Assessment of milk fat content in fat blends by 13 C NMR spectroscopy analysis of butyrate

Butyric acid (butyrate) is a candidate marker of milk fat in complex fat blends, since it is exclusive of milk triacylglycerols (TAGs) from different ruminant species. In this work, we determined the amount of milk fat used for the preparation of fat blends by 13C Nuclear Magnetic Resonance (13C NMR) spectroscopy-based quantification of butyrate. When tested on fat samples spiked with known amounts of reference bovine milk fat (BCR-519 certified material), the relative composition of the mixtures was reliably assessed through the integration of the diagnostic 13C NMR carbonyl (C1) or α-carbonyl methylene (C2) resonances of butyrate. NMR data exhibited strict correlation with high resolution-gas chromatography (GC) of fatty acid methyl esters (R2 = 0.99), which was used as an independent and well-established method for the determination of butyrate. Thus, 13C NMR can be used for the direct assessment of milk fat content in fat mixtures, at a limit of detection lower than 5%, with clear advantages over the traditional GC methods in terms of speed, robustness and minimal sample handling. The natural variability of butyrate in milk has been taken into account to estimate the uncertainty associated with the milk fat content in unknown fat blends.


Introduction
According to the current European legislation (EC Reg., 1994;2007), the milk fat content 16 in commercial fat mixtures must be clearly indicated on the label. Since milk fat has the 17 highest price among fats, there is a realistic possibility that milk fat could be partly or totally 18 replaced with other fats, such as lard, beef tallow or vegetable oils, to obtain industrial 19 blends. 20 The International Dairy Federation (IDF) developed an accurate and reliable method to 21 determine the composition of mixed spreads based on the fatty acid profile of milk and non-22 milk fat ingredients (Muuse & Martens, 1993). The Official EU method of analysis for the 23 authenticity assessment of butter (EC Reg., 2001) is based on the gas-liquid 24 chromatography (GC) evaluation of triacylglycerols (TAGs) (Precht, 1992). Several other 25 strategies or methods have been proposed to assess the content of milk fat in blended 26 fats. Details of these methods and related drawbacks have been recently reviewed by 27

Derewiaka et al. (2011). 28
Butterfat TAGs can be effectively separated, characterized and quantified by capillary GC 29 coupled to electron ionization (EI) mass spectrometry (MS) (Kempinnen & Lalo, 2006). 30 Foreign fats, such as for instance tallow, have been quantified by fatty acid analysis 31 coupled to multivariate statistical techniques (Ulbert, 1995). Yoshinaga et al. (2013) applied 32 liquid chromatography coupled with atmospheric-pressure chemical ionization tandem 33 mass spectrometry (LC-APCI-MS/MS) to quantify milk fat in blends and butter biscuits, 34 using 1,2-dipalmitoyl-3-butyroyl-glycerol as a specific analytical target. Indeed, milk fat 35 contains major amounts of short-chain fatty acids, biosynthesized from acetic (as acetyl-36 CoA) and β-hydroxy-butyric acids derived from rumen fermentation. In particular, butyric 37 foodstuff, such as chocolate, cakes, pastries, ice-creams (Molkentin & Precht, 1997;41 Ulberth, 1998). 42 The analytical determination of BA generally relies on GC analysis of fatty acid methyl 43 esters (Molkentin & Precht 2000). For instance, BA has been specifically targeted to detect 44 milk fat in cocoa butter and chocolate fats (Buchgraber et al., 2007). 45 Based on the stereospecific esterification of BA on the sn-3 position of glycerol backbone 46 in milk fat TAGs, the study of BA sn-regioisomerism using Nuclear Magnetic Resonance 47 esterified TAGs in authentic butterfat (Picariello et al., 2013). 49 In the last decades, NMR has been recognized as a very powerful tool for food analysis 50 (Alberti et al., 2002). NMR boasts several advantages over other techniques, including its 51 non-destructive nature, the possibility to preserve food structure, the high informative level 52 obtainable from complex food systems, minimal sample handling required, speed of 53 analysis and good reproducibility (Alberti et al., 2002;Sacchi & Paolillo, 2007). 54 Milk fat lipids can be distinguished by other hard fats, such as lard, using 1 H or 13 C-NMR, 55 the latter maximizing the resolution, because of the wider range of chemical shifts 56 Several organic compounds in whole milk were also simultaneously quantified by one-( 1 H, 67 13 C) and two-dimensional NMR spectroscopy (Hu et al., 2007). 68 The well-resolved 13 C-NMR resonances of C1 (carbonyl carbon) of BA with respect to C1 69 of long-chain fatty acids (Pfeffer et al., 1977;Andreotti et al., 2000Andreotti et al., , 2002, were also 70 suggested as a diagnostic probe to discriminate genuine butter from mixtures with 71 synthetic TAGs with many advantages related to the robustness and minimal sample  In this work, we aimed at developing a direct and reliable NMR method to assess milk fat 76 in fat blends. Therefore, butyrate as a marker of milk fat was monitored by 13 C-NMR in 77 blends containing butter along with lard and vegetable margarine, which simulate 78 commercial "mixed fats" and other spreads used as cheaper butter surrogates. To this 79 purpose, we compared the direct NMR determination of butyrate in fat blends containing 80 5 known amount of milk fat to the capillary GC analysis of trans-methylated fatty acids, 81 chosen as a robust and well-established control analytical method. 82

Standards and reagents 83
Chloroform-d (with 0.03 % v/v internal tetramethylsilane, TMS) was obtained from Aldrich Belgium). Prior to blending, fat samples were melted at 40°C for 1 h under N2 to prevent 91 possible auto-oxidation and homogenized. Melted lard and margarine were dehydrated 92 using sodium sulfate. Fat blends were prepared by spiking a 1:1 (w/w) mixture of melted 93 lard and margarine with BCR-519 milk fat at 5, 10, 25 and 50 % (w/w) relative amount. 94 Although butyrate can be even detected at 1% (w/w) and quantified at 2.5 % (w/w) of 95 butterfat in complex fats (Picariello et al., 2013, see also herein below), blends containing 96 less than 5% (w/w) of milk fat were not investigated, since they are scarcely relevant under 97 a commercial standpoint. 98

Capillary gas-liquid chromatograpy (GC) 99
GC analysis of fatty acid methyl esters (FAMEs) was carried out by cold trans-methylation 100 in KOH/methanol (Ichihara 1996;Christie, 2003). Fat aliquots (100 mg to the nearest 0.1 101 mg) were dissolved in 10 mL n-hexane, mixed with 0.5 mL of 2 N KOH in methanol and 102 shaken vigorously for 1 min using a vortex mixer. The resulting solution was centrifuged 103 for 1 min. The clear supernatant was used for GC analysis. FAME analysis was performed 104 by using a Shimadzu GLC17A gas chromatography (Shimadzu Italia, Milan, Italy) equipped 105 with split/splitless injection port, flame ionization detector and a 60 m fused-silica capillary 106 column (Quadrex Corporation, New Heaven, U.S.A.) coated with cyanopropyl methyl 107 silicone (0.25μm film thickness). Samples (1μL) were introduced using a Shimadzu AOC-108 20i automatic injector (Shimadzu Italia). The temperature of both the injector and detector 6 was 250 °C. The initial oven temperature was set at 70 °C for 4 min. The temperature rate 110 was set on 10 °C min -1 up to 170 °C for 10 min, and an increase at a rate of 10 °C min -1 111 was followed to finally reach a temperature of 220 °C that was maintained for 10 min. 112 Helium was the carrier gas at a 1.8 mL min -1 flow rate. The split ratio was 1:60. The GC 113 method was calibrated using a mixture of 37 FAME standards fatty acids (Sigma), C4-C24, 114 and the calibration for butyrate had R 2 =0.992. Samples were analyzed in triplicate and 115 results were averaged. 116

High Resolution 13 C-NMR spectroscopy 117
For the 13 C-NMR analysis, 50 mg of fat samples were dissolved in 0.5 mL of chloroform-d 118  2.5 Quantitative spectral analysis 7 NMR signals were fitted to a sum of Lorentzian curves using a nonlinear least-squares 142 algorithm and intensity of peaks was quantified using the Linesim (Bruker) software. The 143 relative concentration of BA was calculated as detailed in the section 3.1. 13 C-NMR spectra 144 were acquired in triplicate and signal integration values were averaged. The 13 C-NMR-145 based limit of detection (LOD) and limit of quantification (LOQ) of butterfat in complex fats 146 were 1% (w/w) and 2.5% (w/w), respectively. LOD and LOQ were determined as previously 147 detailed (Picariello et al., 2013). 148

NMR analysis 149
The 13 C-NMR spectral regions relevant to carbonyls (spectra acquired in high resolution In order to obtain a reliable quantitative response, 13 C-NMR spectra have to be acquired 179 under experimental conditions ensuring a satisfactory digital resolution and a linear 180 recovery of NMR resonances. The linearity between the NMR signal intensity and the 181 concentration of the components is, generally, distorted by different relaxation rates and 182 NOE effects of different carbons (Wollenberg, 1990;Ng, 2000). 183 In routine 13 C-NMR qualitative analysis, most spectra are usually recorded using broad 184 band (BB) proton decoupling mode (complete saturation of the proton transitions in order 185 to produce decoupled spectra, eliminating the multiplicity of carbon signals) and using a 186 short delay time (D1) between two subsequent pulses. For quantitative 13 C-NMR analysis, 187 when carbons have different relaxation behavior, the longitudinal relaxation times (T1) 188 have to be known for all carbons to ensure that all carbons are fully relaxed between two 189 following pulses. For this reason, spectra for quantitative purpose were acquired using 190 experimental conditions that permit a complete relaxation of carbon nuclei between two 191 subsequent pulses, taking into account the known T1 values for different acyl carbons 192 (Wollenberg, 1990;Ng, 2000). T1 values ranged between 9 and 11 s for carbonyls and 193 were shorter than 1 sec for all methylene carbons. Based on these T1 values, and using a 194 45° pulse, a short relaxation delay of 2 sec was used for recording quantitative full spectra. 195 For methylene carbon C2 signals both BB and inverse-gated full spectra were acquired. 196 The advantage of the BB mode is related to its increased sensitivity because of both the 197 NOE enhancement, yielding higher S/N ratio, and the faster recycle delay between pulses 198 (2.37 s) with respect to those used in inverse gated NOE suppressed spectra (22.37 s). 199 For this latter reason, a higher number of scans per minute (25.3 scans/min) is recorded 200 in the BB accumulation mode with respect to the inverse gated recording (2.7 scans/min), 201 giving rise to spectra with higher signal-to-noise (S/N) ratio. To obtain the same S/N for 9 methylene carbons, 20-30 times shorter accumulation time is required in BB mode than 203 the inverse-gated mode. The complete relaxation of C1 carbonyls was guaranteed by the 204 high acquisition time (12-20 sec) requested for high digital resolution (0.04 Hz/pt) 205 (Wollenberg, 1990). Considering the small NOE effect on carbonyl signal intensities 206 (Wollenberg, 1990;Ng, 2000), the BA content was also inferred from the carbonyl high-207 resolution 13 C NMR spectra acquired using the BB mode. 208

Comparison between NMR and GC 209
The quantitative data obtained from the NMR measurements were compared with those 210 obtained by high resolution-gas chromatography (GC) determinations. In Table 1 Herein, to prepare blends spiked with milk fat we used anhydrous BCR-519 certified 247 material, which is representative of the most common cow breeds and averages the 248 fluctuation of barn (winter) and pasture (summer) feeding, containing butyrate at 3.49% 249 (w/w) (Molkentin & Precht, 1997;1998). 250

Conclusions
A rapid and accurate 13 C-NMR spectroscopic method was developed to detect and quantify 251 milk-derived fat in butter-like spreads composed by different kinds of fats, namely 252 vegetable margarine and pork fat. The effectiveness of the method was assessed by 253 comparative GC determination of butyrate, as a molecular marker of milk fat, in milk fat-254 containing blends. The correlation between the two independent methods was excellent. 255 However, the 13 C-NMR spectroscopy detection and quantification of butyrate was 256 advantageous for several reasons, primarily because no sample pre-treatment was 257 required. In addition, NMR analysis does not require the use of chemical standard for signal 258 identification and quantitative calibration. The intensity of selected diagnostic signals, in 259 fact, allows the direct quantitative determination of BA content and, hence, the assessment 260 of the milk fat content. 261 In general, the NMR instrumentation and maintenance costs, as well as the operative skills 262 required, could represent limiting factors to routine application for authenticity assessment 263 of edible fats. However, in a perspective of a rational and integrated control policy, NMR 264 spectroscopy can be considered as a powerful and versatile method, especially suitable 265 for confirmatory purposes. In the last years, the high magnetic fields available (600-800 266 has strikingly enhanced the instrumental sensitivity, reducing the acquisition time to a few 268 minutes. Therefore, it is expected in the future that NMR will be applied widely at both 269 control laboratories and industrial levels for assessing fat authenticity and food quality in 270 general. 271 The method here discussed can help to control fraudulent practices and to certify the