Rapid Quantitative Profiling of Lipid Oxidation Products in a Food

Mar 5, 2018 - Lipid oxidation is one of the most important reasons for the compromised shelf life of food emulsions. A major bottleneck in unravelling...
0 downloads 0 Views 929KB Size
Subscriber access provided by UNIV OF NEW ENGLAND ARMIDALE

Rapid quantitative profiling of lipid oxidation products in a food emulsion by H NMR 1

Donny W.H. Merkx, G.T. Sophie Hong, Alessia Ermacora, and John P. M. van Duynhoven Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.8b00380 • Publication Date (Web): 05 Mar 2018 Downloaded from http://pubs.acs.org on March 11, 2018

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 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 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.

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 26 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

1

Rapid quantitative profiling of lipid oxidation products in a food

2

emulsion by 1H NMR

3

Donny W.H. Merkxa,b,c, G.T. Sophie Honga, Alessia Ermacoraa and John P.M. van

4

Duynhovena,c,*

5 6

a

7

b

8

Wageningen, The Netherlands.

9

c

Unilever R&D Vlaardingen, Olivier van Noortlaan 120, 3133 AT, Vlaardingen, The Netherlands. Wageningen University & Research, Laboratory of Food Chemistry, Bornse Weilanden 9, 6708 WG,

Wageningen University & Research, Laboratory of Biophysics, Stippeneng 4, 6708 WE,

10

Wageningen, The Netherlands.

11

*Corresponding

12

[email protected]

13

Abstract

14

Lipid oxidation is one of the most important reasons for the compromised shelf life of food

15

emulsions. A major bottleneck in unravelling the underlying mechanisms is the lack of

16

methods that provide a rapid, quantitative and comprehensive molecular view on lipid

17

oxidation in these heterogeneous systems. In this study, the unbiased and quantitative nature

18

of 1H NMR was exploited to assess lipid oxidation products in mayonnaise, a particularly

19

oxidation-prone food emulsion. An efficient and robust procedure was implemented to

20

produce samples where the 1H NMR signals of oxidation products could be observed in a well

21

resolved and reproducible manner. 1H NMR signals of hydroperoxides were assigned in a

22

fatty acid and isomer specific way. Band-selective 1H NMR pulse excitation allowed

23

immediate quantification of both hydroperoxides and aldehydes with high throughput and

24

high dynamic range at levels of 0.03 mmol/kg with high precision (RSDR = 5.9%).

author.

Tel:

+31

10460

5534

ACS Paragon Plus Environment

+

E-mail

address:

John-

Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

25

Explorative multivariate data modelling of the quantitative 1H NMR profiles revealed that

26

shelf life temperature has a significant impact on lipid oxidation mechanisms.

27 28

Introduction

29

Unsaturated fatty acids are essential to the human diet and are for a major part consumed in

30

emulsified form. In food emulsions such as mayonnaise and dressings, unsaturated fatty acids

31

are prone to lipid oxidation, a chemical process that consists of two phases. In the primary

32

oxidation phase, hydroperoxides are formed via radical mechanisms. These hydroperoxides

33

then further degrade to form secondary oxidation products, such as aldehydes, hydroxides and

34

epoxides.1-3 Many of these secondary oxidation products are volatile and perceived by the

35

consumer as rancid, which compromises the sensorial quality of food emulsions.

36

Compared to bulk oil, lipid oxidation mechanisms in oil-in-water food emulsions become

37

more complex due to the presence of oil-water interfaces.2,3 In these emulsions, the presence

38

of transition metals and antioxidants at the droplet interface can respectively enhance and

39

retard lipid oxidation.4 Additionally, fatty acids become more hydrophilic after primary

40

oxidation, making them more exposed at the interface and thereby more vulnerable for further

41

oxidation reactions.5-6 Furthermore, colloidal transport of reaction intermediates can modulate

42

oxidation mechanisms.7-9 The interplay of lipid oxidation chemistry with these colloidal

43

effects and environmental conditions, such as temperature and oxygen pressure, is poorly

44

understood. To better understand the mechanisms at play, a comprehensive view on lipid

45

oxidation is required. One-dimensional methods,10 such as Peroxide Value (PV)11 and

46

Anisidine Value (AV)12, are unable to unravel different oxidation mechanisms. Therefore, we

47

deem it necessary to adopt a ‘foodomics’ approach,13 where the broad coverage of lipid

48

oxidation products is combined with explorative multivariate modelling. Currently, no

49

methods provide the required rapid, quantitative and comprehensive molecular view on both

ACS Paragon Plus Environment

Page 2 of 26

Page 3 of 26 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

50

primary and secondary lipid oxidation. LC-MS7 can detect and identify a wide range of

51

primary and secondary oxidation products in a single run, but suffers from low throughput

52

and the need of authentic standards for quantification. Volatile secondary lipid oxidation

53

products can sensitively be monitored by headspace GC,14 but quantification is compromised

54

by partitioning over sample and headspace and the need for authentic standards.

55

Quantitative 1H-NMR (qNMR) profiling has been introduced as a high-throughput method

56

that does not suffer from these drawbacks.1,8,9 NMR has been shown to identify15-16 and

57

quantify17-18 primary and secondary lipid oxidation products in bulk oils by using a binary

58

CDCl3:DMSO-d6 solvent. These investigations, however, focussed on bulk oils in advanced

59

stages of oxidation, and did not address the structural complexity of oil-in-water emulsions. In

60

these systems, the presence of susceptibility mismatches at the water-oil interfaces leads to

61

significant line broadening and thus compromises chemical shift resolution. Furthermore,

62

levels of non-oxidized lipids and hydroperoxides differed four-to-five orders in magnitude,

63

which poses a challenge to the dynamic range of the receiver. In this paper, we describe the

64

development and optimization of an 1H-NMR based profiling method that covers a wide

65

range of primary and major secondary oxidation products in food emulsions. The method

66

makes use of band selective pulses19-20, is quantitative, sensitive, unbiased and allows for high

67

sample throughput. The suitability of the 1H NMR profiling approach for distinguishing

68

temperature induced shifts on the lipid oxidation mechanism was demonstrated in a

69

mayonnaise shelf-life test.

70 71

Experimental section

72

Materials

73

CDCl3, DMSO-d6 and 4Å molsieves were purchased from Euriso-top (Saint-Aubin, France).

ACS Paragon Plus Environment

Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

74

Methyl oleate, methyl linoleate, methyl α-linolenate, glyceryl trioleate, glyceryl trilinoleate

75

and glyceryl trilinolenate were all purchased from Sigma Aldrich (Zwijndrecht, the

76

Netherlands). Medium chained tryglyceride (MCT) oil was prepared in-house by stripping

77

coconut oil.

78

Mayonnaise formulation and processing

79

Mayonnaise was manufactured by emulsification in a colloid mill. The aqueous phase was

80

prepared first, with the following ingredients: egg yolk (4.2% w/w), starch (1% w/w), sorbic

81

acid (0.01% w/w), sodium chloride (1.1% w/w), sugar (1.3% w/w), water (20.1% w/w).

82

Subsequently, rapeseed oil (65% w/w) and spirit vinegar 12% (2.6% w/w) were slowly added

83

to form the emulsion.

84

Mayonnaise storage

85

Aliquots of 1 g mayonnaise were stored in 20 mL screwcap vials in the dark at 20 ºC and 50

86

ºC. At 20ºC, five aliquots were removed from storage each week for 29 weeks. At 50 ºC, five

87

aliquots were removed from storage two to three times a week for 6 weeks. The aliquots were

88

then stored at -20ºC until further analysis.

89

Nuclear Magnetic Resonance (NMR) sample preparation

90

For NMR, two aliquots were used per formulation and timepoint. These were stored at -20 ºC

91

for 16h and subsequently thawed at room temperature, resulting in phase separation of the oil

92

and water phases. 250 µL of the upper layer (oil) was collected and transferred to a 2 mL

93

Eppendorf tube. To this, 750 µL of 5:1 CDCl3:DMSO-d6 (Euriso-top, France) was added.

94

This solvent was dried using 4Å molsieves (Euriso-top, France). After homogenizing the

95

solution, 550 µL was transferred to a 5-mm NMR tube.

96

NMR acquisition

97

Spectra were recorded on a Bruker Avance HD 700 MHz NMR spectrometer (Bruker

98

BioSpin, Switzerland) equipped with a 5-mm BBI-probe or an Avance III 600 MHz

ACS Paragon Plus Environment

Page 4 of 26

Page 5 of 26 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

99

spectrometer equipped with a 5-mm cryo-probe. The internal temperature of the probe was set

100

at 295 K. For each sample, a single pulse and a band-selective experiment were performed.

101

The single pulse experiments were recorded with 4 scans, a relaxation time of 5 sec and an

102

acquisition time of 4 seconds. The 90º pulse length was determined automatically (~7.2 µs),

103

the receiver gain was set to the maximum value where no digitizer overflow occurred. Two

104

band selective experiments were used. The first one used a hard 90º pulse followed by a

105

selective gradient enhanced Gaussian inversion pulse.21 The second one used a double echo

106

with gradients using selective refocussing with a RE-BURP shaped pulse and a 90 º pulse in

107

between the 180º pulses to refocus J-evolution.22-23 The length of the shaped pulses was 1-2

108

ms, and band-selective spectra were recorded with 16-32 scans. The relaxation and

109

acquisition times were respectively set to 5 and 2.7 s, the 90º pulse length was determined

110

automatically (~7.2 µs). The data was processed with Bruker TopSpin 3.2 software. Before

111

Fourier transformation, an exponential window function with a line broadening factor of 0.3

112

was applied, followed by automatic baseline correction and phase correction. Band-selective

113

NOESY and TOCSY experiments were used for spectral assignments.21

114

Calculations

115

The levels of peroxides and aldehydes (c ) are both expressed in mmol/kg and are calculated

116

by Equation 1 for a single pulse experiment recorded under quantitative24 acquisition

117

conditions: c (mmol kg) =

I N 10 ∙ ∙ I N MW

Eq. 1

118 119

and by Equation 2 for spectra recorded with band selective excitation pulses: c mmol kg =

I N ns sin"p $ RG 10 ∙ ∙ ∙ ∙ ∙ ∙ SelFactor I N ns sin(p ) RG MW

ACS Paragon Plus Environment

Eq. 2

Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

120

Here, - is the NMR signal intensity after signal integration, N is the number of protons, MW is

121

the molecular weight (average triglyceride ~ 880 g·mol-1), ns is the number of scans, RG is the

122

receiver gain, p is the pulse angle in degrees. Index ox stands for oxidized lipid product, TG

123

for triglyceride, zg for the direct pulse experiment and sel for the band-selective

124

experiment. SelFactor stands for the empirically determined factor which corrects for the

125

signal intensity loss due to relaxation differences between the single pulse and band-selective

126

experiments. The spectral integration region to determine I is δ 4.4-4.0 ppm, with N

127

corresponding to the 4 outer protons of the glycerol backbone. The glycerol backbone is not

128

expected to undergo changes upon oxidation, hence its 1H NMR signal can be used as internal

129

standard.18 The integrals I are determined from the spectral integration regions listed in

130

Table 1, all with N corresponding to 1.

131

Peroxide Value (PV)

132

PV values were determined according to the international standard procedure ISO 3960.11 The

133

method in short: 5.0 g oil was added to a 25 mL iso-octane/acetic acid solution and a 1.0 mL

134

saturated potassium iodide solution. After 5 minutes, the reaction was quenched with 100 mL

135

water and 2.0 mL starch solution was added. The mixture was titrated with a sodium

136

thiosulfate solution until the dark blue solution became transparent.

137

Headspace Gas Chromatography (HS-GC)

138

For method comparison, mayonnaise samples were measured with both Headspace GC as

139

well as 1H NMR. To semi-quantitatively assess the hexanal content, a hexanal in MCT-oil

140

standard sample was included in the measurement series. The headspace GC analyses were

141

performed on a 7890A gas chromatograph (Agilent Technologies, United States). The

142

screwcap vials were placed in an automated headspace sampler at 60 ºC and allowed to

143

equilibrate for at least 30 minutes. Headspace samples of 0.5 mL were injected with a gas-

144

tight syringe in split mode (ratio 1:40) onto a DB Wax column (J&W Scientific, Belgium)

ACS Paragon Plus Environment

Page 6 of 26

Page 7 of 26 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

145

with specifications: length 20 m, ID 0.18 mm and film thickness of 0.3 µm. The GC was

146

equipped with a 5975C MS detector (Agilent Technologies, United States) using an electron

147

source of 70 eV, m/z 33-150 and solvent delay of 0.75 minutes. Temperatures at the inlet and

148

detector were respectively 200 °C and 230 °C. The temperature program for GC was: 35 °C

149

(2 min) – 35 °C/min to 200 °C – 200 °C (2 min). The helium gas flow through the column

150

was 1.0 mL/min. Peak areas (target response) of hexanal were monitored and compared with

151

the standard sample (hexanal in MCT oil) to semi-quantify the hexanal content.

152

Validation

153

Repeatability (r) and within-laboratory reproducibility (R) were determined by measuring

154

three samples in duplicate for four days (24 measurements in total).25 The samples were

155

stored in the freezer (-20 ºC) prior to analysis. For hydroperoxides and aldehydes, respectively

156

three and two different concentration levels were analyzed. The r and R for each

157

concentration level were obtained using an ANalysis Of VAriance (ANOVA) model. Results

158

were expressed as relative standard deviations of repeatability (RSDr) and within-laboratory

159

reproducibility (RSDR). The limit of detection (LoD) and limit of quantification (LoQ) were

160

respectively based on estimates of ten and thirty times the standard deviation from ten

161

samples containing small amounts of oxidation products. The trueness of results obtained by

162

application of the SelFactor was determined by measuring a set of 5 oil sample in duplicate

163

with both the 1H NMR method and PV titration method. Method offset and proportional bias

164

were determined by linear regression.

165

Principal Component Analysis (PCA)

166

Principal component analysis (PCA) was performed using SIMCA 14 (Umetrics, Sweden).

167

Twenty quantified concentrations of the oxidation products were used as PCA input.

168

Oxidation products below the detection limit were set at 0. The first three principal

ACS Paragon Plus Environment

Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

169

components were calculated. Pareto scaling without variable centering (ParN) was applied to

170

the data set.

171

172

Results and discussion

173

NMR sample preparation

174

For the analysis of lipid oxidation products in food emulsions we used a binary solvent

175

mixture, composed of 5:1 CDCl3:DMSO-d6.18 The binary solvent was used because DMSO

176

forms hydrogen bonds with hydroperoxides, which in turn narrowed the width of

177

hydroperoxide signals and moved them downfield, resulting in well dispersed peaks.

178

However, when dissolving a full emulsion into this binary solvent, the present water partially

179

dissolved as well. This water then broadened the hydroperoxide peaks significantly and

180

effectively countered the effect DMSO had on line width and dispersion. To avoid this, we

181

extracted the lipids prior to analysis. A quick and convenient method to achieve this was by

182

breaking the emulsion in a freeze-thaw step. This resulted into a clear separation of the

183

aqueous (bottom) and oil (top) phases, after which the latter could conveniently be transferred

184

for further analysis. The freeze-thaw step complied well with the logistic requirements of

185

shelf-life tests where samples are typically stored in a freezer at either -20ºC or -80ºC between

186

sampling and actual measurement.

187

To investigate whether the oil-layer correctly represented the oil in the food emulsion, we

188

compared the freeze-thaw extraction with the established, but labour-intensive, chloroform-

189

methanol extraction.26 We found no significant differences in hydroperoxide concentrations

190

between the two extraction methods (see Supporting Info, Figure S1). Compared to the

191

samples obtained by chloroform extraction, the linewidths of the hydroperoxide signals in

192

samples produced by freeze-thawing were narrower. This can be attributed to the presence of

ACS Paragon Plus Environment

Page 8 of 26

Page 9 of 26 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

193

water in the chloroform extracts, which compromises the line narrowing effect of the binary

194

solvent.

195

Dynamic range improvement

196

In the early stages of lipid oxidation, primary oxidation products are typically present at ppm

197

level. Therefore, in CDCl3:DMSO-d6 lipid extracts, signals of these oxidation products could

198

be up to five orders of magnitude smaller than the abundant background signals of non-

199

oxidized lipids. Even with a modern receiver, digitizer noise compromised the detection of

200

such low levels (Figure 1A). To reduce the digital noise in the hydroperoxide and aldehyde

201

regions, the signal of the abundant non-oxidized lipids needed to be suppressed. For this

202

purpose, we excited the signals of the hydroperoxides and aldehydes with a 1D band selective

203

gradient pulse using a shaped Gaussian inversion pulse21. This pulse sequence required

204

separate acquisitions where respectively the hydroperoxide region (δ 11.5-10.5 ppm) and the

205

aldehyde region (δ 10.0-9.0 ppm) were excited. By using a RE-BURP shaped pulse,22

206

however, we were able to uniformly excite both the hydroperoxide (δ 11.5-10.5 ppm) and

207

aldehyde signals (δ 10.0-9.0 ppm). This pulse program effectively reduced the overall

208

measurement time by a factor two, while retaining signal-to-noise ratio.

ACS Paragon Plus Environment

Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Figure 1. The

1

H NMR spectrum (700 MHz) of the hydroperoxide region of a

CDCl3:DMSO-d6 lipid extract prepared from mildly oxidized mayonnaise sample. The top spectrum (A) is recorded with a regular excitation with a hard 900 pulse, the bottom spectrum (B) with a band-selective pulse. 209

The use of band-selective pulses allowed for a significant increase of receiver gain, resulting

210

in improved digitization and a threefold improved signal-to-noise ratio (Figure 1B). The use

211

of a 1H NMR frequency of 700 MHz relative to 600 MHz did not yield better resolved

212

hydroperoxides and aldehydes signals. This can be attributed to the linewidths of these signals

213

which are typically 1-2 Hz. Hence, we proceeded by recording quantitative NMR profiles

214

using a 600 MHz NMR instrument equipped with a 5-mm cryoprobe.

215

1

216

Detailed assignments of 1H NMR aldehyde signals in the binary CDCl3:DMSO-d6 solvent

217

were derived from literature15, 17, 27. The chemical shifts of these signals (Table 2) were stable

218

within large sample series, rendering them useful for assessment in profiling mode. To date,

219

hydroperoxide 1H NMR signals in the binary CDCl3:DMSO-d6 solvent have been assigned for

220

three commonly occurring unsaturated fatty acids: oleic acid (O, 18:1), linoleic acid (LA,

H NMR spectral assignments of hydroperoxide and aldehyde signals

ACS Paragon Plus Environment

Page 10 of 26

Page 11 of 26 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

221

18:2) and α-linolenic acid (αLn, 18:3).18 These 1H NMR assignments were for a major part

222

derived from the oxidation products of the methyl esters of these fatty acids. To obtain more

223

specific assignments, we pursued a more a detailed structural elucidation of these products.

224

However, established 2D homo- and heteronuclear NMR techniques could not be used due to

225

the unfavorable dynamic range of the levels of oxidation products vs non-oxidized lipids.

226

Hence, 1D band-selective NOESY and TOCSY pulse sequences21 were deployed to obtain

227

assignments specific for hydroperoxide position in the fatty acid chain and cis-trans

228

isomerism. 1D NOESY experiments were used to transfer the magnetization from the

229

hydroperoxide (OOH) to the neighboring allylic proton (CH). After identifying the allylic

230

proton, we used 1D TOCSY experiments to further transfer magnetization to neighboring

231

olefinic protons (HC=CH). The J-coupling of these olefinic protons indicated whether the

232

originally excited hydroperoxide is adjacent to a cis- or trans-configured double bond. This

233

allowed us to not only identify the fatty acid that was attached to the hydroperoxide, but also

234

the conformation of the double bonds. These assignments were performed for oleic, linoleic

235

and α-linolenic acid. For oleic and linoleic acid methyl ester hydroperoxides, different

236

geometric isomers (cis/trans) were annotated. The α-linolenic acid hydroperoxides were

237

assigned based on hydroperoxide position (inner vs. outer vs. cyclized).

238

The OOH-signals of the oxidized methyl esters all displayed resonances between δ 11.5-10.5

239

ppm. For oleic acid, two peak-clusters were sufficiently base-line separated to be excited by a

240

band-selective 1D NOESY pulse sequence. The high and low field signals respectively

241

corresponded to cis and trans-isomers. In a similar way, cis-trans- and trans-trans-isomers

242

could be distinguished for the linoleic acid methyl ester. We were not able to further elucidate

243

whether the hydroperoxides were attached to the 8- and 11-positions for oleic acid or the 9-

244

and 13-positions for linoleic acid, since the chemical shift of the allylic protons (next to the

245

hydroperoxide) was not significantly influenced by the position in the fatty acid alkyl chain.

ACS Paragon Plus Environment

Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

246

In the case of α-linolenic acid methyl esters, the major distinction between the hydroperoxides

247

formed by autoxidation, was not the isomeric configuration of the double bonds, but rather the

248

position of the hydroperoxide moiety on the chain. The hydroperoxides were either formed on

249

the ‘inside’ or ‘outside’ of the three double bonds. Moreover, the ‘inner’-isomers can cyclize

250

internally, which generates additional cyclized hydroperoxides.28 These cyclization

251

mechanisms are inhibited by α-tocopherol,28 which enabled us to identify the ‘inner’-isomers

252

and cyclized-isomers, since these were only present in samples without α-tocopherol.

253

Since the above described analyses were done on the methyl esters, we needed to translate

254

this to the triglycerides. To this end, we oxidized model triglycerides that consist of only one

255

type of fatty acid (O, LA, α-Ln) and compared their 1H NMR spectra with those of the

256

corresponding methyl esters. The spectra of hydroperoxides in triglycerides were almost

257

identical to those in methyl esters. The chemical shifts of these signals did not significantly

258

change, which enabled straightforward assignments of hydroperoxides signals originating

259

from oxidized fatty acids in their esterified triglyceride form. In Figure 2 and Table 1 the

260

assignments of hydroperoxides signals from oxidized fatty acids are summarized.

ACS Paragon Plus Environment

Page 12 of 26

Page 13 of 26 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Figure 2. 1H NMR spectra obtained with band selective excitation, with assignment coding, according to Table 2. The sample contained 25% oxidized oil in 75% 5:1 CDCl3:DMSO-d6. Top: Hydroperoxide region (δ 11.24-10.57 ppm); bottom: Aldehyde region (δ 9.81-9.44 ppm).

ACS Paragon Plus Environment

Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47

261

Page 14 of 26

Table 1. 1H NMR signal assignments of the peroxides and aldehydes annotated in Figure 2. Hydroperoxides

Aldehydes Peak

Chem. shift (ppm)

Annotation

Cyclized

K

δ 9.79-9.755

Small n-alkanals + unassigned

αLn

Cyclized

L

δ 9.75-9.73

n-alkanals (n>5)

δ 11.05-11.00

αLn

Cyclized

M

δ 9.71-9.69

unassigned

D

δ 10.99-10.96

αLn

12-OOH, 13-OOH

N

δ 9.65-9.64

E

δ 10.95-10.91

αLn

9-OOH, 16-OOH

O

δ 9.635-9.62

F

δ 10.91-10.87

LA

9-trans,cis-OOH, 13-cis,trans-OOH

P

δ 9.62-9.61

G

δ 10.87-10.82

LA

9-trans,transOOH, 13trans,trans-OOH

Q

δ 9.61-9.58

H

δ 10.82-10.80

unassigned

R

δ 9.58-9.54

4-hydro(pero)xy-trans-2alkenals,

I

δ 10.80-10.77

9-trans-OOH, 13trans-OOH

S

δ 9.53-9.50

Trans,trans-2,4-alkadienals, 4,5-epoxy-trans-2-alkenals

Peak

Chem. shift (ppm)

FA

Annotation

A

δ 11.25-11.12

αLn

B

δ 11.10-11.05

C

Partial chemical structure

unassigned

unassigned

O

unassigned

unassigned

ACS Paragon Plus Environment

Chemical structure

Page 15 of 26 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47

Analytical Chemistry

J

δ 10.74-10.71

O

9-cis-OOH, 13cis-OOH

T

δ 9.50-9.475

ACS Paragon Plus Environment

Trans-2-alkenals

Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 26

262

Quantification of hydroperoxides and aldehydes

263

For samples with high concentrations of oxidized lipids (>>1 mmol/kg), the hydroperoxide,

264

aldehyde and triglyceride signals can all be excited by a single pulse experiment with a

265

sufficient signal-to-noise ratio. For such samples, the methylene signals at 4.4-4.0 ppm were

266

used as an internal reference to quantify the total amount of hydroperoxides and aldehydes in

267

mmol/kg in a straightforward manner by using Equation 1. Samples with low amounts of

268

oxidized lipids required the use of band selective experiments to acquire hydroperoxide and

269

aldehyde 1H NMR signals with sufficient sensitivity. For their quantification, single pulse

270

experiments were used to acquire the methylene signals. To account for the different

271

acquisition schemes of these two spectra, we implemented an empirical correction factor. This

272

‘SelFactor’ was determined by comparing signal intensities of spectra acquired by

273

respectively single pulse and band-selective excitation for samples with high levels of

274

oxidation products.21 In the calculations of concentrations of oxidation products, according to

275

Equation 2, we also accounted for differences in excitation pulse angle, receiver gain and

276

number of scans between the single pulse and band selective experiments. The band-selective

277

and single-pulse experiments were recorded with a recycle delay of 5 seconds, the SelFactor

278

also incorporated the effect of partial saturation by comparison of signal intensities of spectra

279

recorded under fully relaxed conditions (recycle delay of 30 seconds).

280

The accuracy of the introduced correction factors in the 1H NMR approach on hydroperoxides

281

was verified by comparison of results with the classical PV titration method.

282

regression between the total hydroperoxide values obtained by NMR and PV methods

283

resulted in an off-set bias and proportional bias that were not statistically significant (see

284

Supporting Info, Table S1). For 1H NMR band-selective measurements performed within five

285

minutes, the limits of detection and quantification for individual hydroperoxide and aldehydes

286

were respectively 0.01 and 0.03 mmol/kg (see Table 2). This is comparable to the

ACS Paragon Plus Environment

Linear

Page 17 of 26 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

287

conventional and more cumbersome PV method, which only measures total hydroperoxide

288

concentration.

289

We also compared the aldehyde concentrations determined by 1H NMR with the headspace

290

GC method, which is commonly used in shelf-life monitoring of food emulsions since it

291

measures a perceivable volatile and rancid secondary oxidation product. Figure 3 shows a

292

good correlation (R2 = 0.93) between hexanal semi-quantified in the head space of a freeze-

293

thawed mayonnaise by GC-MS and the total n-alkanal concentrations as quantified directly in

294

the extracted oil phase by 1H NMR. The linear relationship is in line with vegetable oil being

295

a near ideal solvent for hexanal, with a gas-oil partitioning coefficient K ≈ 1·10-3.29 It is

296

important to note that hexanal concentrations cannot be determined by 1H NMR due to signal

297

overlap with other n-alkanals (n>5), which likely explains the small offset on the horizontal

298

axis. The summation of n-alkanal levels as determined by 1H NMR also partially explains the

299

proportional bias versus hexanal levels determined in the headspace. Another factor that can

300

explain the proportional bias is the elevated measurement temperature for the headspace GC-

301

MS methods compared to the ambient condition for 1H NMR. The low gas-oil partitioning

302

coefficient K implies that most hexanal is still dissolved in the oil layer and only a small

303

amount is present in the headspace. This makes the GC method very prone to subtle variations

304

in K, whereas this hardly impact on concentrations as determined by the 1H NMR method.

305

Furthermore, the headspace GC method assumes that the gas-matrix partition coefficient does

306

not change during oxidation and therefore ideally requires the analyst to reestablish the gas-

307

matrix coefficients of each analyte at different oxidation stages.30 In summary, even though

308

the methods intrinsically differ in what and how they measure alkanals, they both show the

309

same trend, with the main advantage of the 1H NMR method is that an absolute quantification

310

of n-alkanals in the oil phase can be provided, not biased by variation in partition coefficients.

ACS Paragon Plus Environment

Analytical Chemistry

0.35 0.3

hexanal GC (mmol/kg)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

0.25 0.2 0.15 0.1 y = 0.154x - 0.015 R² = 0.9297

0.05 0 0

0.5

1

1.5

2

2.5

n-alkanals by NMR (mmol/kg)

Figure 3. Comparison of hexanal concentrations in the headspace vials with freeze-thawed mayonnaise measured by headspace GC after 30 min equilibration at 60 ºC with concentrations of n-alkanals in oil extracted from mayonnaise stored in vials determined by 1

H NMR at ambient temperature.

311

312

The precision of the method for the assessment of hydroperoxides and aldehydes was

313

determined by performing duplicate experiments over different measurement days and

314

concentration levels and expressed in terms of repeatability and reproducibility (respectively

315

between- and within-day variation) by ANOVA (Table 2). Pooled relative standard deviations

316

of repeatability (RSDr) and within-laboratory reproducibility (RSDR) were respectively 5.7%

317

and 5.9%. The repeatability and within-laboratory reproducibility variations are close,

318

indicating that the method is stable over time. The aldehyde concentrations, which are closest

319

to the limit of quantification, have the highest relative standard deviations, but are still within

320

the limits to allow for quantitative profiling studies.

321

ACS Paragon Plus Environment

Page 18 of 26

Page 19 of 26

322 323

Table 2. Overview of the validation results for hydroperoxide and aldehyde quantification using band-selective excitation (32 scans). Sel

s

Factor

LoD

LoQ

Conc.

RSDr

RSDR

(mmol

(mmol

(mmol

(%)

(%)

/kg)

/kg)

/kg)

1.3 0.01

Aldehydes

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

1.2

0.03

24.20

2.5

2.5

5.76

1.8

2.0

0.72

1.3

1.4

0.14

8.2

8.2

0.06

7.0

7.6

324

325

Application: assessment of oxidative stability of mayonnaise

326

We used Principal Component Analysis (PCA) to assess the impact of storage temperature on

327

oxidative mechanisms in a food emulsion. A conventional mayonnaise formulation was used,

328

stored at two different temperatures: room temperature (22 ºC) for 200 days and elevated

329

temperature (50 ºC) for 53 days. The oxidation products formed in these mayonnaise samples

330

were quantified by 1H NMR, after which PCA was used to map compositional variations in

331

time.

ACS Paragon Plus Environment

Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

A

Page 20 of 26

B

C

Figure 4. Oxidation product profiles of mayonnaise stored at two different temperatures and followed in time. A) Score plot of PC 2 versus PC 3. The arrows in the score plots indicate the trajectory through time, the base of the arrow is timepoint zero, the arrowhead is located at the last timepoint. Samples at elevated ( ) and ambient ( ) temperature are respectively stored for 53 and 203 days. B/C) Loading plots of PC2/PC3, where grey bars represent hydroperoxides and black bars aldehydes. A-T correspond to the annotations in Table 1. 332

The first three principal components explained 97% of the total variation in the model. Most

333

of the variation was explained by PC 1 (75.9%), followed by PC 2 (18.5%) and PC 3 (2.5%).

334

PC 1 captured the overall variation of primary and secondary lipid oxidation products over

335

time. Here, all oxidation products showed positive loadings (see Supporting Info, Figure S2).

336

Since the variation in PC1 could be explained by the overall time-dependence of primary and

337

secondary lipid oxidation, this principal component could not discriminate between different

338

oxidation mechanisms. For PC 2 (Figure 4B), the positive loadings could be assigned to the

339

major hydroperoxides, while the negative loadings primarily corresponded to the aldehydes.

ACS Paragon Plus Environment

Page 21 of 26

340

This difference prompted us to map formation of hydroperoxides versus aldehydes (Figure

341

5A). At elevated T, both hydroperoxides and aldehydes were formed. However, at ambient T,

342

the formation of hydroperoxides exceeds the formation of aldehydes. The enhanced

343

generation of aldehydes at elevated T may be attributed to instability of hydroperoxides.

344

Consequently, this may bias the translation of accelerated shelf life experiments performed at

345

elevated temperature to ambient conditions. A

B 40

trans-trans LA-OOH (mmol/kg)

60

Sum of aldehydes (mmol/kg)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

50 40 30 20 10 0

35 30 25 20 15 10 5 0

0

50

100

Sum of hydroperoxides (mmol/kg)

0

5

10

15

20

25

cis-trans LA-OOH (mmol/kg)

Figure 5. Map of oxidation trajectories at elevated ( ) and ambient ( ) temperature, respectively stored for 53 and 203 days. (A) Plot of aldehydes vs. hydroperoxides (PC2 visualized) (B) plot of trans-trans-LA-OOH vs. cis-trans-LA-OOH (PC3 visualized). 346 347

For PC3 (2.5%), the loading plot (Figure 4C) showed opposite loadings for different lipid

348

hydroperoxides isomers (F vs. G, A-D vs. E). To verify and visualize whether storage

349

temperature had an impact on isomeric preferences of hydroperoxide formation, we plotted

350

the trans-trans-LA-OOH versus cis-trans-LA-OOH for both storage conditions (Figure 5B).

351

At ambient T, relatively small amounts of trans-trans-LA-OOH were formed, whereas at

ACS Paragon Plus Environment

Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

352

elevated T these trans-trans-LA-OOH levels exceed those of cis-trans-LA-OOH. This could

353

be attributed to temperature induced cis-trans isomerization of the native cis bonds in LA.

354

Although this isomerization reaction is known to occur at elevated temperatures, it has not

355

been well appreciated that this may also impact the formation of different primary oxidation

356

products. Furthermore, we conjecture that the formation of volatile secondary oxidation

357

products may be impacted by storage at accelerated elevated temperature conditions, which

358

may bias translation of results to ambient conditions.

359

In this work, the main purpose of the unsupervised explorative modelling by PCA was to

360

demonstrate the quality and potential of quantitative profiling of primary and secondary

361

oxidation products by 1H NMR. In future work, we intend to exploit the longitudinal nature

362

of the oxidation profiles acquired during food emulsion shelf-life tests by more sophisticated

363

multivariate analysis techniques, such as batch Partial Least Squares (PLS).31-32 Furthermore,

364

we expect to deploy multivariate ANOVA31 and multilevel multivariate approaches to

365

disentangle sources of variation in longitudinally acquired oxidation profiles such as storage

366

conditions (temperature, O2 pressure) and (pro- and anti-oxidant) compositions of the food

367

emulsions.33-35

368 369

Conclusion

370

1

371

food emulsions. By using band selective NOESY and TOCSY experiments, up to nine

372

primary lipid oxidation products were identified at level of common fatty acid types, cis/trans

373

isomers and chain position of hydroperoxide moiety. Band-selective excitation allowed us to

374

quantify hydroperoxides and aldehydes precisely (RSDR = 5.9%) with a detection limit of

375

0.01 mmol/kg in less than five minutes. In comparison to conventional one-dimensional

376

methods, such as PV or headspace GC, 1H NMR profiling provided both a sensitive and

H NMR can be used as a convenient and fast method to assess lipid oxidation products in

ACS Paragon Plus Environment

Page 22 of 26

Page 23 of 26 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

377

simultaneous quantification of primary and secondary oxidation products. Explorative

378

modelling of the quantitative 1H NMR profiles of primary and secondary oxidation products

379

revealed the modulation of lipid oxidation mechanism by temperature.

380 381

Acknowledgements

382

Martijn Brandt, Ruud Poort and Hans-Gerd Janssen are gratefully acknowledged for

383

performing the headspace GC measurements. We thank Christiaan Beindorff for carefully

384

preparing the mayonnaise and overseeing the shelf life experiments. Ewoud van Velzen is

385

gratefully acknowledged for his expert advice and input on multivariate analysis techniques.

386 387

References

388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410

1. Coupland, J. N.; McClements, D. J., Lipid oxidation in food emulsions. Trends in Food Science & Technology 1996, 7 (3), 83-91. 2. Jacobsen, C., Sensory impact of lipid oxidation in complex food systems. European Journal of Lipid Science and Technology 1999, 101 (12), 484-492. 3. McClements, D.; Decker, E., Lipid oxidation in oil‐in‐water emulsions: Impact of molecular environment on chemical reactions in heterogeneous food systems. Journal of Food Science 2000, 65 (8), 1270-1282. 4. Mozuraityte, R.; Rustad, T.; Storrø, I., The Role of Iron in Peroxidation of Polyunsaturated Fatty Acids in Liposomes. Journal of Agricultural and Food Chemistry 2008, 56, 537-543. 5. Berton-Carabin, C. C.; Ropers, M.-H.; Genot, C., Lipid Oxidation in Oil-in-Water Emulsions: Involvement of the Interfacial Layer. Comprehensive Reviews in Food Science and Food Safety 2014, 13 (5), 945-977. 6. Ghorbani Gorji, S.; Smyth, H. E.; Sharma, M.; Fitzgerald, M., Lipid oxidation in mayonnaise and the role of natural antioxidants: A review. Trends in Food Science & Technology 2016, 56, 88-102. 7. Laguerre, M.; Bayrasy, C.; A., P.; Weiss, J.; Mcclements, D. J.; Lecomte, J.; Decker, E. A.; Villeneuve, P., What makes food antioxidatns in lipid-based systems? The enxt theories beyond the polar paradox. Critical Reviews in Food Science and Nutrition 2015, 55, 183-201. 8. Laguerre, M.; Bily, A.; Roller, M.; Birtic, S., Mass Transport Phenomena in Lipid Oxidation and Antioxidation. Annu Rev Food Sci Technol 2017. 9. Budilarto, E. S.; Kamal-Eldin, A., The supramolecular chemistry of lipid oxidation and antioxidation in bulk oils. Eur J Lipid Sci Technol 2015, 117 (8), 1095-1137.

ACS Paragon Plus Environment

Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460

10. Frankel, E. N.; Meyer, A. S., The problems of using one‐dimensional methods to evaluate multifunctional food and biological antioxidants. Journal of the Science of Food and Agriculture 2000, 80 (13), 1925-1941. 11. ISO 3960:2017 In Animal and vegetable fats and oils -- Determination of peroxide value -- Iodometric (visual) endpoint determination, International Organization for Standardization: Geneva, 2017; p 10. 12. ISO 6885:2016 In Animal and vegetable fats and oils -- Determination of anisidine value, International Organization for Standardization: Geneva, 2016; p 7. 13. Laghi, L.; Picone, G.; Capozzi, F., Nuclear magnetic resonance for foodomics beyond food analysis. 2014; Vol. 59. 14. Frankel, E. N.; Hu, M.-L.; Tappel, A. L., Rapid headspace gas chromatography of hexanal as a measure of lipid peroxidation in biological samples. Lipids 1989, 24 (11), 976. 15. Guillen, M. D.; Goicoechea, E., Oxidation of corn oil at room temperature: Primary and secondary oxidation products and determination of their concentration in the oil liquid matrix from 1H nuclear magnetic resonance data. Food Chemistry 2009, 116 (1), 183-192. 16. Ibargoitia, M. L.; Sopelana, P.; Guillen, M. D., (1)H Nuclear Magnetic Resonance monitoring of the degradation of margarines of varied compositions when heated to high temperature. Food Chem 2014, 165, 119-28. 17. Skiera, C., Steliopoulos, P., Kuballa, T., Holzgrabe, U. & Diehl, B. , 1H NMR approach as an alternative to the classical p-anisidine value method. European Food Research and Technology 2012, 235, 1101-1105. 18. Skiera, C.; Steliopoulos, P.; Kuballa, T.; Holzgrabe, U.; Diehl, B., 1H-NMR Spectroscopy as a New Tool in the Assessment of the Oxidative State in Edible Oils. Journal of the American Oil Chemists' Society 2012. 19. Dugo, G.; Rotondo, A.; Mallamace, D.; Cicero, N.; Salvo, A.; Rotondo, E.; Corsaro, C., Enhanced detection of aldehydes in Extra-Virgin Olive Oil by means of band selective NMR spectroscopy. Physica A: Statistical Mechanics and its Applications 2015, 420, 258264. 20. Rastrelli, F., Schievano, E., Bagno, A. & Mammi, S. , NMR quantification of trace components in complex matrices by band-selective excitation with adiabatic pulses. . Magnetic Resonance in Chemistry 2009, 47, 868-872. 21. Berger, S., NMR techniques employing selective radiofrequency pulses in combination with pulsed field gradients. Progress in Nuclear Magnetic Resonance Spectroscopy 1997, 30 (3), 137-156. 22. Adams, R. W.; Holroyd, C. M.; Aguilar, J. A.; Nilsson, M.; Morris, G. A., "Perfecting" WATERGATE: clean proton NMR spectra from aqueous solution. Chemical Communications 2013, 49 (4), 358-360. 23. Magiatis, P.; Melliou, E.; Killday, B., A NEW ULTRA RAPID SCREENING METHOD FOR OLIVE OIL HEALTH CLAIM EVALUATION USING SELECTIVE PULSE NMR SPECTROSCOPY. 2015; p 84-92. 24. van Duynhoven, J.; van Velzen, E.; Jacobs, D. M., Quantification of Complex Mixtures by NMR. In Annual Reports on NMR Spectroscopy, 2013; Vol. 80, pp 181-236. 25. Miller, J. N.; Miller, J. C., Statistics and Chemometrics for Analytical Chemistry. 6th ed.; Pearson Education Limited: 2010. 26. Folch, J.; Lees, M.; Sloane Stanley, G. H., A simple method for the isolation and purification of total lipides from animal tissues. The Journal of biological chemistry 1957, 226 (1), 497-509. 27. Guillén, M. D.; Ruiz, A., Monitoring of heat-induced degradation of edible oils by proton NMR. European Journal of Lipid Science and Technology 2008, 110 (1), 52-60. 28. Frankel, E. N., Hydroperoxide formation. 2012, 25-50.

ACS Paragon Plus Environment

Page 24 of 26

Page 25 of 26 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481

Analytical Chemistry

29. Landy, P.; Pollien, P.; Rytz, A.; Leser, M. E.; Sagalowicz, L.; Blank, I.; Spadone, J.C., Model Studies on the Release of Aroma Compounds from Structured and Nonstructured Oil Systems Using Proton-Transfer Reaction Mass Spectrometry. Journal of Agricultural and Food Chemistry 2007, 55 (5), 1915-1922. 30. Savary, G.; Guichard, E.; Doublier, J.-L.; Cayot, N., Mixture of aroma compounds: Determination of partition coefficients in complex semi-solid matrices. 2006; Vol. 39, p 372379. 31. Smilde, A. K.; Westerhuis, J. A.; Hoefsloot, H. C.; Bijlsma, S.; Rubingh, C. M.; Vis, D. J.; Jellema, R. H.; Pijl, H.; Roelfsema, F.; van der Greef, J., Dynamic metabolomic data analysis: a tutorial review. Metabolomics 2010, 6 (1), 3-17. 32. Wold, S.; Kettaneh, N.; Fridén, H.; Holmberg, A., Modelling and diagnostics of batch processes and analogous kinetic experiments. Chemometrics and Intelligent Laboratory Systems 1998, 44 (1), 331-340. 33. Antti, H.; Bollard, M. E.; Ebbels, T.; Keun, H.; Lindon, J. C.; Nicholson, J. K.; Holmes, E., Batch statistical processing of 1H NMR-derived urinary spectral data. Journal of Chemometrics 2002, 16 (8-10), 461-468. 34. Jonsson, P.; Stenlund, H.; Moritz, T.; Trygg, J.; Sjöström, M.; Verheij, E. R.; Lindberg, J.; Schuppe-Koistinen, I.; Antti, H., A strategy for modelling dynamic responses in metabolic samples characterized by GC/MS. Metabolomics 2006, 2 (3), 135-143. 35. Nomikos, P.; MacGregor, J. F., Multivariate SPC Charts for Monitoring Batch Processes. Technometrics 1995, 37 (1), 41-59.

482 483

Supporting information

484

Hydroperoxide regions of the 1H NMR spectra (700 MHz) of extracts of an oxidized

485

mayonnaise, obtained by different extraction methods. Results of linear regression between

486

the total hydroperoxide values obtained by NMR and the PV method. First component (PC1)

487

of a principal component analysis of oxidation product profiles of mayonnaise stored at two

488

different temperatures and followed in time.

ACS Paragon Plus Environment

Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

489

For ToC only

490

ACS Paragon Plus Environment

Page 26 of 26