Subscriber access provided by - Access paid by the | UCSF Library

Article

Vibrational Spectroscopy and Chemometrics for Rapid, Quantitative Analysis of Bitter Acids in Hops (Humulus lupulus) Daniel Patrick Killeen, Dave H. Andersen, Ron A. Beatson, Keith C. Gordon, and Nigel B. Perry J. Agric. Food Chem., Just Accepted Manuscript • DOI: 10.1021/jf5042728 • Publication Date (Web): 08 Dec 2014 Downloaded from http://pubs.acs.org on December 14, 2014

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 34

Journal of Agricultural and Food Chemistry

Page 1 of 28

1

Vibrational Spectroscopy and Chemometrics for Rapid, Quantitative Analysis of Bitter

2

Acids in Hops (Humulus lupulus)

3

Daniel P. Killeen,†,‡ David H. Andersen,§ Ron A. Beatson,§ Keith C. Gordon†,‡ and Nigel B.

4

Perry*,†,┴

5 6 7 8 9 10 11



Department of Chemistry, University of Otago, P. O. Box 56, Dunedin, New Zealand



MacDiarmid Institute for Advanced Materials and Nanotechnology, University of Otago,

Dunedin, New Zealand §

The New Zealand Institute for Plant & Food Research Limited, 55 Old Mill, RD 3, Motueka

7198, New Zealand ┴

The New Zealand Institute for Plant & Food Research Limited, Department of Chemistry,

University of Otago, P.O. Box 56, Dunedin, New Zealand

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 2 of 28 12

ABSTRACT: Hops, Humulus lupulus, are grown worldwide for use in the brewing industry

13

to impart characteristic flavour and aroma to finished beer. Breeders produce many varietal

14

crosses with the aim of improving and diversifying commercial hops varieties. The large

15

number of crosses critical to a successful breeding program imposes high demands on the

16

supporting chemical analytical laboratories. With the aim of reducing the analysis time

17

associated with hops breeding, quantitative partial least squares regression (PLS-R) models

18

have been produced, relating reference data, acquired by the industrial standard HPLC and

19

UV methods, to vibrational spectra of the same, chemically diverse hops sample set. These

20

models, produced from rapidly acquired IR, NIR and Raman spectra, were appraised using

21

standard statistical metrics. Results demonstrated that all three spectroscopic methods could

22

be used for screening hops for α-acid, total bitter acids and cohumulone concentrations in

23

powdered hops. Models generated from Raman and IR spectra also showed potential for use

24

in screening hops varieties for xanthohumol concentrations. NIR analysis was performed

25

using both a standard bench-top spectrometer and a portable NIR spectrometer, with

26

comparable results obtained by both instruments. Finally, some important vibrational features

27

of cohumulone, colupulone and xanthohumol were assigned using DFT calculations allowing

28

more insightful interpretation of PLS-R latent variable plots.

29

KEYWORDS: hops, Humulus, Raman, infrared, chemometrics, α-acids

ACS Paragon Plus Environment

Page 2 of 34

Page 3 of 34

Journal of Agricultural and Food Chemistry

Page 3 of 28

30

■ INTRODUCTION

31

Hops (Humulus lupulus L., Cannabaceae) are a high value crop grown worldwide for use in

32

the brewing industry.1 Female plants are cultivated for their inflorescences, which produce

33

large quantities of terpenes and bitter acids (Figure 1) in extracellular trichomes known as

34

lupulin.1 The terpenes and their oxidation products impart aroma to finished beer while the

35

bitter acids are precursors to compounds responsible for beer’s distinctive bitter taste.2 The

36

trichomes also produce xanthohumol, a prenyl chalcone with potential health benefits (Figure

37

1).3 Commercial hops are traditionally divided into two categories: bittering hops with high

38

α-acid contents; and aroma hops which impart volatile terpenes with favourable aromas.

39

Some hops varieties are used for both purposes.4

40

Hops bitter acids, the focus of this work, comprise two structurally related families: α-

41

acids (humulones) and β-acids (lupulones) (Figure 1). The latter exist in solution in major

42

tautomeric forms (Figure 1).5,6 The α-acid concentration of a hops variety is of primary

43

concern to the brewing industry because these compounds undergo thermal isomerization,

44

forming the intensely bitter iso-α-acids.7 Total bitter acids (α plus β) and cohumulone

45

concentrations are also important for breeding selections.8 Bitter acid concentrations of hop

46

varieties are measured commercially using UV spectrophotometry and/or HPLC.9-11 UV

47

analysis is rapid but requires solvent extraction and serial dilution of hops samples. HPLC

48

shares these drawbacks and additionally requires longer analysis times. However, HPLC is

49

more specific and can be used to quantitate cohumulone concentrations within total α-acids.

50

In non-commercial settings, solvent extracted α-acids have been quantitated by 1H

51

NMR spectroscopy using their distinctive intramolecularly hydrogen-bonded proton signals

52

at 18 – 20 ppm.12,13 NMR has also been combined with HPLC to unambiguously identify the

53

individual bitter acids.14 Several reports show the potential of near infrared (NIR)

54

spectroscopy for the quantitation of bitter acids15-17 but we are not aware of any reports using

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 4 of 28 55

Raman or IR spectroscopy for hops analyses. General advantages of the vibrational

56

spectroscopic approach include the potential for rapid analysis, reduced sample preparation

57

and portable analysis.18-20

58

When relating vibrational spectra to reference data i.e. data acquired using an

59

established technique, the most commonly used chemometric method is partial least squares

60

regression (PLS-R).21-25 PLS-R iteratively maximizes the covariance between the spectral

61

matrix and the reference data and produces highly interpretable latent variables (LVs) which

62

describe diminishing quantities of spectral variance in the context of the reference data.26,27

63

LVs extract problem specific data, which means that a single spectral matrix can be

64

successively tuned to different reference data using PLS-R.28 This approach is particularly

65

relevant to hops, where quantitation of the structurally similar α- and β-acids in addition to

66

the bioactive xanthohumol is desirable.

67

PLS-R has recently been successfully applied to quantitation of iridoid glycosides

68

from Verbena by IR and NIR24, aspalathin in green rooibos by Raman23 and caffeine in

69

roasted coffee beans by NIR.29 As suggested by these examples, studies generally use only

70

one or two vibrational spectroscopy techniques, a practice that risks missing the most suitable

71

technique. Analysis by all three techniques results in complementary datasets which can be

72

appraised relative to one another and in the context of their performances for a given

73

application. In some cases, analytical challenges have been overcome using “fused”

74

spectroscopic datasets.30

75

To better understand the PLS-R models, cohumulone, colupulone and xanthohumol

76

were isolated, their Raman and IR spectra were measured, and the vibrational bands were

77

assigned using density functional theory (DFT) calculations. These assignments allowed the

78

interpretation of LVs in the generated PLS-R models, which in turn permitted a more

79

insightful understanding of model performances.

ACS Paragon Plus Environment

Page 4 of 34

Page 5 of 34

Journal of Agricultural and Food Chemistry

Page 5 of 28 80 81 82

■ MATERIALS AND METHODS Plant Material. Hops were grown at Motueka, South Island, New Zealand (41° 6’ S,

83

172° 58’ E) in Riwaka silt loam. Advanced selections (139 individual plants encompassing

84

39 genetic varieties) from hops breeding trials were harvested in March 2013. Hops cones

85

were kiln dried at ~ 60 °C for 8 h followed by determination of residual water content. Sub-

86

samples, ~ 5 g of cones, were pulverized under N2 (l) and stored at –20 °C prior to analyses.

87

Isolation of Hops Bitter Acids. An aliquot of the hops International Calibration

88

Extract (ICE-3) was purchased from the American Society of Brewing Chemists (ASBC). A

89

solution containing ~ 100 mg mL-1 of ICE-3 in MeOH was filtered before injection onto an

90

Agilent 1260 Infinity semi-preparative HPLC system equipped with a Phenomenex Luna

91

(5µ) C18 column (10 x 250 mm) with UV detection at 210 nm. Injections (50 µL ) were

92

eluted over 18 min in 87.9:12:0.1 MeOH:H2O:formic acid, with a flow rate of 5 mL min-1.

93

Fractions were collected at retention times of 7.3, 8.4, 10.8 and 12.9 min and identified by 1H

94

NMR as cohumulone, a mixture of humulone/adhumulone, colupulone and a mixture of

95

lupulone/adlupulone respectively by comparison to published spectra.6 Using the same

96

methodology, xanthohumol was isolated from a MeOH extract of hops inflorescences. This

97

compound had a retention time of 5.5 min. The compound was identified by 1H NMR by

98

comparison to published spectra.31

99

Quantitation of Bitter Acids by UV Spectrophotometry. Hops cones (10 g) were

100

added to 100 mL of toluene for extraction in an Omni mixer. Over a 10 min period, samples

101

were sequentially blended and rested for 30 s periods. Samples were filtered and 50 µL was

102

diluted to 25 mL with alkaline MeOH. The UV absorbances at 275, 325 and 355 nm were

103

recorded using a UV spectrophotometer, with 50 µL of toluene in alkaline MeOH (25 mL) as

104

a blank. This was adapted from a standard industry method.10

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 6 of 28 105

Quantitation of Bitter Acids and Xanthohumol by HPLC. Hops were extracted as

106

described above and 3 mL was transferred to a 50 mL volumetric flask and made to volume

107

with MeOH. 10 µL of this solution was injected onto a Shimadzu SCL10Avp HPLC system

108

equipped with a 250 x 4.6 mm, 5 µm Kinetix C18 column at 30 °C at a flow rate of 1.5 mL

109

min-1. The mobile phase was MeOH:H2O:phosphoric acid:0.1 M EDTA (1700:350:5:1) with

110

UV detection at 270 nm. Run time was 18 min. This was adapted from a standard industry

111

method.9

112

IR Analysis. Hops powder (~ 10 mg) was deposited directly onto the ATR crystal

113

and compressed by the anvil mechanism. Spectra were acquired using a Bruker ALPHA

114

Platinum FT-IR spectrometer equipped with a diamond ATR crystal. Spectra were recorded

115

from 376 – 4000 cm-1 with a spectral resolution of 4 cm-1 and were the average of 32 scans.

116

Prior to acquisition of each spectrum, the crystal and anvil were thoroughly cleaned and a

117

background spectrum was acquired to account for any environmental changes throughout the

118

analysis. Spectral acquisition time was ~ 1 min per sample.

119

Raman Analysis. Hops powder (~ 50 mg) was packed into aluminium divots for

120

analysis. Spectra were acquired using a FRA 106/5 Bruker Equinox FT-Raman spectrometer

121

equipped with a Nd:YAG laser emitting at 1064 nm, Equinox 55 interferometer and a

122

germanium detector cooled with N2 (l). The Raman stokes shift was recorded from 200 –

123

3500 cm-1 with a spectral resolution of 4 cm-1, spot size of 0.3 mm and laser power of 120

124

mW. Spectra were the average of 256 scans and the acquisition time was ~ 8 min per sample.

125

Raman analysis with excitation at 830 nm and 785 nm could not distinguish Raman bands

126

from raised baselines due to emission. These Raman systems have previously been described

127

in detail.21

128 129

NIR Analysis. Hops powder (~ 1 g) was added to a sample cup and the surface was smoothed with gentle compression. Background spectra, consisting of diffusely reflected

ACS Paragon Plus Environment

Page 6 of 34

Page 7 of 34

Journal of Agricultural and Food Chemistry

Page 7 of 28 130

light from a reflective surface were taken prior to each sample acquisition. Diffuse

131

reflectance infrared Fourier transform (DRIFT) spectra were recorded, using a FRA 106/5

132

Bruker FT-NIR spectrometer equipped with an Equinox 55 interferometer and germanium

133

detector cooled with N2 (l), from 4000 – 10000 cm-1 with a resolution of 8 cm-1 in Kubelka-

134

Munk units. Spectral acquisition time was ~ 1 min per sample.

135

Portable NIR Analysis. Hops powder (~ 1 g) were presented to the MicroNIR®

136

spectrometer (JDS Uniphase Corporation) in clear, 4 mL glass vials and reflectance spectra

137

were acquired through the base of the vial. A spectral acquisition time of 1 s was used and

138

analysis was performed in triplicate, shaking the vial between acquisitions. Two tungsten

139

filaments were used to produce polychromatic radiation and diffusely reflected light was

140

dispersed with a linear variable filter onto a 128-pixel uncooled InGaAs photodiode array

141

detector. Spectra were recorded from 4545 – 8696 cm-1 (1150 – 2100 nm ) with a spectral

142

resolution of 32 cm-1.

143

Density Functional Theory (DFT) Calculations. The molecular geometries of

144

xanthohumol, cohumulone and both colupulone conformers (Figure 1) were optimized with

145

DFT using the 6-31+G(d) basis set. A correction factor of 0.960 was applied to the predicted

146

vibrational frequencies to correct for their systematic overestimation.32-34 The mean absolute

147

deviation between the calculated and experimental vibrational bands was 13 cm-1. Quantum

148

chemical calculations were performed using Gaussian 09 and, where possible, initiated from

149

published crystal structures.31,35

150

Regression Modeling. Quantitative IR spectroscopy. The IR spectrum of each sample

151

was imported into the multivariate software “The Unscrambler®”. A Standard Normal Variate

152

(SNV) transformation was applied to each spectrum to compensate for scattering effects

153

arising from experimental factors such as varying optical path lengths, sample inhomogeneity

154

and background effects.36 Models were generated from the spectral region from 850 – 1700

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 8 of 28 155

cm-1 (the spectral region below 850 cm-1 and 1810 – 2700 cm-1 were omitted as IR spectra of

156

purified bitter acids (Figure 3(a)) did not contain any distinct and/or intense vibrational bands

157

in these regions, and the spectral region above 2700 cm-1 was also omitted because the C-H

158

and O-H stretching vibrations of many biological compounds contribute intensity in this

159

region). One third of the sample spectra were selected at random to be used as a test set for

160

model validation. The remaining two thirds were used as a calibration set to generate PLS-R

161

models (non-iterative, PLS127) relating the selected spectral regions to the reference values

162

from the UV and HPLC analyses (Figure 2). This independent ‘test-set’ validation

163

methodology is considered robust.37 Models were used to predict test set values from their IR

164

spectra with selected results in Table 1. The number of LVs selected was the optimum for

165

predicting values in the independent validation test-set. Data for all models are included in

166

supplementary information Table S1.

167

Quantitative Raman spectroscopy. The spectral region from 1050 – 1750 cm-1 was

168

used to generate PLS-R models relating Raman spectra to UV and HPLC reference values

169

(Figure 2). This spectral region was selected by comparison to the spectra of purified bitter

170

acids (Figure 3(b)). The same validation methodology was used, with the same samples

171

comprising the calibration and test sets. Selected results of the test set analyses are shown in

172

Table 1. Data for all models are included in supplementary information Table S1.

173

Quantitative NIR spectroscopy. PLS-R models were generated from 4000 – 7300 cm-1

174

from the DRIFT NIR spectra. This spectral region was chosen based on visual inspection and

175

correlation loadings, which suggested that the noisy spectral region from 7300 – 10000 cm-1

176

could not effectively be related to the reference data. For the NIR spectra measured with the

177

portable NIR spectrometer, regression models were generated from entire spectral range

178

acquired (4545 – 8696 cm-1). In both cases, spectra were subjected to SNV and used to

179

generate PLS-R models relating them to UV and HPLC reference data (Figure 2). The

ACS Paragon Plus Environment

Page 8 of 34

Page 9 of 34

Journal of Agricultural and Food Chemistry

Page 9 of 28 180

calibration and test sets were comprised of the same samples used for in the Raman and IR

181

analyses and selected model performances are presented in Table 1. Data for all models are

182

included in supplementary information Table S1.

183 184

■ RESULTS AND DISCUSSION

185

Thirty nine hops varieties, with 1-9 replicate samples each (139 total) were used in this study,

186

including both well-known commercial varieties and genetically novel selections chosen to

187

encompass broad chemotypical differences. The novel selections were designated

188

sequentially based on their associated breeding trials i.e. Alpha, Aroma and Brewing (Brew)

189

(Figure 2).

190

Reference Analyses. Bitter acid concentrations for all 139 hops samples were

191

determined by the standard industry methods of HPLC9 and UV spectrophotometry.10 The

192

average α-acid and β-acid concentrations for each variety by UV are shown in Figure 2, plus

193

HPLC results for the concentrations of the α-acids cohumulone and the coeluting

194

adhumulone and humulone, the β-acids colupulone and the coeluting adlupulone and

195

lupulones, and the chalcone xanthohumol (Figure 1). α-Acid concentrations were 2 – 17 %

196

w/w and β-acids 2 – 8 %, close to the full range of values reported for wild and domesticated

197

hops varieties.17,38 Xanthohumol concentrations of 0.1 – 0.8 % (Figure 2) also covered almost

198

the full reported range.3

199

Total α-acids and total β-acids measured by the HPLC and UV methods showed good

200

agreement between the results of both analyses, with correlation coefficients of 0.99 and 0.95

201

respectively. However, the absolute values for α- and β-acids by UV spectrophotometry were,

202

on average, 10% larger than values from HPLC analyses. It has been reported that the UV

203

analysis method is subject to interference from bitter acid oxidation products, which absorb at

204

analytically relevant wavelengths and cause this discrepancy.39 Total α-acid concentrations

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 10 of 28 205

did not correlate strongly with total β-acid concentrations: r2 = 0.36 (UV) and 0.39 (HPLC).

206

Likens et al. found that α-acid concentrations were negatively correlated to β-acid

207

concentrations in lupulin from different hops varieties but that the variation in lupulin content

208

between hops varieties confounded this correlation.39 The key quality indicators for breeding

209

selections i.e. total α-acids, total bitter acids (α plus β) and cohumulone concentrations, were

210

strongly correlated to one another. The correlations for the HPLC results were: total α-acids

211

versus total bitter acids, r2 = 0.95; total α-acids versus cohumulone, r2 = 0.82. The α-acids

212

were the main bitter acids in most hops varieties, and the proportion of cohumulone in the α-

213

acids was 17 – 40 % in these varieties (Figure 2). Vibrational Spectroscopic Analyses. These same 139 hops samples were analysed

214 215

by the different vibrational spectroscopy techniques of IR, Raman and NIR (two instruments)

216

to assess which was best suited to the application of rapid screening of bittering hops. PLS-R

217

models were generated, relating spectral matrixes to each of the UV and HPLC reference

218

data, and these models were categorized using the Ratio of Prediction to Deviation (RPD)

219

methodology described by Williams (Table 1).40 “Prediction” refers to the standard error of

220

prediction (SEP) of the test set values predicted by the calibration model and “Deviation”

221

refers to the standard deviation of the corresponding reference data. The scale is used to

222

broadly categorizing PLS-R models by their fitness for purpose and was useful for our data

223

set, which contained an unwieldy number of models with a large range of performances. The

224

RPD scale defines models with values of ≥3.1 as suitable for quantitative screening

225

applications.40 Details of hops models which met this criterion are shown in Table 1. These

226

models predicted the concentrations of the most important chemical values for breeding

227

selections i.e. total α-acids, cohumulone, total bitter acids (α plus β) and the bioactive

228

xanthohumol. Results from the other models can be found in supporting information (Table

229

S1).

ACS Paragon Plus Environment

Page 10 of 34

Page 11 of 34

Journal of Agricultural and Food Chemistry

Page 11 of 28 230

The RPD values (Table 1) showed that spectra from all three vibrational spectroscopy

231

techniques could be used to predict total α-acids and total bitter acids as determined by

232

industry standard UV analyses. However, the portable NIR analysis, which had a RPD value

233

of 3.0 for total α-acids, fell just short of the suitability criterion. The spectroscopic techniques

234

could also be used to predict total α-acids and total bitter acids as determined by HPLC

235

analyses, with the exception of the IR models, which fell short of the criterion. The IR spectra

236

could be used to predict concentrations of the minor α-acid cohumulone, as could Raman

237

spectra, but NIR spectral analyses failed (Tables 1 and S1).

238

We had expected that xanthohumol would be readily quantifiable by Raman

239

spectroscopy as the compound was predicted to show strong Raman scattering (see below).

240

However, the RPD values suggested that both Raman and IR spectra would not be

241

appropriate for quantitative analysis (Table 1), highlighting the difficulty of predicting which

242

vibrational spectroscopic technique will be most suitable and illustrating how different

243

vibrational techniques can be appropriate for different analytes. Models predicting total and

244

individual β-acids failed for all spectroscopic techniques (RPD values 1.0 – 2.4), but these

245

components are not considered critical for the screening of hops varieties (Table S1).

246

Assignment of IR and Raman Spectra of Bitter Acids and Xanthohumol. DFT

247

calculations were used to predict the optimized geometries of cohumulone, colupulone and

248

xanthohumol. The predicted IR and Raman vibrational modes were compared to the

249

experimental IR (Figure 3(a)) and Raman (Figure 3(b)) spectra of cohumulone and

250

colupulone. Calculated values were in good agreement with the experimental, with a mean

251

absolute deviation value of 13 cm-1 for the assigned bands.

252

Most of the IR spectral differences occurred in the region from 1700 – 1000 cm-1

253

(Figure 3(a)). The ring carbonyl band occurred at 1663 cm-1 in cohumulone and at 1652 cm-1

254

in colupulone. These were in good agreement with the DFT calculations which predicted the

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 12 of 28 255

bands at 1663 and 1660 cm-1 respectively. A medium intensity band was observed as a

256

shoulder in cohumulone at 1579 cm-1 and an equivalent colupulone band was observed at

257

1581 cm-1. These bands were assigned to the keto-enol carbonyl stretching vibration,

258

predicted at 1603 and 1608 cm-1 for cohumulone and colupulone respectively. The relatively

259

low frequencies of these carbonyl bands were consistent with previous vibrational

260

spectroscopic and DFT studies performed on compounds bearing similar intramolecularly

261

hydrogen bonded groups.41,42 The intense band at ~1525 cm-1 in both compounds was due to

262

a conjugated mode, involving large displacement of the intramolecularly bonded proton.

263

Lastly, a number of overlapping C-H bending modes gave rise to a strong band at ~1440 cm-1

264

in both compounds.

265

Selected regions of the Raman spectra of cohumulone and colupulone are shown in

266

Figure 3(b). The prenyl alkene stretching vibration at 1674 ± 1 cm-1 was prominent in the

267

spectra of both compounds (predicted at 1681 ± 1 cm-1). The cohumulone and colupulone

268

spectra deviated from one another in the region from 1660 – 1500 cm-1 (Figure 3(b)). Four

269

bands were predicted for each compound in this region with variable intensities. The

270

vibrations are complex and involve displacement throughout the conjugated backbones of the

271

molecules. The bands were loosely assigned at their maxima as diene stretches. The predicted

272

spectrum of the colupulone tautomer (T2) (Figure 1) had no intense vibrational modes in this

273

region, implying that the lower energy (T1) structure (Figure 1) was abundant in the solid

274

state of the compound. This was consistent with a published crystal structure.35 The other

275

noteworthy bands in the Raman spectra occured at 1452, 1383 and 1352 cm-1 in both

276

cohumulone and colupulone. These were in good agreement with the calculations for

277

overlapping C-H twisting, wagging and bending modes.

278 279

IR PLS-R models predicting UV reference data. The PLS-R models predicting total α and β-acid concentrations (by UV) from IR and Raman spectral data were interpreted

ACS Paragon Plus Environment

Page 12 of 34

Page 13 of 34

Journal of Agricultural and Food Chemistry

Page 13 of 28 280

visually by comparing the first three LVs of each model to the relevant bitter acid spectra

281

(Figure 4). When a LV is similar to spectra of the pure compound it implies that the model is

282

spectrally meaningful, but when the LVs no longer contain the features of the analyte, the

283

model may no longer be extracting spectrally relevant information and overfitting becomes a

284

consideration.43 This approach complements the standard approach for model interpretation

285

based on statistical data i.e. the validation data (Tables 1 and S1).

286

In Figure 4(a), the IR spectrum of cohumulone is overlaid with the first three LVs

287

from the PLS-R model for α-acids by IR spectroscopy. The first LV explained 73% of the

288

reference value variance and was almost identical to the profile of cohumulone – a sure

289

indication of effective modeling. Successive LVs described less variance and deviated

290

increasingly from the spectrum of the pure compound. In Figure 4(b), LVs from the IR β-

291

acids models are overlaid with colupulone. In this case the first LV did not mimic the

292

colupulone spectrum and explained only 46% of the reference data variance. In particular,

293

one of the key lupulone bands, at 1580 cm-1, was almost absent, making the profile more like

294

cohumulone. Subsequent LVs related more specifically to β-acid concentrations and the third

295

LV, which explained 9% of the variance in the reference data, had high loading coefficients

296

around 1580 cm-1. This inspection of LVs implied that the α-acids, which were nearly always

297

present in greater concentrations (Figure 2), were interfering with chemometric quantitation

298

of β-acids, explaining the poor performance of these models (Table S1).

299

Raman PLS-R models predicting UV reference data. The Raman PLS-R LVs are

300

compared to the relevant IR and Raman spectra in Figures 4(c) and 4(d). The first LV in the

301

α-acid model had high loading coefficients around 1673, 1452, 1382 and 1352 cm-1 (Figure

302

4(c)). These bands were all observed in the spectrum of cohumulone (Figure 3(b)),

303

suggesting that the model was extracting spectrally relevant data. However, the first LV also

304

had high loading coefficients around 1608 and 1626 cm-1 which were not observed in the

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 14 of 28 305

spectrum of cohumulone (Figure 3(b)). Xanthohumol, the conjugated chalcone responsible

306

for the yellow colour of lupulin (Figure 1), had intense Raman bands at both these

307

frequencies (Figure 5) and was present in the hops samples at concentrations ranging from

308

0.13 - 0.85 % (by HPLC, Figure 2). Despite its relatively low concentration, the contribution

309

of xanthohumol to the Raman spectrum of hops was pronounced due to the large and highly

310

polarizable π-conjugated system which makes the compound extremely Raman active (Figure

311

1). The relative intensities of xanthohumol and cohumulone (intensity × 10) are shown in

312

Figure 5, highlighting the extreme Raman activity of the bioactive chalcone. Figure 5 also

313

shows the vibrational displacement vectors of the most intense xanthohumol bands,

314

highlighting the complexity and delocalization of these vibrations.

315

After initial studies, it was feared that these intense xanthohumol bands would

316

interfere with PLS-R models and prohibit the use of Raman spectroscopy for the analysis of

317

hops bitter acids. This did not prove to be the case, as evidenced by the performance metrics

318

in Table 1. This was not altogether surprizing given the fact that xanthohumol concentrations

319

were correlated to α-acids (r2 = 0.68) and total bitter acids (r2 = 0.73). In the case of β-acids

320

models, xanthohumol had a more detrimental effect since concentrations of these components

321

were poorly correlated to one another (r2 = 0.46). This, in tandem with the interfering α-acids,

322

which were more concentrated in nearly all varieties, could explain the poor performance of

323

the Raman PLS-R models for β-acids (Table S1).

324

We have shown that NIR, IR and Raman spectroscopy could be used to quantitatively

325

screen for the key hops components α-acids, total bitter acids and cohumulone. It has also

326

been shown that portable NIR technology could be used for this purpose. Some of the

327

limitations of these techniques are also reported, including the fact that quantitation of β-

328

acids is unlikely to be achieved using this approach. Although some analytical challenges

329

have been overcome using “fused” spectroscopic datasets30,44, our preliminary investigation

ACS Paragon Plus Environment

Page 14 of 34

Page 15 of 34

Journal of Agricultural and Food Chemistry

Page 15 of 28 330

suggested that this approach did not improve the performance of the reported models. The

331

spectroscopic techniques required less sample preparation and, in most cases, greatly reduced

332

analysis times compared with traditional analyses. Solvent use was also eliminated. Some

333

vibrational features of cohumulone, colupulone and xanthohumol have been assigned.

334

Finally, the PLS-R LVs have been related to the pure spectra of the compounds for which

335

they are predicting, allowing a more insightful appraisal of the generated PLS-R models.

336 337

■ ASSOCIATED CONTENT

338

*S Supporting Information

339

Table S1: PLS-R model summaries (all models). This material is available free of charge via

340

the Internet at http://pubs.acs.org.

341

342

■ AUTHOR INFORMATION

343

Corresponding Author

344

*Phone: +64 3 4798354. E-mail: [email protected].

345

Funding

346

This research was supported by a University of Otago Doctoral Scholarship, in collaboration

347

with Plant & Food Research Capability Funding.

348

Notes

349

The authors declare no competing financial interest.

350

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 16 of 28

351

■ ACKNOWLEDGMENTS

352

We thank R. Rushton Green, L. Graham, D. Graham and C. Sansom for their technical

353

assistance and O. Watkins and E. Burgess for advice.

354 355

■ REFERENCES

356

(1) Neve, R. A. Hops; Chapman and Hall: London, 1991.

357

(2) Malowicki, M. G.; Shellhammer, T. H. Isomerization and degradation kinetics of hop

358

(Humulus lupulus) acids in a model wort-boiling system. J. Agric. Food Chem. 2005, 53,

359

4434-4439.

360

(3) Stevens, J. F.; Page, J. E. Xanthohumol and related prenylflavonoids from hops and beer:

361

to your good health! Phytochemistry 2004, 65, 1317-1330.

362

(4) Zanoli, P.; Zavatti, M. Pharmacognostic and pharmacological profile of Humulus lupulus

363

L. J. Ethnopharmacol. 2008, 116, 383-396.

364

(5) Molyneux, R. J.; Wong, Y. Nuclear magnetic resonance spectroscopic determination of α-

365

and β-acid homolog composition in hops. J. Agric. Food Chem. 1975, 23, 1201-1204.

366

(6) Intelmann, D.; Haseleu, G.; Hofmann, T. LC-MS/MS quantitation of hop-derived bitter

367

compounds in beer using the ECHO technique. J. Agric. Food Chem. 2009, 57, 1172-1182.

368

(7) Hughes, P. The significance of iso-α-acids for beer quality - Cambridge prize paper. J.

369

Inst. Brew. 2000, 106, 271-276.

370

(8) Beatson, R.; Ansell, K.; Graham, L. Breeding, development, and characteristics of the hop

371

(Humulus lupulus) cultivar ‘Nelson Sauvin’. N. Z. J. Crop Hort. Sci. 2003, 31, 303-309.

372

(9) American Society of Brewing Chemists. Hops 14. α-Acids and β-Acids in Hops and Hop

373

Extracts by HPLC (International Method). 2010, http://methods.asbcnet.org/summaries/hops-

374

14.aspx.

ACS Paragon Plus Environment

Page 16 of 34

Page 17 of 34

Journal of Agricultural and Food Chemistry

Page 17 of 28 375

(10) American Society of Brewing Chemists. Hops 6. α-Acids and β-Acids in Hops and Hop

376

Pellets by Spectrophotometry. 2010, http://methods.asbcnet.org/summaries/hops-6.aspx.

377

(11) European Brewery Convention. EBC 7.7 Alpha- and Beta-acids in hops and hop

378

products by HPLC; Brussels, 2010.

379

(12) Holtzel, A.; Schlotterbeck, G.; Albert, K.; Bayer, E. Separation and characterisation of

380

hop bitter acids by HPLC-H-1 NMR coupling. Chromatographia 1996, 42, 499-505.

381

(13) Hoek, A. C.; Hermans-Lokkerbol, A. C. J.; Verpoorte, R. An improved NMR method

382

for the quantification of α-acids in hops and hop products. Phytochem. Anal. 2001, 12, 53-57.

383

(14) Pusecker, K.; Albert, K.; Bayer, E. Investigation of hop and beer bitter acids by coupling

384

of high-performance liquid chromatography to nuclear magnetic resonance spectroscopy. J.

385

Chromatogr. A 1999, 836, 245-252.

386

(15) Halsey, S. A. Near Infrared reflectance analysis of whole hop cones. J. Inst. Brew. 1987,

387

93, 399-404.

388

(16) Axcell, B.; Tulej, R.; Murray, J. The determination of alpha-acids and moisture by near

389

infrared reflectance spectroscopy. Brew. Dig. 1981, 41, 18-19.

390

(17) Garden, S. W.; Pruneda, T.; Irby, S.; Hysert, D. W. Development of near-infrared

391

calibrations for hop analysis. Am. Soc. Brew. Chem. 2000, 58, 73-82.

392

(18) Foley, W. J.; McIlwee, A.; Lawler, I.; Aragones, L.; Woolnough, A. P.; Berding, N.

393

Ecological applications of near infrared reflectance spectroscopy–a tool for rapid, cost-

394

effective prediction of the composition of plant and animal tissues and aspects of animal

395

performance. Oecologia 1998, 116, 293-305.

396

(19) Carron, K.; Cox, R. Qualitative Analysis and the Answer Box: A Perspective on Portable

397

Raman Spectroscopy. Anal. Chem. 2010, 82, 3419-3425.

398

(20) Van Vuuren, J.; Meyer, J.; Claassens, A. Potential use of near infrared reflectance

399

monitoring in precision agriculture. Commu. Soil Sci. Plan. 2006, 37, 2171-2184.

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 18 of 28 400

(21) Killeen, D. P.; Sansom, C. E.; Lill, R. E.; Eason, J. R.; Gordon, K. C.; Perry, N. B.

401

Quantitative Raman Spectroscopy for the Analysis of Carrot Bioactives. J. Agric. Food

402

Chem. 2013, 61, 2701-2708.

403

(22) Baranska, M.; Schulz, H.; Siuda, R.; Strehle, M. A.; Rosch, P.; Popp, J.; Joubert, E.;

404

Manley, M. Quality control of Harpagophytum procumbens and its related

405

phytopharmaceutical products by means of NIR-FT-Raman spectroscopy. Biopolymers 2005,

406

77, 1-8.

407

(23) Baranska, M.; Schulz, H.; Joubert, E.; Manley, M. In situ flavonoid analysis by FT-

408

Raman spectroscopy: identification, distribution, and quantification of aspalathin in green

409

rooibos (Aspalathus linearis). Anal. Chem. 2006, 78, 7716-7721.

410

(24) Schonbichler, S. A.; Bittner, L. K. H.; Pallua, J. D.; Popp, M.; Abel, G.; Bonn, G. K.;

411

Huck, C. W. Simultaneous quantification of verbenalin and verbascoside in Verbena

412

officinalis by ATR-IR and NIR spectroscopy. J. Pharm. Biomed. Anal. 2013, 84, 97-102.

413

(25) Santos, J. R.; Sarraguça, M. C.; Rangel, A. O.; Lopes, J. A. Evaluation of green coffee

414

beans quality using near infrared spectroscopy: A quantitative approach. Food Chem 2012,

415

135, 1828-1835.

416

(26) Geladi, P.; Kowalski, B. R. Partial least-squares regression - a tutorial. Anal. Chim. Acta.

417

1986, 185, 1-17.

418

(27) Wold, S.; Sjöström, M.; Eriksson, L. PLS-regression: a basic tool of chemometrics.

419

Chemometr. Intell. Lab. 2001, 58, 109-130.

420

(28) Perez-Marin, D.; Garrido-Varo, A.; Guerrero, J. Non-linear regression methods in NIRS

421

quantitative analysis. Talanta 2007, 72, 28-42.

422

(29) Zhang, X.; Li, W.; Yin, B.; Chen, W.; Kelly, D. P.; Wang, X.; Zheng, K.; Du, Y.

423

Improvement of near infrared spectroscopic (NIRS) analysis of caffeine in roasted Arabica

ACS Paragon Plus Environment

Page 18 of 34

Page 19 of 34

Journal of Agricultural and Food Chemistry

Page 19 of 28 424

coffee by variable selection method of stability competitive adaptive reweighted sampling

425

(SCARS). Spectrochim. Acta. A Mol. Biomol. Spectrosc. 2013, 114, 350-356.

426

(30) Khaleghi, B.; Khamis, A.; Karray, F. O.; Razavi, S. N. Multisensor data fusion: A

427

review of the state-of-the-art. Inform. Fusion 2013, 14, 28-44.

428

(31) Chadwick, L. R.; Nikolic, D.; Burdette, J. E.; Overk, C. R.; Bolton, J. L.; van Breemen,

429

R. B.; Frohlich, R.; Fong, H. H. S.; Farnsworth, N. R.; Pauli, G. F. Estrogens and congeners

430

from spent hops (Humulus lupulus). J. Nat. Prod. 2004, 67, 2024-2032.

431

(32) Andersson, M. P.; Uvdal, P. New scale factors for harmonic vibrational frequencies

432

using the B3LYP density functional method with the triple-ζ basis set 6-311+ G (d, p). J.

433

Phys. Chem. A 2005, 109, 2937-2941.

434

(33) Clarke, T. M.; Gordon, K. C.; Officer, D. L.; Hall, S. B.; Collis, G. E.; Burrell, A. K.

435

Theoretical and spectroscopic study of a series of styryl-substituted terthiophenes. J. Phys.

436

Chem. A 2003, 107, 11505-11516.

437

(34) Scott, A. P.; Radom, L. Harmonic vibrational frequencies: an evaluation of Hartree-

438

Fock, Møller-Plesset, quadratic configuration interaction, density functional theory, and

439

semiempirical scale factors. J. Phys. Chem. 1996, 100, 16502-16513.

440

(35) Archer, R. P.; Tyrrell, E.; Singh, K. Colupulone. Acta Crystallogr. E-Structure Reports

441

Online 2007, 63, o1511-o1512.

442

(36) Barnes, R. J.; Dhanoa, M. S.; Lister, S. J. Standard normal variate transformation and de-

443

trending of near-infrared diffuse reflectance spectra. Appl. Spectrosc. 1989, 43, 772-777.

444

(37) Snee, R. D. Validation of regression models: methods and examples. Technometrics

445

1977, 19, 415-428.

446

(38) Krofta, K. Comparison of quality parameters of Czech and foreign hop varieties. Plant

447

Soil Environ. 2003, 49, 261-268.

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 20 of 28 448

(39) Likens, S.; Nickerson, G.; Haunold, A.; Zimmermann, C. Relationship between alpha

449

acids, beta acids, and lupulin content of hops. Crop Science 1978, 18, 380-386.

450

(40) Williams, P. C. Variables affecting near-infrared reflectance spectroscopic analysis. In

451

Near Infrared Technology in the Agriculture and Food Industries; Hill, L., Ed.; Am. Cereal

452

Assoc. Cereal Chem.: St. Paul, MN, 1987; Vol. 2, pp 145-169.

453

(41) Teslova, T.; Corredor, C.; Livingstone, R.; Spataru, T.; Birke, R. L.; Lombardi, J. R.;

454

Canamares, M. V.; Leona, M. Raman and surface-enhanced Raman spectra of flavone and

455

several hydroxy derivatives. J. Raman Spectrosc. 2007, 38, 802-818.

456

(42) Galasso, V. Probing the molecular and electronic structure of the lichen metabolite usnic

457

acid: a DFT study. Chem.Phys. 2010, 374, 138-145.

458

(43) Hawkins, D. M. The problem of overfitting. J. Chem. Inf. Comp. Sci. 2004, 44, 1-12.

459

(44) Biancolillo, A.; Bucci, R.; Magrì, A. L.; Magrì, A. D.; Marini, F. Data-fusion for

460

multiplatform characterization of an Italian craft beer aimed at its authentication. Anal. Chim.

461

Acta 2014, 820, 23-31.

462 463

ACS Paragon Plus Environment

Page 20 of 34

Page 21 of 34

Journal of Agricultural and Food Chemistry

Page 21 of 28

Table 1. Comparison of Hops Analyses by Vibrational Spectroscopy with UV and HPLC Analyses using Partial Least Squares Regression (PLS-R)

Spectroscopic PLS-R Technique Parameters r2 (RMSEC)a IR r2 (RMSEP)b # of LVs RPDc

UV Analyses Total αTotal Acids Acids 0.92 (1.2) 0.90 (1.6) 0.91 (1.3) 0.91 (1.8) 5 5 3.4 3.3

HPLC Analyses Alpha 0.86 (1.4) 0.89 (1.4) 3 2.9

Total 0.90 (1.5) 0.89 (1.8) 5 3.0

Cohumulone 0.92 (0.3) 0.91 (0.4) 5 3.4

Xanthohumol 0.85 (0.09) 0.88 (0.07) 5 2.9

Raman

r2 (RMSEC)c r2 (RMSEP)d # of LVs RPDe

0.88 (1.5) 0.91 (1.3) 4 3.3

0.87 (1.9) 0.90 (1.8) 4 3.1

0.86 (1.5) 0.90 (1.3) 4 3.2

0.84 (1.9) 0.90 (1.7) 4 3.1

0.90 (0.4) 0.91 (0.4) 7 3.3

0.87 (0.08) 0.87 (0.08) 4 2.8

NIR DRIFT

r2 (RMSEC)c r2 (RMSEP)d # of LVs RPDe

0.87 (1.5) 0.91 (1.4) 4 3.3

0.91 (1.6) 0.93 (1.5) 4 3.7

0.86 (1.4) 0.91 (1.2) 4 3.3

0.89 (1.6) 0.92 (1.5) 4 3.6

0.82 (0.5) 0.82 (0.5) 5 2.4

0.85 (0.09) 0.78 (0.10) 6 2.1

r2 (RMSEC)c 0.82 (1.8) 0.85 (2.0) 0.81 (1.7) 0.82 (2.0) 0.76 (0.6) 0.77 (0.10) 2 d r (RMSEP) 0.88 (1.5) 0.90 (1.8) 0.90 (1.3) 0.91 (1.6) 0.76 (0.6) 0.63 (0.13) # of LVs 7 7 7 7 7 6 RPDe 3.0 2.0 1.5 3.3 3.2 3.5 a b c Root mean square error of calibration. Root mean square error of prediction. Ratio of prediction to deviation.

NIR Micro

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 22 of 28

Figure 1. Structures of hops bitter acids and xanthohumol.

Figure 2. Concentrations of α-acids, β-acids and xanthohumol by HPLC and UV spectrophotometry.

Figure 3. IR and Raman spectral profiles of pure cohumulone and colupulone.

Figure 4. Overlay of spectral profiles of the pure bitter acids with latent variables from the Raman and IR PLS-R models predicting α-acids (a and b) and β-acids (c and d) by UV spectrophotometry. Traces are scaled for ease of comparison. Dashed lines highlight discussed bands.

Figure 5. Raman spectra of xanthohumol and cohumulone (intensity × 10) with the DFT predicted vibrational displacement vectors of the most intense xanthohumol modes inset (lengths of arrows are proportional to predicted displacements).

ACS Paragon Plus Environment

Page 22 of 34

Page 23 of 34

Journal of Agricultural and Food Chemistry

Page 23 of 28

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 24 of 28

ACS Paragon Plus Environment

Page 24 of 34

Page 25 of 34

Journal of Agricultural and Food Chemistry

Page 25 of 28

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 26 of 28

ACS Paragon Plus Environment

Page 26 of 34

Page 27 of 34

Journal of Agricultural and Food Chemistry

Page 27 of 28

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Page 28 of 28

Graphical Abstract

ACS Paragon Plus Environment

Page 28 of 34

Page 29 of 34

Journal of Agricultural and Food Chemistry

164x101mm (96 x 96 DPI)

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Structures of hops bitter acids and xanthohumol 325x251mm (72 x 72 DPI)

ACS Paragon Plus Environment

Page 30 of 34

Page 31 of 34

Journal of Agricultural and Food Chemistry

Concentrations of α-acids, β-acids and xanthohumol by HPLC and UV spectrophotometry 605x1246mm (100 x 100 DPI)

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

107x220mm (100 x 100 DPI)

ACS Paragon Plus Environment

Page 32 of 34

Page 33 of 34

Journal of Agricultural and Food Chemistry

Overlay of spectral profiles of the pure bitter acids with latent variables from the Raman and IR PLS-R models predicting α-acids (a and b) and β-acids (c and d) by UV spectrophotometry. Traces are scaled for comparison. Dashed lines highlight discussed bands. 459x323mm (300 x 300 DPI)

ACS Paragon Plus Environment

Journal of Agricultural and Food Chemistry

Raman spectra of xanthohumol and cohumulone (intensity × 10) with the DFT predicted vibrational displacement vectors of the most intense xanthohumol modes inset 632x357mm (72 x 72 DPI)

ACS Paragon Plus Environment

Page 34 of 34

Vibrational spectroscopy and chemometrics for rapid, quantitative analysis of bitter acids in hops (Humulus lupulus).

Hops, Humulus lupulus, are grown worldwide for use in the brewing industry to impart characteristic flavor and aroma to finished beer. Breeders produc...
3MB Sizes 0 Downloads 6 Views