Journal of Food Composition and Analysis 28 (2012) 31–39

Original Research Article

Gas chromatography–mass spectrometry approach to study fatty acid profiles infried potato crisps

Pilar Manzano, Juan Carlos Diego, Marıa Jesus Nozal, Jose Luis Bernal, Jose Bernal *

IU CINQUIMA, Analytical Chemistry Group, University of Valladolid, E-47071 Valladolid, Spain

A R T I C L E I N F O

Article history:

Received 7 December 2011

Received in revised form 27 June 2012

Accepted 16 July 2012

Keywords:

Fatty acids profiles

Frying oils

GC–MS

Potato crisps

Principal component analysis

Soxhlet extraction

Two-way ANOVA

Food analysis

Food composition

A B S T R A C T

The fatty acid profiles of crisp samples from Hermes and Mustang potato varieties were obtained using a

validated GC–MS method. Among the four different solid–liquid extraction procedures tested, Soxhlet

extraction with diethyl ether was chosen due to the higher recovery percentages obtained (�90%) and

the shorter sample treatment time required. The proposed method was applied to analyze 28 crisps

samples from both potato varieties, which were fried with two batches of similar vegetable oils, and the

majority fatty acids in potato crisps (�33.3 g/100 g) were C18:0, C16:0, C18:2(n6) and C18:1(n9). The

results of a two-way ANOVA test have demonstrated that the potato variety caused more significant

differences (11) than the frying oil batch (4) in relation to the fatty acid content. A PCA analysis has made

it possible to relate the potato variety with the fatty acid content; it was observed that the first four

principal components represented 85.5% of the variability, with a consequent reduction in the

dimensions of the data from 36 variables to 4 components. Finally, it was found that the Hermes variety

contained lower amounts of total fatty acids (�34.4 g/100 g) and v-6/v-3 ratios (�17.3) than the

Mustang variety (�35.6 g/100 g and �18.6).

� 2012 Elsevier Inc. All rights reserved.

Contents lists available at SciVerse ScienceDirect

Journal of Food Composition and Analysis

jo u rn al ho m epag e: ww w.els evier . c om / lo cat e/ j fc a

1. Introduction

Potato crisps are a widely extended snack food, andnowadays their consumption is having a greater influence onthe dietary habits of populations around the world. Therefore, anevaluation of their composition is essential, as they are foodswith a high calorie intake, mainly due to their considerable fattyacid (FA) composition (Fernandez-San Juan, 2000). The presenceof FAs in the final product could be influenced by several factors,such as the manufacturing process, but this also depends on thefrying oil type used or the potato variety, as will be discussedthroughout this work.

Regarding the oil, many food manufacture companies arecurrently increasing their commitment to consumers’ health bycarrying out research into the quality of the oils used for crispfrying. In this regard, it must be pointed out that vegetable oils arethe most used. Many types of vegetable oils can be used, andsubsequently a large amount of different FAs can be added to the

Abbreviations: PCA, principal component analysis; PC, principal component; FA,

fatty acid; FAME, fatty acid methyl esther; FID, flame ionization detector; SFE,

supercritical fluid extraction; ASE, accelerated solvent extraction; ICH, Internation-

al cooperation on harmonization.

* Corresponding author. Tel.: +34 983 186347; fax: +34 983 423013.

E-mail address: jose.bernal@qa.uva.es (J. Bernal).

0889-1575/$ – see front matter � 2012 Elsevier Inc. All rights reserved.

http://dx.doi.org/10.1016/j.jfca.2012.07.003

final product. But, at the same time, the variety of potato shouldalso be considered, because during frying, the water present in theraw material evaporates and is partially replaced by oil; thisconstitutes up to 40% of the finished product and consequentlyinfluences its properties. This affects not only the flavor and aromaof the product, but also the texture, in accordance with thequantity of oil absorbed during frying (Kita et al., 2007). Therefore,the selection of the potato variety according to its own physical–chemical characteristics is related with the oil content and thesubsequent final fatty acid content in the potato crisps. Concern-ing this issue, a study has recently been published (Ooko andKabira, 2011) in which five different potato varieties wereevaluated as raw materials for producing French fries and potatocrisps in Kenya, and one of the parameters which, according to theauthors, showed some effect on the oil content was the dry matterinherent in each potato variety. In the existing literature, severalstudies have been published where the effect of the frying oil (typeor temperature) on the FA content in potato crisps was evaluated(Aro et al., 1998; Fernandez-San Juan, 2000; Kita et al., 2007; Ookoand Kabira, 2011; Sanches-Silva et al., 2004; Wagner et al., 2008),but to our knowledge no study has been published which has alsomade a detailed analysis of the relation between the potatovariety alone, or in combination with the frying oil, and the fattyacid profile. For this reason, it was decided to perform the studysummarized in this manuscript.

P. Manzano et al. / Journal of Food Composition and Analysis 28 (2012) 31–3932

The analytical technique most often employed to study FAs isgas chromatography (GC), with flame ionization (FID) usuallyselected for the identification of total fatty acids (Fernandez-SanJuan, 2000; Kita et al., 2007; Lo Scalzo et al., 2007; Wagner et al.,2008); however, mass spectrometry (MS) detectors have beenpreferred to determine minority FAs (Hauff and Vetter, 2009;Jimenez et al., 2009; Toribio et al., 2011; Zu et al., 2009), andnowadays the use of multidimensional gas chromatography isgaining in importance in this field (Hejazi et al., 2009; Herreroet al., 2009; Manzano et al., 2011). As a particular case, a reversed-phase high performance liquid chromatographic (RP-HPLC)method with UV detection has also been used for FAs determina-tion (Sanches-Silva et al., 2004). It must be pointed out that theconversion of FAs into more volatile compounds than free acidcomponents, usually in their methyl esters derivatives (FAMEs) (Liand Watkins, 2001; Morrison and Smith, 1964), is essential prior toGC analysis. Consequently, the FAs derivatization procedure andthe reagents employed in this process played a significant rolewhen identifying and determining FAs. Several derivatizationreagents have been used, KOH in methanol catalyzed with BF3

being widely used (Hauff and Vetter, 2009; Jimenez et al., 2009;Kita et al., 2007; Li and Watkins, 2001; Lo Scalzo et al., 2007;Morrison and Smith, 1964; Toribio et al., 2011), due to the fact thatBF3 provoked rapid methylation of FAs, and that when they werefreshly made and stored properly, BF3 solutions could last nearlytwo years (Li and Watkins, 2001).

One of the main problems when determining FAs compositionis the fat extraction step, as it depends to a great deal on the natureof the sample and the diversity of the FAs in terms of their chainlength, branching, degree of unsaturation and the position andgeometry of double bonds (Aro et al., 1998; Shahidi, 2001). So far,within the huge variety of extraction methods employed, the mostcommonly used were: solid–liquid extraction (Fernandez-SanJuan, 2000; Folch et al., 1957; Jimenez et al., 2009; Kita et al., 2007;Li and Watkins, 2001; Shahidi, 2001), acid digestion (Shahidi,2001), supercritical fluid extraction (SFE) (Catchpole et al., 2009;Hauff and Vetter, 2009; Sahena et al., 2009; Toribio et al., 2011),accelerated solvent extraction (ASE) (Hauff and Vetter, 2009;Wagner et al., 2008) and microwave procedures (Zu et al., 2009).One of the most commonly used extraction methods was describedby Folch et al. (1957). This has been modified several times (Blighand Dyer, 1959; Fernandez-San Juan, 2000; Li and Watkins, 2001;Shahidi, 2001), the best known version being performed by Blighand Dyer (1959), but always based on the use of a solvent mixture(chloroform and methanol), to obtain the simultaneous extractionof neutral and polar lipids. Furthermore, Soxhlet extraction hasalso usually been employed in order to extract fats and oils in foodmatrices (using different ethers (Jimenez et al., 2009; Kita et al.,2007; Shahidi, 2001), together with other solvents like hexane(Pedneault et al., 2008).

To summarize, the main aim of this study was to analyze indetail the influence of the potato variety alone or combined withthe effect of frying oil on the fatty acid content of commercialpotato crisps, as this had not been done before. To achieve thisaim and taking into account the existing scientific literatureconcerning the determination of FAs in food matrices and, morespecifically, in potato derived products, it was decided to useGC–MS as an analytical separation tool. In order to extract asmany FAs as possible from the potato crisp samples, fourdifferent solid–liquid extraction procedures were tested. Oncethe analytical method was optimized, it was validated andapplied to analyze 28 potato crisps samples. Finally, and toevaluate the influential factors mentioned above, namely potatovariety and frying oil, statistical tools such as two-way analysisof variance (ANOVA) and principal component analysis (PCA)were used.

2. Materials and methods

2.1. Chemicals and standard solutions

A standard mixture of FAMEs in dichloromethane (reference47885-u), and reference standards of methyl palmitate (C16:0),octadecanoate (C18:0), cis-9 oleate (C18:1(n9)) and linoleate(C18:2(n6)), were purchased from Supelco (Bellefonte, PA, USA).Fatty acids were named using the formula Cx:y(nz;catb), where ‘‘x’’is the number of carbon atoms, ‘‘y’’ the number of double bondsand ‘‘z’’ is the position of the first double bond beginning at themethyl terminal group; ‘‘a’’ and ‘‘b’’ were the conventionalpositions of the double bonds with cis, ‘‘c’’, or trans, ‘‘t’’,stereoisomerism, which were omitted in the formula when allthe double bonds of FAs were cis-type.

Standard stock solutions were prepared in dichloromethane(Labscan, Dublin, Ireland) at a concentration of 1000 mg/L. Thesestandard stock solutions were diluted daily with dichloromethaneto produce a set of working standards. All standards and stocksolutions were kept in the dark at +4 8C and were stable for at least1 month.

Deionized water was obtained in a Milli-RO plus systemtogether with a Milli-Q system from Millipore (Bedford, MA, USA).Chloroform and hexane were obtained from Labscan (Dublin,Ireland). Diethyl ether was supplied by Panreac (Barcelona, Spain),while 1-propanol, hydrochloric acid (HCl) 32%, potassium hydrox-ide (KOH) 1 M in methanol and boron triflouride (BF3) 14% (w/w) inmethanol were purchased from Sigma–Aldrich (St. Louis, MO,USA). All the reagents used were of analytical grade.

2.2. Samples

A total of 28 different potato crisp samples directly obtainedafter the frying process were analyzed. All samples were providedby Facundo S.A. (Villada, Palencia, Spain) and belonged to twodifferent potato varieties, Hermes and Mustang, both of which arelargely used for industrial food processing. Some of the moreimportant physico-chemical characteristics of both potato varie-ties are summarized in the website of Agrico UK Ltd. (Agrico, 2009).Two batches of a similar mixture of vegetable oils for frying thepotatoes were employed: vegetable oil 1, which was composed ofolive oil:sunflower oil (80:20), and vegetable oil 2, composed ofolive oil:sunflower oil (82:18).

In the factory the potatoes were processed in an automaticcontinuous snack fryer, where they were washed, trimmed and cutinto slices of 1.48 � 0.18 mm in thickness. After removal of the starchin hot water at 70–90 8C for 3–4 min and superficial drying, thepotatoes were placed in a fryer (3000 L capacity) with oil heated to170–180 8C for 3–4 min, until the moisture content was below 2%.Finally, sampling was carried out as will be below mentioned, andcrisps samples were taken to the laboratory, where they were finelyground and stored at 4 8C until analysis.

A batch sampling was carried out. Once the industrial process ofmaking the potato crisps was completed, sample portions (15 g)from 10 different points of the same industrial batch were taken. Itmust be pointed out that seven batches of every possiblecombination of potato variety and frying oil (4 combinations)were used to perform the sampling. Following a homogenizationprocess, a large sample portion (150 g) was stored at 4 8C in a darkatmosphere. All samples were treated in the same conditionsregarding washing, trimming, cutting, starch removal, tempera-ture and other frying conditions, the differential factors repre-sented by the potato variety and the oil frying. A sample from eachof the batches of every possible combination of potato variety andfrying oil was analyzed (28 samples in total), and every sample wasanalyzed in triplicate not later than 48 h after their sampling.

P. Manzano et al. / Journal of Food Composition and Analysis 28 (2012) 31–39 33

2.3. Sample treatment

A sample of 0.5 g of ground crisps was weighed in celluloseextraction thimbles and placed in the extractor section of theSoxhlet apparatus, along with 90 mL of diethyl ether. Extractionwas carried out continuously for 4 h at the solvent boilingtemperature. Following this, the solvent was removed by RE-111 (Buchi, Flawil, Switzerland) rotary evaporation until dry. Thesample was then transferred with 6 mL of hexane to a 15 mL glasstube, and which the sample was concentrated until dry with anitrogen sample concentrator (Alltech, IL, USA). Subsequently, thedry extract was hydrolyzed in sealed tubes with 1 mL of KOH 1 Min methanol at 100 8C for 5 min. After cooling of the tube, 1 mL ofBF3/methanol solution was added and heated again for 5 min at100 8C. Following the cooling procedure, 8.5 mL of deionized waterwas added and the solution was extracted with 2 mL of hexane.The extract was stirred for 10 min and immediately centrifuged for5 min at 104.7 rad/s and 20 8C with a 5810R refrigerated centrifuge(Eppendorf; Hamburg, Germany). Afterwards, the hexane layerwas withdrawn, filtered with 0.45 mm nylon syringe filters(Nalgene, Rochester, NY, USA) and collected in vials of 2 mL.Finally, the hexane was removed with a nitrogen sampleconcentrator, and the extracts were reconstituted in 1.5 mL ofdichloromethane.

Due to the large differences in concentrations of FAMEs in thesamples, it was necessary to dilute all the treated samples prior tochromatographic analysis. Therefore, two dilutions of each samplewere consecutively performed with dichloromethane, at a dilutionof 1:400 (v/v), for the identification of majority FAMEs, that is,those that appeared in concentrations greater than 1 g/100 g,C16:0, C18:0, C18:1(n9) and C18:2(n6); and at a dilution of 1:4(v/v), for minority FAMEs, whose presence in samples was lowerthan 1 g/100 g. Finally, both solutions were submitted to GC–MSanalysis.

2.4. GC–MS instrumentation and operating conditions

FAMEs were analyzed by means of an Agilent Technologies (CA,USA) 7890A gas chromatograph coupled to an Agilent Technolo-gies 5975C mass spectrometer equipped with an ALS 7683Bautosampler and a MS ChemStation E 01.00.237 software (AgilentTechnologies, CA, USA). The chromatographic column was an SP-2560 cyano fused silica capillary column (100 m � 0.25 mm ID,0.2 mm) from Supelco (Bellefonte, PA, USA). It must be mentionedthat the initial GC–MS conditions selected were those previously

Table 1Percentages of recovery and relative standard deviations values (n = 5) obtained using the

and 30 g/100 g for majority FAMEs; 0.2 mg/100 g and 100 mg/100 g for minority FAMEs

Extraction procedure Sample (g) % Recovery (%RSD)

Majority FAMEs

C16:0 C18:1(n9)

5 g/100 g 30 g/100 g 5 g/100 g

Modification of the

procedure described

by Folch et al. (Li and

Watkins, 2001)

0.5 66.8 (4.0) 68.1 (3.7) 69.5 (3.5)

0.1 69.7 (2.7) 70.5 (3.1) 61.3 (2.1)

Soxhlet diethyl ether

(Jimenez et al., 2009)

0.5 92.9 (2.0) 93.6 (1.8) 89.5 (0.9)

0.1 76.0 (13) 77.1 (11) 67.8 (17)

Acid digestion

(Shahidi, 2001)

0.5 54.3 (10) 55.5 (9.8) 57.1 (7.4)

Soxhlet n-propanol/water

(Shahidi, 2001)

0.5 35.5 (26) 34.7 (22) 44.8 (15)

optimized in the analysis of FAME in corn and soybean seeds(Jimenez et al., 2009). However, we made some slight modifica-tions in order to improve the separation.

The GC was operated under programmed temperature condi-tions from 50 8C (1 min) to 205 8C (17 min) at 6 8C/min, and thenincreased to 250 8C (10 min) at 10 8C/min. An injection volume of1 mL was employed using the autosampler in pulsed splitlessmode. The inlet temperature was set at 250 8C and the carrier gaswas helium, supplied by Carburos Metalicos (Barcelona, Spain) at aflow rate of 1.1 mL/min.

The MS scan parameters included a mass range of 40–500 m/z, operating in positive electron impact mode with an ionizationenergy of 70 eV. The ion source and quadrupole temperatureswere 240 8C and 180 8C, respectively. Analyses were performedwith selected ion monitoring (SIM) mode, with one target andtwo qualifier ions for each of the FAMEs (see Table 1). Theanalytes were identified and confirmed by comparison of theirretention times and mass spectra with a Mass Spectra Library,Wiley 7N edition (Agilent Part No. G1035B) and referencecompounds.

In Fig. 1, a GC–MS separation of the 36 FAMEs using the above-described conditions is shown. As can be seen, a well-definedseparation was achieved between the different compounds, asalmost all the peaks were baseline resolved.

2.5. Statistical analysis of data

The whole set of data was submitted to two-way ANOVAand principal component analysis (PCA), with the aim ofproviding a better understanding of how to describe the largedata matrix obtained. By means of a correlation matrix, strong orweak relationships between variables can be detected. In thiscase, if responses to the 36 FAs displayed some redundancy, itwould be advantageous to somehow reduce the number ofvariables in the data set, from 36 FA to a few artificial variablestermed principal components. The components were extractedmathematically from the original standardized variablesthrough the correlation matrix. The PCA results were discussedin terms of scores and loadings. Scores are defined as thetransformed variable values corresponding to a particularsample data, while the definition of loadings is the weight bywhich each original variable should be multiplied to obtain thecomponent score. Statistics were carried out using StatgraphicsCenturion XVI version 16.1.03 from Statpoint Technologies Inc.(Warrenton, VA, USA).

four extraction procedures assayed at two different concentration levels (5 g/100 g

).

Minority FAMEs

C14:0 C18:3(n3)

30 g/100 g 0.2 mg/100 g 100 mg/100 g 0.2 mg/100 g 100 mg/100 g

68.5 (2.9) 72.5 (4.5) 71.7 (3.8) 78.2 (4.3) 80.1 (4.6)

63.0 (2.4) 88.3 (6.1) 87.2 (5.4) 84.6 (9.2) 85.2 (8.6)

90.6 (1.2) 96.2 (2.6) 95.6 (2.2) 90.5 (3.8) 91.3 (3.3)

68.8 (16) 80.9 (17) 82.1 (18) 78.2 (15) 79.1 (14)

57.9 (8.0) 41.6 (13) 40.5 (13) 44.9 (8.5) 45.4 (9.0)

45.8 (14) 54.1 (7.8) 53.1 (8.7) 42.4 (8.8) 44.1 (7.9)

Fig. 1. GC–MS chromatograms of a standard mixture of 36 FAMEs. The numbers refer to FAMEs listed in Table 1, and their concentrations were 10 mg/L (FAMEs number: 4, 6,

8–10, 12–14, 16, 18–21, 23–29, 31–34, 36), 20 mg/L (FAMEs number: 1–3, 5, 7, 15, 17, 22, 30, 35) and 30 mg/L (FAME number 11). Chromatographic conditions are

summarized in Section 2.4.

P. Manzano et al. / Journal of Food Composition and Analysis 28 (2012) 31–3934

3. Results and discussion

3.1. Sample treatment

In order to obtain the most complete FA profiles of crispssamples, four different solid–liquid extraction procedures weretested with always the same weight of potato crisp (0.5 g): (i) amodification of the procedure proposed by Folch et al. (Li andWatkins, 2001). Briefly, a 2:1 (v:v) chloroform/methanol mixturewas selected as the organic solvent to extract the FAs from thepotato crisps. The mixture was stirred (10 min) and centrifuged for5 min (104.7 rad/s, 20 8C). The extract was filtered, the lowerorganic phase isolated and, finally, the organic solvent wasremoved with a nitrogen sample concentrator; (ii) acid digestion(Shahidi, 2001) by previous hydrolysis with HCl 6 M at 75 8C for30 min. Following this, fatty acids were extracted with hexane. Theorganic phase was isolated and removed with a nitrogen sampleconcentrator. (iii and iv) Two Soxhlet methods were assessed usingn-propanol/water 3:1 (v/v) (Shahidi, 2001) and diethyl ether(Jimenez et al., 2009) as extractants. Both of them were carried outin a Soxhlet apparatus for 4 h at the temperature corresponding tothe boiling point of each solvent, approximately 85 8C (n-propanol/water) and 35 8C (diethyl ether). Afterwards, solvents wereremoved in a rotary evaporator until dry. The four extractionstudies were always carried out with the same factors, namely, thevariety of potato crisps, derivatization treatment (Jimenez et al.,2009), amount of dichloromethane to reconstitute the final dryextract (1.5 mL) and GC–MS method, with the aim of comparingthe results obtained with these four procedures. The resultssummarized in Table 1 were obtained at two different concentra-tion levels for majority (5 and 30 g/100 g) and minority (0.2 and100 mg/100 g) FAMEs.

After studying the results obtained in the above-mentionedpreliminary experiments (Table 1), two extraction procedures(acid digestion and n-propanol/water Soxhlet extraction) werediscarded due to the low recoveries (<65%) and high relativestandard deviation (%RSD) values (7–25%), in comparison withthose obtained with the other two procedures (recoveries >65%and %RSD <5%) for the different FAME concentration levelsassayed. It must be remarked that in Table 1 the results for onlytwo majority and minority FAMEs are given, as the results forthe other FAME were comparable with these. Therefore, theFolch modification and the Soxhlet extraction with diethyl etherwere repeated by employing a smaller sample weight (0.1 g).After an examination of the results summarized in Table 1, thebest results based on the percentage recoveries in all the

experiments were obtained when using Soxhlet extraction withdiethyl ether and a sample of 0.5 g, with recovery percentageshigher than 89% and %RSD values lower than 4%. As an addedadvantage of the Soxhlet procedure, it must be pointed out thatthe preparation time was shorter.

The derivatization procedure was also optimized by analyzingone single variety of crisps. Initial conditions selected were thoseemployed in previous publications (Jimenez et al., 2009; Manzanoet al., 2011; Toribio et al., 2011), as they provided good results. Thederivatization reagents were exactly the same as in the latterpublications (KOH and BF3 in methanol), and the oven temperatureand derivatization time were varied in order to obtain optimalconditions. To study the influence of each parameter on FAMEanalysis, one was varied while the other remained constant. Theoven temperature was tested at three temperatures (90, 95 and100 8C), whilst oven derivatization time was studied in a rangebetween 5 and 30 min (5 min steps). The results showed that thisdid not play an important role in FAME analysis, as the amount ofFAME was quite similar in all the experiments. Regarding the oventemperature, it must be pointed out that a slight increase in FAMEsignals was observed when using 100 8C. For these reasons, it wasdecided to perform oven derivatization for 5 min at 100 8C.

Finally, the volume of dichloromethane necessary to reconsti-tute the dry extract was evaluated (0.5 and 2.0 mL). Recoveries ofFAME increased in line with volume up to 1.5 mL (�90%), but nosignificant improvement was observed for higher volumes. As aresult, 1.5 mL was selected as optimal dichloromethane volume. InFig. 2, the GC–MS chromatograms are shown for a crisp sample of(A) majority FAMEs with a 1:400 dilution and (B) minority FAMEswith a 1:4 dilution. As might be expected, the separation of thedifferent FAMEs was not as clearly defined as in the standardmixture (Fig. 1), due to the presence of inherent matrixcomponents and the different number of FAs in the crisp samplesand the standard mixture. However, analyzing individual FAMEcontent would not be a problem, as a target ion (Table 2) wasemployed for quantifying each compound.

3.2. Method validation

Method validation was performed according to InternationalCooperation on Harmonization (ICH) guidelines for validation ofanalytical procedures (ICH, 2005).

3.2.1. Linearity

Due to the presence of FAs in this matrix, it was necessary toconstruct matrix matched standards to carry out the standard

Fig. 2. GC–MS chromatograms for a crisp sample of (A) majority FAMEs with a 1:400 dilution and (B) minority FAMEs with a 1:4 dilution. The numbers refer to FAMEs listed in

Table 1. Chromatographic conditions are summarized in Section 2.4.

Table 2FAME parameters and validation data obtained in five replicates of a potato crisp sample.

FAMEs Identification parameters Calibration data Limits Repeatability (n = 5)

Majority RT (min) T Q1 Q2 Slope R2 LOD (mg/100 g) LOQ (mg/100 g) Meana (g/100 g) RSD (%)

11 C16:0 34.3 74.0 227.2 270.2 657670 0.9985 0.12 0.40 3.52 0.2

15 C18:0 37.6 74.0 298.3 255.2 655184 0.9967 0.13 0.42 1.36 2

17 C18:1(n9) 38.8 55.1 264.2 296.2 194502 0.9947 0.65 2.1 25.3 0.1

19 C18:2(n6) 41.1 67.1 294.3 263.2 203492 0.9988 0.43 1.4 4.84 0.5

FAMEs Identification parameters Calibration data Limits Repeatability (n = 5)

Minority RT (min) T Q1 Q2 Slope R2 LOD

(mg/100 g)

LOQ

(mg/100 g)

Meana

(mg/100 g)

RSD (%)

1 C6:0 19.6 74.0 87.0 99.1 836,596 0.9979 0.01 0.02 0.03 7

2 C8:0 22.8 74.0 87.1 127.1 473,915 0.9969 0.08 0.3 7.34 5

3 C10:0 25.9 74.0 143.0 155.1 1,221,267 0.9933 0.06 0.2 3.10 4

4 C11:0 27.3 74.0 157.1 169.1 1,233,815 0.9948 0.01 0.04 0.14 3

5 C12:0 28.7 74.0 171.1 183.2 1,291,932 0.9953 0.09 0.3 8.44 4

6 C13:0 30.1 74.0 185.2 228.3 1,277,773 0.9947 0.01 0.02 0.05 1

7 C14:0 31.5 74.0 199.1 242.2 1,362,089 0.9992 0.08 0.3 115 4

8 C14:1(n5) 32.7 55.1 166.1 208.2 489,550 0.9992 0.03 0.09 3.24 0.9

9 C15:0 32.9 74.0 213.2 256.2 1,367,981 0.9993 0.07 0.2 8.79 5

10 C15:1(n5) 34.1 55.0 222.1 254.2 546,595 0.9973 0.1 0.3 26.8 1

12 C16:1(n7) 35.5 55.1 236.2 268.3 416,950 0.9967 0.09 0.3 20.5 2

13 C17:0 35.8 74.0 241.2 284.3 1,274,223 0.9967 0.08 0.3 7.39 0.9

14 C17:1(n7) 37.1 55.1 250.2 282.2 442,127 0.9991 0.1 0.4 5.51 2

16 C18:1(n9;t9) 38.4 55.1 264.2 296.2 393,889 0.9995 0.1 0.4 25.6 0.5

18 C18:2(n6;t9t12) 39.9 67.1 294.3 263.2 439,113 0.9989 0.09 0.3 <LOQ <LOQ

20 C20:0 42.0 74.0 326.4 283.3 1,032,503 0.9997 0.1 0.5 58.9 0.1

21 C18:3(n6) 42.9 79.0 292.2 194.0 427,997 0.9970 0.1 0.3 22.5 0.1

22 C20:1(n9) 43.7 55.0 292.3 324.4 382,390 0.9982 0.1 0.4 26.1 0.1

23 C18:3(n3) 44.1 79.0 292.2 261.2 537,912 0.9964 0.3 0.9 255 0.3

24 C21:0 44.7 74.0 340.3 297.2 901,757 0.9973 0.06 0.2 0.75 2

25 C20:2(n6) 46.1 81.0 322.2 291.1 397,873 0.9976 0.08 0.3 1.67 0.08

26 C22:0 46.9 74.0 354.4 311.3 848,183 0.9978 0.2 0.8 105 0.4

27 C20:3(n6) 47.8 79.1 320.2 222.1 388,462 0.9983 0.05 0.2 0.75 0.05

28 C22:1(n9) 48.3 55.1 320.3 352.4 313,767 0.9976 0.1 0.4 0.70 2

29 C20:3(n3) 48.5 79.1 320.3 289.4 462,020 0.9981 0.06 0.2 <LOQ <LOQ

30 C20:4(n6) 48.8 79.0 150.0 203.0 422,275 0.9985 0.02 0.09 0.19 2

31 C23:0 49.0 74.0 368.4 325.4 763,669 0.9980 0.05 0.2 4.70 2

32 C22:2(n6) 50.4 67.1 350.3 319.2 352,145 0.9954 0.01 0.03 0.07 7

33 C24:0 51.0 74.0 382.4 339.3 713,670 0.9976 0.06 0.2 63.0 0.8

34 C20:5(n3) 51.3 79.0 91.0 201.1 467,841 0.9969 0.01 0.02 0.01 7

35 C24:1(n9) 52.4 55.1 348.3 306.3 226,888 0.9934 0.09 0.3 <LOQ <LOQ

36 C22:6(n3) 56.6 79.1 119.0 199.1 325,067 0.9981 0.04 0.1 0.17 0.7

RT, retention time; T, target ion; Q1, Q2: qualifier ions.a Mean concentration values for a Mustang potato crisp.

P. Manzano et al. / Journal of Food Composition and Analysis 28 (2012) 31–39 35

P. Manzano et al. / Journal of Food Composition and Analysis 28 (2012) 31–3936

addition method. However, the matrix effect was also studied inorder to check if it is possible to employ solvent standardcalibration curves to determine FAMEs in crisps samples.Subsequently, a comparison was made of the slopes of both typesof calibration curves (solvent standard calibration and the matrixmatched standard addition method) obtained for six concentrationlevels, in duplicate, between 30 g/100 g (majority FAMEs) and0.01 mg/100 g (minority FAMEs).

As the confidence intervals of both slopes overlapped, it couldbe deduced that the matrix had no effect on the FAMEs signal, andit was possible to employ solvent standard calibration to quantifyFAMEs in crisps. Consequently, the calibration data summarizedin Table 2 refers to the solvent standard calibration curves.Linearity was evaluated from the same calibration curvesobtained by linear least-squares regression of target ion responsefor all FAMEs. The graphs obtained were straight lines with anintercept not significantly different from zero (p < 0.05); thesewere linear across the range studied (R2 values higher than 0.99;see Table 2), and the lack of bias was confirmed.

Table 3Two-Way ANOVA data and mean FAMEs content values in crisps samples grouped as a fu

vegetable oil 2).

FAME g/100g � CI

Majority H 1 H 2 M 1

C16:0 3.80 � 0.06 3.17 � 0.03 3.52 � 0.0

C18:0 1.73 � 0.02 2.08 � 0.01 1.36 � 0.0

C18:1(n9) 23.0 � 0.04 23.5 � 0.04 25.3 � 0.0

C18:2(n6) 4.81 � 0.02 5.08 � 0.01 4.84 � 0.0

FAME mg/100g � CI

Minority H 1 H 2 M 1

C6:0 0.2 � 0.1 0.2 � 0.1 0.03 �C8:0 18.5 � 0.2 7.57 � 0.12 7.34 �C10:0 3.53 � 0.09 1.98 � 0.04 3.10 �C11:0 0.14 � 0.01 0.15 � 0.01 0.14 �C12:0 5.93 � 0.16 23.9 � 0.2 8.44 �C13:0 0.03 � 0.01 0.06 � 0.01 0.05 �C14:0 130 � 1 72.1 � 0.2 115 �C14:1(n5) <LOQc <LOQc 3.24 �C15:0 9.51 � 0.11 10.1 � 0.1 8.79 �C15:1(n5) 37.7 � 0.2 27.7 � 0.5 26.8 �C16:1(n7) 27.2 � 0.3 1.26 � 0.01 20.5 �C17:0 8.77 � 0.20 2.98 � 0.13 7.39 �C17:1(n7) 4.78 � 0.12 6.80 � 0.11 5.51 �C18:1(n9;t9) 20.0 � 0.1 24.0 � 0.6 25.6 �C18:2n(6;t9t12) 6.37 � 0.02 6.96 � 0.12 <LOQ

C18:3(n6) 24.9 � 0.1 16.1 � 0.1 22.5 �C18:3(n3) 279 � 1 294 � 4 255 �C20:0 67.1 � 0.1 83.2 � 0.8 58.9�C20:1(n9) 25.7 � 0.2 24.0 � 1.0 26.1 �C20:2(n6) 1.62 � 0.01 2.23 � 0.02 1.67 �C20:3(n6) 0.80 � 0.01 0.84 � 0.01 0.75 �C20:3(n3) <LOQe <LOQe <LOQ

C20:4(n6) <LOQc <LOQc 0.19 �C20:5(n3) <LOQb <LOQb <LOQ

C21:0 0.66 � 0.01 3.3 � 0.2 0.75 �C22:0 98.5 � 0.4 150 � 1 105 �C22:1(n9) 0.68 � 0.01 1.8 � 0.1 0.70 �C22:2(n6) 0.04 � 0.01 0.1 � 0.1 0.07 �C22:6(n3) 0.15 � 0.01 0.17 � 0.01 0.17 �C23:0 3.88 � 0.05 10.5 � 0.2 4.70 �C24:0 59.0 � 0.4 68.8 � 1.2 63.0 �C24:1(n9) 0.37 � 0.01 1.2 � 0.1 <LOQ

CI, confidence interval at 95% of significance.a Main effects and interactions: P, potato variety; O, frying oil; P � O, interaction. Effb Limit of quantification (LOQ) < 0.02 g/100 g.c LOQ < 0.09 g/100 g.d LOQ < 0.3 g/100 g.e LOQ < 0.2 g/100 g

3.2.2. Accuracy

Accuracy was evaluated in terms of relative standard deviation(%RSD) to analyze FAME content of a selected crisp sample, theMustang variety, analyzed in quintuplicate with the sampletreatment presented in this manuscript. Results are shown inTable 2, and it can be concluded that the proposed method wasaccurate insofar as the %RSD was between 0.1% and 7%. It shouldbe added that the accuracy results were comparable for theHermes variety.

3.2.3. Detection and quantification limits

The detection limits (LODs) and quantification limits (LOQs) incrisps samples were experimentally determined for each FAME as3 and 10 times the signal-to-noise ratio, respectively. As may beobserved in Table 2, for minority FAMEs LOD values were between0.01 and 0.3 mg/100 g, whereas LOQ values were between 0.02and 0.9 mg/100 g. For majority FAMEs these values were higher,ranging from 0.12 to 0.65 mg/100 g for LODs and 0.40 to 2.1 for theLOQs.

nction of potato variety and frying oil (H, Hermes; M, Mustang; 1, vegetable oil 1; 2,

Effectsa

M 2 P O P � O

1 3.48 � 0.01 *

4 1.40 � 0.01 * *

4 25.0 � 0.07 * *

4 4.68 � 0.02

Effectsa

M 2 P O P � O

0.02 0.09 � 0.01 * *

0.55 6.52 � 0.05 * *

0.20 1.59 � 0.10

0.01 0.16 � 0.01

0.57 25.0 � 1.0 * *

0.02 <LOQb

8 93.9 � 0.5 * *

0.05 3.85 � 0.09

0.70 9.90 � 0.67

0.4 23.1 � 1.2 *

0.6 2.67 � 0.18 * *

0.10 4.20 � 0.15 *

0.15 5.51 � 0.13 * *

0.2 24.8 � 0.3 * *d 6.74 � 0.35

0.6 14.4 � 2.8

14 257 � 4. * *

0.3 90.6 � 1.3 *

0.5 27.3 � 1.7 * *

0.04 1.60 � 0.2

0.04 0.49 � 0.02 * *e <LOQe

0.01 0.19 � 0.02b <LOQb

0.02 0.81 � 0.02 * *

1 181 � 1 *

0.02 0.53 � 0.04

0.01 0.1 � 0.1 * *

0.01 0.16 � 0.01

0.15 3.44 � 0.10

0.8 62.6 � 0.5d 0.82 � 0.02

ects flagged with an asterisk (*) are significant for p < 0.05.

Table 4Weight and percentage average values of fatty acids with health interest in crisps samples (H, Hermes; M, Mustang; 1, vegetable oil 1; 2, vegetable oil 2).

Variety FAMEs

g/100 g % Ratio

Mean Saturated trans v-3 v-6 v-9 v-6/v-3

H 1 34.2 17.4 0.08 0.82 14.2 67.4 17.4

H 2 34.7 16.4 0.09 0.85 14.7 67.9 17.3

M 1 35.8 14.7 0.07 0.71 13.6 70.8 19.0

M 2 35.4 15.1 0.09 0.73 13.3 70.7 18.3

P. Manzano et al. / Journal of Food Composition and Analysis 28 (2012) 31–39 37

3.3. Analysis of crisps samples

The optimized validated method described above was appliedto analyze various crisps samples in order to identify and quantify36 FAMEs. Samples were grouped by means of their variety, andalso by considering the frying oil used in each case. The FAMEconcentrations found in those samples along with their confidenceintervals (a = 0.05) are reported in Table 3.

These results were used to establish FAME profiles for eachcrisp type, and they were also employed in order to obtainnutritional and statistical interpretations. It should be mentionedthat the results were consistent with those expected from a typicalvegetable frying oil, where the most common FAs are C18:1(n9),C18:2(n6), C16:0 and C18:0 (Fernandez-San Juan, 2000; Sanches-Silva et al., 2004). It should be also pointed out that trace levels ofother fatty acids such as C12:0, C14:0, C16:1(n7), C18:3(n3), C20:0,C20:1(n9) and C22:0 were also found. In general, food with lowerFA content is better in terms of human nutrition. Some of the FAMEgroups with health associations are shown in Table 4: saturated,trans, omega 3 (v-3), and omega 6/omega 3 ratio (v-6/v-3). Afterthe results were analyzed it was concluded that the Hermes potatovariety contained lower total FAME concentrations and v-6/v-3ratios; therefore, moderate consumption would not be so harmfulfor human health. However, those values did not balance theirtrans and, especially, saturated FAME percentages, compared withthe Mustang potato variety.

A two-way ANOVA test was done to assess if the potato variety orthe frying oil had a significant influence on FAME content (see Table3), and at the same time the interactions between both factors wereevaluated. Thus, regarding majority FAMEs, significant differenceswere found only for C18:0 and C18:1(n9) in terms of the potatovariety, and for C16:0 in relation to the interaction between the

Fig. 3. Principal component analysis of FAMEs in crisps samples f

frying oil and the potato variety. Meanwhile, for minority FAMEsdifferences regarding the type of potato or the frying oil were foundfor 9 and 4 FAMEs, respectively, while effects due to the interactionbetween the frying oil and variety of potato were observed for 14FAMEs. The larger number of differences found for minorityFAMEs could probably be due to the fact that vegetable oils donot have such a homogeneous composition for minority FAMEs asmajority ones. Nevertheless, it can be concluded that the frying oilwas an influential factor in the fatty acid profiles of potato crisps, ashas been previously observed (Aro et al., 1998; Kita et al., 2007;Wagner et al., 2008). Furthermore, it can also be postulated that thelargest number of significant differences found in the final FAMEscontent of the different crisp samples analyzed was mainly due tothe potato variety. Moreover, the influence of the latter in the fattyacid profiles, the main goal of this study, has been demonstrated bymeans of the data presented in Table 3. As has been previouslystated, to our knowledge this study is the first devoted to analyzingthe influence of the potato variety on the fatty acid profile incommercial crisps; consequently, it is not possible to make a realcomparison with other research. However, a study has recentlybeen published (Ooko and Kabira, 2011) in which an assessmenthas been made of the suitability of five different potato varieties(Purple gold, Kenya Mpya, Sherekea, Tigoni and Dutch Robjin) as rawmaterials for French fries and potato crisps processing. It was foundthat the oil content of the crisps varied significantly (p � 0.05)among the different potato varieties. This finding is consistent withthe results obtained in the present work.

Subsequent to the two-way ANOVA test, a comprehensivestatistical interpretation of the results by means of a principalcomponent analysis (PCA) was carried out. The probabilityassociated with Bartlett’s test of sphericity revealed that thecorrelation matrix was suitable for applying a PCA to describe the

or the first two factors: (A) loadings plot and (B) scores plot.

Fig. 4. Principal component analysis of FAMEs in crisps samples for the third and fourth factors: (A) loadings plot and (B) scores plot.

P. Manzano et al. / Journal of Food Composition and Analysis 28 (2012) 31–3938

data. Statistics of the PCA following varimax rotation showed thatthe first two principal components (PCs) accounted for 54.0% of thetotal variance, while with the first four PCs this represented 85.5%of the variability, with a consequent reduction in the dimensions ofthe data from 36 variables to 4 components.

In order to demonstrate whether the variability of the originaldata could be the result of a specific factor, Figs. 3 and 4 wereconstructed. A PCA can establish the variables that involve morevariance, and it is often represented by a loading plot. Neverthe-less, this variability in the original data could be due to aheterogeneity of samples, rather than differences explained by, forinstance, the potato variety. Furthermore, the scores of everysample in the new space of the components must be plotted in ascores figure. Since PCA is an unsupervised pattern recognitiontechnique and the samples in a score plot are labeled according totheir characteristics, a global distribution may be studied if thesamples can be grouped by any common property, such as potatovariety or frying oil. In this way, as can be observed in Fig. 3A, thevariables with the highest and the lowest values for loadings in PC1corresponded with FAMEs C22:2(n6), C12:0, C22:0 and C18:3(n6),C16:1(n7), C16:0, respectively. Regarding PC2, the more represen-tative variables were C22:1(n9), C21:0 and C20:2(n6). TheseFAMEs, with loadings close to the edge of a unit radius circle, arethe best defined when extracting the first two PCs. Once the mostimportant loadings were selected, an interpretation was made ofthe scores represented in Fig. 3B. As may be observed, there is noclear distribution of scores according to the variety of potatoes,whereas if the frying oil is taken into consideration moredifferences among the samples of crisps can be seen. This situationcould be explained by the fact that C16:0, as one of majority fryingoil components, presented different values for the two oil batchesstudied, despite both oils possessing similar characteristics;however, as was demonstrated, a small variation could provokescore differences for PC2 according to the oil type.

Similarly, these majority FAMEs in vegetable frying oils, withhigh loadings in PC3 and PC4, such as, C18:1(n9) and C18:2(n6), ledto higher scores for crisps from the Mustang variety than for thosefrom the Hermes variety (Fig. 4A). It has been also observed that forFAMEs with a content lower than 1% in the frying oil, such asC18:3(n3), the actual effect, despite their possible high loadings, oncrisps differentiation according to the variety of potatoes was notso great as in the case of majority FAMEs. Furthermore, a cleardifferentiation between both potato varieties could be observed

and associated with the most abundant FAMEs in the oil. Finally, itwas not possible to group the crisps according to the oil fryingbatch within a homogeneous lot, because if the crisps sampleswere labeled with their respective frying oil, grouped sampleswould not be observed (Fig. 4B).

4. Conclusions

In this study, the fatty acid profiles of potato crisp samples wereobtained from two different potato varieties (Hermes andMustang), produced with the same frying experiments by meansof a validated GC–MS method. Soxhlet extraction with diethylether, followed by a derivatization procedure with KOH and BF3 for5 min at 100 8C, provided the best results in terms of recoverypercentages and preparation time. According to the resultssummarized in this study, majority fatty acids in potato crispswere C18:0, C16:0, C18:2(n6) and C18:1(n9). The sum of all FAMEsled to an average of 35% of the gross weight of the sample, withsimilar data for crisp samples from both potato varieties. A studyhas also been made of the influence of the potato variety and thefrying oil on the fatty acid profiles. The results of two-way ANOVAstudies have demonstrated that the potato variety caused moresignificant differences than the frying oil batch in terms of fattyacid content. In fact, the interaction between both factors hasprovided more differences between Hermes and Mustang potatovarieties, especially for minority fatty acids. A comparison betweenMustang and Hermes potato crisps, by means of a PCA, hasestablished a relationship between the potato variety and fattyacid content, showing, for example, that the Hermes varietycontained lower total FAME concentrations and v-6/v-3 ratios inthe potato crisp samples analyzed.

Acknowledgments

The authors wish to thank Facundo S.A. (Villada, Palencia,Spain) for providing the samples. P.M. acknowledges the Junta deCastilla y Leon for her PhD grant.

References

Agrico UK Ltd., 2009. Potato Varieties Physico-chemical Characteristics. RetrievedMay 21, 2012, from: http://www.agrico.co.uk/products/varieties/

Aro, A., Amaral, E., Kesteloot, H., Rimestad, A., Thamm, M., Van Poppel, G., 1998.Trans fatty acids in French fries, soups, and snacks from 14 European

P. Manzano et al. / Journal of Food Composition and Analysis 28 (2012) 31–39 39

countries: the TRANSFAIR study. Journal of Food Composition and Analysis11, 170–177.

Bligh, E.G., Dyer, W.J., 1959. A rapid method of total lipid extraction and purification.Canadian Journal of biochemistry and physiology 37, 911–917.

Catchpole, O.J., Tallon, S.J., Eltringham, W.E., Grey, J.B., Fenton, K.A., Vagi, E.M.,Vyssotski, M.V., MacKenzie, A.N., Ryan, J., Zhu, Y., 2009. The extraction andfractionation of specialty lipids using near critical fluids. Journal of SupercriticalFluids 47, 591–597.

Fernandez-San Juan, P.M., 2000. Fatty acid composition of commercial Spanish fastfood and snack food. Journal of Food Composition and Analysis 13, 275–281.

Folch, J., Lees, M., Stanley, G.H.S., 1957. A simple method for the isolation andpurification of total lipids from animal tissues. Journal of Biological Chemistry226, 497–509.

Hauff, S., Vetter, W., 2009. Quantification of fatty acids as methyl esters andphospholipids in cheese samples after separation of triacylglycerides andphospholipids. Analytica Chimica Acta 636, 229–235.

Hejazi, L., Ebrahimi, D., Guilhaus, M., Hibbert, D.B., 2009. Determination of thecomposition of fatty acid mixtures using GC � FI-MS: a comprehensive two-dimensional separation approach. Analytical Chemistry 81, 1450–1458.

Herrero, M., Ibanez, E., Cifuentes, A., Bernal, J., 2009. Multidimensional chromatog-raphy in food analysis. Journal of Chromatography A 1216, 7110–7129.

ICH, 2005. ICH Topic Q2, Validation of Analytical Procedures: Text and Methodolo-gy. International Conference on Harmonization, Geneva.

Jimenez, J.J., Bernal, J.L., Nozal, M.J., Toribio, L., Bernal, J., 2009. Profile and relativeconcentrations of fatty acids in corn and soybean seeds from transgenic andisogenic crops. Journal of Chromatography A 1216, 7288–7295.

Kita, A., Lisinska, G., Gołubowska, G., 2007. The effects of oils and frying tempera-tures on the texture and fat content of potato crisp. Food Chemistry 102, 1–5.

Li, Y., Watkins, B.A., 2001. Analysis of fatty acids in food lipids. In: Current Protocolsin Food Analytical Chemistry, John Wiley & Sons, Inc., Hoboken, NJ, pp. D1.2.1–D1.2.15, Wiley Online Library.

Lo Scalzo, R., Bianchi, G., Genna, A., Summa, C., 2007. Antioxidant properties andlipidic profile as quality indexes of cauliflower (Brassica oleracea L. var. botrytis)in relation to harvest time. Food Chemistry 100, 1019–1025.

Manzano, P., Arnaiz, E., Diego, J.C., Toribio, L., Garcıa-Viguera, C., Bernal, J.L., Bernal,J., 2011. Comprehensive two-dimensional gas chromatography with capillaryflow modulation to separate FAME isomers. Journal of Chromatography A 1218,4952–4959.

Morrison, W.R., Smith, L.M., 1964. Preparation of fatty acid methyl esters anddimethylacetals from lipids with boron fluoride-methanol. Journal of LipidResearch 5, 600–608.

Ooko, G.A., Kabira, J.N., 2011. Suitability of three newly released Kenyan potatovarieties for processing into crisps and French fries. African Journal of Food,Agriculture, Nutrition and Development 11, 5266–5281.

Pedneault, K., Angersa, P., Gosselina, A., Tweddell, R.J., 2008. Fatty acid profiles ofpolar and neutral lipids of ten species of higher basidiomycetes indigenous toeastern Canada. Mycological Research 112, 1428–1434.

Sahena, F., Zaidul, I.S.M., Jinap, S., Karim, A.A., Abbas, K.A., Norulaini, N.A.N., Omar,A.K.M., 2009. Application of supercritical CO2 in lipid extraction—a review.Journal of Food Engineering 95, 240–253.

Sanches-Silva, A., Rodrıguez-Bernaldo de Quiros, A., Lopez-Hernandez, J., Paseiro-Losada, P., 2004. Comparison between high-performance liquid chromatogra-phy and gas chromatography methods for fatty acid identification and quanti-fication in potato crisps. Journal of Chromatography A 1032, 7–15.

Shahidi, F., 2001. Extraction and measurement of total lipids. In: Current Protocolsin Food Analytical Chemistry, John Wiley & Sons, Inc., Hoboken, NJ, USA, pp.D1.1.1–D1.1.11, Wiley Online Library.

Toribio, L., Bernal, J.L., Nozal, M.J., Arnaiz, E., Bernal, J., 2011. Sequential supercriticalfluid extraction of lipids. Application to the obtention of the fatty acid profileof some genetically modified varieties of corn. Food Analytical Methods 4,196–202.

Wagner, K.H., Plasser, E., Proell, C., Kanzler, S., 2008. Comprehensive studies on thetrans fatty acid content of Austrian foods: convenience products, fast food andfats. Food Chemistry 108, 1054–1060.

Zu, Y., Zhang, S., Fu, Y., Liu, W., Liu, Z., Luo, M., Efferth, T., 2009. Rapid microwave-assisted transesterification for the preparation of fatty acid methyl esters fromthe oil of yellow horn (Xanthoceras sorbifolia Bunge). European Food Researchand Technology 229, 43–49.