A

ofihp(hhaciac©

K

1

mw[d6od

b

0d

Journal of Chromatography A, 1129 (2006) 21–28

Analysis of eicosapentaenoic and docosahexaenoic acid geometricalisomers formed during fish oil deodorization

Veronique Fournier a, Pierre Juaneda a, Frederic Destaillats b, Fabiola Dionisi b,Pierre Lambelet b, Jean-Louis Sebedio c, Olivier Berdeaux a,∗

a UMR FLAVIC Department, INRA, 17 Rue Sully BP 86510, 21065 Dijon, Cedex, Franceb Nestle Research Center, Vers-chez-les-Blancs, Switzerland

c INRA, Clermont-Ferrand, France

Received 18 May 2006; received in revised form 18 June 2006; accepted 26 June 2006Available online 7 August 2006

bstract

Addition of long-chain polyunsaturated fatty acids (LC-PUFAs) from marine oil into food products implies preliminary refining procedures of theil which thermal process affects the integrity of LC-PUFAs. Deodorization, the major step involving high temperatures, is a common process usedor the refining of edible fats and oils. The present study evaluates the effect of deodorization temperature on the formation of LC-PUFA geometricalsomers. Chemically isomerized eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) were used as reference samples. Fish oil samplesave been deodorized at 180, 220 and 250 ◦C for 3 h and pure EPA and DHA fatty acid methyl esters (FAMEs) were chemically isomerized using-toluenesulfinic acid as catalyst. FAMEs prepared from fish oil were fractionated by reversed-phase high-performance liquid chromatographyRP-HPLC). Geometrical isomers produced by both processes were fractionated by silver-ion thin-layer chromatography (Ag-TLC) and silver-ionigh-performance liquid chromatography (Ag-HPLC). The FAME fractions were subsequently analyzed by gas chromatography (GC) on a 100 mighly polar cyanopropylpolysiloxane coated capillary column, CP-Sil 88. Our results show that thermally induced geometrical isomerizationppears to be a directed reaction and some ethylenic double bond positions on the hydrocarbon chain are more prone to stereomutation. Only minor

◦

hanges were observed in the EPA and DHA trans isomers content and distribution after deodorization at 180 C. The analyses of EPA and DHAsomer fractions revealed that it is possible to quantify EPA geometrical isomers by GC using the described conditions. However, we notice thatmono-trans isomer of DHA, formed during both chemical and thermal treatments, co-elute with all-cis DHA. This feature should be taken into

onsideration for the quantification of DHA geometrical isomers.2006 Elsevier B.V. All rights reserved.

oil; S

i[tftpL

eywords: Deodorization; Docosahexaenoic acid; Eicosapentaenoic acid; Fish

. Introduction

Unsaturated fatty acids have been recommended for replace-ent of saturated fatty acids in human diet since saturated fatsere demonstrated to be correlated with cardiovascular diseases

1]. This lead in an increase of unsaturated fatty acids in theiet. However this increase occurred mainly to the benefit of n-

fatty acids, those being more widely available in commodity

ils. The ratio n-6/n-3 at that moment increased in the humaniet and this imbalance could be the source of chronic health

∗ Corresponding author. Tel.: +33 3 80 69 35 40; fax: +33 3 80 69 32 23.E-mail addresses: frederic.destaillats@rdls.nestle.com (F. Destaillats),

erdeaux@dijon.inra.fr (O. Berdeaux).

tabsda[b

021-9673/$ – see front matter © 2006 Elsevier B.V. All rights reserved.oi:10.1016/j.chroma.2006.06.089

ilver-ion chromatography; Trans fatty acids

mpairment such as cardiovascular and auto-immune diseases2]. This poor availability can be compensated by the addi-ion of n-3 long-chain polyunsaturated fatty acids (LC-PUFAs)rom marine oil into food products. However, particular atten-ion have to be given at the quality of marine oil as the refiningrocess involves thermal treatment which affect the integrity ofC-PUFAs [3]. Deodorization, which implies exposure to high

emperatures (180–220 ◦C) under low pressure (1–10 mbar), iscommon step of the refining of edible fats and oils [4]. It haseen shown that the temperatures used for deodorization areufficient to induce geometrical isomerization of the natural cis

ouble bonds of polyunsaturated fatty acids (e.g. �-linoleniccid) to the more thermodynamically stable trans configuration5]. This fact is of great interest because trans fatty acids haveeen shown to be incorporated into body tissues and fluids of

2 mato

hm

(samfdli

afuFteabbm[e1t2

milmmlftsf

2

2

Ki3dtmF2(t21E

(1Dwi

2

uA2erown0

2

(8uFt3ta1w(

2m

c[artwt3

2t

sA

2 V. Fournier et al. / J. Chro

umans, namely brain, liver, adipose tissue, spleen, plasma andilk [6].Formation of trans isomers during food manufacturing

refining, hydrogenation) or simply during frying have beencrutinized [4,5,7–20]. In consequence, formation of mono-, di-nd triunsaturated fatty acid geometrical isomers are well docu-ented. The proneness of a methylene-interrupted unsaturated

atty acid to geometrical isomerization upon heat treatmentepends on its number of double bonds. Adlof in 1994 has ana-yzed chemically isomerized arachidonic acid [21], but few stud-es were dedicated to LC-PUFAs geometrical isomers [3,22–23].

Results from a study by Loı et al. [24] suggest that EPA withtrans double bond in the �11 position might be responsible

or the different physiological effects of the trans polyunsat-rated fatty acids as compared to their cis homologue (EPA).urthermore, it was demonstrated that a trans double bond at

he terminal position of a n-3 fatty acid is not recognized bynzymes involved in lipid metabolism and that such compoundsre rather recognized as their n-6 counterpart [24,25]. In fact,ecause of their spatial orientation, trans double bonds mighte recognized as a single bond by enzymes involved in lipidetabolism. Effectively, cis-9,cis-12,trans-15 18:3 acid isomer

25], is incorporated (acyltransferase) into rat cardiolipins at thexpense of linoleic (cis-9,cis-12 18:2) acid while cis-5,cis-8,cis-1,cis-14,trans-17 20:5 acid isomer seems to be recognized byhe enzymatic system as arachidonic (cis-5,cis-8,cis-11,cis-140:4) acid [24].

Number of countries have or are in the way to legislate on theaximum trans fatty acid concentration allowed in oils and fats

ntended for human consumption. In order to be applied, theseaws and recommendations have to rely on consistent analytical

ethods. Methodologies for quantification of trans isomers ofono-, di- and tri-unsaturated fatty acids were developed in the

ast decade [4,5,7–20]. However, such work has not been doneor trans isomers of LC-PUFAs potentially formed during heatreatment of fish oils. The methodology described in the presenttudy aims at analyzing geometrical isomers of EPA and DHAormed during fish oil deodorization, a crucial step of refining.

. Material and methods

.1. Samples and reagents

Deodorization of semi-refined fish oil (Nippon Suisanaisha, Tokyo, Japan) was performed with a laboratory deodor-

zer. Briefly, fish oil was heated at either 180, 220 or 250 ◦C forh under a pressure of 1.5 mbar and with 2%/h (based on oil)irect steam injection. The control sample was not submittedo any deodorization. Hexane, toluene, dichloromethane,

ethanol, acetonitrile were purchased from SDS (Peypin,rance). Standards of fatty acid methyl esters and 2-amino--methyl-1-propanol were purchased from Sigma–AldrichSaint-Quentin Fallavier, France). Pure standards were used

o identify retention times of cis-5,cis-8,trans-11,cis-14,cis-170:5 acid isomer (shortened as trans-11 EPA), cis-5,cis-8,cis-1,cis-14,trans-17 20:5 acid isomer (shortened as trans-17PA), cis-5,cis-8,trans-11,cis-14,trans-17 20:5 acid isomer

MiPt

gr. A 1129 (2006) 21–28

shortened as trans-11,trans-17 EPA) and cis-4,cis-7,cis-10,cis-3,cis-16,trans-19 22:6 acid isomer (shortened as trans-19HA). The isomer of DHA and the three isomers of EPAere synthesised earlier by Vatele et al. [26,27] for biological

nvestigations purpose.

.2. Fatty acid methyl esters (FAMEs) preparation

Prior to GC analysis, the acylglycerols were transesterifiedsing basic catalyst (0.5 M sodium methanolate in methanol).bout 20 mg of fish oil were diluted in 1 mL of toluene, andmL of sodium methanolate solution was added and the trans-sterification reaction was conducted at 50 ◦C for 5 min. Theeaction was stopped by adding 0.1 mL of acetic acid and 5 mLf distilled water. The organic phase was extracted successivelyith 5 and 3 mL of hexane. The solvents were evaporated underitrogen and the FAMEs diluted in hexane to a concentration of.1 mg/mL.

.3. Gas chromatography (GC) analysis

FAMEs were analyzed on a 5890 gas chromatographHewlett-Packard, Palo Alto, CA, USA) equipped with a CP-Sil8 highly polar cyanopropylpolysiloxane coated capillary col-mn (100 m × 0.25 mm I.D., 0.2 �m film, Varian, Courtaboeuf,rance). The instrument was equipped with a split/splitless injec-

ion port (splitless for 0.5 min). Linear velocity of hydrogen was7.0 cm/s at 60 ◦C. The temperature was held at 60 ◦C for 5 min,emperature programmed to 165 at 15 ◦C/min and held for 1 minnd then to 225 ◦C at 2 ◦C/min and finally held at 225 ◦C for7 min [28]. The injection port was held at 250 ◦C and detectionas achieved using flame ionization detection (FID) at 250 ◦C

hydrogen at 35 mL/min and air at 350 mL/min).

.4. Chemical isomerization of EPA and DHA fatty acidethyl esters

p-Toluenesulfinic acid was prepared by acidification of theorresponding sodium salt according to Snyder and Schofield29]. Geometrical isomerization of about 20 mg of all-cis EPAnd all-cis DHA FAMEs in 2.5 mL dioxane was performed undereflux with 2.5 mg of p-toluenesulfinic acid for 5 (mild condi-ions) or 10 min (strong conditions). The acid was neutralizedith 7.5 mL, 1 M NaOH then the mixture was extracted three

imes with 7.5 mL hexane. The organic phase was washed with× 7.5 mL of NaCl saturated water.

.5. Separation of geometrical isomers by argentationhin-layer chromatography (Ag-TLC)

FAMEs fractions containing isomers of EPA and DHA wereeparated according to their number of trans double bonds byg-TLC [30]. Briefly, TLC plates (silica gel, 20 cm × 20 cm,

erck, Darmstadt, Germany) were impregnated by immersion

n a silver nitrate solution (10%, w/v in acetonitrile) for 30 min.lates were eluted with toluene:methanol (85:15, v/v). Frac-

ions containing mono-, di-, tri-, tetra-, penta-trans for EPA

matogr. A 1129 (2006) 21–28 23

aam2

2h

wbamCdUpdtgp

2r(

rQwcrRtcft

3

mcoiomwFiiiacDas

1ttg

3m

t[tsmoalratts

V. Fournier et al. / J. Chro

nd DHA and hexa-trans for DHA were scraped off the platend the FAMEs recovered by adding 5 mL of a 1% NaCl inethanol:water 90:10 (v/v) solution, then extracted twice withmL of hexane.

.6. Separation of geometrical isomers by argentationigh-performance liquid chromatography (Ag-HPLC)

FAMEs fractions of EPA and DHA geometrical isomersere separated according to their number of trans doubleonds by Ag-HPLC. Briefly, the isocratic mode was used withbinary mixture hexane:acetonitrile (99.7:0.3,v/v). The chro-atographic system was constituted by a Varian 9010 pump, ahromspher 5 lipids column (250 mm × 4.6 mm, Varian, Mid-leburgh, The Netherlands), a UV–vis Varian 9050 (Varian, Leslis, France) detector and a Linseis L250E recorder. FAME sam-les (0.08 mg) were separated into fractions containing mono-,i-, tri-, tetra-, penta-trans EPA and DHA isomers and hexa-rans for DHA. The solvent evaporated under a stream of nitro-en and the FAME fractions were diluted in 0.1 mL of hexanerior to GC analysis.

.7. Separation of EPA and DHA FAMEs fraction byeversed-phase high-performance liquid chromatographyRP-HPLC)

FAMEs prepared from deodorized oil samples were sepa-ated using a Kromasil-C18 25 cm × 10 mm I.D., 5 �m (Thermouest, Courtaboeuf, France) column. FAMEs (about 20 mg)ere dissolved in 100 �l of acetone and analyzed under iso-

ratic mode using distilled and filtered acetonitrile at a flowate of 4 mL/min as described by Juaneda and Sebedio [31]. InP-HPLC, the FAMEs with the same carbon equivalent elute

ogether following this rule: carbon equivalent = (number ofarbons)−2 × (number of double bonds). EPA and DHA FAMEraction (EC = 10) was collected to recover enough material (7imes at around 3 mg per run) for further separation by Ag-TLC.

. Results and discussion

Fish oils contain an important number of different saturated,ono- and polyunsaturated fatty acids. Heat treated fish oils

ould contain different degradation products due to the presencef thermally sensitive LC-PUFAs [3]. To avoid possible analyt-cal interferences between LC-PUFAs geometrical isomers andther degradation products (e.g. cyclic fatty acid monomers),ethods for the separation of LC-PUFAs geometrical isomersere developed using chemically isomerized EPA and DHAAMEs as references. The used analytical workflow is providedn Fig. 1. Briefly, samples containing all possible geometricalsomers of EPA and DHA were prepared by chemical isomer-zation of EPA and DHA FAMEs. Chemically isomerized EPAnd DHA were fractionated by Ag-TLC and Ag-HPLC. Optimal

onditions allowing analysis of chemically produced EPA andHA geometrical isomers were then applied to separate EPA

nd DHA FAMEs prepared from different deodorized fish oilample. Deodorization temperatures selected for our study were

wmna

Fig. 1. Analytical workflow used in the present study.

80, 220 and 250 ◦C. Control oil was not submitted to any heatreatment. It was previously established that 220 ◦C seems to behe critical temperature at which LC-PUFAs undergo excessiveeometrical isomerization [3,22].

.1. Analysis of chemically isomerized EPA and DHAethyl esters

Pure EPA and DHA FAMEs were isomerized using p-oluenesulfinic acid as catalyst according to literature procedure29]. Respectively 32 (25) and 64 (26) geometrical isomers canheoretically be produced from all-cis EPA and DHA. We havehown in a previous study [3] that mainly mono- and di-trans iso-ers of EPA and DHA are formed during deodorization of fish

ils. Therefore, the reaction conditions with p-toluenesulfiniccid was optimized to obtain principally mono- and di-trans andow amount of all-trans isomer (see Fig. 2A and C). A longereaction time favored the formation of highly isomerized EPAnd DHA FAMEs with low amount of residual all-cis and mono-rans isomers (see Fig. 2B and D). It can be noticed in Fig. 2hat a single GC analysis is not sufficient to achieve baselineeparation of individual geometrical isomers.

Pre-separation of these two mixtures of geometrical isomers

as achieved by Ag-TLC and Ag-HPLC. Mixtures of trans iso-ers of EPA or DHA FAMEs were separated according to their

umber of trans double bonds by Ag-TLC (Fig. 3). All-cis EPAnd all-cis DHA FAMEs were the most strongly retained isomers

24 V. Fournier et al. / J. Chromato

Fmc

watoci(tBsuar

gmobpiitruio1mtptotFTvc

FDrAsTtc1asamiptL

rbiaomssDh

ig. 2. Gas chromatographic analysis of EPA (A, B) and DHA (C, D) fatty acidethyl esters submitted to mild (A, C) and strong (B, D) chemical isomerization

onditions (see Section 2 for experimental procedure).

ith RF corresponding to 0.22 and 0.07, respectively (bands A1nd B1). As DHA has one additional double bond comparedo EPA, it interacts therefore more strongly with the silver ionsn the stationary phase (see Fig. 3). Second bands (A2 and B2)ontained mono-trans isomers of EPA (RF = 0.4) or mono-transsomers of DHA (RF = 0.2), third bands (A3 and B3) di-transRF = 0.6, RF = 0.4), then tri-trans (A4 RF = 0.7, B4 RF = 0.6),etra-trans (A5 RF = 0.8, B5 RF = 0.7) penta-trans (A6 RF = 0.9,6 RF = 0.8) and hexa-trans (B7 = 0.9). EPA and DHA kept the

ame pattern of elution as obtained previously for the less unsat-rated fatty acids [30]. This pattern of separation was recentlylso obtained and visualized using a two-dimensional fatty acidetention indices map [23].

fFwo

gr. A 1129 (2006) 21–28

GC analysis of mono-trans of EPA FAME isomers (Fig. 4A2)ave four peaks that contain the five possible configurations ofono-trans geometrical isomers. Similar results were recently

btained by Mjos [23] on similar cyanopropyl phases and earliery Wijesundera [22] on Supelcowax-10. We showed by using aure standard, that trans-17 EPA is the less retained mono-transsomer. The area of the peak corresponding to the trans-11 EPAsomer (identified using a synthesized standard) was roughlywo-fold higher than the other peaks, which indicates occur-ence of a co-elution with another mono-trans isomer under thesed GC conditions. Likewise, eight out of ten possible di-transsomers only are resolved; peaks 1 and 2 appear to correspond toverlapping peaks of two isomers each (see Fig. 4A3). The trans-1, trans-17 EPA isomer was identified as well using a syntheticolecule. Resolution of the tri-trans FAME fraction that poten-

ially contains 10 isomers was less efficient since only seveneaks were separated (see Fig. 4A4). The tetra-trans FAME frac-ion was also less resolved than the mono-trans fraction showingnly three peaks out of five. Peak 3 (Fig. 4A5) contains probablyhree tetra-trans isomers. The all-trans isomers of EPA and DHAAMEs seem to be the first geometrical isomers eluted in GC.his feature was also shown by Mjos [23] and it had been pre-iously shown for C18 polyunsaturated fatty acids using polarolumn Silar 10C [32].

Analogies between the separation patterns of EPA and DHAAME geometrical isomers could be noticed. Indeed, trans-19HA, identified using a pure standard (Fig. 4B2) had the lower

etention time as previously shown for trans-17 EPA (Fig. 4A2).dditionally, the mono-trans geometrical isomers of DHA are

eparated in five peaks out of six possible isomers (Fig. 4B2).he area of peak 4 is roughly two times more important than

he area of other peaks and it is quite probable that this peakorresponds to the co-elution of the cis-4,cis-7,cis-10,trans-3,cis-16,cis-19 22:6 (shortened as trans-13 DHA) acid withnother mono-trans DHA isomer. Di-trans DHA isomers areeparated in 11 peaks out of 15, and three peaks are twice as bigs the others what appears to indicate the co-elution of two iso-ers (see Fig. 4B3). The tri-trans DHA isomers chromatogram

s much more complex and 20 isomers are separated in eleveneaks. Further investigations have to be done in order to iden-ify if a more suitable stationary phase useful for separation ofC-PUFAs geometrical isomers is commercially available.

Ag-HPLC fractionation was used both to separate geomet-ical isomers of EPA and DHA and to confirm results obtainedy Ag-TLC. Separation of EPA and DHA FAME fractions hav-ng the same number of trans ethylenic double bonds could bechieved by Ag-HPLC [23]. Ag-HPLC was optimized and webserved that the separation of EPA FAMEs geometrical iso-ers obtained by chemical isomerization of all-cis EPA gives

ix peaks, and DHA FAMEs analysis allowed the resolution ofeven peaks (see Fig. 5). Mono-, di-, tri- and tetra-trans EPA andHA FAME fractions were partially resolved (see Fig. 5) andave been collected separately (a and b). The collected FAMEs

ractions were further analyzed by GC and results are provided inig. 6. Interestingly, trans-17 EPA and trans-19 DHA (identifiedith synthetic molecules), are both isolated from their groupsf mono-trans in Ag-HPLC (Fig. 5A: mono-trans a corresponds

V. Fournier et al. / J. Chromatogr. A 1129 (2006) 21–28 25

EPA a

t1Wc

Fa

Fig. 3. Ag-TLC separation of chemically isomerized (A)

o trans-17 EPA; Fig. 5B: mono-trans a corresponds to trans-9 DHA) and in gas chromatography (see Fig. 6A and 6B).olff [10] found similar results for cis-9,trans-12 18:2 and

is-9,cis-12,trans-15 18:3 on the same GC column. Similarly,

WEc1

ig. 4. Partial gas chromatograms of EPA (A2 to A6) and DHA (B2 to B7) geometricll-cis EPA and DHA fatty acid methyl esters (see Section 2 for experimental proced

nd (B) DHA (see Section 2 for experimental procedure).

ijesundera [22] found that trans-17 EPA eluted before all-cisPA on Supelcowax-10 (100% polyethylene glycol) capillaryolumn. Marty et al. [33] have shown that the cis-4,cis-7,cis-0,trans-13 22:4 acid isomer elutes separately from other isomer

al isomer fractions obtained by Ag-TLC separations of chemically isomerizedure).

26 V. Fournier et al. / J. Chromatogr. A 1129 (2006) 21–28

(B) D

cc[ci

p

Dbi

Fa

Fig. 5. Ag-HPLC chromatograms of chemically isomerized (A) EPA and

lasses on Ag-HPLC. Separation of �-linolenic acid geometri-al isomers by Ag-HPLC was previously investigated by Adlof21] and it was noticed the occurrence of an isolated peak, that

ould probably correspond to the cis-9,cis-12,trans-15 18:3 acidsomer.

Ag-TLC is a chromatographic method that combines sim-licity, low cost and efficiency for the separation of EPA and

acce

ig. 6. Partial gas chromatograms of EPA (A) and DHA (B) geometrical isomer fracnd DHA fatty acid methyl esters (see Section 2 for experimental procedure).

HA fatty acid methyl esters (see Section 2 for experimental procedure).

HA geometrical isomers according to their number of doubleonds in the trans configuration. Ag-HPLC analysis, resultedn baseline resolution and this absence of overlapping allowed

confirmation of the results obtained by Ag-TLC. Geometri-al isomers with the same number of double bonds in the transonfiguration showed a partial resolution in Ag-HPLC. How-ver, Ag-HPLC important drawbacks are the life of the column

tions obtained by Ag-HPLC separations of chemically isomerized all-cis EPA

mato

aa

3D

alutaofRTfdsi2td(scTmdtistli

iiwtad

maaibiFDcc[tt

smAdedfitau

Fe

V. Fournier et al. / J. Chro

nd the difficulty to obtain reproducible retention times betweennalyses.

.2. Analysis of geometrical isomerization of EPA andHA during deodorization of fish oil

Three fish oil samples deodorized at 180, 220 and 250 ◦Cnd one non-deodorized sample, named control oil, were ana-yzed in the present study. The four samples were transesterifiednder basic conditions and submitted to RP-HPLC fractiona-ion. The first eluting fraction (EC = 10), which contained EPAnd DHA FAME geometrical isomers was collected. Tracesf octadecatetraenoic acid were also present in this RP-HPLCraction but did not interfere during the GC analysis. The fourP-HPLC fractions of fish oil samples were submitted to Ag-LC fractionation. No geometrical isomer of EPA and DHA was

ound by Ag-TLC analysis in the control oil and in the fish oileodorized at 180 ◦C. This is in agreement with previous resultshowing that deodorization at 180 ◦C did not induce fatty acidsomerization in fish oil [3]. For samples deodorized at 220 and50 ◦C, four bands were scraped-off the TLC plates and fur-her analyzed by GC. Partial GC traces of the fish oil sampleeodorized at 220 ◦C are provided in Fig. 7. The first fractionRF = 0.1) was composed mainly of all-cis DHA (Fig. 7A). Theecond fraction (RF = 0.2) was a mixture of all-cis EPA, all-is DPA and the six mono-trans isomers of DHA (Fig. 7B).he third fraction (RF = 0.4) contained the five mono-trans iso-ers of EPA and the di-trans of DHA (Fig. 7C). For the sample

eodorized at 220 ◦C, it was possible to detect by GC analysisrace amount of di-trans EPA and tri-trans DHA geometricalsomers in the fourth fraction (RF = 0.6, see Fig. 7D). GC–MS

pectra of 4,4-dimethyloxazoline derivatives of peak identifiedo be geometrical isomers of EPA and DHA exhibited simi-ar fragmentation patterns as pure all-cis EPA and all-cis DHAsomers. Moreover, comparing the retention times of chemically

p

ta

ig. 7. Partial gas chromatograms of EPA and DHA geometrical isomer fractionsxperimental procedure).

gr. A 1129 (2006) 21–28 27

somerized fatty acids, which are known to contain no positionalsomers [29], revealed that peaks found in deodorized fish oilsere geometrical isomers of EPA and DHA. Besides, we note

hat the tri-trans isomers of DHA and di-trans isomers of EPAre formed only when the fish oil was submitted to relativelyrastic conditions, i.e. 250 ◦C.

This result is in accordance with a previous study on the ther-al isomerization of �-linolenic acid showing that the trienoic

cid with two trans double bonds is only formed when the fattycid is submitted to a 10 h heating at 240 ◦C [7]. Thermal isomer-zation during deodorization at 240 ◦C and 220–230 ◦C seem toe the critical points above which dienes and trienes isomer-zation, respectively, increases very strongly [12]. As shown inig. 7B, a mono-trans isomer of DHA co-elute with the all-cisHA isomer. Different GC conditions were tried but we con-

lude as previously mentioned by Mjos [23] that this co-elutionould not be avoided by varying temperature program. Mjos23] observed the same co-elution on similar cyanopropyl sta-ionary phase. This co-elution has to be taken into account forhe quantification of geometrical isomers of DHA.

The analysis of the geometrical isomers of EPA in the fish oilample deodorized at 220 ◦C revealed that the main products areono-trans isomers (possibly trans-11 + trans-14, see Fig. 7C).ccordingly, one may consider that thermal isomerization isirected (trans configuration of a specific double bond beingnergetically favored). Wijesundera [22] obtained similar resultsuring thermal isomerization of EPA and he tentatively identi-ed, using calculation of theoretical equivalent chain length,

his peak as cis-5, cis-8, cis-11, trans-14, cis-17 20:5 (shorteneds trans-14 EPA). Additionally, we have previously shown bysing a standard that the trans-11 EPA co-elutes with another

eak that could be the trans-14 EPA (Fig. 7C).

The distribution of mono-trans isomers formed by thermalreatment of DHA show two important peaks, which elute afterll-cis DHA. By analogy with observations performed on mono-

from fish oil deodorized at 220 ◦C obtained by Ag-TLC (see Section 2 for

2 mato

tm1(sbos7t

owtDn�otci

mtlottobiDDadt

4

dtnWdasaifEEtipc

fwii

A

a

R

[[

[

[

[

[

[[

[[

[

[[

[[

[

[

[

[[[

8 V. Fournier et al. / J. Chro

rans EPA isomers, we can hypothesize that the most importantono-trans DHA isomer peak represents the co-elution of trans-

3 and cis-4,cis-7,cis-10,cis-13,trans-16,cis-19 22:6 acid isomershortened as trans-16 DHA, see Fig. 7B). It could be hypothe-ized that the most predominant mono-trans isomers could haveeen formed by stereomutation of the two central double bondsf the molecule (i.e. �10 and �13 double bonds). Therefore, theecond most important peak could correspond to the cis-4,cis-,trans-10,cis-13,cis-16,cis-19 22:6 acid isomer (shortened asrans-10 DHA, see Fig. 7B).

However, opposite observations have been obtained earliern �-linolenic acid thermally induced geometrical isomers [7],here the central double bond seemed to be not affected by

hermal treatment. All-cis PUFAs such as �-linolenic, EPA andHA adopt a curved spatial conformation due to the importantumber of double bonds in the cis configuration. Compared to-linolenic acid, all-cis EPA and DHA show a significant bendf their olefinic chains and this feature could have a substan-ial impact on the cis/trans equilibrium occurring in thermalonditions. These hypotheses have to be confirmed by formaldentification of each mono-trans isomer.

The possibility to quantify EPA and DHA geometrical iso-ers in fish oils by Ag-HPLC was evaluated. The control oil and

he samples deodorized at 180, 220 and 250 ◦C were methy-ated and then analyzed by Ag-HPLC (data not shown). Webserved that retention times varied widely from an injectiono another, compromising any automation of fraction collec-ion. Oil deodorized at 180 ◦C showed only a small appearancef mono-trans from DHA. Deodorization at 220 ◦C, as showedefore, is harmful to EPA and DHA. At this temperature, anmportant peak was observed, corresponding to mono-transHA isomers. Results showed interference between di-transHA isomers, which appear at a temperature of 220 ◦C, and

ll-cis EPA isomer. Therefore, this technique cannot be used toetermine isomers in fish oils submitted to heat treatment higherhan 180 ◦C.

. Conclusion

All possible geometrical isomers of EPA and DHA were pro-uced by reaction with p-toluenesulfinic acid. We showed thathese isomers could be readily fractionated according to theirumber of trans double bond by either Ag-HPLC or Ag-TLC.e also showed that few of these are formed in heated fish oil at

eodorization temperatures ranging from 180 to 250 ◦C. In fact,t 250 ◦C, which corresponds to drastic condition giving rise toignificant decreases of all-cis EPA and DHA [3], only mono-nd di-trans EPA isomers and mono-, di- and tri-trans DHAsomers are formed. trans-11 and trans-14 EPA co-elute and areormed in a more important concentration than other mono-transPA isomers. We hypothesize that the central double bonds inPA (�11) are preferentially isomerized, and, by analogy, that

he two central double bonds of DHA (�10 and �13) DHAsomers are also prone to stereomutation. We show that it isossible to quantify EPA isomers by GC using the describedonditions. However, we noticed that a mono-trans isomer

[[

[

gr. A 1129 (2006) 21–28

ormed during both chemical and thermal treatments co-eluteith the all-cis DHA isomer and that this feature should be taken

nto consideration for the quantification of DHA geometricalsomers.

cknowledgement

The authors kindly acknowledge Mr. E. Semon for GC–MSnalysis (Analytical platform, PPM, UMR-FLAVIC, Dijon).

eferences

[1] U. Keil, Basic Res. Cardiol. 95 (Suppl. 1) (2000) I52.[2] B. Delplanque, Ol. Corps Gras Lipids 11 (2004) 98.[3] V. Fournier, F. Destaillats, P. Juaneda, F. Dionisi, P. Lambelet, J.L. Sebedio,

O. Berdeaux, Eur. J. Lipid Sci. Technol. 108 (2006) 33.[4] G. Henon, Zs. Kemeny, K. Recseg, F. Swobada, K. Kovari, J. Am. Oil

Chem. Soc. 74 (1997) 1615.[5] R.G. Ackman, S.N. Hooper, D.L. Hooper, J. Am. Oil Chem. Soc. 51 (1974)

42.[6] E. Larque, S. Zamora, A. Gil, Early Hum. Dev. 65 (2001) S31.[7] A. Grandgirard, J.L. Sebedio, J. Fleury, J. Am. Oil Chem. Soc. 61 (1984)

1563.[8] A. Grandgirard, F. Julliard, Rev. Fr. Corps Gras 34 (1987) 213.[9] R.L. Wolff, Sci. Ailment 13 (1993) 155.10] R.L. Wolff, J. Am. Oil Chem. Soc. 70 (1993) 425.11] P. Juaneda, J.L. Sebedio, W.W. Christie, J. High Resolut. Chromatogr. 17

(1994) 321.12] G. Henon, Zs. Kemeny, K. Recseg, F. Swobada, K. Kovari, J. Am. Oil

Chem. Soc. 76 (1999) 73.13] G. Henon, P.Y. Vigneron, B. Stoclin, J. Caigniez, Eur. J. Lipid Sci. Technol.

103 (2001) 467.14] R.L. Wolff, in: S.G. Pandalai (Ed.), Recent Research and Developments in

Oil Chemistry, Transworld Research Network, Trivandrum, India, 1997, p.127.

15] K. Eulitz, M.P. Yurawecz, N. Sehat, J. Fritsche, J.A.G. Roach, M.M.Mossoba, J.K.G. Kramer, R.O. Adlof, Y. Ku, Lipids 34 (1999) 873.

16] D. Precht, J. Molkentin, M. Vahlendieck, Nahrung 43 (1999) 25.17] J. Cmolik, W. Schwarz, Z. Svoboda, J. Pokorny, Z. Reblova, M. Dolezal,

H. Valentova, Eur. J. Lipid Sci. Technol. 102 (2000) 15.18] J. Cmolik, J. Pokorny, Eur. J. Lipid Sci. Technol. 102 (2000) 472.19] Zs. Kemeny, K. Recseg, G. Henon, K. Kovari, F. Swobada, J. Am. Oil

Chem. Soc. 78 (2001) 973.20] M. Leon-Camacho, M.V. Ruiz-Mendez, M. Graciani-Constante, E.

Graciani-Constante, Eur. J. Lipid Sci. Technol. 103 (2001) 85.21] R.O. Adlof, J. Chromatogr. A 659 (1994) 95.22] R.C. Wijesundera, W.M.N. Ratnayake, R.G. Ackman, J. Am. Oil Chem.

Soc. 66 (1989) 1822.23] S.A. Mjos, J. Chromatogr. A 1100 (2005) 185.24] C. Loı, J.M. Chardigny, O. Berdeaux, J.M. Vatele, D. Poullain, J.P. Noel,

J.L. Sebedio, Thromb. Haemostatis 80 (1998) 656.25] R.L. Wolff, N.A. Combe, B. Entressangles, J.L. Sebedio, A. Grandgirard,

Biochim. Biophys. Acta 1168 (1993) 285.26] J.M. Vatele, H. Dong Doan, J.M. Chardigny, J.L. Sebedio, A. Grandgirard,

Chem. Phys. Lipids 74 (1994) 185.27] J.M. Vatele, H. Dong Doan, B. Fenet, J.M. Chardigny, J.L. Sebedio, Chem.

Phys. Lipids 78 (1995) 65.28] F. Dionisi, P. Golay, L. Fay, Anal. Chim. Acta 465 (2002) 395.29] J.M. Snyder, C.R. Scholfield, J. Am. Oil Chem. Soc. 59 (1982) 469.30] J.L. Morris, J. Lipid Res. 7 (1966) 717.

31] P. Juaneda, J.-L. Sebedio, J. Chromatogr. B 724 (1999) 213.32] A. Grandgirard, F. Julliard, J. Prevost, J.L. Sebedio, J. Am. Oil Chem. Soc.

64 (1987) 1434.33] Y. Marty, P. Soudant, S. Perrotte, J. Moal, J. Dussauze, J.F. Samain, J.

Chromatogr. A 839 (1999) 119.