Advertisement

When detection of dairy food fraud fails: An alternative approach through proton nuclear magnetic resonance spectroscopy

  • Anamaria Hanganu
    Affiliations
    Faculty of Chemistry, Department of Organic Chemistry, Biochemistry and Catalysis, Research Centre of Applied Organic Chemistry, University of Bucharest, 90-92 Panduri Street, RO-050663 Bucharest, Romania

    Institute of Organic Chemistry “C.D. Nenitescu” of the Romanian Academy, 202B Spl. Independentei, 060023 Bucharest, Romania
    Search for articles by this author
  • Nicoleta-Aurelia Chira
    Correspondence
    Corresponding author
    Affiliations
    Faculty of Applied Chemistry and Materials Science, “C. Nenitescu” Organic Chemistry Department, University “Politehnica” of Bucharest, 1-7 Polizu Str., 011061 Bucharest, Romania
    Search for articles by this author
Open AccessPublished:April 29, 2021DOI:https://doi.org/10.3168/jds.2020-19883

      ABSTRACT

      This paper investigated the limits of the current approach for the determination of the fatty acids profile of milk fats from proton nuclear magnetic resonance data based on the hypothesis that the signal at 0.96 ppm, currently assigned in the literature as a marker for the “short chain fatty acids,” is generated only by the butyric moiety (not by all of the short-chain fatty acids, which also include C6:0—caproic acid). The hypothesis was tested and experimentally confirmed. Moreover, the triplet at 0.96 ppm can also be due to n-3 fatty acids such as linolenic acid (C18:3); therefore, a previously reported methodology for the fatty acids profiling of dairy products—considered as general in the literature—cannot be used in fraud-detection approaches because it allows linolenic acid to be mistaken for butyric acid, consequently leading to misclassification of adulterated samples as nonadulterated. To support our opinion, we have applied the current literature approach for the determination of the fatty acids composition of 3 synthetic nondairy fat blends and have obtained fatty acid compositions similar to milk fats, allowing for their misclassification as genuine milk fats. However, in reality, the blends had very different compositions, as confirmed by gas chromatography. Consequently, we have highlighted the weaknesses of the existing methodology for the detection of dairy food adulteration. In return, new proton nuclear magnetic resonance descriptors based on various integral ratios of signals associated with CH2 moiety versus signals associated with butyric and n-3 fatty acids were proposed to detect adulterations.

      Key words

      INTRODUCTION

      Milk fat (MF) is one of the most expensive food products; therefore, it may be subjected to fraud (replacement with cheaper fats or oils;
      • Pereira C.G.
      • Leite A.I.N.
      • Andrade J.
      • Bell M.J.V.
      • Anjos V.
      • Nunes Leite C.A.I.
      • Andrade J.
      • Bell M.J.V.
      • Anjos V.
      Evaluation of butter oil adulteration with soybean oil by FT-MIR and FT-NIR spectroscopies and multivariate analyses.
      ;
      • Leite A.I.N.
      • Pereira C.G.
      • Andrade J.
      • Vicentini N.M.
      • Bell M.J.V.
      • Anjos V.
      FTIR-ATR spectroscopy as a tool for the rapid detection of adulterations in butter cheeses.
      ). In some countries (particularly the developing countries), adulteration of dairy products is recognized as a serious issue (
      • Handford C.E.
      • Campbell K.
      • Elliott C.T.
      Impacts of milk fraud on food safety and nutrition with special emphasis on developing countries.
      ). For example, in Romania, the state institutions are aware of the practice of counterfeiting dairy products on a considerable scale; therefore, an agreement between the National Authority for Consumers Protection and the Producers' Association of Romanian Milk Industry was established on January 21, 2020, aiming at the collaboration of the 2 parties to reduce the counterfeiting practice on dairy products (
      • ANPC. (Autoritatea Nationala pentru Protectia Consumatorilor)
      Intalnire de lucru.
      ). The agreement proves that fraud in the dairy industry is a problem acknowledged by the authorities; consequently, developing efficient methods for the detection of fake/adulterated products is necessary.
      Milk fat is distinguished from other fats and oils by the short- and medium-chain fatty acids (FA) content. The standard method for the accurate determination of the fatty acids profile (FAP) of milk and dairy products, GC (
      • Månsson H.L.
      Fatty acids in bovine milk fat.
      ;
      • Laučienė L.
      • Andrulevičiute V.
      • Sinkevičienė I.
      • Sederevičius A.
      • Musayeva K.
      • Šernienė L.
      Analysis of fatty acid composition and healthy lipids indices in raw and processed milk.
      ), requires harmful chemicals and is time-consuming (almost 2 h/sample with an experienced operator). Therefore, faster alternative methods are welcome. Recently, spectroscopic methods coupled with multivariate data analysis have received attention, as they are considerably faster and more convenient from the viewpoint of the practical procedure. For example, modified partial least square regression applied on near infrared spectroscopic data allowed quantification of 19 FA in cheese (from C8:0 to C20:0, including saturated and unsaturated species) with individual confidence levels of up to 96% (
      • Gonzáles-Martin M.I.
      • Vivar-Quintana A.M.
      • Revilla I.
      • Salvador-Esteban J.
      The determination of fatty acids in cheeses of variable composition (cow, ewe's, and goat) by means of near infrared spectroscopy.
      ). The capabilities, advantages, limitations, and drawbacks of other chromatographic (GC-MS, HPLC, LC-MS) and spectral [fluorescence, Fourier-transform infrared, near infrared, mid infrared, nuclear magnetic resonance (NMR)] techniques combined with multivariate data analysis to detect adulteration of dairy products were systematically discussed in excellent reviews (
      • Kamal M.
      • Karoui R.
      Analytical methods coupled with chemometric tools for determining the authenticity and detecting the adulteration of dairy products: A review.
      ;
      • Alexandri E.
      • Ahmed R.
      • Siddiqui H.
      • Choudhary M.
      • Tsiafoulis C.G.
      • Gerothanassis I.P.
      High Resolution NMR Spectroscopy as a Structural and Analytical Tool for Unsaturated Lipids in Solution.
      ).
      The standard method for the specific determination of the butyric moiety of MF or MF-containing fat blends is based on the saponification of fat with KOH. The corresponding FA are liberated from the resulting soaps by acidification with H3PO4, before the separation of the water-insoluble and water-soluble FA. Butyric acid is subsequently directly determined by GLC in the presence of valeric acid as an internal standard (
      • Dieffenbacher A.
      • Pocklington W.D.
      Section 2: Oils and Fats.
      ). Although accurate, the method is laborious, and reagent- and time-consuming.
      • Sacchi R.
      • Paduano A.
      • Caporaso N.
      • Picariello G.
      • Romano R.
      • Addeo F.
      Assessment of milk fat content in fat blends by 13C-NMR spectroscopy analysis of butyrate.
      recently developed a faster method for the determination of butyric acid in MF blends with lard and vegetable margarine through 13C-NMR spectroscopy, based on the distinct C1 and C2 resonances of the butyryl backbone, used as markers of milk triacylglycerols. A modern approach based on Raman spectroscopy coupled with chemometrics (
      • Gómez-Mascaraque L.G.
      • Kilcawley K.
      • Hennessy D.
      • Tobin J.T.
      • O'Callaghan T.F.
      Raman spectroscopy: A rapid method to assess the effects of pasture feeding on the nutritional quality of butter.
      ) has allowed quantification of 28 FA from butter, including C4:0, as influenced by the cow feeding systems.
      A recent paper (
      • Tociu M.
      • Todasca M.-C.
      • Bratu A.
      • Mihalache M.
      • Manolache F.
      Fast approach for fatty acid profiling of dairy products fats using 1H-NMR spectroscopy.
      ) aiming at the determination of the FAP of MF through chemometric equations on 1H-NMR data reported a rapid quantification of the FA on 4 classes: short-chain, polyunsaturated, monounsaturated, and long-chain saturated FA. The equations resulted from the integrals of the 1H-NMR signals of MF, based on the contribution of specific protons from each FA class. For example, the short-chain FA were determined from the integral of the triplet at 0.96 ppm (assigned to the terminal methyl of C4-C8 acyl chains) weighted against all methyl terminal protons of all fatty acyl chains. The method was further applied for the differentiation of cheese substitutes (made of vegetable ingredients) from genuine bovine milk cheeses and the detection of adulteration of cheeses with “exogeneous” fats (mainly palm oil) on the basis of the ratio of long-chain:short-chain FA, which is typically lower than 6.89 for bovine MF (
      • Tociu M.
      • Todasca M.-C.
      • Bratu A.
      • Mihalache M.
      • Manolache F.
      Fast approach for fatty acid profiling of dairy products fats using 1H-NMR spectroscopy.
      ). The adulteration of dairy products with other fats and oils will result in an increase of the above-mentioned ratio, considering the only source of short-chain FA is MF. The major drawback of this approach is that n-3 FA (such as linolenic acid) also have a distinctive 1H-NMR resonance (also triplet) at 0.96 ppm (
      • Brescia M.A.
      • Mazzilli V.
      • Sgaramella A.
      • Ghelli S.
      • Fanizzi F.P.
      • Sacco A.
      1H-NMR characterization of milk lipids: A comparison between cow and buffalo milk.
      ), overlapping the triplet from the short-chain FA; therefore, these 2 species cannot be determined independently. Moreover, only C4:0 contributes to the 0.96 ppm triplet; the methyl from C6:0 resonates at 0.85 ppm together with the rest of the fatty acyl chains. According to the proposed chemometric equations, linolenic acid may be counted as a short-chain FA, falsely increasing the short-chain FA content of MF blends with other fats/oils. The issue becomes serious because the reported method plays a key role in the detection of adulteration.
      The present work aimed to discuss the limits of the 1H-NMR spectroscopy for the determination of the FAP of dairy products, with emphasis on adulteration scenarios. From the hypothesis that only the methyl terminal protons from butyric moiety resonate at 0.96 ppm, the short-chain FA cannot be determined as a class. Moreover, because the marker signals of butyric and linolenic moieties overlap at 0.96 ppm, these species cannot be determined independently, and thus may be misleading in fraud-detection approaches. Hence, the novelty of our study was such: (1) to evaluate the previously reported 1H-NMR methods of determining the FAP of dairy fats, pointing out their limitations and possibly misleading aspects in authenticity issues, (2) to showcase the weaknesses of a previously reported method by applying it on synthetic nondairy fat blends simulating the butter fat (BF) composition, leading to sample misclassification as genuine dairy fat, and (3) to propose new 1H-NMR fat descriptors based on various integral ratios of signals associated with CH2 moiety versus signals associated with butyric and n-3 FA, which can be useful in authenticity approaches. Therefore, our reported results are useful in the food control area to assess dairy product authenticity and to detect fraud.

      MATERIALS AND METHODS

      Reagents and Samples

      Supelco 37 Component FAME Mix was purchased from Supelco; methanol, CH2Cl2 (HPLC purity), and anhydrous MgSO4 were purchased from Sigma Aldrich; tributyrin (97%) and caproic acid (p.a.) were purchased from Merck; and BF3-MeOH (10–14%) complex was purchased from Alpha-Aesar. Sodium methoxide was freshly prepared from Na (metallic) and methanol. Butter samples of certified bovine origin (n = 35) were obtained from Romanian ISO 22000:2018 licensed dairy companies (Napolact and Covalact). Butter fat was extracted from butter samples with CH2Cl2, dried on anhydrous MgSO4, and followed by evaporation of the solvent (
      • Kontson A.
      • Tamsma A.
      • Kurtz F.E.
      Method for extracting fat from dry whole milk.
      ). Linseed and Lallemantia iberica seeds were obtained from the National Agricultural Research and Development Institute of Fundulea, Romania. Oil was extracted from seeds according to standard Soxhlet protocol (
      • Shahidi F.
      Unit 1.1: Extraction and Measurement of Total Lipids.
      ). Beef and sheep tallow were extracted according to
      • Shahidi F.
      Unit 1.1: Extraction and Measurement of Total Lipids.
      , with small modifications: the subcutaneous adipose tissue was homogenized with quartz sand before Soxhlet extraction with CH2Cl2, dried on anhydrous MgSO4, and followed by evaporation of the solvent. Coconut oil was purchased from Trio Verde S.R.L. (distributor), and palm stearin and palm kernel oil were purchased from Scintilla Silk (distributor).

      Preparation of the Methyl Caproate

      Methyl caproate was prepared from caproic acid and methanol, using H2SO4 as a catalyst (
      • Li Y.
      • Watkins B.A.
      Unit 1.2: Analysis of Fatty Acids in Food Lipids.
      ). The 1H-NMR signals were assigned as follows: 0.85 ppm (triplet, –CH3, 3H), 1.25 ppm (multiplet, –CH2–CH2–CH3, 4H), 1.64 ppm (quintet, –OOC–CH2–CH2–, 2H), 2.26 ppm (triplet, –OOC–CH2–CH2–, 2H), 3.60 ppm (singlet, H3C–OOC–, 3H), where hydrogen atoms that contributed to the referred signals are in italics.

      Sample Derivatization

      The FAME of vegetable oils (linseed, coconut, Lallemantia iberica, palm stearin, and palm kernel oil) were transesterified with methanol, using BF3-MeOH complex as a catalyst, according to the standard protocol (
      • Li Y.
      • Watkins B.A.
      Unit 1.2: Analysis of Fatty Acids in Food Lipids.
      ). Methylation was carried out under alkaline conditions in methanol with MeONa as a catalyst because isomerization may occur to FA containing conjugated double bonds (such as in dairy and ruminant meat products, which may contain conjugated linoleic acids) in the case of animal fats (BF, beef and sheep tallow) under acidic conditions and high temperature (
      • Li Y.
      • Watkins B.A.
      Unit 1.2: Analysis of Fatty Acids in Food Lipids.
      ).

      Gas Chromatograms

      The gas chromatograms of FAME were recorded on an Agilent Technologies 7890A instrument equipped with autosampler and triple axis MS detector model 5975 C VL MSD. The operating conditions were previously described (
      • Anastasiu A.-E.
      • Chira N.-A.
      • Banu I.
      • Ionescu N.
      • Stan R.
      • Roşca S.-I.
      Oil productivity of seven Romanian linseed varieties as affected by weather conditions.
      ). The FAP on 5 categories (i.e. C4:0, n-3, n-6, monounsaturated, and saturated FA) are presented in Table 1.
      Table 1Fatty acid profiles
      Expressed as g/100 g. But = butyric acid; n-3 = total amount of n-3 fatty acids (mainly linolenic acid, Ln); n-6 = total amount of n-6 fatty acids (mainly di-unsaturated such as linoleic acid, Di); MUFA = total amount of monounsaturated fatty acids; SFA = total amount of saturated fatty acids, except for butyric acid.
      of fats and oils (mean values
      Data were processed from the indicated references.
      )
      Oil or fatButn-3 (Ln)n-6 (Di)MUFASFAReference
      Butter fat
      Determined in our laboratory (n = number of analyzed samples; all samples were analyzed in triplicate) and in agreement with the indicated references.
      (from bovine milk)
      4.620.731.8826.7965.98n = 35,
      • Månsson H.L.
      Fatty acids in bovine milk fat.
      ; Laucčienė et al. (2019)
      Goat milk2.490.552.6822.3271.96
      • Yurchenko S.
      • Sats A.
      • Tatar V.
      • Kaart T.
      • Mootse H.
      • Jõudu I.
      Fatty acid profile of milk from Saanen and Swedish Landrace goats.
      Sheep milk2.571.312.3229.7564.05
      • Markiewicz-Kęszycka M.
      • Czyżak-Runowska G.
      • Lipińska P.
      • Wójtowski J.
      Fatty acid profile of milk - A Review.
      Coconut oil
      Determined in our laboratory (n = number of analyzed samples; all samples were analyzed in triplicate) and in agreement with the indicated references.
      001.857.1690.99n = 4
      Corn oil00.7648.9733.6716.60
      • Dorni C.
      • Sharma P.
      • Saikia G.
      • Longvah T.
      Fatty acid profile of edible oils and fats consumed in India.
      Cotton seeds oil00.3551.8119.6628.18
      • Dorni C.
      • Sharma P.
      • Saikia G.
      • Longvah T.
      Fatty acid profile of edible oils and fats consumed in India.
      Groundnut oil0026.9653.7719.27
      • Dorni C.
      • Sharma P.
      • Saikia G.
      • Longvah T.
      Fatty acid profile of edible oils and fats consumed in India.
      Mustard oil00083.6916.31
      • Jamwal R.
      • Amit S.
      • Kumari S.
      • Balan B.
      • Dhaulaniya A.S.
      • Kelly S.
      • Cannavan A.
      • Singh D.K.
      Attenuated total Reflectance–Fourier transform infrared (ATR–FTIR) spectroscopy coupled with chemometrics for rapid detection of argemone oil adulteration in mustard oil.
      Brazil nut kernel oil00.0130.8637.2231.92
      • Cornelio-Santiago H.P.
      • Mazalli M.R.
      • Rodrigues C.E.C.
      • de Oliveira A.L.
      Extraction of Brazil nut kernel oil using green solvents: Effects of the process variables in the oil yield and composition.
      Palm olein00.311.2343.6244.85
      • Dorni C.
      • Sharma P.
      • Saikia G.
      • Longvah T.
      Fatty acid profile of edible oils and fats consumed in India.
      Safflower oil00.1376.5814.219.08
      • Dorni C.
      • Sharma P.
      • Saikia G.
      • Longvah T.
      Fatty acid profile of edible oils and fats consumed in India.
      High oleic safflower oil00.2512.7879.697.28
      • Knothe G.
      • Kenar J.A.
      Determination of the fatty acid profile by 1H-NMR Spectroscopy.
      Mid oleic sunflower oil00.4433.4457.558.57
      • Knothe G.
      • Kenar J.A.
      Determination of the fatty acid profile by 1H-NMR Spectroscopy.
      Soybean05.1654.1624.7715.91
      • Dorni C.
      • Sharma P.
      • Saikia G.
      • Longvah T.
      Fatty acid profile of edible oils and fats consumed in India.
      Sunflower0062.6925.9211.39
      • Dorni C.
      • Sharma P.
      • Saikia G.
      • Longvah T.
      Fatty acid profile of edible oils and fats consumed in India.
      Rapeseed oil09.2918.9763.178.57
      • Dorni C.
      • Sharma P.
      • Saikia G.
      • Longvah T.
      Fatty acid profile of edible oils and fats consumed in India.
      Beef tallow
      Determined in our laboratory (n = number of analyzed samples; all samples were analyzed in triplicate) and in agreement with the indicated references.
      00.342.4640.4156.79n = 5
      Pork lard00.5612.9845.2141.25
      • Indrasti D.
      • Che Man Y.B.
      • Mustafa S.
      • Hashim D.M.
      Lard detection based on fatty acids profile using comprehensive gas chromatography hyphenated with time-of-flight mass spectrometry.
      Sheep tallow
      Determined in our laboratory (n = number of analyzed samples; all samples were analyzed in triplicate) and in agreement with the indicated references.
      00.396.2543.7849.58n = 4,
      • Aali M.
      • Moradi-Shahrbabak H.
      • Moradi-Shahrbabak M.
      • Sadeghi M.
      • Kohram H.
      Polymorphism in the SCD gene is associated with meat quality and fatty acid composition in Iranian fat- and thin-tailed sheep breeds.
      Chicken fat0017.3056.2026.50
      • Mohammed H.H.H.
      • Jin G.
      • Ma M.
      • Khalifa I.
      • Shukat R.
      • Elkhedir A.E.
      • Zeng Q.
      • Noman A.E.
      Comparative characterization of proximate nutritional compositions, microbial quality and safety of camel meat in relation to mutton, beef, and chicken.
      Palm stearin
      Determined in our laboratory (n = number of analyzed samples; all samples were analyzed in triplicate) and in agreement with the indicated references.
      005.1024.8470.06n = 5,
      • Liu C.
      • Meng Z.
      • Chai X.
      • Liang X.
      • Piatko M.
      • Campbell S.
      • Liu Y.
      Comparative analysis of graded blends of palm kernel oil, palm kernel stearin and palm stearin.
      Palm kernel oil
      Determined in our laboratory (n = number of analyzed samples; all samples were analyzed in triplicate) and in agreement with the indicated references.
      002.5816.8780.55n = 6,
      • Naksuk A.
      • Sabatini D.A.
      • Tongcumpou C.
      Microemulsion-based palm kernel oil extraction using mixed surfactant solutions.
      Lallemantia iberica
      Determined in our laboratory (n = number of analyzed samples; all samples were analyzed in triplicate) and in agreement with the indicated references.
      068.459.0911.6410.82n = 3,
      • Amini R.
      • Ebrahimi A.
      • Nasab A.D.M.
      Moldavian balm (Dracocephalum moldavica L.) essential oil content and composition as affected by sustainable weed management treatments.
      Linseed
      Determined in our laboratory (n = number of analyzed samples; all samples were analyzed in triplicate) and in agreement with the indicated references.
      060.2216.2116.037.54n = 5,
      • Anastasiu A.-E.
      • Chira N.-A.
      • Banu I.
      • Ionescu N.
      • Stan R.
      • Roşca S.-I.
      Oil productivity of seven Romanian linseed varieties as affected by weather conditions.
      ,
      • Xie Y.
      • Yan Z.
      • Niu Z.
      • Coulter J.A.
      • Niu J.
      • Zhang J.
      • Wang B.
      • Yan B.
      • Zhao W.
      • Wang L.
      Yield, oil content, and fatty acid profile of flax (Linum usitatissimum L.) as affected by phosphorus rate and seeding rate.
      1 Expressed as g/100 g. But = butyric acid; n-3 = total amount of n-3 fatty acids (mainly linolenic acid, Ln); n-6 = total amount of n-6 fatty acids (mainly di-unsaturated such as linoleic acid, Di); MUFA = total amount of monounsaturated fatty acids; SFA = total amount of saturated fatty acids, except for butyric acid.
      2 Data were processed from the indicated references.
      3 Determined in our laboratory (n = number of analyzed samples; all samples were analyzed in triplicate) and in agreement with the indicated references.

      The 1H-NMR Spectra

      The 1H-NMR experiments were recorded in a field of 6.9 T using a Bruker Fourier spectrometer operating at 1H Larmor frequency of 300.18 MHz. The 1H-NMR spectral acquisition characteristics were as follows: 45° pulse, 5.37-s acquisition time, 6.1-kHz spectral window, 16 scans, 20 K data points, 1-s delay time; all spectra were recorded at 25°C. Fat samples were dissolved in CDCl3 (isotopic purity 99.8% D, Sigma Aldrich).

      Statistical Data Processing

      All samples were analyzed in triplicate. We used XLStat (Addinsoft, 14 d free trial) for data processing and 3-dimensional data visualization.

      RESULTS AND DISCUSSION

      Comparative 1H-NMR Spectral Characterization of BF and Other Fats and Oils

      Figure 1 presents the superimposed spectra of BF (a), tributyrin (b), linseed oil (c), and methyl caproate (d). The 1H-NMR spectra of oils and fats are similar in terms of signals. They usually display 10 signals (illustrated in Figure 1c for the case of linseed oil) that are characteristic to specific structural groups of the fatty acyl chains and glycerol backbone, as described in Table 2 (
      • Knothe G.
      • Kenar J.A.
      Determination of the fatty acid profile by 1H-NMR Spectroscopy.
      ). Linseed oil was chosen to illustrate the specific 1H-NMR signals because it displays all the main signals with high intensity and contains large amounts of C18:3.
      Figure thumbnail gr1
      Figure 1Superimposed proton nuclear magnetic resonance (1H-NMR) spectra of methyl caproate (a), linseed oil (b), tributyrin (c), and butter fat (d). f1 refers to the chemical shift.
      Table 2Chemical shifts and peak assignment of proton nuclear magnetic resonance spectra of fats and oils (
      • Knothe G.
      • Kenar J.A.
      Determination of the fatty acid profile by 1H-NMR Spectroscopy.
      )
      SignalChemical shift (ppm)Proton
      The protons contributing to each specific signal are in italics.
      Compound
      A0.85−CH2–CH2–CH2–CH3All acids except butyric acid and linolenic acid
      B0.96−OOC–CH2–CH2–CH3Butyric acid
      B′0.96−CH=CH–CH2–CH3Linolenic acid
      C1.24−(CH2)nAll fatty acids
      D1.64−CH2–CH2–COO–All fatty acids
      E2.02−CH2–CH=CH–All UFA
      F2.26−CH2–COO–All fatty acids
      G2.76−CH=CH–CH2–CH=CH–n-6 (linoleic) acid and n-3 (linolenic) acid
      H4.19−CH2OCORH in the sn-1/3 position of the glycerol backbone
      I5.15−CHOCORH in the sn-2 position of the glycerol backbone
      J5.29−CH=CHAll UFA
      1 The protons contributing to each specific signal are in italics.
      It is generally accepted that BF differs from other fats and oils due to its specific FA composition with characteristic short-chain FA moiety (C4:0 and C6:0), as well as important amounts of medium-chain FA (C8:0–C12:0). It appeared that BF had a distinctive signal (triplet B, see Figure 1) that corresponded to the terminal –CH3 group from butyric acid moiety that was shifted downfield compared with the rest of the FA due to the proximity of the ester group. Caproic acid moiety on the other hand—although considered as a short-chain FA—contributed through its methyl group to signal A (Figure 1d), together with the rest of the FA, except for the linolenic acid, whose terminal methyl 1H-NMR signal B′ (Figure 1c) was also shifted downfield (due to its proximity to the C=C double bond) and overlapped with signal B from butyric acid. Therefore, we disagreed with
      • Tociu M.
      • Todasca M.-C.
      • Bratu A.
      • Mihalache M.
      • Manolache F.
      Fast approach for fatty acid profiling of dairy products fats using 1H-NMR spectroscopy.
      on 2 issues: (1) signal B (0.96 ppm) cannot be assigned in a bulky manner to the short-chain FA class (because it results only from the contribution of C4:0), and (2) signal B cannot be considered as a distinctive signal (marker) for C4:0 (and even less for the short-chain FA class;
      • Tociu M.
      • Todasca M.-C.
      • Bratu A.
      • Mihalache M.
      • Manolache F.
      Fast approach for fatty acid profiling of dairy products fats using 1H-NMR spectroscopy.
      ) because it completely overlaps the terminal methyl from any n-3 FA (e.g., linolenic). Even if BF contains small amounts of C18:3 (up to 1–2%), its contribution to signal B should not be neglected. Moreover, for the detection of adulteration in dairy products, considering signal B as a marker for dairy fats can be misleading, falsely allowing linolenic acid to account for butyric acid.
      On the other hand, it should be noticed that at the same amplification degree (calibrated on signal H, Figure 1a and c), 1H-NMR spectrum of a genuine BF sample differs from spectra of other oils and fats due to the very weak signal G, corresponding to bis-allylic protons from n-3 and n-6 FA. Indeed, BF contains low amounts of linolenic acid (0.19–1.45%) depending on species, diet, lactation stage, or breed (
      • Månsson H.L.
      Fatty acids in bovine milk fat.
      ;
      • Markiewicz-Kęszycka M.
      • Czyżak-Runowska G.
      • Lipińska P.
      • Wójtowski J.
      Fatty acid profile of milk - A Review.
      ) and linoleic acid (2–3%), respectively, which explains the very small bis-allylic moiety contributing to signal G. In addition, other signals associated with unsaturated FA (i.e., signal J: unsaturated protons –CH=CH– and signal E: allylic protons), although present due to the large amount of oleic acid (C18:1), are less intense than in the case of more unsaturated oils. Therefore, when facing 1H-NMR spectra of fats from uncertain dairy provenience, before proceeding to calculations, analysts should first investigate possible structural characteristics (such as the ratios of the integral values of signals J, G, B, and A, respectively) that might indicate adulteration with cheaper fats and oils.

      Can FAP of Milk and Nonmilk Fat Blends Be Determined from the 1H-NMR Data?

      In the case of most oils and fats (of both animal and vegetal origin), 1H-NMR spectra may be used as raw data for the determination of the FAP on 4 classes: saturated, monounsaturated, di-unsaturated, and tri-unsaturated FA (
      • Knothe G.
      • Kenar J.A.
      Determination of the fatty acid profile by 1H-NMR Spectroscopy.
      ;
      • Chira N.A.
      • Todasca M.C.
      • Nicolescu A.
      • Rosu A.
      • Nicolae M.
      • Rosca S.I.
      Evaluation of the computational methods for determining vegetable oils composition using 1H-NMR spectroscopy.
      ). The equations leading to the FA composition derive from the integrals of the 1H-NMR spectra of oils, based on the contributions of specific protons from each class of FA. The tri-unsaturated FA may be computed assuming this class is mainly constituted of n-3 FA such as the α-linolenic acid, whose terminal methyl group has a distinctive triplet (signal B′, Figure 1c) shifted downfield (0.96 ppm) compared with the rest of the FA (signal A, 0.85 ppm). (
      • Knothe G.
      • Kenar J.A.
      Determination of the fatty acid profile by 1H-NMR Spectroscopy.
      ). Indeed, α-linolenic acid is the main representative of this class, occurring in considerably larger amounts than its corresponding γ-linolenic acid or other tri-unsaturated n-6 FA, that can therefore be neglected.
      In the case of genuine BF, butyric acid has also a distinctive signal (triplet) corresponding to the terminal –CH3 group, shifted downfield compared with the rest of the FA currently occurring in the MF, including caproic acid. As previously mentioned, the short-chain FA class of BF is composed of butyric acid and caproic acid, of which only butyric acid may be determined independently, based on its distinctive signal (triplet B, Figure 1). Caproic acid, on the other hand, contributes through its methyl group to signal A, together with the rest of the FA, except for the linolenic acid, whose terminal methyl signal overlaps signal B from butyric acid. In these conditions, the balance of specific proton contribution leads to the following equations:
      IA=3k(Sat+Mono+Di)Sat+Mono+Di=IA3k,
      [1]


      IB=3k(But+Ln)But+Ln=IB3k.
      [2]


      Because all fatty acyl chains (including butyric moiety, Figure 1b) have methylene groups contributing to signals F (adjacent to –COO) and D (in β position from –COO),
      ID=2k(But+Ln+Sat+Mono+Di)But+Ln+Sat+Mono+Di=ID2k,
      [3]


      IF=2k(But+Ln+Sat+Mono+Di)But+Ln+Sat+Mono+Di=IF2k,
      [4]


      IE=4k(Mono+Di+Ln)Mono+Di+Ln=IE4k,
      [5]


      IG=k(2Di+4Ln)=2k(Di+2Ln)Di+2Ln=IG2k.
      [6]


      Signals I (corresponding to the single proton in the sn-2 position of the glycerol backbone) and J (–CH=CH–) are partially overlapping each other, and thus cannot be integrated individually; however, their integrals may be computed from signal H (corresponding to 4 protons in sn-1/3 position from glycerol):
      II=k×IH/4,
      [7]


      IJ=II+Jk×IH/4=k(2Mono+4Di+6Ln)=2k(Mono+2Di+3Ln)Mono+2Di+3Ln=II+Jk×IH/42k.
      [8]


      The sum of all molar fractions equals 1, leading to another equation:
      But+Ln+Sat+Mono+Di=1,
      [9]


      where IA, IB, … IJ are the integrals of A, B, …and J signals, respectively; But, Sat, Mono, Di, and Ln = molar ratios of butyric, SFA (except for butyric acid, MUFA, n-6 FA (di-unsaturated FA, i.e. linoleic acid), and n-3 FA (tri-unsaturated FA, i.e., linolenic acid), respectively.
      Additionally, k is a constant that is specific for each spectrometer/software; it is a computed coefficient that correlates the integral with the number of protons that generate the signal. Consequently, k value may be determined based on different signals. For example, replacing equation 9 into equation 3 or 4 leads to
      k=ID2
      [10]


      or
      k=IF2.
      [11]


      From equation [1] + equation [2], we get the following:
      k=IA+IB3.
      [12]


      Constant k can also be computed as a mean of the values obtained from equations 10 to 12.
      Because protons leading to signals A, B, D, and F have different relaxation times (especially longitudinal T1), the corresponding integral value per one proton is slightly different; therefore, k value will differ when computed with equations 10 to 12. From our previous experience working with vegetable oils, the best results (in terms of the best correlation with GC data) were obtained when k was computed according to equation 12 (
      • Chira N.A.
      • Todasca M.C.
      • Nicolescu A.
      • Rosu A.
      • Nicolae M.
      • Rosca S.I.
      Evaluation of the computational methods for determining vegetable oils composition using 1H-NMR spectroscopy.
      ).
      Equations 1 through 9 form a system of 9 equations with 6 unknowns (i.e., But, Ln, Sat, Mono, Di, and k). However, equations such as (1, 2, and 3) or (5, 6, and 8) are circular, and thus interdependent (e.g., equation 5 + equation 6 = equation 8). Consequently, only equations 1, 2, 5, 6, and 9 are independent, leading to an undetermined system with 5 equations and 6 unknowns that cannot be resolved. Moreover, there are not 2 independent equations with But and Ln as unknowns from which But and Ln could be specifically determined.
      However, given that pure BF contains low amounts of C18:3, there is temptation to neglect its contribution to signal B as in the approach proposed by
      • Tociu M.
      • Todasca M.-C.
      • Bratu A.
      • Mihalache M.
      • Manolache F.
      Fast approach for fatty acid profiling of dairy products fats using 1H-NMR spectroscopy.
      , allowing for the assessment of the butyric, di-unsaturated, monounsaturated, and saturated moieties as molar ratios, respectively:If we replace equation 12 in equation 2:
      But=IBIA+IB.
      [13]


      If we replace equation 12 in equation 6:
      Di=32×IGIA+IB.
      [14]


      If we subtract equation 14 from equation 5:
      Mono=34×IE2IGIA+IB.
      [15]


      Saturated (molar %):
      Sat=1ButMonoDi.
      [16]


      This method may be useful for the rapid determination of the FAP of genuine dairy products; however, in addition to the inherent lack of accuracy caused by the deliberate neglection of linolenic acid, the approach is risky for authentication matters.
      On the other hand, if spectra are recorded on instruments operating at frequencies higher than 500 MHz, the indetermination of the original chemometric equation system (i.e. equations 1, 2, 5, 6, and 9) could be resolved.
      • Brescia M.A.
      • Mazzilli V.
      • Sgaramella A.
      • Ghelli S.
      • Fanizzi F.P.
      • Sacco A.
      1H-NMR characterization of milk lipids: A comparison between cow and buffalo milk.
      suggested a different approach for the determination of C18:3 in buffalo MF from 1H-NMR data, based on the bis-allylic signal G. At frequencies higher than 500 MHz, the 2 triplets corresponding to the bis-allylic moieties from linolenic (signal G1, shifted downfield at 2.83 ppm) and linoleic acids (signal G2, shifted upfield at 2.79 ppm) appear almost separated (Figure 2a). Consequently, the authors proposed to separately integrate the 2 triplets, allowing the determination of C18:3 and C18:2 moieties as follows:
      IG1=4k×LnLn=IG14k,
      [17]


      IG2=2k×DiDi=IG22k.
      [18]


      Consequently, the rest of the molar fractions (But, Mono, and Sat) may be determined based on equations 2, 5, and 1, respectively.
      Figure thumbnail gr2
      Figure 2Expanded spectral window of the bis-allylic region: (a) capture from
      • Brescia M.A.
      • Mazzilli V.
      • Sgaramella A.
      • Ghelli S.
      • Fanizzi F.P.
      • Sacco A.
      1H-NMR characterization of milk lipids: A comparison between cow and buffalo milk.
      (reprinted with permission from the Journal of the American Oil Chemists' Society) of buffalo milk fat recorded on a 500-MHz instrument; (b) bovine butter fat, recorded on a 300-MHz instrument.
      Compared with the neglection of the C18:3 fraction discussed in equations 13 through 16, this approach allows for the determination of linolenic acid, increasing the FAP accuracy. However, errors in the assessment of the Ln and Di moieties may still be considerable because the separate integration of signals G1 and G2 does not respect the general rule of signals integration (i.e. from baseline to baseline). In addition, on instruments operating at frequencies lower than 500 MHz, the 2 triplets appear partially overlapped (Figure 2b); therefore, their individual integration is not possible. Consequently, butyric acid cannot be unequivocally determined from 1H-NMR data with acceptable accuracy.
      However, the butyric acid moiety may be determined from quantitative 13NMR data. Compared with 1Hw-NMR spectra, the electronic configuration of the 13C nucleus determines the spreading of the chemical shifts over ∼200 ppm frequency range, resulting in minor signal overlapping and, as a consequence, enabling more refined structural information to determine the butyric fraction. Butyric acid displays 2 distinctive 13C resonances (1 peak at 173.13 ppm, corresponding to sn-1,3 carbonyl carbon, and 1 peak at 35.34 ppm, assigned to the methyl group), which appear well-separated from adjacent signals. These 2 resonances may therefore be integrated separately and represent markers for butyric acid. According to
      • Sacchi R.
      • Paduano A.
      • Caporaso N.
      • Picariello G.
      • Romano R.
      • Addeo F.
      Assessment of milk fat content in fat blends by 13C-NMR spectroscopy analysis of butyrate.
      , any of these signals may be exploited to assess the amount of butyric acid by referring their areas to the corresponding signals of long-, medium- and short-chain FA (C6–C22). This approach has led to good results in the evaluation of the MF content in fat blends, based on the 13C-NMR analysis of C4:0. However, for quantitative 13C-NMR analysis, when carbons have different relaxation behavior (as in the case of the carbonyl carbons and methylene/methyl carbons of the fatty acyl chains), the longitudinal relaxation times (T1) must be known for all carbons to ensure they are completely relaxed between 2 consecutive pulses. Consequently, q13C-NMR spectra should be acquired using experimental conditions that allow for complete relaxation of carbon nuclei between 2 subsequent pulses, based on the known T1 values for different acyl carbons (
      • Ng S.
      Quantitative analysis of partial acylglycerols and free fatty acids in palm oil by 13C nuclear magnetic resonance spectroscopy.
      ). Long relaxation times and high number of scans necessary to obtain higher signal/noise ratios for quantitative determinations with acceptable accuracy make q13C-NMR spectroscopy a time-consuming technique. From time perspective, the method becomes comparable with GC, which, although lengthy, offers an accurate and detailed sample FAP.

      Investigating Dairy Product Adulteration Through 1H-NMR: Traps and Challenges

      Food fraud has been intensively studied through several methods. Generally, these methods are based on extensive databases of authentic samples which are analyzed for their composition or other specific features (e.g. fingerprinting through spectroscopic techniques), leading to mathematical models. In the case of an unknown sample (suspicioned of adulteration with cheaper fats/oils), the use of 1H-NMR to assess its FAP may prove misleading, due to the difficulty of correctly assessing the linolenic acid content and the possibility of confusing linolenic acid for butyric acid.
      • Tociu M.
      • Todasca M.-C.
      • Bratu A.
      • Mihalache M.
      • Manolache F.
      Fast approach for fatty acid profiling of dairy products fats using 1H-NMR spectroscopy.
      have tested their proposed system of chemometric equations on vegetable cheese produced from palm oil and obtained a 100% correct classification of the investigated samples of vegetable provenience based on the “short-chain fatty acids content,” which was lower than the typical value for bovine MF. In this case, the correct classification was due to the fact that palm oil does not contain linolenic acid, leading to the absence of the triplet at 0.96 ppm, considered by the authors as specific for the short-chain FA.
      However, in the case of dairy products adulterated with cheaper oils/fats containing convenient amounts of linolenic acid, the approach proposed by
      • Tociu M.
      • Todasca M.-C.
      • Bratu A.
      • Mihalache M.
      • Manolache F.
      Fast approach for fatty acid profiling of dairy products fats using 1H-NMR spectroscopy.
      may lead to sample misclassification. It is even possible that in the absence of GC and based only on the 1H-NMR profile, appropriate vegetable and animal fats/oils compositions (of nonmilk origin) can imitate the 1H-NMR profiles of MF. Indeed, relatively small amounts (6–7%) of oils rich in linolenic acid such as linseed oil or Lallemantia iberica seed oil will ensure the triplet at 0.96 ppm, and the rest up to 100% will be constituted from fats with very low amounts of linoleic acid (n-6) and rich in monounsaturated and saturated FA (such as the coconut oil, beef tallow, palm kernel oil, palm stearin, and mustard oil, Table 1). Such “smart adulterations” can simulate the 1H-NMR spectra of genuine butter samples. Figure 3 illustrates, as an example, the superimposed 1H-NMR spectra of a simulated butter of vegetal origin (Lallemantia iberica oil 6% + coconut oil 40% + beef tallow 54%) opposite to an authentic BF. As it can be seen, the 2 spectra are similar and easy to mistake in the attempt of determining their FAP from 1H-NMR data. Indeed, applying the method proposed by
      • Tociu M.
      • Todasca M.-C.
      • Bratu A.
      • Mihalache M.
      • Manolache F.
      Fast approach for fatty acid profiling of dairy products fats using 1H-NMR spectroscopy.
      for the determination of the FA composition of 3 synthetic nondairy fat blends leads to FAP similar to MF, allowing for their misclassification as genuine MF; although, in reality, the blends had very different compositions, as confirmed by GC (Table 3).
      Figure thumbnail gr3
      Figure 3Superimposed spectra of butter fat (BF; blue) and simulated BF (ternary mixture of Lallemantia iberica oil 6% + coconut oil 40% + beef tallow 54%; red).
      Table 3Simulated butter from nonmilk fats
      No.Simulated butter compositionFatty acid profile
      The proton nuclear magnetic resonance (1H-NMR) composition was computed according to the method proposed by Tociu et al. (2018). But = butyric acid; n-3 = total amount of n-3 fatty acids (mainly linolenic acid, Ln); n-6 = total amount of n-6 fatty acids (mainly di-unsaturated such as the linoleic acid, Di); MUFA = total amount of monounsaturated fatty acids; SFA = total amount of saturated fatty acids, except for butyric acid.
      (g/100 g, mean value
      Three replicates.
      ± SD)
      Oil or fatAmount (%)Butn-3 (Ln)n-6 (Di)MUFASFA
      1H-NMRGC1H-NMRGC1H-NMRGC1H-NMRGC1H-NMRGC
      1Lallemantia iberica oil64.46 ± 0.324.11 ± 0.282.78 ± 0.193.05 ± 0.2125.86 ± 1.0926.91 ± 0.6266.90 ± 1.8765.93 ± 0.83
      Coconut oil40
      Beef tallow54
      2L. iberica oil64.36 ± 0.524.17 ± 0.263.05 ± 0.233.83 ± 0.1922.06 ± 0.7223.11 ± 0.5770.53 ± 2.1068.89 ± 0.95
      Sheep tallow20
      Palm kernel oil74
      3Linseed oil7.54.23 ± 0.294.01 ± 0.234.63 ± 0.415.07 ± 0.2420.02 ± 0.9221.30 ± 0.6171.12 ± 2.0569.62 ± 0.93
      Palm kernel oil46
      Palm stearin46.5
      1 The proton nuclear magnetic resonance (1H-NMR) composition was computed according to the method proposed by
      • Tociu M.
      • Todasca M.-C.
      • Bratu A.
      • Mihalache M.
      • Manolache F.
      Fast approach for fatty acid profiling of dairy products fats using 1H-NMR spectroscopy.
      . But = butyric acid; n-3 = total amount of n-3 fatty acids (mainly linolenic acid, Ln); n-6 = total amount of n-6 fatty acids (mainly di-unsaturated such as the linoleic acid, Di); MUFA = total amount of monounsaturated fatty acids; SFA = total amount of saturated fatty acids, except for butyric acid.
      2 Three replicates.

      How Can 1H-NMR Data Help in the Detection of “Smart Adulterations”?

      Although we have demonstrated that the FA composition of dairy fats cannot be reliably determined from 1H-NMR data and subsequently used for authentication purposes, it would be wrong to deny the potential of 1H-NMR data in assessing dairy authenticity and fraud detection. An essential characteristic of the milk FAP is its important amount of short- and medium-chain FA. This feature is also reflected in the 1H-NMR spectra, through signal C, which is generated by –CH2– groups that have not been specifically assigned. Therefore, signal C is correlated with the FA chain lengths. Consequently, key features that needs to be further explored in the detection of dairy products adulteration approaches are the ratios of integrals of signals assigned to –CH2– groups (including signal C, allylic, and bis-allylic groups) and specific signals related to butyric/n-3 FA and other structural features (such as unsaturation and bis-allylic groups). Therefore, we propose the integral ratios (C+D+E+F+G)/F, (C+D+E+F+G)/(I+J), and B/(A+B) as 3 new 1H-NMR fat descriptors (where A, B, C, D, E, F, G, I, and J refer to the integral values of the corresponding signals, according to Table 2). Table 4 presents their mean values for several fats and oils and for the 3 simulated BF compositions described in Table 3.
      Table 4Proton nuclear magnetic resonance (1H-NMR) descriptors for fats and oils
      No.Oil or fatn1H-NMR descriptor
      Mean values ± SD, computed based on the integral values of the indicated signals. A, B, C, D, E, F, G, I, and J refer to the integral values of the corresponding signals, according to Table 2.
      (C+D+E+F+G)/F (descriptor 1)(C+D+E+F+G)/(I+J) (descriptor 2)B/(A+B) (descriptor 3)
      1Butter fat3511.80 ± 0.3310.80 ± 0.800.12 ± 0.02
      2Sheep tallow414.50 ± 0.1911.24 ± 0.150.02 ± 0.01
      3Lallemantia iberica seed311.21 ± 0.130.99 ± 0.080.59 ± 0.01
      4Coconut411.09 ± 0.3514.13 ± 0.080.01 ± 0.01
      5Beef tallow513.82 ± 0.1812.15 ± 0.260.02 ± 0.01
      6Palm kernel oil611.50 ± 0.1611.05 ± 0.29 ND
      ND = not detected (assumed as 0 for graphical data visualization).
      7Linseed oil511.73 ± 0.151.17 ± 0.170.57 ± 0.02
      8Palm stearin512.69 ± 0.119.97 ± 0.17ND
      9Simulated butter 1
      Compositions corresponding to Table 3.
      312.45 ± 0.188.00 ± 0.230.10 ± 0.02
      10Simulated butter 2
      Compositions corresponding to Table 3.
      311.52 ± 0.128.83 ± 0.100.10 ± 0.01
      11Simulated butter 3
      Compositions corresponding to Table 3.
      311.91 ± 0.147.97 ± 0.090.09 ± 0.01
      1 Mean values ± SD, computed based on the integral values of the indicated signals. A, B, C, D, E, F, G, I, and J refer to the integral values of the corresponding signals, according to Table 2.
      2 ND = not detected (assumed as 0 for graphical data visualization).
      3 Compositions corresponding to Table 3.
      As it can be seen, fats and oils have specific values of the proposed descriptors, according to their specific compositions. For example, linseed oil and Lallemantia iberica seed oil (which are highly unsaturated oils) differ significantly from BF with respect to descriptors 2 and 3, and saturated fats show closer values. On the other hand, the simulated BF compositions differ from genuine BF with respect to descriptor 2. The orthogonal representation of descriptor 1 versus descriptor 2 (Figure 4) allows for the data visualization. It can be observed that samples cluster according to the species. Not surprisingly, the unsaturated oils group distinctly and away from the BF class, and the more saturated samples group closer to BF. Among these, coconut, beef, and sheep tallow form distinct clusters from BF, and palm kernel oil and palm stearin classes partially overlap BF group. The simulated BF samples (represented in red) place distinctly, but close, to the BF samples. Consequently, the first 2 descriptors are not sufficient for a satisfactory separation of palm kernel oils, palm stearin, and the 3 simulated BF samples from the BF group. In this case, the B:(A+B) ratio (i.e. descriptor 3) can help in the discrimination according to the 3-dimensional orthogonal plot represented in Figure 5.
      Figure thumbnail gr4
      Figure 4Orthogonal representation of descriptor 1 versus descriptor 2 (analytical formulas corresponding to ). C, D, E, F, G, I, and J refer to the integral values of the corresponding signals, according to .
      Figure thumbnail gr5
      Figure 5Three-dimensional orthogonal representation of descriptors 1, 2, and 3 for butter fat, palm kernel oil, palm stearin, and the simulated butter fat samples (analytical formulas corresponding to ). A, B, C, D, E, F, G, I, and J refer to the integral values of the corresponding signals, according to .
      As it can be observed, palm kernel oils and palm stearin samples group distinctively from the BF samples in the lower side of the plot, and the simulated BF samples place in the upper side, still apart from the BF group, allowing for a clear separation. The separation of fat samples of dairy products that originate from oils and fats of different vegetal or animal origin based on the NMR data was also reported in the literature. For example, according to a recent paper (
      • Guyader S.
      • Thomas F.
      • Portaluri V.
      • Jamin E.
      • Akoka S.
      • Silvestre V.
      • Remaud G.
      Authentication of edible fats and oils by non-targeted 13C INEPT NMR spectroscopy.
      ), the application of a refocused adiabatic 13C Insensitive Nuclei Enhanced by Polarization Transfer sequence enables discrimination of BF from 30 different oils and fats of animal and vegetal origins (including margarines), proving that butter could clearly be distinguished from plant oil. On the other hand, the separation of butter or margarine into polar (water soluble) and nonpolar fractions (soluble in CDCl3) and subsequent analysis of these fractions by 1H-NMR spectroscopy allows for a comprehensive analysis of the fat constituents (i.e., quantification of conjugated linoleic acids, rumenic acid, diglycerides, and linoleic acid) and a clear differentiation of butter samples from margarines (
      • Schripsema J.
      Comprehensive analysis of polar and apolar constituents of butter and margarine by nuclear magnetic resonance, reflecting quality and production processes.
      ).

      CONCLUSIONS

      In this paper, we demonstrated that—contrary to a previously reported method—in the case of the dairy fat samples, the butyric and linolenic moieties cannot be accurately determined from 1H-NMR spectra due to the overlapping resonances of methyl signals at 0.96 ppm. Moreover, short-chain FA cannot be determined as a class. We have highlighted the weaknesses of the respective method by applying it on 3 synthetic nondairy fat blends by obtaining FA profiles similar to BF, which lead to their misclassification as genuine dairy provenience. In return, we have proposed 3 new 1H-NMR fat descriptors based on various integral ratios of signals associated with CH2 moiety versus signals associated with butyric and n-3 FA that allowed the correct classification of both BF samples and simulated BF compositions as well as other vegetable (linseed, coconut, palm stearin, palm kernel oil, Lallemantia iberica seed oil) and animal fats (beef and sheep tallow). The 2-dimensional and 3-dimensional data visualization by means of orthogonal representation of the new descriptors allowed discrimination of MF from the rest of the fat samples. Consequently, 1H-NMR spectroscopy can be used to detect dairy fraud, but not through FAP determination.

      ACKNOWLEDGMENTS

      Anamaria Hanganu acknowledges project no. CNFISFDI-2020-0355, “Consolidarea și creșterea competitivității instituționale a Universtății din București ca hub de proiecte interdisciplinare și instituție-gazdă pentru cercetători de top din țară și din străinătate,” Bucharest, Romania. The University of Bucharest–UniRem project no. 244 is gratefully acknowledged. The authors thank Constantinescu (Chemont S. A., Mouscron, Belgium) for providing certified Lallemantia iberica seed samples. The authors have not stated any conflicts of interest.

      REFERENCES

        • Aali M.
        • Moradi-Shahrbabak H.
        • Moradi-Shahrbabak M.
        • Sadeghi M.
        • Kohram H.
        Polymorphism in the SCD gene is associated with meat quality and fatty acid composition in Iranian fat- and thin-tailed sheep breeds.
        Livest. Sci. 2016; 188: 81-90
        • Alexandri E.
        • Ahmed R.
        • Siddiqui H.
        • Choudhary M.
        • Tsiafoulis C.G.
        • Gerothanassis I.P.
        High Resolution NMR Spectroscopy as a Structural and Analytical Tool for Unsaturated Lipids in Solution.
        Molecules. 2017; 22 (28981459): 1663-1735
        • Amini R.
        • Ebrahimi A.
        • Nasab A.D.M.
        Moldavian balm (Dracocephalum moldavica L.) essential oil content and composition as affected by sustainable weed management treatments.
        Ind. Crops Prod. 2020; 150112416
        • Anastasiu A.-E.
        • Chira N.-A.
        • Banu I.
        • Ionescu N.
        • Stan R.
        • Roşca S.-I.
        Oil productivity of seven Romanian linseed varieties as affected by weather conditions.
        Ind. Crops Prod. 2016; 86: 219-230
        • ANPC. (Autoritatea Nationala pentru Protectia Consumatorilor)
        Intalnire de lucru.
        https://anpc.ro/articol/1383/intalnire-de-lucru-1
        Date: 2020
        Date accessed: September 25, 2020
        • Brescia M.A.
        • Mazzilli V.
        • Sgaramella A.
        • Ghelli S.
        • Fanizzi F.P.
        • Sacco A.
        1H-NMR characterization of milk lipids: A comparison between cow and buffalo milk.
        J. Am. Oil Chem. Soc. 2004; 81: 431-436
        • Chira N.A.
        • Todasca M.C.
        • Nicolescu A.
        • Rosu A.
        • Nicolae M.
        • Rosca S.I.
        Evaluation of the computational methods for determining vegetable oils composition using 1H-NMR spectroscopy.
        Rev. Chim. 2011; 62: 42-46
        • Cornelio-Santiago H.P.
        • Mazalli M.R.
        • Rodrigues C.E.C.
        • de Oliveira A.L.
        Extraction of Brazil nut kernel oil using green solvents: Effects of the process variables in the oil yield and composition.
        J. Food Process Eng. 2019; 42e13271
        • Dieffenbacher A.
        • Pocklington W.D.
        Section 2: Oils and Fats.
        in: Standard Methods for the Analysis of Oils, Fats and Derivatives – 1st Supplement. 7th ed. Oxford: Blackwell Scientific Publications, 1992: 2.310/1-4
        • Dorni C.
        • Sharma P.
        • Saikia G.
        • Longvah T.
        Fatty acid profile of edible oils and fats consumed in India.
        Food Chem. 2018; 238 (28867107): 9-15
        • Gómez-Mascaraque L.G.
        • Kilcawley K.
        • Hennessy D.
        • Tobin J.T.
        • O'Callaghan T.F.
        Raman spectroscopy: A rapid method to assess the effects of pasture feeding on the nutritional quality of butter.
        J. Dairy Sci. 2020; 103 (32773315): 8721-8731
        • Gonzáles-Martin M.I.
        • Vivar-Quintana A.M.
        • Revilla I.
        • Salvador-Esteban J.
        The determination of fatty acids in cheeses of variable composition (cow, ewe's, and goat) by means of near infrared spectroscopy.
        Microchem. J. 2020; 156104854
        • Guyader S.
        • Thomas F.
        • Portaluri V.
        • Jamin E.
        • Akoka S.
        • Silvestre V.
        • Remaud G.
        Authentication of edible fats and oils by non-targeted 13C INEPT NMR spectroscopy.
        Food Control. 2018; 91: 216-224
        • Handford C.E.
        • Campbell K.
        • Elliott C.T.
        Impacts of milk fraud on food safety and nutrition with special emphasis on developing countries.
        Compr. Rev. Food Sci. Food Saf. 2016; 15 (33371582): 130-142
        • Indrasti D.
        • Che Man Y.B.
        • Mustafa S.
        • Hashim D.M.
        Lard detection based on fatty acids profile using comprehensive gas chromatography hyphenated with time-of-flight mass spectrometry.
        Food Chem. 2010; 122: 1273-1277
        • Jamwal R.
        • Amit S.
        • Kumari S.
        • Balan B.
        • Dhaulaniya A.S.
        • Kelly S.
        • Cannavan A.
        • Singh D.K.
        Attenuated total Reflectance–Fourier transform infrared (ATR–FTIR) spectroscopy coupled with chemometrics for rapid detection of argemone oil adulteration in mustard oil.
        Lebensm. Wiss. Technol. 2020; 120108945
        • Kamal M.
        • Karoui R.
        Analytical methods coupled with chemometric tools for determining the authenticity and detecting the adulteration of dairy products: A review.
        Trends Food Sci. Technol. 2015; 46: 27-48
        • Knothe G.
        • Kenar J.A.
        Determination of the fatty acid profile by 1H-NMR Spectroscopy.
        Eur. J. Lipid Sci. Technol. 2004; 106: 88-96
        • Kontson A.
        • Tamsma A.
        • Kurtz F.E.
        Method for extracting fat from dry whole milk.
        J. Dairy Sci. 1973; 56: 144-145
        • Laučienė L.
        • Andrulevičiute V.
        • Sinkevičienė I.
        • Sederevičius A.
        • Musayeva K.
        • Šernienė L.
        Analysis of fatty acid composition and healthy lipids indices in raw and processed milk.
        J. Food Nutr. Res. 2019; 7: 386-390
        • Leite A.I.N.
        • Pereira C.G.
        • Andrade J.
        • Vicentini N.M.
        • Bell M.J.V.
        • Anjos V.
        FTIR-ATR spectroscopy as a tool for the rapid detection of adulterations in butter cheeses.
        Lebensm. Wiss. Technol. 2019; 109: 63-69
        • Li Y.
        • Watkins B.A.
        Unit 1.2: Analysis of Fatty Acids in Food Lipids.
        in: Whitaker J. Current Protocols in Food Analytical Chemistry. John Wiley & Sons Inc, 2001: D1.2.1-D1.2.15
        • Liu C.
        • Meng Z.
        • Chai X.
        • Liang X.
        • Piatko M.
        • Campbell S.
        • Liu Y.
        Comparative analysis of graded blends of palm kernel oil, palm kernel stearin and palm stearin.
        Food Chem. 2019; 286 (30827657): 636-643
        • Månsson H.L.
        Fatty acids in bovine milk fat.
        Food Nutr. Res. 2008; 52 (19109654): 1821-1823
        • Markiewicz-Kęszycka M.
        • Czyżak-Runowska G.
        • Lipińska P.
        • Wójtowski J.
        Fatty acid profile of milk - A Review.
        Bull. Vet. Inst. Pulawy. 2013; 57: 135-139
        • Mohammed H.H.H.
        • Jin G.
        • Ma M.
        • Khalifa I.
        • Shukat R.
        • Elkhedir A.E.
        • Zeng Q.
        • Noman A.E.
        Comparative characterization of proximate nutritional compositions, microbial quality and safety of camel meat in relation to mutton, beef, and chicken.
        Lebensm. Wiss. Technol. 2020; 118108714
        • Naksuk A.
        • Sabatini D.A.
        • Tongcumpou C.
        Microemulsion-based palm kernel oil extraction using mixed surfactant solutions.
        Ind. Crops Prod. 2009; 30: 194-198
        • Ng S.
        Quantitative analysis of partial acylglycerols and free fatty acids in palm oil by 13C nuclear magnetic resonance spectroscopy.
        J. Am. Oil Chem. Soc. 2000; 77: 749-755
        • Pereira C.G.
        • Leite A.I.N.
        • Andrade J.
        • Bell M.J.V.
        • Anjos V.
        • Nunes Leite C.A.I.
        • Andrade J.
        • Bell M.J.V.
        • Anjos V.
        Evaluation of butter oil adulteration with soybean oil by FT-MIR and FT-NIR spectroscopies and multivariate analyses.
        Lebensm. Wiss. Technol. 2019; 107: 1-8
        • Sacchi R.
        • Paduano A.
        • Caporaso N.
        • Picariello G.
        • Romano R.
        • Addeo F.
        Assessment of milk fat content in fat blends by 13C-NMR spectroscopy analysis of butyrate.
        Food Control. 2018; 91: 231-236
        • Schripsema J.
        Comprehensive analysis of polar and apolar constituents of butter and margarine by nuclear magnetic resonance, reflecting quality and production processes.
        J. Agric. Food Chem. 2008; 56 (18355021): 2547-2552
        • Shahidi F.
        Unit 1.1: Extraction and Measurement of Total Lipids.
        in: Whitaker J. Current Protocols in Food Analytical Chemistry. John Wiley & Sons Inc, 2001: D1.1.1-D1.1.11
        • Tociu M.
        • Todasca M.-C.
        • Bratu A.
        • Mihalache M.
        • Manolache F.
        Fast approach for fatty acid profiling of dairy products fats using 1H-NMR spectroscopy.
        Int. Dairy J. 2018; 83: 52-57
        • Xie Y.
        • Yan Z.
        • Niu Z.
        • Coulter J.A.
        • Niu J.
        • Zhang J.
        • Wang B.
        • Yan B.
        • Zhao W.
        • Wang L.
        Yield, oil content, and fatty acid profile of flax (Linum usitatissimum L.) as affected by phosphorus rate and seeding rate.
        Ind. Crops Prod. 2020; 145112087
        • Yurchenko S.
        • Sats A.
        • Tatar V.
        • Kaart T.
        • Mootse H.
        • Jõudu I.
        Fatty acid profile of milk from Saanen and Swedish Landrace goats.
        Food Chem. 2018; 254 (29548460): 326-332