significant criteria in both OPLS-DA and SAM. Discriminant

343

markers were chosen independently for the Arabica and Robusta coffee.

344

To confirm whether these selected markers were generated as a result of

345

animal digestion, we investigated cause-effect relationships by quantitating

346

the discriminant marker candidates in green and roasted coffee beans from

347

controlled processing, pre- and post- animal digestion (Samples No. 5 and

348

No. 11 experimental sets). The results are displayed in Figure S3. In both the

349

raw and roasted beans, citric acid was present in higher concentration after

350

animal digestion, exhibiting a significant value difference (p < 0.05) between

351

Kopi Luwak and regular coffee. The concentration of caffeine was also

352

increased after digestion, but the difference was insignificant (p > 0.05). As a

353

result of roasting, the glycolic acid concentration increased dramatically (p
|0.2|. Triangles indicate peaks detected by GC/MS. Figure 3. PCA score plot of validation coffee set. Data plotted inside dashed circle indicates authentic Kopi Luwak. Figure 4. Box plots of peak intensity and concentrations of selected discriminant markers for validation.

ACS Paragon Plus Environment

Page 28 of 35

Page 29 of 35

Journal of Agricultural and Food Chemistry

Tables

Table 1. Candidates for discriminant markers from OPLS-DA and SAM, and analytical parameters for quantitation Discriminant marker

Glycolic acid Malic acid Pyroglutamic acid Citric acid Caffeine Inositol

RT (min)

4.96 9.05 9.43 11.61 12.18 13.45

VIP

3.93 5.53 1.7 5.6 2.28 4.47

RSD [%] (n = 3) RT Areaa 0.12 0.05 0.05 0.04 0.04 0.03

1.87 2.29 3.36 3.29 3.51 5.09

Linearity R2 0.9999 0.9996 0.9992 0.9997 0.9961 0.9974

Range (µ µM) 1-1000 1-1000 1-750 1-1000 100-2000 1-1000

a

at 100 µM

ACS Paragon Plus Environment

LOD (mg/kg)

LOQ (mg/kg)

0.021 0.043 0.054 0.504 1.531 0.082

0.066 0.132 0.164 1.526 4.638 0.247

Journal of Agricultural and Food Chemistry

Figure graphics

Figure 1

ACS Paragon Plus Environment

Page 30 of 35

Page 31 of 35

Journal of Agricultural and Food Chemistry

B

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

D

Figure 2

ACS Paragon Plus Environment

Page 32 of 35

Page 33 of 35

Journal of Agricultural and Food Chemistry

Figure 3

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Figure 4

ACS Paragon Plus Environment

Page 34 of 35

Page 35 of 35

Journal of Agricultural and Food Chemistry

Graphic for table of contents (TOC)

ACS Paragon Plus Environment