Evaluation of FT-NIR and ATR-FTIR Spectroscopy ... - ACS Publications

Subscriber access provided by DUKE UNIV

Article

EVALUATION OF FT-NIR AND ATR-FTIR SPECTROSCOPY TECHNIQUES FOR DETERMINATION OF MINOR ODD AND BRANCHED CHAIN SATURATED AND TRANS UNSATURATED MILK FATTY ACIDS Ivan Valentinov Stefanov, Vincent Paul Baeten, Ouissam Abbas, Bruno Vlaeminck, Bernard De Baets, and Veerle Fievez J. Agric. Food Chem., Just Accepted Manuscript • DOI: 10.1021/jf304515v • Publication Date (Web): 19 Feb 2013 Downloaded from http://pubs.acs.org on March 11, 2013

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

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

Page 1 of 44

Journal of Agricultural and Food Chemistry

1

EVALUATION OF FT-NIR AND ATR-FTIR SPECTROSCOPY TECHNIQUES FOR

2

DETERMINATION OF MINOR ODD AND BRANCHED CHAIN SATURATED AND

3

TRANS UNSATURATED MILK FATTY ACIDS

4

I. STEFANOV1, V. BAETEN2, O. ABBAS2, B. VLAEMINCK1, B. DE BAETS3 and V.

5

FIEVEZ*1

6

1

7

Bioscience Engineering, Ghent University, Proefhoevestraat 10, B-9090 Melle, Belgium

8

2

9

Agricultural Research Centre. Chaussée de Namur, 24. B-5030 Gembloux, Belgium

Laboratory for Animal Nutrition and Animal Product Quality (LANUPRO), Faculty of

Food and Feed Quality Unit, Valorisation of Agricultural Products Department, Walloon

10

3

11

Engineering, Ghent University, Coupure links 653, B-9000 Ghent, Belgium

Department of Mathematical Modeling, Statistics and Bioinformatics, Faculty of Bioscience

12 13 14 15 16

Corresponding author: Veerle Fievez, Tel: +32 9 264 90 01, Fax: +32 9 264 90 99

17

E-mail address: [email protected]

18 19 20 21 22 23

1 ACS Paragon Plus Environment


Page 2 of 44

24

ABSTRACT

25

Determination of nutritionally important trans MUFA, CLA and OBCFA milk fatty acids (often

26

present in amounts lower than 1.0 g/100g of total fat) using fast and non-destructive analytical

27

methods would enhance their use as diagnostic tools in dairy herd and human health

28

management. Here, PLS regression using ATR/FTIR spectra indicated potential for

29

determination of trans-11 C18:1 and trans-12 C18:1 (Rcv2≥0.80), trans-9 C18:1 in very minor

30

concentration (Rcv2>0.82), as well as for anteiso C15:0 (Rcv2=0.57) and iso C17:0 (Rcv2=0.61).

31

Furthermore, the main cis-9, trans-11 CLA isomer was predicted well despite the high trans

32

MUFA concentration. Differentiation between the CLA and the trans MUFA signals was evident

33

(based on specific cis/trans bands), as well as branched chain saturated fatty acid methyl esters

34

revealed specific iso and anteiso ATR/FTIR absorbance bands. None of the minor FA PLS

35

results with FT-NIR showed interesting potential, except satisfactory predictions for trans-9

36

C18:1 and cis-9, trans-11 CLA. Overall ATR/FTIR resulted in better calibrations and provided

37

more specific information for determination of minor milk fatty acids.

38 39

KEYWORDS: ATR, FTIR, FT-NIR, milk fat, trans fat, CLA, OBCFA, spectroscopy

40 41 42 43 44 45 46


Page 3 of 44


47

INTRODUCTION

48

In addition to the already established nutritional concerns for trans monounsaturated fatty acids

49

(trans MUFA) and the beneficial biological effects of conjugated linoleic acids (CLA) for human

50

health 1, these two groups of milk fatty acids together with another group of specific odd and

51

branched chain saturated fatty acids (OBCFA) have emerged as important nutritional biomarkers

52

for dairy herds

53

duodenal flow of microbial biomass, as well as the potential of specific OBCFA isomers to

54

monitor the nutrients produced during digestive processes would allow their utilization as

55

excellent diagnostic tools of rumen function

56

MUFA, CLA and OBCFA for early diagnosis of digestive disorders in cows, such as diagnosis

57

of subacute ruminal acidosis (SARA) have recently been identified as isomer specific 3. For

58

example, sudden increase in trans-10 C18:1, which is part of the trans-10, cis-12 C18:2

59

biohydrogenation pathway 7, has been associated with low ruminal pH, thus both fatty acids

60

might be indicative of SARA. In addition, the cis-9, trans-11 C18:2 isomer was recently

61

discovered to be an effective SARA predictor 3. Moreover, drops in the iso C13:0 to iso C16:0

62

concentrations in milk fat were also indicative of SARA induction. The latter is thought to be

63

caused by an OBCFA productivity shrinkage (production of total OBCFA) from ruminal

64

cellulolytic microbiota and/or lower competitiveness of the cellulolytic bacteria relative to the

65

rest of the microbial community (decrease in the cellulolytic/total bacteria ratio).

66

Thus, the ability to follow the variations of these interesting isomers in milk fat would be a

67

valuable non-invasive strategy for detection of altered ruminal environment. Still implementation

68

of the latter requires reliable and effective analytical techniques, which can simultaneously

69

quantify and differentiate minor milk fatty acids.

2, 3

. The microbial origin of OBCFA and their direct relationship with the

4, 5, 6

. Further, the monitoring potential of trans



Page 4 of 44

70

It would not be trivial to mention that gas chromatography (GC) and high-performance liquid

71

chromatography are the reference methods, as they are still the golden standard and the most

72

widely used analytical techniques for determination of almost any lipid species. However, the

73

effectiveness for disease prevention using this strategy based on quantification of minor fatty

74

acids would be drastically enhanced by the availability of fast, cheap and non-destructive

75

analytical methods, such as spectrophotometry techniques. Developing spectroscopy based

76

methods for the simultaneous determination of several different types of minor fatty acids with

77

different chemical structures would be of major importance for animal as well as for human

78

nutrition.

79

Fourier transform near-infrared (FT-NIR) and attenuated total reflectance infrared (ATR/FTIR)

80

are well established spectroscopy techniques, both of which are widely used in the food industry

81

8

82

official AOCS (Cd 14d-99) method for quantifying total trans fats 9, 10, due to unique property to

83

detect the absorption of the C-H deformation vibration from a trans C=C bond in the vicinity of

84

966 cm-1 11. The sensitivity of the ATR/FTIR technique allows accurate determination of the total

85

monounsaturated trans content in various naturally occurring non-ruminant fats, vegetable and

86

hydrogenated oils and fats (e.g. canola, margarine and lard), but it has not been extensively

87

tested with more complex fat matrices, such as dairy fat.

88

FT-NIR is the more widely used technique in various niches of the food industry, due to its

89

property of a strong absorption signal from many types of molecules and the ability for

90

transmission of the NIR radiation by optical fibres over long distances. However, the former

91

feature has also brought FT-NIR the legacy of low specificity and inability to differentiate

92

between individual components in various homogeneous matrices, due to combinations and

. ATR/FTIR is famous as the only spectroscopy analytical technique to be approved as an


Page 5 of 44


93

overlaps of broad bands from fundamental and overtone chemical vibrations (generally

94

overlapping vibrational bands that are non specific and non-resolved)

95

NIR was previously reported as a viable technique for determination of the complete fatty acid

96

profile in soybean, olive and flaxseed oils, as well as in different mixtures of shortening and lard

97

14

98

might allow trans FA determinations for regulatory labelling purposes 14. The use of FT-NIR in

99

bovine milk fatty acid analyses has not been investigated.

12, 13

. Nevertheless, FT-

. The latter has permitted FT-NIR to be used for rapid screening/monitoring of fat products and

100

Milk fat is a complex matrix and in addition to the many different types of fatty acids,

101

correlations between biohydrogenation intermediates are inevitable. Thus, having a technique

102

with differentiation ability is of great importance. Here we evaluated both FT-NIR and

103

ATR/FTIR spectroscopy techniques for the quantification of individual and grouped trans

104

MUFA, CLA and OBCFA in a dataset of 75 milk samples sub-selected from a collection of more

105

than 1000 milk samples to cover the full biological concentration span of these milk fatty acids.

106 107

MATERIALS AND METHODS

108

Sample Selection

109

The sample storage and sample selection methodology was previously described

110

total of 100 milk samples were selected from a sample bank (n = 1033) of six different cow

111

feeding experiments aiming at changing the fatty acid (FA) profile of dairy products. The sample

112

subset was selected using a genetic algorithm applied to cover the naturally occurring range of

113

several milk fatty acids of interest (different odd and branched chain saturated fatty acids and

114

different trans-C18:1 and cis/trans-C18:2 unsaturated isomers) in high, mid, and low

15

. Briefly, a



16

Page 6 of 44

115

concentration ranges

. The milk fat was extracted using a previously described methodology

116

involving dichloromethane-ethanol 17.

117

Fatty Acid Standards

118

Two trans monounsaturated trans-9 and trans-10 C18:1 fatty acid methyl ester (ME) standards,

119

three branched chain saturated iso C14:0, iso C15:0 & anteiso C15:0 ME, one odd chain

120

saturated C15:0 ME and two triacylglyceride (TAG) standards (trielaidin and tristearin) were

121

purchased from Larodan (Larodan Fine Chemicals, Malmo, Sweden) and Nu-Check (Nu-Check

122

Prep, Inc., MN, U.S.A.).

123

GC Reference Data

124

Quantification of trans fatty acids (TFA) and fatty acid groups using spectral data requires

125

precise Gas Liquid Chromatography (GC) reference data for the construction of mathematical

126

models. Identification and quantification of TFA through GC have been greatly improved with

127

new highly polar, long capillary columns, but direct GC without prior fractionation could show

128

overlapping between different trans-n and cis-n C18:1 positional isomers and trans-n C16:1

129

coelution with specific branched chain saturated and cis-n C16:1 mono-unsaturated FAs, which

130

might result in an underestimation of the total TFA content

131

dependency of the polarity of cyanopropyl phases 18 to mathematically deduce concentrations of

132

overlapping fatty acids using two different temperature programs without prior fractionation. A

133

similar approach was described before

134

for 92 of the 100 selected milk fat samples) were methylated

135

(FAME) were analyzed by GC according to Vlaeminck et al.2 (first temperature program) and by

136

an isothermal (T=180oC) (second) temperature program. A different separation with the second

137

temperature program allowed the quantification of most individual FA 15, 20 (Table 1). Due to an

15, 20, 21

18, 15

. Here, we used the temperature

. After extraction, all samples (sufficient quantity 17

and the fatty acid methyl esters


Page 7 of 44


138

influence from the cows’ diet the total trans mono-unsaturated fatty acids (trans MUFA) were

139

present in very high concentrations and ranged from 1.5 to 21 g/100g in the milk samples

140

Total Conjugated Linoleic fatty acids (CLA) maximally represented 2.0 g/100g of milk fat

141

(Table 1). Further, insufficient GC resolution prevented reliable separation of individual trans-6

142

C18:1, trans-7 C18:1 and trans-8 C18:1 FA isomers, but allowed accurate determination of the

143

trans-6+7+8 C18:1 sum. The cross-correlation results between the individual fatty acids and the

144

major fatty acid groups indicated a high correlation (R2>0.95) between trans-10 & trans-11

145

C18:1 and trans MUFA, between cis-9, trans-11 C18:2 and CLA, as well as between C15:0 and

146

OBCFA and C15:0 and ODD.

147

Vibrational Spectroscopy Analysis

148

ATR/FTIR Spectroscopy

149

All attenuated total reflectance Fourier transform mid-infrared (ATR/FTIR) spectra were

150

acquired on a Vertex 70 – RAM II Bruker spectrometer (Bruker Analytical, Madison, WI)

151

operating with a Golden Gate

152

internal reflection element was a small, non-temperature-controlled Type IIa diamond prism

153

allowing a sampled diameter of approximately 2.0 mm. The optically dense medium was in

154

contact with two ZnSe focusing lenses, one used to focus the incident infrared radiation and the

155

second one to collect the reflected infrared radiation. The optical bench included an

156

interferometer with a RockSolid configuration, KBr substrate beam splitter and a deuterated

157

triglycerin sulfate (RT-DLaTGS) detector. The OPUS 6.5 software for Windows of Bruker

158

Instruments was used for instrument management, spectra acquisition and OPUS JCAMP to

159

JCAMP-DX file transformation. All milk fat samples were stored in vials selected by CRA-W in

160

previous Fourier transform Raman analysis

TM

23

.

diamond ATR accessory (Specac Ltd., Slough, UK). The

22

. Prior to ATR/FTIR analysis, the milk fat vials



Page 8 of 44

161

were taken out of storage and the milk fat was kept in a melted state at a minimum of 30 minutes

162

prior to spectra acquisition in a 38±1oC water bath, accompanied by gently shaking by hand to

163

ensure homogenization. One drop of each milk fat sample in liquid form was placed on top of the

164

optical medium and spectra were immediately acquired [room temperature (∼24±2oC)], as well

165

as selected FA standards (∼0.1 g/sample) were analyzed after freezing at -80oC. The spectra

166

were collected against air as a background over the wavenumber range of 4498-500 cm-1 with a

167

resolution of 4 cm-1. For each spectrum, 64 scans were co-added and averaged to obtain a good

168

signal-to noise ratio. A total of 2074 data points were recorded from 500 to 4498 cm-1. Because

169

of very low milk fat quantity in 14 and extraction solvent contamination in 3 of all 92 selected

170

samples 15, ATR/FTIR spectra were available for 75 milk fat samples. For each milk fat sample,

171

ATR/FTIR spectra were acquired in duplicate by taking a new milk fat drop and the average of

172

the two spectra was used for chemometrical analysis.

173

FT-NIR Spectroscopy

174

The Fourier transform near-infrared (NIR) spectra of the 75 milk fat samples were acquired

175

using a Bruker Optics Matrix-F instrument equipped with a thermoelectrically cooled InGaAs

176

diode detector and a Bruker Optics N261 diffuse reflection fiber optic probe (Bruker Analytical,

177

Madison, WI). The microprocessor controlled optical bench included an interferometer with a

178

RockSolid configuration, quartz substrate beam splitter with proprietary coating and an air

179

cooled tungsten NIR excitation source (12V, 20W) with a HeNe laser at a frequency of 633 nm.

180

Milk fat samples and pure FA standards (∼0.1-0.5 g/sample) were taken out of storage, left to

181

rest until reaching room temperature (∼21±2oC) and analyzed in vials selected by CRA-W in

182

previous Fourier transform Raman spectroscopy analysis 22 with PE-caps (Klaus Ziemer GmbH,

183

Mannheim, Germany). The spectra of all milk fat samples were collected against air as a


Page 9 of 44


184

background with a resolution of 8 cm-1 and a total of 780 data points were recorded over the

185

wavenumber range of 4000 to 10,000 cm-1 (2500-1000 nm). For each milk fat sample, FT-NIR

186

spectra were acquired in duplicate and the average of the two spectra was used for

187

chemometrical analysis.

188

Data Treatment

189

Data incorporation, treatment and partial least squares (PLS) regression were performed

190

similarly to previous methodology 15. The OPUS spectra files were converted to the JCAMP-DX

191

file system, which was imported into The UnscramblerTM v. 9.1 software (CAMO, Trondheim,

192

Norway). The ATR/FTIR spectra were reduced from 2074 to 1241 variables, by removing

193

selected regions, e.g. noisy region between 500-560 cm-1, regions identified as carriers of -OH

194

chemical information (1020-1070 cm-1 and 3050-4000 cm-1), as well as minor absorbance from

195

the background (2000-2450 cm-1). The observed -OH absorption signal did not originate from

196

the milk fat samples, but was attributed to moisture/trace of ethanol solvent used to clean the

197

surface of the ATR crystal between analyses. Although a drying fan was used, it appeared that

198

some ethanol moisture/vapours could be trapped between the milk fat and the crystal surface,

199

which was apparent in the ATR/FTIR spectra of six samples. As a precaution, these regions not

200

related to FA information were removed and all mathematical treatments or investigations were

201

performed with the remaining 1241 variables. The pre-processing methods of choice were a

202

combination of linear baseline correction and Multiplicative Scatter Correction (MSC) or a linear

203

baseline correction and Standard Normal Variate (SNV). Both MSC and SNV compensate for

204

undesired variation in the scattering intensity caused by various multiplicative effects, thus the

205

choice of using either pre-treatment would be subjective 24. Standard partial least squares (PLS)

206

regression was carried out in combination with the uncertainty testing “jackknife” procedure



Page 10 of 44

207

from The UnscramblerTM program, which is used to select significant variables that correlated

208

with the measured parameter 24, 25, 26. Further, PLS regression was performed after ATR/FTIR or

209

FT-NIR spectra transformation according to three different strategies- after MSC or SNV pre-

210

processing only, MSC/SNV followed by 1st Savitzky-Golay derivative or MSC/SNV followed by

211

2nd Savitzky-Golay derivative (using a 1,1,2 derivative treatment, which indicates a smoothing

212

factor with the number of data points in a running average on the left, the secondary smoothing

213

on the right, and the polynomial order of the derivative, respectively). A total of six prediction

214

models for each of the individual trans isolated and conjugated fatty acids (trans-4 C18:1, trans-

215

5 C18:1, trans-6+7+8 C18:1, trans-9 C18:1, trans-10 C18:1, trans-11 C18:1, trans-12 C18:1,

216

trans-15 C18:1, cis-9, trans-11 C18:2 and trans-10, cis-12 C18:2), both trans MUFA and CLA

217

FA groups, as well as for each of the individual OBCFA (iso C13:0, anteiso C13:0, iso C14:0,

218

iso C15:0, anteiso C15:0, C15:0, iso C16:0, iso C17:0,anteiso C17:0 and C17:0), as well as for

219

ODD (sum of C5:0, C7:0, C9:0, C11:0, C13:0, C15:0, C17:0, C19:0, C21:0, C23:0), ISO (sum of

220

iso C13:0, iso C14:0, iso C15:0, iso C16:0, iso C17:0 and iso C18:0), ANTE (sum of anteiso

221

C13:0, anteiso C15:0 and anteiso C17:0), BRANCHED (sum of ISO and ANTE) and OBCFA

222

(sum of ODD and BRANCHED) FA groups were built when using either ATR/FTIR or FT-NIR

223

spectra of all 75 milk samples. In addition, to better evaluate the prediction performance for

224

minor fatty acids in their most common concentration range (below 1.0 g/100g total fat) and to

225

avoid overfitting based on discontinuous dataset distribution, milk fat samples with high amounts

226

of specific individual fatty acids (such as trans-10 C18:1, trans-11 C18:1 and trans-12 C18:1)

227

were removed and new PLS models were constructed using only the remainder. For all PLS

228

models validation was carried out using systematic cross-validation with 3 folds and 25 units per

229

segment. The optimal number of PLS factors used for the regression was determined from the


Page 11 of 44


230

minimum residual validation variance. The cross-validation coefficient of determination (Rcv2),

231

the root mean square error of cross-validation (RMSECV, g/100g FAME), the number of PLS

232

factors, intercept, slope and bias parameters of the best prediction models for individual and

233

grouped FA are presented in Tables 2 and 3 for ATR/FTIR and FT-NIR, respectively. All other

234

results are available as Supporting Material.

235 236

RESULTS AND DISCUSSION

237

While manufacturers are implementing different dietary strategies for the increase in the amounts

238

of beneficial fats

239

prevention of productivity and quality drops. The latter requires fast and cost effective analytical

240

methods, which could determine fatty acids of interest ex situ or desirably in situ. For this

241

purpose, attenuated total reflectance Fourier transform infrared (ATR/FTIR) and Fourier

242

transform near-infrared (FT-NIR) spectroscopy techniques were considered. Looking at the

243

number of scientific publications it becomes evident that FT-NIR has a predominant impact in

244

the food industry. However, ATR/FTIR would seem a more logical choice, since it possesses

245

enhanced spectral absorbance features, such as from specific unsaturated C=C bonds, and has

246

already been established as an official method for the quantification of total trans fat content in

247

foods and dietary supplements

248

several fatty acids with similar structures present in the complex milk fat matrix. Thus, the

249

evaluation of both most widely used spectrophotometrical methods for their ability to produce

250

fatty acid specific signals and in turn to determine various fatty acids present in different high,

251

mid and low concentration combinations is necessary.

252

Fatty Acid Standards

27

, following the health status of animals is always of importance for quick

11

. Here, the interest is in the simultaneous determination of



253

Page 12 of 44

trans MUFA standards

254

The general differences between the FT-NIR spectra of pure triolein and trielaidin standards in

255

the vicinity of 4500-4700 cm-1 (2200-2130 nm) and 5600-5900 cm-1 (1785-1695 nm) (attributed

256

to the differences in signals from the cis and trans C=C bonds) were previously reported 14. Here

257

two trans monounsaturated fatty acid methyl ester standards, as well as trielaidin and tristearin

258

pure standards were analyzed using FT-NIR and ATR/FTIR spectroscopy techniques. The 2nd

259

Savitzky-Golay derivative FT-NIR spectra of pure tristearin and trielaidin standards were

260

compared in order to establish specific contributions from the isolated trans C=C bond (Figure

261

1). A striking difference between the spectra of both TAG was in the vicinity of 4705 cm-1,

262

where trielaidin and not tristearin showed increased absorption. The latter is a combination

263

region, which might be specific to absorbance from C-H vibrations in isolated trans C=C bonds.

264

Further, analyses of the pure standards with ATR/FTIR spectra confirmed an enhanced

265

absorbance feature from non-conjugated trans C=C bonds in the vicinity of 966 cm-1 (from the

266

C-H out of a plane deformation band), but no differences in the peak position or absorbance

267

intensity were evident between the two trans-9 and trans-11 methyl ester positional isomers

268

(Figure 2). Other bands around 1245 cm-1 and 1000-1130 cm-1 showed clear absorbance

269

intensity and shift perturbations, which might be isomer specific. In addition, the shape of the

270

966 cm-1 band was very broad when liquid spectra of the standards was considered, thus partial

271

overlap with previously reported absorption bands from conjugated cis and trans C=C bonds in

272

cis-9, trans-11 and trans-10, cis-12 CLA is suspected

273

determinations of CLA isomers in low quantities, especially in the presence of high trans MUFA

274

content (corresponding to an increase in band broadness and intensity) 29.

28

. These features might complicate


Page 13 of 44


275

Moreover, all standards showed drastic differences between the ATR/FTIR absorption spectra

276

obtained at liquid or at solid physical phases. The crystalline spectra of the trielaidin indicated

277

the appearance of new peaks, peak shifts and absorption increase in specific ATR/FTIR

278

wavebands, which are believed to contain extensive structural and vibrational information

279

(Supporting Materials, Figure SM1). The most striking feature (believed to be of most

280

importance) was the concurrent absorbance increase and peak shift of the isolated trans C=C

281

waveband in the vicinity of 966 cm-1 (shifted to 960 cm-1). Moreover, a closer examination of the

282

latter revealed narrower band shape compared to the liquid phase band. Detailed investigation is

283

required for origin identification of all newly reported vibrational features (Supporting

284

Materials, Table SM1).

285

OBCFA standards

286

Limited information is available for specific FT-NIR and ATR/FTIR absorbance features related

287

to branched (ISO and ANTE) and odd numbered (ODD) saturated fatty acids. Here pure iso

288

C14:0, iso C15:0, anteiso C15:0 and C15:0 fatty acid methyl ester standards were analyzed at

289

room temperature. Overall the ATR/FTIR spectra of the iso C15:0, anteiso C15:0 and C15:0

290

standards were similar, except the appearance of two new peaks in the 700-920 cm-1 region and

291

shifting of bands with absorbance intensity increases in the 1330-1400 and 1475-1435 cm-1

292

regions (Figure 3). Unique signals related to iso and anteiso branching were observed in the

293

vicinity of 920, 1366 & 1466 cm-1 and 770, 1375 & 1462 cm-1, respectively (Figure 3).

294

Literature in regards to the origin of the vibrational modes in the lower wavenumber region

295

(below 920 cm-1) for branched alkanes is scarce, however they are believed to be caused by the

296

FA chain C-C skeletal rocking vibrations [the band position difference between anteiso C15:0

297

and iso C15:0 (770 and 920 cm-1, respectively) might be caused by the difference in the position



Page 14 of 44

298

of the extra CH3 methyl group at the end of the FA chain, which in turn influences the mobility

299

of the FA chain]

300

1365 cm-1) is known to be associated with the C-H symmetric bending vibrations in -CH3

301

The appearance of this more intense absorbance band in the ATR/FTIR spectra of branched, but

302

not in straight chain standards is due to the presence of an extra methyl group at the end of the

303

fatty acid chain - within branched methyl ester standards, the iso and anteiso configuration

304

position of this extra methyl group causes differences in absorbance intensity and band position

305

at 1366 cm-1 and 1375 cm-1 for iso C15:0 and anteiso C15:0, respectively (Figure 3). The 1475-

306

1435 cm-1 absorption region is related to the asymmetrical deformation vibration of -CH3 in

307

branched alkanes 30, 31. Further, no shifts in these iso bands between the spectra of iso C14:0 and

308

iso C15:0 were evident and only minor differences between the spectra of both fatty acids in C-H

309

(higher in iso C15:0) and C=O (higher in iso C14:0) vibrations due to differences in the fatty

310

acid chain length were present (Supporting Materials, Figure SM2). These results might help

311

in differentiation between anteiso, iso and straight chain fatty acids, but specificity for saturated

312

fatty acids of similar nature within the ISO, ANTE or ODD groups apparently seems low.

313 314 315 316

30, 31

. Absorbance in the vicinity of 1350-1395 cm-1 (often in the vicinity of 30, 31

.

Fatty Acid Determinations Most, trans isolated (trans MUFA) and conjugated (CLA) unsaturated, as well as odd and

317

branched chain saturated (OBCFA) fatty acids (FA) of interest are present in minor

318

concentrations (lower than 1.0 g/100g of total fat) in bovine milk fat. Hence their quantification

319

using a spectroscopy technique is a challenge. This is mainly due to the property of

320

spectrophotometrical

321

physical/molecular characteristics of all FA present in the sample. Thus FA with similar

322

chemical structure characteristics, such as the trans isolated and conjugated FA of interest, could

methods

producing spectral

information

based

strictly on

the


Page 15 of 44


323

have a significant overlap of spectral bands issued from the wide range of vibrations types from

324

the different fatty acid isomers. These specific absorbances are dependent on the types and

325

concentrations in which the corresponding fatty acids are present. For instance, previous research

326

indicated that both trans-10 C18:1 and cis-9, trans-11 C18:2 CLA are important biomarkers,

327

which might reflect the viability status of the ruminal microbial population. The latter is

328

influenced by drop in ruminal pH values, which in turn could drive the animal into acidotic

329

status 3. For this, it would be important to possess a spectroscopy methodology, which is able to

330

differentiate between both types of fatty acid groups (trans MUFA versus CLA) when they are

331

present in low, as well as in higher than “normal” quantities. Typical average concentrations of

332

CLA in milk fat are 0.3-0.6 g/ 100g with cows fed a total mixed diet, but the levels of CLA in

333

milk can vary widely among herds 32, 33 and could reach an average of 1.1 g/100g of CLA in milk

334

fat with cows consuming pasture only 34. Also, it has been shown that the concentrations of CLA

335

in milk are affected by different dietary fatty acid sources and could drastically vary between

336

individual animals, reaching a maximum of 5.17 g / 100 g in milk fat with cows consuming a

337

sunflower oil based diet

338

trans MUFA (~21 g/100g) and high CLA content (~ 2.5 g/100g), thus we attempted to evaluate

339

the prediction performance within those naturally occurring fatty acid concentration ranges. In

340

such cases, it has already been established that in the presence of at least 1.0 g/100g conjugated

341

unsaturation in total fat, the official trans FA determination method is not applicable anymore 9.

342

However, that does not necessarily mean that quantification is impossible, but rather not

343

comparable to officially applied methods. Thus, determinations with lower performance could

344

still have implications for disease screening or monitoring.

345

PLS accuracy

35

. The current dataset contained both milk fat samples with very high



Page 16 of 44

346

In general, FT-NIR spectroscopy with partial least squares (PLS) regression resulted in complex

347

prediction models (high number of PLS factors), performed poorly in the determination of

348

individual fatty acids and the ATR/FTIR technique was superior (higher cross-validation

349

coefficient of determination, Rcv2, and lower root mean square error of cross-validation,

350

RMSECV, g/100g FAME).

351

trans MUFA determinations

352

The ATR/FTIR spectroscopy method predicted well (Rcv2≥0.90) all major individual trans

353

monounsaturated fatty acids (trans MUFA) as well as the trans MUFA group in high, mid and

354

low concentration (Table 2). The regression coefficients plots for all models were similar and

355

indicated variables in the 960-970 cm-1 band as main predictors. While derivative transformation

356

did not improve the high concentration range models, the best low trans MUFA model (less than

357

4.6 g/100g in total fat) was with 2nd Savitzky-Golay derivative spectra transformation. No

358

variables in the 930-950 and 980-990 cm-1 wavenumber ranges (associated with absorbance from

359

conjugated C=C) were picked as contributors (Figure 4). It seems that the milk fat matrix does

360

not cause major difficulties for prediction of trans MUFA fatty acids in this concentration range,

361

even in the presence of CLA.

362

Further, satisfactory performance was evident for some of the individual fatty acids in minor

363

concentrations (lower than 1.0 g/100g in total fat), such as trans-11 C18:1 (Rcv2=0.795,

364

RMSECV=0.097 with 5 PLS factors) and trans-12 C18:1(Rcv2=0.814, RMSECV=0.066 with 1

365

PLS factors). While trans-11 C18:1 was mainly determined from a specific signal in the vicinity

366

960-970 cm-1 (Figure 5) without the need for derivative transformation, the trans-12 C18:1

367

prediction was due to a high reference data cross-correlation with the total trans MUFA group.

368

Except for trans-9 C18:1 and the trans-6-7-8 C18:1 sum (Table 2), all other ATR/FTIR models


Page 17 of 44


369

for fatty acids in very minor concentrations (lower than 0.50 g/100g in total fat) resulted in

370

Rcv2

Evaluation of FT-NIR and ATR-FTIR Spectroscopy ... - ACS Publications

Recommend Documents