Investigations into the total antioxidant capacities of cultivars of gluten-free grains using near-infrared spectroscopy

Investigations into the total antioxidant capacities of cultivars of gluten-free grains using near-infrared spectroscopy

Accepted Manuscript Investigations into the total antioxidative capacities of cultivars of gluten-free grains using near-infrared spectroscopy Verena ...

745KB Sizes 0 Downloads 40 Views

Accepted Manuscript Investigations into the total antioxidative capacities of cultivars of gluten-free grains using near-infrared spectroscopy Verena Wiedemair, Reinhold Ramoner, Christian W. Huck PII:

S0956-7135(18)30382-7

DOI:

10.1016/j.foodcont.2018.07.045

Reference:

JFCO 6258

To appear in:

Food Control

Received Date: 2 May 2018 Revised Date:

5 July 2018

Accepted Date: 27 July 2018

Please cite this article as: Wiedemair V., Ramoner R. & Huck C.W., Investigations into the total antioxidative capacities of cultivars of gluten-free grains using near-infrared spectroscopy, Food Control (2018), doi: 10.1016/j.foodcont.2018.07.045. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Investigations into the total antioxidative capacities of

2

cultivars of gluten-free grains using near-infrared

3

spectroscopy

4

RI PT

1

Verena WIEDEMAIRa, Reinhold RAMONERb, Christian W. HUCK*a

6 7

a

Institute of Analytical Chemistry and Radiochemistry, CCB – Center for Chemistry and Biomedicine, University of Innsbruck, Innrain 80-82, 6020 Innsbruck, Austria

8

b

SC

5

9

M AN U

Health University of Applied Sciences Tyrol, Innrain 98, 6020 Innsbruck, Austria

Mag. Verena Wiedemair, MA MSc

11 12 13 14

E-Mail: [email protected] Tel.: +43 512 507 57372 Adress: Innrain 80/82, 6020 Innsbruck, Austria

15

Priv.Doz. Reinhold Ramoner

16 17 18 19

E-Mail: [email protected] Tel.: +43 50 8648 4779 Adress: Innrain 98, 6020 Innsbruck

20

Univ.-Prof. Christian W. Huck (corresponding author)

21 22 23

E-Mail: [email protected] Tel.: +43 512 507 57304 Adress: Innrain 80/82, 6020 Innsbruck, Austria

AC C

EP

TE D

10

24

*

Corresponding author

1

ACCEPTED MANUSCRIPT

Abstract

25

26 Millet, buckwheat and oat are considered to be minor crops, hence chemical profiles for

28

different cultivars are rare. The examination of a sum parameter, like the total antioxidative

29

capacity (TAC) can thus be a first step to systematically assess the quality of different cultivars of

30

mentioned gluten-free grains and thereby serve as an indicator for the selection of cultivars for

31

food processing.

32

TAC of 20 common buckwheat, 14 proso millet and six common oat cultivars was analysed using

33

Folin-Ciocalteu assay and an optimized NIRS methodology. PLS regressions for milled and intact

34

samples were established and yielded a R2(CV) of 0.892 and 0.929. The SEPs for milled and intact

35

samples were approximately 1.7 and 1.6 mgGAE/g. Multivariate LOD and LOQ were also

36

calculated. LODmax for intact and milled samples was 1.72 and 2.80 mgGAE/g.

37

TAC varies considerably among cultivars of one species, emphasising the need for full chemical

38

profiles. Values for LOD and LOQ show that established PLS-R models can be used to quantify

39

TAC of buckwheat and oat cultivars. TAC of millet cultivars can be detected, however not

40

quantified.

42 43

SC

M AN U

TE D

EP

AC C

41

RI PT

27

44 45 46

Keywords: NIRS, total antioxidative capacity, gluten-free, buckwheat, millet, oat, Folin-Ciocalteu

2

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

47

3

ACCEPTED MANUSCRIPT

1. Introduction

49

In the past, only patients suffering from coeliac disease used to be on a gluten-free diet, as it is

50

the only cure for their illness (Alaedini & Green, 2005; Ludvigsson et al., 2013). Coeliac disease is

51

a chronic autoimmune disorder, which is triggered in genetically sensitive individuals by

52

exposure to gluten. The result are small intestinal inflammations and a severe ablation of

53

intestinal villi (Alaedini & Green, 2005; Ludvigsson et al., 2013).

54

Nowadays however, the gluten-free market benefits from the notion that going gluten-free is

55

healthier (Kim et al., 2016), which is why many people who do not suffer from coeliac disease

56

choose to cut out or reduce gluten in their diets. A study by the National Health and Nutrition

57

Examination found that in 2014 more than three times more people without coeliac disease

58

were avoiding gluten in their diets than in 2009 (Kim et al., 2016). Another study indicates that

59

around 22% of adults in North America are trying to avoid gluten in their diet, thus creating a $

60

8,8 billion market (Fromartz, 2015). Even more so, the market of gluten-free products is

61

expected to grow 10.2% over each year between 2015–2019 (Markets and Markets Web site,

62

2014). This market growth resulted in an increase in availability of gluten-free products in stores,

63

however many customers following this health trend are unaware of its downsides. Studies on

64

patients with coeliac disease, who follow a strict diet, have shown that many of them suffer from

65

a poor vitamin and mineral status (Hallert et al., 2002; Thompson, 2000; Udachan & Sahoo,

66

2017).

67

In order to improve the quality of gluten-free food, new cultivars with more favourable

68

properties must not only be bred, but also be analysed thoroughly and consequently be

69

implemented in foodstuff production. However, exhaustive chemical profiles of different

70

cultivars of gluten-free grains are still missing, leaving companies often clueless about the quality

71

of their primary products. The establishment of a complete nutritional profile is a tedious task.

AC C

EP

TE D

M AN U

SC

RI PT

48

4

ACCEPTED MANUSCRIPT That is why sum parameters, like the antioxidant capacity, can be used in a first step to assess

73

the quality of different cultivars. It is the hypothesis of this study that cultivars of one grain

74

species differ significantly from one another and thus it is important to review various cultivars,

75

in order to make informed decisions, when buying grains for food processing or breeding.

76

Antioxidants are substances, which inhibit oxidation, reactions promoted by oxygen or reactive

77

oxygen species (ROS) (Halliwell & Gutteridge, 2015), thus reducing the risk of several diseases,

78

including chronic inflammation, cardiovascular diseases, type 2 diabetes and cancer (Adom,

79

Sorrells, & Liu, 2005; Kris-Etherton et al., 2002; McKeown, Meigs, Liu, Wilson, & Jacques, 2002;

80

Scalbert, Johnson, & Saltmarsh, 2005). But natural ROS also have important function in the body

81

as second messengers, meaning that a delicate equilibrium of radical production and scavenging

82

is in existence (Clara et al., 2016; Mittler, 2002). Typical antioxidant substances are polyphenols,

83

vitamins and phenolic acids (Choi, Jeong, & Lee, 2007; Pietta, 2000; Scalbert et al., 2005).

84

The total antioxidative capacity can be estimated by various reactions and methods, which can

85

be divided into two groups according to their reaction mechanism: hydrogen atom transfer

86

(HAT) or single electron transfer (SET) (Prior, Wu, & Schaich, 2005). One widely used technique

87

for the determination of total antioxidative capacity is the Folin-Ciocalteu method, which

88

exhibits a SET mechanism. It was developed by Folin and Ciocalteu (Folin & Ciocalteu, 1927) and

89

later improved by Singleton et al. (Singleton, Orthofer, & Lamuela-Raventós) to monitor total

90

phenolic content. Later on studies showed that this assay is sensitive towards a variety of

91

molecules besides phenols, like vitamins, thiols and nitrogen containing moleclues (Amorati &

92

Valgimigli, 2015; Everette et al., 2010; Ikawa, Schaper, Dollard, & Sasner, 2003; Prior et al., 2005;

93

Rao, Kanjilal, & Mohan, 1978). This is why this assay is suitable for the analysis of the total

94

antioxidant capacity. Nevertheless, this method is still often used to determine only phenolic

95

content, as the extraction reagent used is often only sensitive towards phenols. This is not the

96

case in this study, which is why the Folin-Ciocalteu assay was chosen to determine total

AC C

EP

TE D

M AN U

SC

RI PT

72

5

ACCEPTED MANUSCRIPT antioxidant capacity. (Amorati & Valgimigli, 2015; Everette et al., 2010; Huang, Ou, & Prior,

98

2005; Prior et al., 2005).

99

Folin-Ciocalteu’s method is a wet chemical analysis, thus sample preparation as well as the

100

measurements themselves are quite time-consuming. Samples have first to be extracted and

101

normalized to a certain volume, before they can be exposed to Folin-Ciocalteu’s reagent and

102

lastly be measured with a UV/VIS-photometer at 750 nm. This is why in recent years, many

103

studies using near-infrared spectroscopy to determine the antioxidative capacity have been

104

published (Clara et al., 2016; Lu et al., 2011; Lucas, Andueza, Rock, & Martin, 2008; Schmutzler &

105

Huck, 2016; Wang, Yu, Fan, & Qu, 2009; Zhang, Shen, Chen, Xiao, & Bao, 2008). Near-infrared

106

spectroscopy provides a non-invasive and fast tool and is due to its broad bands often used for

107

the determination of sum parameters. Furthermore, overtones and combination vibrations

108

appearing in NIR spectra hold additional information (Blanco & Villarroya, 2002), which can be

109

interpreted using multivariate data analysis (MVA) – a known powerful tool for establishing

110

calibration and validation models (Næs, Isaksson, Fearn, & Davies, 2002). However, near-infrared

111

spectroscopy also has its limits which are usually estimated using well-established quality

112

parameters like the standard error of prediction. This study uses a multivariate, yet UIPAC

113

conform approach on LOD and LOQ prediction, in order to more accurately assess the

114

possibilities and limits for the use of near-infrared spectroscopy for grain analysis. The

115

application of near-infrared spectroscopy to estimate the antioxidative capacity of different

116

cultivars of gluten-free buckwheat, millet and oat grains is new, as differences within one species

117

might be too small to be recognized by NIRS.

118

Knowing which cultivar holds more favourable properties is very important to breeders, as well

119

as companies producing gluten-free foods. Thereby the study also investigates differences of

120

hulled and dehulled grains of one cultivars, as grains need to be dehulled for food production.

121

Additionally, the partial least square regression (PLS-R) model established using NIR data may be

AC C

EP

TE D

M AN U

SC

RI PT

97

6

ACCEPTED MANUSCRIPT 122

used to estimate the antioxidative capacity for other cultivars, not surveyed in this study as well

123

and has thus the potential to replace the time consuming and invasive wet-chemical Folin-

124

Ciocalteu measurements.

RI PT

125 126

2. Methods and Materials

128

2.1. Samples management

129

20 buckwheat, 14 millet and six oat cultivars were provided by Research Center Laimburg

130

(Bolzano, Italy) and Dr. Schär AG / SPA (Burgstall, Italy). All cultivars were provided as grains with

131

husk, but for some also dehulled grains were received. The supplementary material gives a

132

detailed overview over all cultivars used in this study and further shows in what state and from

133

what year they were. In total, 77 samples was analysed:

TE D

M AN U

SC

127

Twelve buckwheat samples with husk, harvested in 2016

135

Twelve dehulled buckwheat samples, harvested in 2016

136

Sixteen buckwheat samples with husk, harvested in 2016

137

Six oat samples with husk, harvested in 2016

139

AC C

138

EP

134

Ten millet samples with husk, harvested in 2016 Five millet samples with husk, harvested in 2015

140

Five dehulled millet samples, harvested in 2015

141

Eleven millet samples with husk, harvested in 2016

7

ACCEPTED MANUSCRIPT Upon receipt, samples were organized and later milled for extraction at 12000 rpm with a 6.0

143

grid in a Retsch mill ZM200 (Haan, Germany). Both milled and intact samples were analysed

144

using NIR spectroscopy. 1 g of milled samples were further used for acidified methanol

145

extraction and consequently for Folin-Ciocalteu analyses.

RI PT

142

146

2.2 Extraction

148

In order to measure the total antioxidative capacity, samples had to be extracted first. The

149

extraction reagent consisted of 95% methanol (≥99.9%) and 5% hydrochloric acid (35%)

150

(Chethan & Malleshi, 2007; Pradeep & Guha, 2011), and was always prepared on the same day

151

the extraction was performed. 1 g of milled sample was weighed into a 50 mL Falcon tube and

152

then 10 mL of extraction reagent was added. The tube was put into an ultrasonic bath at 60 °C

153

for 20 min. Next, the sample was centrifuged at 3500 rpm (25473 rcf) for 10 min. Preliminary

154

extraction studies showed that usually six repetitions are needed to extract all compound

155

contributing to the antioxidative capacity. Only for buckwheat samples with husk a total of nine

156

extraction cycles was needed. Lastly, the volume of the extracts was adjusted to 60 and 90 mL

157

with acidified methanol, respectively. All samples were stored at -80 °C overnight then Folin-

158

Ciocalteu measurements were performed.

M AN U

TE D

EP

AC C

159

SC

147

160

2.3. Folin-Ciocalteu measurements

161

The previously prepared extracts were thawed and then 1.5 mL distilled water, 100 µL extract,

162

100 µL Folin-Coicalteu’s phenol reagent (2N, Sigma Aldrich) and 1.3 mL sodium carbonate (50

163

mg/mL; made of sodium carbonate 99.0%, Sigma Aldrich) were mixed together in PMMA

164

(Polymethyl(methacrylate)) cuvettes (d = 1 cm) (Brand, Wertheim, Germany). Next, the cuvettes

165

were put in an oven at 60°C for 30 minutes. After that, the samples cooled down to room 8

ACCEPTED MANUSCRIPT temperature for 35 minutes and were measured at 750 nm with an UV/VIS spectrometer

167

(Genova Plus, Jenway, Stone, Staffordshire, UK). Three replicates were prepared for each

168

extract. The standard calibration curve was established using gallic acid in concentrations

169

between 0 and 250 mg/L gallic acid in equidistant 25 mg/L steps. Each concentration was

170

prepared as a triplicate and the calibration curve yielded a R2 of 0.995. The standard error of the

171

regression was 0.02 mg/L.

172

SC

RI PT

166

2.4. Near-infrared spectroscopy (NIRS)

174

All samples were measured with NIRS in intact and milled form at 21°C air-conditioner controlled

175

room temperature. The device used was a NIRFlex N-500 (Büchi, Flawil, Switzerland) with the

176

solids add-on. Samples were filled in a rotating cylindrical quartz cuvette (h = 25 mm, inner

177

diameter = 31.6 mm) and measured six times with 64 scans in the wavelength range between

178

10000–4000 cm-1 (1000–2500 nm) in diffuse reflection mode. The spectral resolution was 8 cm-1,

179

whereas the digital resolution was 4 cm-1.

TE D

EP

180

M AN U

173

2.5. Multivariate data analysis

182

All NIRS spectra recorded were exported and then analysed using the multivariate data analysis

183

software The Unscrambler X 10.4 (Camo, Oslo, Norway). First, spectra were transformed into -

184

logR spectra, then all of them were averaged by a factor of six, in order to get one spectrum per

185

sample. The descriptive statistics tool was applied, in order to determine necessary spectral pre-

186

treatments. Consequently, Savitzky Golay 2nd derivation was applied to remove additive baseline

187

offset, enhance small peaks and differences and smooth spectra. The number of smoothing

188

points was adjusted, according to the state of the sample. For milled samples the number of

189

smoothing points was chosen to be 17, and when intact grains were measured 13 smoothing

AC C

181

9

ACCEPTED MANUSCRIPT points were needed. Then, standard normal variate (SNV) was employed in the range between

191

8948–4032 cm-1 (1118–2480 nm) to reduce multiplicative scatter effects. Data outside the

192

selected wavenumber range was noise and thus not included in further spectral pre-treatments

193

or the establishment of PLS-R models. Spectra were then split into a calibration and a test set by

194

Kennard-Stone sample selection (Galvão et al., 2005). This algorithm maximizes the Euclidean

195

distance between the instrumental response vectors of the selected samples. This means that

196

the algorithm first generates a group consisting of the two most distant objects in the sample

197

set. Then, the object with the largest minimal distance to the group is added and so on (Galvão

198

et al., 2005; Rajer-Kanduč, Zupan, & Majcen, 2003). Last, orthogonal signal correction (OSC)

199

according to Fearn (Fearn, 2000) was used on the calibration set in the same wavenumber range

200

to filter out variables, which are orthogonal to the y data, i.e. the reference data. The established

201

OSC model was then used on the test set.

202

The calibration set consisted of two thirds of the samples (ncal=51). 35 of these samples are

203

unique hulled or dehulled cultivars. The remaining 16 spectra stem from eight unique cultivars.

204

and was used to build a PLS-R model, which was first validated using random leave one out cross

205

validation (LOOCV) with 20 segments and then using the test set, which consisted of the

206

remaining third of the samples (nval=26). The latter is called test set validation (TV). For the

207

establishment of the PLS-R model for intact and milled grains two and four factors were needed,

208

respectively.

209

Lastly, the limit of detection (LOD) was determined in a multivariate way, as proposed by

210

Allegrini and Olivieri (Allegrini & Olivieri, 2014). The two authors explain that since the matrix of

211

each sample is a little different, a univariate LOD is unfit to cope with a multivariate problem.

212

Hence, they introduced an LOD range, with a LODmin and an LODmax, which are calculated

213

according to equation (1) and (2)

AC C

EP

TE D

M AN U

SC

RI PT

190

10

ACCEPTED MANUSCRIPT 214

(1)

= 3.3

+ℎ

215

(2)

= 3.3

+ℎ

/

+ℎ

/

+ℎ

SEN refers to the sensitivity, which in a PLS-R is given by the inverse length of the regression

217

coefficient. var(x) is the variance of the x data, hence the variance in the spectral data, whereas

218

var(ycal) refers to the variance in the calibration data. h refers to the sample leverage and can be

219

calculated according to equations (3) and (4):

220

(3) ℎ

!" &

= ∑) #$%!&

(4) ℎ

= max ℎ

M AN U

221

SC

(*+ (

RI PT

216

Values below LODmin indicate that no antioxidative capacity can be detected. If a value above

223

LODmax is reached, the sample is certain to hold antioxidative capacity. Between LODmin and

224

LODmax the existence of any antioxidative capacity is not certain, as different values may be due

225

to matrix effects. Additionally, the limit of quantification (LOQ) can also be calculated in a

226

multivariate way, by tripling LODmin and LODmax. It is not possible to quantify antioxidative

227

capacity, if it is below LOQmin. A value above LOQmax indicates that quantification is possible.

228

Values between LOQmin and LOQmax cannot be quantified with certainty. If the matrix is very

229

homogenous throughout a sample set, then the values LODmin and LOQmin will be close to their

230

respective maxima.

232

EP

AC C

231

TE D

222

233

3. Results and Discussion

234

3.1 Folin-Ciocalteu measurements

235

As mentioned before, gallic acid was used to establish a calibration curve for all Folin-Ciocalteu

236

measurements. Consequently, the results are given in mg gallic acid equivalents per g sample 11

ACCEPTED MANUSCRIPT (mgGAE/g). The samples ranged from 1.4 to 18.8 mgGAE/g, which corresponds to 23 to 212

238

mg/L. The sample data collected had a mean of 7.19 mgGAE/g, a median of 6.18 mgGAE/g and a

239

standard deviation of 4.86 mgGAE/g. On average oat samples had an antioxidative capacity of

240

5.76 mgGAE/g (SD=1.29) and buckwheat and millet samples with husk an antioxidative capacity

241

of 12.90 mgGAE/g (SD=2.49) and 2.74 mgGAE/g (SD=0.36), respectively. Dehulled buckwheat

242

samples had an antioxidative capacity of 6.43 mgGAE/g (SD=0.55) and dehulled millet samples a

243

capacity of 1.45 mgGAE/g (SD=0.04). Figure 1 shows more detailed results of the Folin-Ciocalteu

244

measurements.

245

Samples with husk generally had a higher total antioxidative capacity than their dehulled

246

counterparts. Buckwheat samples with husk hold 2.3 times more antioxidants on average than

247

their complement. Millet samples with husk have an average 1.8 times higher total antioxidative

248

capacity than their dehulled counterpart. This poses a challenge, as most of the husk is removed

249

for food production. Buckwheat has the highest antioxidative capacity and the values for

250

dehulled buckwheat are still in the same range as oat samples with husk. Millet samples on the

251

other hand have a very low total antioxidative capacity. Additionally, millet cultivars had a very

252

homogenous distribution of antioxidative capacity throughout all samples. Buckwheat cultivars

253

on the other hand showed greater variance, even when just looking at samples with husk or

254

dehulled samples. Oat cultivars have quite similar total antioxidative capacities, however cv.

255

Irina and cv. ITA-7 seem to underperform in comparison with the other oat samples.

256

To compare the results with similar studies of antioxidative capacities of grains is difficult, as

257

most previously performed studies do either not state the exact cultivar studied or do not use

258

the same species. Nevertheless, a comparison of the data at hand with similar literature showed,

259

that the results for millet cultivars are in good agreement with previous, similar studies

260

examining other millet species in respect to their antioxidative capacity (Dykes & Rooney, 2006;

261

Pradeep & Guha, 2011). Little millet (Panicum sumatrense) has a reported antioxidative capacity

AC C

EP

TE D

M AN U

SC

RI PT

237

12

ACCEPTED MANUSCRIPT of 4.30 mgGAE/g (Pradeep & Guha, 2011), which is higher than the average TAC of 2.74

263

mgGAE/g of the investigated proso millets with husk. The cultivar Tiroler Rispenhirse has the

264

highest TAC of the examined proso millets with a value of 3.68 mgGAE/g. Another study showed

265

that different species of millet vary greatly in regard to their total antioxidative capacity with

266

proso millet only holding 0.5–1.0 mgCatechin/g and finger millet containing 5.5–5.9

267

mgCatechin/g (Dykes & Rooney, 2006).

268

The values for buckwheat samples are also in good accordance with previously performed

269

similar studies (Li, Yuan, Yang, Tao, & Ming, 2013). The reported antioxidant capacities for

270

buckwheat flours ranged from 8.05–15.11 mgGAE/g (Li et al., 2013). The buckwheat cultivars

271

with husk analysed in this study ranged from 9.08–18.78 mgGAE/g.

272

Looking at oat studies it becomes clear that these grains have rarely been investigated towards

273

their total antioxidant capacity. Furthermore, it seems that the oat cultivars used in this study

274

carry more antioxidants, than the cultivars used in other studies (Fardet, Rock, & Rémésy, 2008).

275

This may be due to investigating other oat species, other cultivars or because of this study using

276

oat samples with husk.

277

In general, it becomes clear that the intraspecific variance of total antioxidant capacity of

278

buckwheat and oat is high, thus establishing chemical profiles of each cultivar is important. The

279

selection of oat and buckwheat varieties with favourable properties is also important for

280

breeders, as high quality grains will lead to higher quality food products. Millet samples have less

281

variance, meaning that the selection of favourable grains is even more important. By effectively

282

choosing millet cultivars the quality can be improved and thus support individuals living on a

283

gluten-free diet.

AC C

EP

TE D

M AN U

SC

RI PT

262

284

13

ACCEPTED MANUSCRIPT 3.2. Near-infrared spectroscopy

286

Spectra were recorded between with the NIRFlex N-500 (Büchi, Flawil, Switzerland) between

287

10000–4000 cm-1. The samples rotated in the cylindrical quartz cuvettes during measurements.

288

An average spectrum of all intact samples is shown in figure 2. Table 1 lists important maxima

289

and their respective vibration according to Workman and Weyer (Workman & Weyer, 2008). It is

290

evident that the upper part of the NIR spectrum is determined by overtones, whereas

291

combination vibrations mainly contribute to the lower part of the spectrum.

RI PT

285

Table 1: Important vibrations in the NIR spectrum of millet, buckwheat and oat grains (Workman & Weyer, 2012).

Vibration C-H str. 2 overtone O-H & N-H str. 1st overtone C-H str. 1st overtone C-H str. 1st overtone O-H str. & O-H def. N-H str. C-H str. & C-H def. C-H str. & C-H def. comb.

M AN U

293

SC

292

Wavenumber / cm-1

294

TE D

nd

8312 6808 5788 5664 5184 4748 4380 4280

Wavelength / nm 1203 1469 1728 1766 1929 2106 2283 2336

3.3. Prediction of antioxidative capacity using NIRS

296

The data collected during Folin-Ciocalteu measurements was used as reference data, in order to

297

establish a prediction model for the total antioxidative capacity using NIRS. Necessary spectral

298

pre-treatments were identified using the descriptive statistics tool implemented in the software

299

The Unscrambler X 10.4 (see section 2.5). Two and four PLS-R factors were used for intact and

300

milled samples, respectively. Table 2 gives an overview over the statistical quality parameters

301

regression coefficient (R2), bias, standard error of cross validation (SECV), and standard error of

302

prediction (SEP) (Williams, 1987) of the established PLS-R models. Figure 3 shows the PLS-R

303

model of the calibration set for grains. The SEP was calculated using the 26 test set samples.

AC C

EP

295

14

ACCEPTED MANUSCRIPT 304 305

Table 2: Overview over statistical parameters of the established PLS-R models. CV refers to cross validation, whereas depicts test set validation.

State of the grains Intact Milled

SECV / mgGAE/g 1.1794 1.6156

R2 (CV) 0.9288 0.8918

Bias (CV) / mgGAE/g -0.0004 -0.0206

SEP / mgGAE/g 1.6381 1.6925

R2 (TV) 0.8911 0.8831

Bias (TV) / mgGAE/g -0.4534 0.0592

RI PT

306 The regression coefficients for models established with data from milled samples reach a value

308

of 0.89 for cross validation and 0.88 for test set validation, which is satisfactory for NIRS analyses

309

and indicates good correlation between the spectra and the reference values. The regression

310

coefficient for the cross validated model for intact samples reaches a value of 0.93 and for the

311

test set validated model a value of 0.89. This is even higher than for the respective results for

312

milled samples. Consequently, the intact samples also indicate a good correlation between the

313

spectra and the reference data.

314

For cross validation, the bias, which indicates the presence of systematic error, is close to zero,

315

as it should be. Looking at the test set validation, the bias value for intact and milled samples is

316

much higher. This has to be expected as samples used for validation are not part of the

317

calibration model.

318

SECV and SEP values are acceptable, considering that oat and buckwheat samples have an

319

antioxidative capacity between 5.8 and 18.8 mgGAE/g. Only millet samples with an antioxidative

320

capacity between 1.4 and 3.7 mgGAE/g are in the range of SECV and SEP values. When looking at

321

the LODmin and LODmax values (Table 3) for each model, it becomes clear that dehulled millet

322

samples can just barely be detected with certainty, as most of them have a value just above the

323

LODmax. The antioxidative capacity of millet samples with husk are well above the LODmax,

324

however quite similar to the LOQmin value, meaning that although the presence of antioxidative

325

capacity is certain, quantification is not yet possible. All of this suggest, that only the total

326

antioxidative capacities of buckwheat and oat can be quantified using NIRS. Proso millet cultivars

AC C

EP

TE D

M AN U

SC

307

15

ACCEPTED MANUSCRIPT seem to be too poor in antioxidants to quantify them. To further lower the LOD, solid phase

328

extraction (SPE) or a reflector, which enhances the signal has to be applied. Interestingly,

329

LODmin/LODmax and LOQmin/LOQmax are lower for intact samples. This seems to be

330

counterintuitive, as milled samples are usually more homogeneous. However, the milling

331

process might have an influence on the matrix, thus effecting some samples more, some less.

332

Furthermore, even intact grains are quite small, meaning that near-infrared radiation can

333

penetrate quite far into the sample and thus give accurate feedback about chemical properties.

RI PT

327

Table 3: LOD and LOQ ranges for the established PLS-R models. All values are given in mgGAE/g.

State of the seeds Intact Milled

LODmin 0.8666 1.2381

336

M AN U

335

SC

334

LODmax

LOQmin

LOQmax

1.7160 2.7969

2.5997 3.7144

5.1480 8.3906

Another way to evaluate SEP values is the calculation of the ratio of performance to deviation

338

(RPD). This value was first proposed by Williams (Williams, 1987) and indicates how accurate a

339

model is and for what purposes it can be used. The RPD is calculated according to equation (5): 23

(5) 01 = 2564

EP

340

TE D

337

The RPD value for intact samples is 2.7 and for milled samples 2.8. According to the quality

342

attribution, a RPD below 3 indicates that the model is adequate for screening analyses (Williams,

343

1987). Intact and milled samples show very similar values for SEP and RPD, indicating that near-

344

infrared radiation is able to penetrate even intact samples well. This shows that no sample

345

preparation is necessary to establish a calibration model for total antioxidative capacity for

346

gluten-free grains. Thus, the implementation of this model is very practical.

AC C

341

16

ACCEPTED MANUSCRIPT In summary, NIRS is able to detect even low amounts of total antioxidative capacity, however

348

only in buckwheat and oat samples they can be accurately quantified. Furthermore, sample

349

milling is not necessary, as it has little effect on the quality of the PLS-R model. By measuring

350

even more samples, it might be possible to further lower the SEP, thus improving the RPD and

351

the robustness of the model.

RI PT

347

352

SC

353

4. Conclusion

355

This study examines 40 cultivars of buckwheat, millet and oat in different forms towards their

356

antioxidative capacity using Folin-Ciocalteu measurements and NIRS, in order to get first insights

357

into the chemical profile of different cultivars of certain species of gluten-free grains. A total of

358

77 samples were investigated. Folin-Coicalteu measurements proved the importance of

359

investigating cultivars of one species separately, as the total antioxidative capacity differs

360

significantly. The differences are especially prominent among buckwheat samples, whereas

361

millet samples show fewer variation.

362

Furthermore, it was possible to establish PLS-R models using Folin-Ciocalteu measurements as

363

reference data with a R2 of about 0.9, which show good correlation between spectra and

364

reference data. By calculating LOD and LOQ in a multivariate way, it was demonstrated that the

365

antioxidative capacity of most millet cultivars can be detected with certainty, however it is too

366

low to be quantified. The antioxidative capacity of buckwheat and oat cultivars on the other

367

hand can be quantified in milled and intact grains using the established PLS-R models.

368

Calculation of RPD showed that the state of the grains – milled or intact – has only little influence

369

on the established models. Thus an approach with no sample preparation is suggested for future

370

studies.

AC C

EP

TE D

M AN U

354

17

ACCEPTED MANUSCRIPT Since the gluten-free market grows, it becomes important to secure the best possible nutritional

372

quality of buckwheat, oat and millet grains. This study showed that cultivars of one species differ

373

in terms of their antioxidative capacites, meaning that the selection of a specific grain has great

374

influence on gluten-free products. But, examining one sum parameter – the total antioxidative

375

capacity – makes it possible to only get an indication for which cultivar to use in breeding, as well

376

as food production. Naturally, also agricultural properties have to be taken into account when

377

breeding, however the balance between favourable agricultural and nutritional properties will

378

become more and more important as gluten-free market grows.

SC

RI PT

371

380

M AN U

379

5. Acknowledgement

382

The authors want to thank the European Union, the European Regional Development Fund and

383

the cross-border programme Interreg V-A Italy-Austria 2014–2020 (project “RE-Cereal”, ITAT

384

1005, P-7250-013-042) for financial support.

385

For sample management we want to thank Dr. Schär AG / SPA (Burgstall, Italy) and Research

386

Center Laimburg (Bolzano, Italy).

388

EP

AC C

387

TE D

381

389

6. Conflicts of Interest

390

The authors declare no conflicts of interest.

391 392 18

ACCEPTED MANUSCRIPT 393

References

394

Adom, K. K., Sorrells, M. E., & Liu, R. H. (2005). Phytochemicals and antioxidant activity of milled

395

fractions of different wheat varieties. Journal of agricultural and food chemistry, 53(6), 2297–

396

2306. https://doi.org/10.1021/jf048456d

398 399

Alaedini, A., & Green, P. H. R. (2005). Narrative review: Celiac disease: understanding a complex

RI PT

397

autoimmune disorder. Annals of internal medicine, 142(4), 289–298.

Allegrini, F., & Olivieri, A. C. (2014). IUPAC-consistent approach to the limit of detection in partial least-squares calibration. Analytical chemistry, 86(15), 7858–7866.

401

https://doi.org/10.1021/ac501786u

M AN U

402

SC

400

Amorati, R., & Valgimigli, L. (2015). Advantages and limitations of common testing methods for

403

antioxidants. Free radical research, 49(5), 633–649.

404

https://doi.org/10.3109/10715762.2014.996146

Blanco, M., & Villarroya, I. (2002). NIR spectroscopy: A rapid-response analytical tool. TrAC

TE D

405 406

Trends in Analytical Chemistry, 21(4), 240–250. https://doi.org/10.1016/S0165-

407

9936(02)00404-1

Chethan, S., & Malleshi, N. (2007). Finger millet polyphenols: Optimization of extraction and the

EP

408

effect of pH on their stability. Food Chemistry, 105(2), 862–870.

410

https://doi.org/10.1016/j.foodchem.2007.02.012

411

AC C

409

Choi, Y., Jeong, H.-S., & Lee, J. (2007). Antioxidant activity of methanolic extracts from some

412

grains consumed in Korea. Food Chemistry, 103(1), 130–138.

413

https://doi.org/10.1016/j.foodchem.2006.08.004

414

Clara, D., Pezzei, C. K., Schönbichler, S. A., Popp, M., Krolitzek, J., Bonn, G. K., & Huck, C. W.

415

(2016). Comparison of near-infrared diffuse reflectance (NIR) and attenuated-total-

416

reflectance mid-infrared (ATR-IR) spectroscopic determination of the antioxidant capacity of

19

ACCEPTED MANUSCRIPT 417

Sambuci flos with classic wet chemical methods (assays). Anal. Methods, 8(1), 97–104.

418

https://doi.org/10.1039/c5ay01314c

420 421

Dykes, L., & Rooney, L. W. (2006). Sorghum and millet phenols and antioxidants. Journal of Cereal Science, 44(3), 236–251. https://doi.org/10.1016/j.jcs.2006.06.007 Everette, J. D., Bryant, Q. M., Green, A. M., Abbey, Y. A., Wangila, G. W., & Walker, R. B. (2010).

RI PT

419

Thorough study of reactivity of various compound classes toward the Folin-Ciocalteu reagent.

423

Journal of agricultural and food chemistry, 58(14), 8139–8144.

424

https://doi.org/10.1021/jf1005935

Fardet, A., Rock, E., & Rémésy, C. (2008). Is the in vitro antioxidant potential of whole-grain

M AN U

425

SC

422

426

cereals and cereal products well reflected in vivo? Journal of Cereal Science, 48(2), 258–276.

427

https://doi.org/10.1016/j.jcs.2008.01.002

431 432 433 434

TE D

430

Systems, 50(1), 47–52. https://doi.org/10.1016/S0169-7439(99)00045-3 Folin, O., & Ciocalteu, V. (1927). On tyrosine and tryptophane determinations in proteins. Journal of biological chemistry, 73(2), 627–650.

Fromartz, S. (2015). The Gluten Enigma: Unraveling the gluten-free trend. Retrieved from

EP

429

Fearn, T. (2000). On orthogonal signal correction. Chemometrics and Intelligent Laboratory

http://www.eatingwell.com/article/285160/unraveling-the-gluten-free-trend/

AC C

428

Galvão, R. K. H., Araujo, M. C. U., José, G. E., Pontes, M. J. C., Silva, E. C., & Saldanha, T. C. B.

435

(2005). A method for calibration and validation subset partitioning. Talanta, 67(4), 736–740.

436

https://doi.org/10.1016/j.talanta.2005.03.025

437

Hallert, C., Grant, C., Grehn, S., Grännö, C., Hultén, S., Midhagen, G.,. . . Valdimarsson, T. (2002).

438

Evidence of poor vitamin status in coeliac patients on a gluten-free diet for 10 years.

439

Alimentary pharmacology & therapeutics, 16(7), 1333–1339.

20

ACCEPTED MANUSCRIPT

441 442 443 444

Halliwell, B., & Gutteridge, J. M. (2015). Free Radicals in Biology and Medicine (5th ed.). chap. 3. p. 77. Oxford: OUP Oxford. Huang, D., Ou, B., & Prior, R. L. (2005). The chemistry behind antioxidant capacity assays. Journal of agricultural and food chemistry, 53(6), 1841–1856. https://doi.org/10.1021/jf030723c Ikawa, M., Schaper, T. D., Dollard, C. A., & Sasner, J. J. (2003). Utilization of Folin-Ciocalteu

RI PT

440

445

phenol reagent for the detection of certain nitrogen compounds. Journal of agricultural and

446

food chemistry, 51(7), 1811–1815. https://doi.org/10.1021/jf021099r

Kim, H.-S., Patel, K. G., Orosz, E., Kothari, N., Demyen, M. F., Pyrsopoulos, N., & Ahlawat, S. K.

448

(2016). Time Trends in the Prevalence of Celiac Disease and Gluten-Free Diet in the US

449

Population: Results From the National Health and Nutrition Examination Surveys 2009-2014.

450

JAMA internal medicine, 176(11), 1716–1717.

451

https://doi.org/10.1001/jamainternmed.2016.5254

M AN U

Kris-Etherton, P. M., Hecker, K. D., Bonanome, A., Coval, S. M., Binkoski, A. E., Hilpert, K. F.,. . .

TE D

452

SC

447

453

Etherton, T. D. (2002). Bioactive compounds in foods: Their role in the prevention of

454

cardiovascular disease and cancer. The American journal of medicine, 113 Suppl 9B, 71S-88S. Li, F.-h., Yuan, Y., Yang, X.-l., Tao, S.-y., & Ming, J. (2013). Phenolic Profiles and Antioxidant

EP

455

Activity of Buckwheat (Fagopyrum esculentum Möench and Fagopyrum tartaricum L. Gaerth)

457

Hulls, Brans and Flours. Journal of Integrative Agriculture, 12(9), 1684–1693.

458

https://doi.org/10.1016/S2095-3119(13)60371-8

459

AC C

456

Lu, X., Wang, J., Al-Qadiri, H. M., Ross, C. F., Powers, J. R., Tang, J., & Rasco, B. A. (2011).

460

Determination of total phenolic content and antioxidant capacity of onion (Allium cepa) and

461

shallot (Allium oschaninii) using infrared spectroscopy. Food Chemistry, 129(2), 637–644.

462

https://doi.org/10.1016/j.foodchem.2011.04.105

21

ACCEPTED MANUSCRIPT 463

Lucas, A., Andueza, D., Rock, E., & Martin, B. (2008). Prediction of dry matter, fat, pH, vitamins,

464

minerals, carotenoids, total antioxidant capacity, and color in fresh and freeze-dried cheeses

465

by visible-near-infrared reflectance spectroscopy. Journal of agricultural and food chemistry,

466

56(16), 6801–6808. https://doi.org/10.1021/jf800615a Ludvigsson, J. F., Rubio-Tapia, A., van Dyke, C. T., Melton, L. J., Zinsmeister, A. R., Lahr, B. D., &

468

Murray, J. A. (2013). Increasing incidence of celiac disease in a North American population.

469

The American journal of gastroenterology, 108(5), 818–824.

470

https://doi.org/10.1038/ajg.2013.60

SC

Markets and Markets Web site. (2014). Gluten-free products market by type (bakery &

M AN U

471

RI PT

467

confectionery, snacks, breakfast cereals, baking mixes & flour and meat & poultry products),

473

sales channel (natural & conventional) & geography - global trends & forecasts to 2019.

474

Retrieved from http://www.marketsandmarkets.com/Market-Reports/gluten-free-products-

475

market-738.html

476

TE D

472

McKeown, N. M., Meigs, J. B., Liu, S., Wilson, P. W. F., & Jacques, P. F. (2002). Whole-grain intake is favorably associated with metabolic risk factors for type 2 diabetes and cardiovascular

478

disease in the Framingham Offspring Study. The American journal of clinical nutrition, 76(2),

479

390–398.

481 482 483 484 485

Mittler, R. (2002). Oxidative stress, antioxidants and stress tolerance. Trends in Plant Science,

AC C

480

EP

477

7(9), 405–410. https://doi.org/10.1016/S1360-1385(02)02312-9 Næs, T., Isaksson, T., Fearn, T., & Davies, T. (2002). A user-friendly guide to multivariate calibration and classification. Chichester: NIR Publications. Pietta, P.-G. (2000). Flavonoids as Antioxidants. Journal of Natural Products, 63(7), 1035–1042. https://doi.org/10.1021/np9904509

22

ACCEPTED MANUSCRIPT 486

Pradeep, S. R., & Guha, M. (2011). Effect of processing methods on the nutraceutical and

487

antioxidant properties of little millet (Panicum sumatrense) extracts. Food Chemistry, 126(4),

488

1643–1647. https://doi.org/10.1016/j.foodchem.2010.12.047

489

Prior, R. L., Wu, X., & Schaich, K. (2005). Standardized methods for the determination of antioxidant capacity and phenolics in foods and dietary supplements. Journal of agricultural

491

and food chemistry, 53(10), 4290–4302. https://doi.org/10.1021/jf0502698

492

RI PT

490

Rajer-Kanduč, K., Zupan, J., & Majcen, N. (2003). Separation of data on the training and test set for modelling: a case study for modelling of five colour properties of white pigment.

494

Chemometrics and Intelligent Laboratory Systems, 65(2), 221–229.

M AN U

495

SC

493

Rao, G. R., Kanjilal, G., & Mohan, K. R. (1978). Extended application of Folin-Ciocalteu reagent in

496

the determination of drugs. The Analyst, 103(1230), 993.

497

https://doi.org/10.1039/AN9780300993

499 500

Scalbert, A., Johnson, I. T., & Saltmarsh, M. (2005). Polyphenols: antioxidants and beyond.

TE D

498

American Journal of clinical nutrition, 81(1), 215–217. Schmutzler, M., & Huck, C. W. (2016). Simultaneous detection of total antioxidant capacity and total soluble solids content by Fourier transform near-infrared (FT-NIR) spectroscopy: A quick

502

and sensitive method for on-site analyses of apples. Food Control, 66, 27–37.

503

https://doi.org/10.1016/j.foodcont.2016.01.026

AC C

EP

501

504

Singleton, V. L., Orthofer, R., & Lamuela-Raventós, R. M. Analysis of total phenols and other

505

oxidation substrates and antioxidants by means of folin-ciocalteu reagent, 299, 152–178.

506

https://doi.org/10.1016/S0076-6879(99)99017-1

507

Thompson, T. (2000). Folate, Iron, and Dietary Fiber Contents of the Gluten-free Diet. Journal of

508

the American Dietetic Association, 100(11), 1389–1396. https://doi.org/10.1016/S0002-

509

8223(00)00386-2

23

ACCEPTED MANUSCRIPT 510

Udachan, I., & Sahoo, A. K. (2017). Quality evaluation of gluten free protein rich broken rice

511

pasta. Journal of Food Measurement and Characterization, 11(3), 1378–1385.

512

https://doi.org/10.1007/s11694-017-9516-3

513

Wang, Y., Yu, L.-y., Fan, X.-h., & Qu, H.-b. (2009). An approach to predicting antioxidative activities of natural products based on near infrared spectroscopy. Spectroscopy and spectral

515

analysis, 9(29), 2401–2404. https://doi.org/10.3964/j.issn.1000-0593(2009)09-2401-04

516

Williams, P. (1987). Variables Affecting Near-Infrared Reflectance Spectroscopic Analysis. In P.

517

Williams & K. Norris (Eds.), Near Infrared Technology in the Agriculture and Food Industries

518

(Chap. 8, 143–167). St. Paul: American Association of Cereal Chemists.

521 522 523

SC

M AN U

520

Workman, J., & Weyer, L. (2008). Practical guide to interpretive near-infrared spectroscopy. Boca Raton: Taylor & Francis.

Workman, J., & Weyer, L. (2012). Practical guide and spectral atlas for interpretive near-infrared spectroscopy (2nd ed.). Boca Raton, FL: CRC Press.

TE D

519

RI PT

514

Zhang, C., Shen, Y., Chen, J., Xiao, P., & Bao, J. (2008). Nondestructive prediction of total phenolics, flavonoid contents, and antioxidant capacity of rice grain using near-infrared

525

spectroscopy. Journal of agricultural and food chemistry, 56(18), 8268–8272.

526

https://doi.org/10.1021/jf801830z

528 529

AC C

527

EP

524

530

24

ACCEPTED MANUSCRIPT 531 532

Figure 1: Overview of the total antioxidative capacities of all samples. H refers to samples with husk, DH to samples without husk.

AC C

EP

TE D

M AN U

SC

RI PT

533

25

ACCEPTED MANUSCRIPT 534

Figure 2: Averaged spectrum of all intact samples after spectroscopic transformation to absorbance spectra.

AC C

EP

TE D

M AN U

SC

RI PT

535

26

ACCEPTED MANUSCRIPT 536

Figure 3: PLS-R model for grains established using the calibration set.

537 538 539

RI PT

540 541

SC

542

AC C

EP

TE D

M AN U

543

27

TE D

10

M AN U

SC

RI PT

15

0

Millet Oat

Athego−H Irina−H ITA−6−H ITA−7−H Rocky−H Scorpion−H

EP

ACCEPTED MANUSCRIPT

Early Bird−DH Early Bird−H−1 Early Bird−H−2 Early Bird−H−3 Gelbhirse−H Gierczyckie−H−1 Gierczyckie−H−2 GL RH 16106−H Horizon−DH Horizon−H−1 Horizon−H−2 Horizon−H−3 Huntsman−DH Huntsman−H−1 Huntsman−H−2 Huntsman−H−3 ITA−4−DH ITA−4−H−1 ITA−4−H−2 ITA−5−H Jagna−H−1 Jagna−H−2 Kornberger−H−1 Kornberger−H−2 Quartett−H Silberhirse−H Sunrise−DH Sunrise−H−1 Sunrise−H−2 Sunrise−H−3 Tiroler Rispenhirse−H

AC C

5

Bamby−DH Bamby−H−1 Bamby−H−2 Billy−H Cebelica−H Chinese Landrace−DH Chinese Landrace−H Devyatka−H Dikul−H ITA−1−DH ITA−1−H ITA−2−DH ITA−2−H ITA−3−DH ITA−3−H Kaerntner Hadn−H Koma−DH Koma−H−1 Koma−H−2 Kora−DH Kora−H−1 Kora−H−2 Koto−DH Koto−H−1 Koto−H−2 La Harpe−DH La Harpe−H−1 La Harpe−H−2 Lileja−DH Lileja−H−1 Lileja−H−2 Panda−DH Panda−H−1 Panda−H−2 Spacinska−H Temp−H VB Nojai−H VB Vokiai−DH VB Vokiai−H−1 VB Vokiai−H−2

Total antioxidant capacity / mgGAE/g

Buckwheat

ACCEPTED MANUSCRIPT

RI PT

1.2

M AN U EP

TE D

0.6

Buckwheat−DH Buckwheat−H Millet−DH Millet−H Oat

AC C

Absorbance / a. u.

SC

0.9

0.3

10000

9000

8000

7000

6000 −1

Wavenumbers / cm

5000

4000

ACCEPTED MANUSCRIPT

15

RI PT

● ●

●●







● ●

SC

● ●







●●

● ● ● ●

● ● ●







● ●● ● ●



● ●



5





● ●

● ● ●

● ● ●

EP





TE D

10





● ●

● ●● ● ● ●● ● ●● ● ● ● ● ● ● ●



● ● ●

0

5

10

Reference TAC / mgGAE/g





M AN U







AC C

Predicted TAC / mgGAE/g





15

Buckwheat Millet Oat

ACCEPTED MANUSCRIPT Highlights: First investigation into different cultivars of millet, buckwheat and oat



Total antioxidative capacity varies significantly between cultivars of one grain species



NIR model development for antioxidants of different cultivars of three grain species



Calculation of UIPAC conform multivariate LOD and LOQ for established PLS-R models

AC C

EP

TE D

M AN U

SC

RI PT