HAL Id: hal-00895411https://hal.archives-ouvertes.fr/hal-00895411

Submitted on 1 Jan 2000

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.

Interactions between β-lactoglobulin and flavourcompounds of different chemical classes. Impact of theprotein on the odour perception of vanillin and eugenol

Jutta Reiners, Sophie Nicklaus, Elisabeth Guichard

To cite this version:Jutta Reiners, Sophie Nicklaus, Elisabeth Guichard. Interactions between β-lactoglobulin and flavourcompounds of different chemical classes. Impact of the protein on the odour perception of vanillin andeugenol. Le Lait, INRA Editions, 2000, 80 (3), pp.347-360. �10.1051/lait:2000130�. �hal-00895411�

Original article

Interactions between β-lactoglobulinand flavour compounds of different chemical classes.

Impact of the protein on the odour perceptionof vanillin and eugenol

Jutta REINERS, Sophie NICKLAUS, Elisabeth GUICHARD*

Laboratoire de Recherches sur les Arômes, INRA, 17 rue Sully, 21034 Dijon Cedex, France

(Received 4 January 1999; accepted 25 November 1999)

Abstract — Interactions between β-lactoglobulin and flavour compounds were analysed by instru-mental and sensory studies. Affinity chromatography was used to determine binding constants offlavour substances belonging to different chemical classes. A systematic study on esters, pyrazinesand phenolic compounds revealed that the increase in the hydrophobic chain length increases the affin-ity for the protein. Concerning the 4 diastereoisomers of 3-oxo-p-menthane-8-thiol, only the cis-transisomery led to different binding constants. In general, binding constants increased with increas-ing log P values, except for the terpenic compounds studied, for which another explanation has to befound, taking into account the geometry of the molecule. Sensory analyses applying a matching testshowed that the addition of β-lactoglobulin had no effect on the odour perception of vanillin, butbrought about a significant decrease in the odour perception of eugenol.

β-lactoglobulin / flavour / interaction / affinity chromatography / sensory analysis

Résumé — Interactions entre β-lactoglobuline et composés d’arôme de différentes classeschimiques. Impact de la protéine sur la perception de la vanilline et de l’eugénol. Les interactionsentre β-lactoglobuline et composés d’arôme ont été étudiées par des méthodes instrumentales etsensorielles. La chromatographie d’affinité a été utilisée pour déterminer les constantes d’affinité demolécules d’arômes de différentes classes chimiques. Une étude systématique sur des esters, pyra-zines et composés phénoliques a montré que l’allongement de la chaîne hydrophobe augmente laconstante d’affinité. Concernant les 4 diastéréoisomères du 3-oxo-p-menthane-8-thiol, seule l’isomérie

Lait 80 (2000) 347–360 347© INRA, EDP Sciences

* Correspondence and reprintsTel.: (33) 3 80 69 32 77; fax: (33) 3 80 69 32 27; e-mail: guichard@arome.dijon.inra.frThese results form part of the European COST Action 96 “Interactions of food matrix with small lig-ands” and the DGAL programme “Physico-chemical interactions between proteins and aroma inaqueous or emulsified media”, partly financed by the French Ministry of Agriculture and Fisheries.

J. Reiners et al.348

1. INTRODUCTION

Factors influencing the release of flavourcompounds during eating have beenreviewed by several authors [3, 17, 20].They outlined the importance of proteinson flavour release due to the binding offlavour compounds. Reactions betweenflavour compounds and proteins have beensummarized by Fischer and Widder [6]emphasizing the possibility of reversibleand irreversible interactions.

β-Lactoglobulin (BLG) is one of the bestcharacterized milk proteins. Its structurewas determined by Papiz et al. [18] by stan-dard crystallographic techniques, showing asimilarity to the retinol binding protein. Apossible binding site for retinol was identi-fied by model-building. Monaco et al. [12],who analyzed the structure of BLG usingX-ray diffraction methods, located the bind-ing site for retinol in a hydrophobic pocketon the surface of BLG. More recently, Wuet al. [27] demonstrated, after co-crystal-lization of the complex, that palmitate wasbound in the central cavity of the protein.Dufour et al. [4] proposed the presence oftwo binding sites based on their competi-tion studies of retinol and protoporphyrinIX, which was confirmed by the results ofNarayan and Berliner [13] on the simulta-neous binding of retinol and fatty acids.

Sostmann and Guichard [25] showed thataffinity chromatography allowed a rapidscreening of flavour compounds (methylketones, alcohols and aldehydes) interact-ing with BLG. The calculated global affin-ity obtained for methyl ketones was in agree-

ment with the values found by O’Neil andKinsella [15]. Using the same methodol-ogy, Pelletier et al. [19] found no interac-tions between BLG and short-chain fattyacids or methyl pyrazine and hydrophobicinteractions with esters from homologousseries. The global affinity calculated wasslightly lower than the value obtained byother methods on the same batch of protein,suggesting some hindrance of binding sitesdue to the immobilization of the protein.

No work has been undertaken until nowon key flavour compounds such as ethyl-3-methyl butyrate and furaneol in strawberryaroma [23], or isopropyl-methoxypyrazinesand vanillin in chocolate [24]. Therefore,these flavour-producing compounds wereanalyzed in the present study, together withrelated molecules in order to assess the influ-ence of structural differences of substanceson the binding to BLG.

Moreover, no previous study has inves-tigated the influence of native BLG onflavour perception. BLG is used as an emul-sifier or fat replacer by the food industry.For example, in ice creams flavoured withvanillin, Ohmes et al. [16] did not find anychange in flavour perception when milk fatwas replaced by whey-based fat replacers.Hansen and Heinis [8] found that the addi-tion of a whey protein concentrate to asucrose solution flavoured with vanillindecreased its flavour perception. In thisstudy, our aim was to assess the real impactof pure BLG on the flavour perception ofvanillin and of a structurally similar com-pound, eugenol.

cis-transconduit à des constantes différentes. De façon générale, l’affinité augmente avec le log P saufpour les composés terpéniques étudiés, pour lesquels une autre explication doit être trouvée, tenantcompte de la géométrie de la molécule. Un test d’appariement en analyse sensorielle a montré quel’ajout de β-lactoglobuline n’a pas d’effet sur la perception de l’odeur de vanilline mais induit une baissesignificative de l’odeur de l’eugénol.

β-lactoglobuline / arôme / interaction / chromatographie d’affinité / analyse sensorielle

β-Lactoglobulin-flavour compound interactions

0.3 mol·L–1 sodium chloride, pH 7.2), andthen filtered (0.45 µm). Immobilization ofthe protein was performed by connectingthe PEEK column to a high-performanceliquid chromatography (HPLC) pump andforcing 67.5 mL of the BLG solution(9.3 mg·mL–1) to circulate at a flow rate of0.5 mL·min–1 for 18 h at room temperature.Then the column was rinsed successivelyat a flow rate of 1 mL·min–1 with the fol-lowing buffers: 0.2 mol·L–1 Tris–HCl con-taining 1 mmol·L–1 dithioerythriol, pH 8.0(180 mL), 0.1 mol·L–1 sodium phosphatecontaining 0.5 mol·L–1 sodium chloride and1 mmol·L–1 dithioerythriol, pH 7.5 (90 mL),0.1 mol·L–1 sodium phosphate containing1 mmol·L–1 dithioerythriol, pH 7.5 (90 mL).The column was then rinsed with water con-taining 25 mmol·L–1 NaCl, pH 3.0 for equi-libration before use. No variation in pH wasobserved during the experiment. The amountof bound protein was calculated from thedifference in BLG concentrations beforeand after immobilization (determinedspectrophotometrically at 278 nm). The4 columns produced contained 8.0, 8.2, 8.3and 8.5 mmol·L–1 BLG. Column produc-tion was thus repeatable under the condi-tions described above. Maximal immobi-lization of protein on the activated silicadiol was indicated by the fact that BLG con-centration in the solution after coupling didnot decrease when increasing the time ofcirculation.

2.2.2. High performance liquidchromatography

The HPLC system consisted of a Waters600 pump coupled to a Waters LC spec-trophotometer model 481. Flavour com-pounds were detected at the following wave-lengths, determined by recording UV spectraof their solutions in water (25 mmol·L–1

NaCl, pH 3.0): ester (208 nm), pyrazines(214 nm), phenols (200 nm), 2-phenyle-thanol (206 nm), vanillin and ethylvanillin(204 nm), norfuraneol, furaneol and ethyl-furaneol (286 nm), mesifurane (278 nm),

2. MATERIALS AND METHODS

2.1. Materials

PEEK columns (5 cm × 4.6 mm i.d.) andsupplements were purchased from Touzart& Matignon (Courtabœuf, France). β-Lac-toglobulin (variants AB, purity 90%) wasobtained from Besnier (Chateaulin, France),silica diol ‘LiChroprep Diol’ (25–40 µm)was from Merck (Darmstadt, Germany) and2,2,2-trifluoroethanesulfonyl chloride (tre-syl chloride) was obtained from Sigma-Aldrich (Deisenhofen, Germany). Flavourcompounds were from IFF (Longvic,France), Givaudan-Roure (Dübendorf,Switzerland), Sigma-Aldrich or Fluka (Neu-Ulm, Germany). Their purity was deter-mined by high resolution gas chromatogra-phy. Purity of flavour compounds was≥ 95% except for 5-ethyl-3-hydroxy-4-methyl-2(5H)furanone (93%), (1R, 4R)-trans- and (1S, 4R)-cis-3-oxo-p-menthane-8-thiol, 4-vinylguaiacol and nerol (90%),3-ethyl-2-methoxypyrazine (87%), ethylisobutyrate and 4-vinylphenol (83%), pule-gone, (1S, 4S)-trans- and (1R, 4S)-cis-3-oxo-p-menthane-8-thiol (80%), (–)-carveol(98%; a mixture of isomers, 52% and 46%).The qualitative purity of vanillin andeugenol was checked by gas chromatogra-phy/olfactometry, applying the on-columninjection technique. All other chemicals usedwere of p.a. quality. For sensory analysis,Evian mineral water was used.

2.2. Affinity chromatography

2.2.1. Preparation of the columns

The method of immobilizing the proteindeveloped by Sostmann and Guichard [25]was modified with reference to Felix andLiu [5]. Silica diol was activated with tresylchloride. The dried activated diol was thenfilled into a PEEK column under vacuum.A BLG solution was prepared by suspend-ing 1.5 g of the protein in 150 mL buffer(0.3 mol·L–1 sodium phosphate containing

349

J. Reiners et al.

sotolon and abhexone (232 nm), pulegone(259 nm), (–)-carvone (242 nm), other ter-penes (200 nm). Aqueous solutions offlavour compounds were injected using aRheodyne injector with a 50-µL loop. Theconcentrations of flavour solutions werebetween 0.05 and 1 mmol·L–1, dependingon water solubility and sensitivity of UVabsorption of the substances. Elution of thecompounds was performed at room temper-ature at a flow rate of 1 mL·min–1. A multi-channel chromatography workstation(Almanza and Mielle, INRA, Dijon, France)was used for data acquisition and treatmentof data. Retention times of flavour sub-stances, at 2 different concentrations, andvoid times of the columns (injection of purewater) were determined for 2 columns con-taining BLG and for 1 column without BLG.

2.2.3. Calculation of binding constants

Binding constants (KB) were calculatedwith reference to Nilsson and Larsson [14]basing on the retention times determined byaffinity chromatography and on the proteinconcentration of the columns:

tR − tKB = (L·mol–1)

cP × t0with:tR: retention time of the compound on the

column containing protein,t: retention time of the compound on the

column without protein,cp: protein concentration (mol·L–1),t0: void time.

This equation supposed that the concen-tration of ligand was small in comparisonto the number of binding sites, which wasthe case. However, we verified that the valueobtained was the same for 2 different con-centrations of ligands.

2.3. Calculation of log P values

As BLG-flavour compound interactionsare considered to be hydrophobic, log P

values of the substances were calculated(Tab. I). If experimental values of relatedmolecules were available, the π-method wasapplied [21]. The log P values for the ter-penes and the furanones were calculatedusing the fragment method [7].

2.4. Sensory analysis

The test panel consisted of 15 experi-enced and inexperienced assessors (11 malesand 4 females, aged from 25 to 70 years).Sensory evaluations were performed in atesting room equipped with computerizedbooths. Data acquisition and data treatmentwere conducted with FIZZ software (Biosys-tèmes, Dijon, France).

Samples (20 mL) were presented inbrown glass bottles (60 mL) closed withscrew caps. After an equilibration time of4 h at 21 ± 1 °C, their odour intensities wereevaluated.

In two training sessions, panelists wereasked to classify 8 concentrations of vanillinand of eugenol. Additionally, they werefamiliarized with the method described inthe following.

Odour intensities of aqueous solutions(Evian water, 25 mmol·L–1 NaCl, pH 3.0) ofvanillin and eugenol in the presence ofβ-lactoglobulin were estimated by applyingthe matching test according to Rousseau etal. [22]. The concentration ranges of BLGand of the flavour compounds had beendetermined in preliminary tests.

During an experimental session, panelistswere first provided with a set of 8 samples ofthe flavour compound analyzed, whichserved as a reference. Concentrations rangedfrom 1.60 to 447 mg·L–1 (vanillin) and from0.16 to 45 µL·L–1 (eugenol), respectively,differing by a concentration step of √5. Thesamples coded from 1 to 8 were spotted ona continuous line scale overlapping the8 reference points, ranging from 0 to over 8.Subjects were asked to sniff the 8 samples inorder of increasing concentration, and to

350

β-Lactoglobulin-flavour compound interactions 351

Table I. Binding constants KB determined by affinity chromatography and log P values of analyzedsubstances (SD = standard variation on 2 columns).Tableau I. Constantes d’affinité KB déterminées par chromatographie d’affinité et valeurs de log Ppour les composés analysés (SD = coefficients de variation sur 2 colonnes).

Flavour compound KB log P(SD %)

Ethyl isobutyrate 132 1.51a

(5)

Ethyl butyrate 136 1.71a

(1)

Ethyl 3-methylbutyrate 284 2.01a

(7)

Ethyl 2-methylbutyrate 288 2.01a

(7)

Ethyl pentanoate 366 2.21a

(7)

2-Methoxypyrazine 47 −0.24b

(16)

3-Methyl-2-methoxypyrazine 62 0.32b

(10)

3-Ethyl-2-methoxypyrazine 171 0.82b

(3)

3-Isopropyl-2-methoxypyrazine 452 1.12b

(8)

3-Isobutyl-2-methoxypyrazine 795 1.62b

(9)

3-sec-Butyl-2-methoxypyrazine 912 1.62b

(9)

2-Phenylethanol 132 1.36c

(5)

4-Methylphenol 440 1.95c

(p-cresol) (3)

4-Ethylphenol 888 2.26c

(5)

2-Methoxyphenol 245 1.33c

(guaiacol) (2)

4-Ethyl-2-methoxyphenol 830 2.38 d

(4-ethylguaiacol) (5)

J. Reiners et al.352

4-Ethenyl-2-methoxyphenol 1 165 2.08d

(4-vinylguaiacol) (1)

2-Methoxy-4-(2-propenyl)phenol 1 360 2.58d

(eugenol) (1)

4-Hydroxy-3-methoxybenzaldehyd 319 1.26c

(vanillin) (3)

3-Ethoxy-4-hydroxybenzaldehyd 475 1.76e

(ethylvanillin) (2)

Linalool 565 2.91f

(9)

Nerol 1 134 3.00f

(10)

α-Terpineol 483 3.15f

(9)

(-)-Carveol 542 2.60f

(14)

(-)-Carvone 748 1.91f

(7)

Pulegone 857 2.46f

(5)

α-Menthone 1 138 3.01f

(4)

cis-3-oxo-p-Menthane-8-thiol 1 208 3.10f

(1)

rans-3-oxo-p-Menthane-8-thiol 1 461 3.10f

(2)

Table I/Tableau I. (Continued/Suite).

Flavour compound KB log P

β-Lactoglobulin-flavour compound interactions

marking their position on the line scale. Ses-sions were duplicated.

Sensory data were subjected to a 3-wayanalysis of variance (3-way ANOVA) test-ing the 3 following factors (concentrationof BLG, concentration of flavour compound,assessor) and all interactions, consideringthe assessor factor as random.

Means and confidence limits (P = 0.05)were calculated. Sample means were com-pared by the multiple range test of New-man-Keuls and of Dunnett (P = 0.05).

memorize the intensity of the odours per-ceived. After this calibration exercise, pan-elists had to estimate the odour intensity of9 samples coded with 3-digit random num-bers, following a presentation order based ona Latin square. The samples were preparedaccording to a complete factorial designwith the flavour compound factor at 3 levels(8, 40 and 200 mg·L–1 vanillin or 0.8, 4.0and 20.0 µL·L–1 eugenol, respectively) andthe BLG factor at 3 levels (0, 0.3 and 1.0%).Their odour intensities were evaluated bycomparing them with the references and by

353

Table I/Tableau I. (Continued/Suite).

Flavour compound KB log P

4-Hydroxy-5-methyl-3(2H) 4 0.24f

furanone (norfuraneol) (100)

4-Hydroxy-2,5-dimethyl-3(2H) 16 0.78f

furanone (furaneol) (2)

2-Ethyl-4-hydroxy-5-methyl-3(2H) 39 1.32f

furanone (ethylfuraneol) (3)

4-Methoxy-2,5-dimethyl-3(2H) 19 1.61f

furanone (mesifurane) (7)

3-Hydroxy-4,5-dimethyl-2(5H) 31 −0.22f

furanone (sotolone) (14)

5-Ethyl-3-hydroxy-4-methyl-2(5H) 82 0.32f

furanone (abhexone) (4)

a Calculated based on the experimental value of ethyl propionate (1.21, [6]) by applying the π-method [15].b Calculated basing on the experimental value of pyrazine (−0.22, [6]) by applying the π-method [15].c Mean of the experimental values cited by Hansch and Leo [6].d Calculated based on the experimental value of guaiacol by applying the π-method [15].e Calculated based on the experimental value of vanillin by applying the π-method [15].f Calculated by applying the fragment method [6].

a Calculé à partir de la valeur expérimentale du propionate d’éthyle (1,21, [6]) en appliquant la π-méthode [15].b Calculé à partir de la valeur expérimentale de la pyrazine (−0,22, [6]) en appliquant la π-méthode [15].c Moyenne des valeurs expérimentales citées par Hansch et Leo [6].d Calculé à partir de la valeur expérimentale du guaiacol en appliquant la π-méthode [15].e Calculé à partir de la valeur expérimentale de la vanilline en appliquant la π-méthode [15].f Calculé en appliquant la méthode des fragments [6].

J. Reiners et al.

3. RESULTS

3.1. Affinity chromatography

After stocking the HPLC columns forseveral weeks, retention times decreasedindependently on the use of the columns.The KB values obtained with a freshly pre-pared and an aged column differed for mostof the substances by 25 to 35%. This wasalso observed by Vidal-Madjar et al. [26],and may be due to protein degradation, con-tamination by organic substances or micro-bial agents. In order to minimize this phe-nomenon, Aubel and Rogers [1] addedsodium azide. As we noticed interactionsbetween sodium azide and BLG, we did notadd this antimicrobial agent. Binding con-stants were calculated from the retentiontimes obtained on 2 freshly preparedcolumns (Tab. I). In these conditions, theobtained values were highly reproducible(CV < 10%). Log P values of the substanceswere calculated as mentioned in Table I.

For the esters, a significant increase inthe binding constants by increasing the num-ber of C-atoms of the acid chain from 3(ethyl butyrate) to 4 (ethyl pentanoate) wasfound, thus confirming previous resultsobtained by Pelletier et al. [19]. The effect ofbranching on a hydrophobic chain was stud-ied for 2 series of esters. Branching on achain length of 3 C-atoms had practicallyno influence on the binding constants of theesters (ethyl butyrate and ethyl isobutyratehad the same KB), whereas the 2- and3-branched esters of pentanoic acid showedlower KB values and lower log P values thanthe corresponding linear ester. Obviously,the position of the methyl group was notimportant, as ethyl 2-methylbutyrate andethyl 3-methylbutyrate had the same bindingconstants.

2-Methoxypyrazine had a rather lowbinding constant, corresponding to its neg-ative log P value. KB and log P increasedwith increasing chain length of the sub-stituent in position 3. Here, differing fromthe esters, the position of branching influ-

enced KB. The binding constant of 3-iso-butyl-2-methoxypyrazine was lower thanthat of 3-sec-butyl-2-methoxypyrazine,although the 2 substances have the sametheoretical hydrophobicity. The differencebetween the constants of the structural iso-mers was smaller than the differencebetween KB of the pyrazines differing by1 C-atom in the substituent chain.

When comparing the phenols with dif-ferent functional groups (4 methyl with4-ethylphenol, guaiacol with 4-ethylguaia-col or 4-vinylguaiacol with eugenol, vanillinwith ethylvanillin), a prolongation of thesubstituent chain length always enhancedthe hydrophobicity and hence the interac-tions of the compound with the protein.2-Phenylethanol had a rather low bindingconstant, especially when compared with4-ethylphenol which has a similar structureand the same number of C-atoms in the side-chain, but with a lower hydrophobicity.Introduction of a methoxy function in posi-tion 2 of 4-ethylphenol had very little influ-ence on KB or on log P. Only a slightdecrease in the binding constant of 4-ethyl-guaiacol was observed. The double bond of4-vinylguaiacol caused an increase in KBcompared to the same structure with a singlebond (4-ethylguaiacol) with lower hydro-phobicity. The vanillin aldehyde group wasresponsible for only a slight but significantincrease in KB of guaiacol.

In general, the terpenes analyzed had highbinding constants compared with the otherclasses of substances, indicating strong inter-actions with β-lactoglobulin. The terpeniccompounds are rather apolar, in spite of theircarbonyl or alcohol functions. For ketoneswith a similar structure, KB values arerelated to log P values. For the alcohols, thediffering binding constants cannot beexplained by the log P values.

Steric effects have been studied by appli-cation of affinity chromatography to a chiralsubstance. Four enantiomerically pure thiols were analyzed: cis-(1R, 4S)- and(1S, 4R)- as well as trans-(1R, 4R)- and

354

β-Lactoglobulin-flavour compound interactions

protein. However, affinity chromatographyonly measures reversible interactions [25].

BLG concentrations were chosen basedon their natural content (0.3%) in milk. A3-way ANOVA was performed in order toestablish the effect of each variable (con-centration of vanillin or eugenol, concen-tration of β-lactoglobulin, assessor) and theirinteractions on perceived odour intensities.

Results are shown in Table II. For theanalysis of the vanillin odour intensity, sig-nificant effects of the vanillin concentra-tion, of the assessors and of the β-lac-toglobulin-assessor interaction were found.For the analysis of the eugenol odour inten-sity, significant effects of the eugenol con-centration, of the assessor and of the inter-actions between eugenol and BLG wereobserved.

Perceived odour intensities of the flavourcompounds in aqueous solutions with orwithout BLG are depicted in Figures 2 and 3.A significant difference between sampleswas observed in only 1 case: at a concen-tration of 20 µL·L–1, eugenol odour intensitywas significantly (P = 0.05) diminished by1% BLG. Performing the Newman-Keuls

(1S, 4S)-3-oxo-p-menthane-8-thiol. Whileretention times on the HPLC column with-out protein were exactly the same for all4 enantiomers, they were different on thecolumn containing BLG for cis- and trans-isomers. KB of the 2 trans-isomers weresimilar, and significantly higher than KB ofthe 2 cis-compounds.

The furanones showed very low oralmost no interactions with β-lactoglobulinexpressed by the lowest binding constants ofall analyzed substances. Log P values werealso rather low. No interaction was foundfor norfuraneol (4-hydroxy-5-methyl-3(2H)-furanone), having only 1 methyl group inthe furanone ring. Increasing the proportionof the apolar substituents led to an increasein hydrophobicity and KB (norfuraneol, fura-neol, ethylfuraneol). Replacement of thehydroxyl group in furaneol by a methoxygroup (mesifurane) doubled log P, but didnot change KB. When comparing sotolon(3-hydroxy-4,5-dimethyl-2(5H)-furanone)with furaneol (4-hydroxy-2,5-dimethyl-3(2H)-furanone) and abhexone (5-ethyl-3-hydroxy-4-methyl-2(5H)-furanone) withethylfuraneol (2-ethyl-4-hydroxy-5-methyl-3(2H)-furanone), KB of the 2(5H)furanoneswas 2-fold the KB of the corresponding3(2H)furanones, though log P valuesbehaved in the reverse manner. Sotolon andethyl furaneol showed the same behaviour toBLG in spite of the ethyl group of the 3(2H)-furanone. The relation between binding con-stant KB and the hydrophobicity of the ana-lyzed substances is depicted in Figure 1.

3.2. Sensory analysis

Vanillin was selected because it is com-monly used in the aromatization of dairyproducts and eugenol because of its similarstructure, different chemical properties andhigher binding constants. The interactionsbetween BLG and vanillin can be ofreversible and irreversible nature, due to thepossibility of the aldehydic function to formSchiff bases with free NH2-groups of the

355

Figure 1. Binding constants KB in relation tohydrophobicity of the analyzed substances.Figure 1. Relation entre hydrophobicité etconstante d’affinité KB pour les composés ana-lysés.

J. Reiners et al.

test at P = 0.1 revealed that the 3 samples

containing 20 µL·L–1 eugenol and 0, 0.3

and 1% BLG were significantly different.

No significant effect was observed for the

vanillin samples.

4. DISCUSSION

4.1. Affinity chromatography

As interactions between BLG and smallmolecules are considered to be of hydrophobic

356

Table II. Matching test of vanillin and eugenol. Results of 3-way ANOVA.Tableau II. Test d’appariement pour la vanilline et l’eugénol. Résultats de l’ANOVA à 3 facteurs.

Vanillin Eugenol

Factor F-value P-value F-value P-value

Concentration of flavour (F) 131.41 < 0.0001* 238.43 < 0.0001*Concentration of β-lactoglobulin (BLG) 1.75 0.1929 2.31 0.1175Assessor (A) 2.35 0.0061* 4.85 < 0.0001* F × BLG 0.19 0.9432 3.01 0.0254*F × A 1.25 0.2023 0.96 0.5234BLG × A 1.57 0.0466* 1.14 0.3074F × BLG × A 1.33 0.0916 0.98 0.5221

* Significant at P ≤ 0.05.

Figure 2. Results of the matching test forvanillin; a, b, c are the groups from the New-man-Keuls test (P = 0.05). Figure 2. Résultats du test d’appariement pour lavanilline. a, b, c sont les groupes du test de New-man-Keuls (P = 0,05).

Figure 3. Results of the matching test foreugenol. a, b, c, d are the groups from the New-man-Keuls test (P = 0.05).Figure 3. Résultats du test d’appariement pourl’eugénol. a, b, c, d sont les groupes du test deNewman-Keuls (P = 0,05).

β-Lactoglobulin-flavour compound interactions

phenylalanine. The same effect was dis-cussed for tryptophan by Papiz et al. [18]for the other binding site of BLG.

Introduction of a methoxy group in thephenol ring led to only a slight decrease inKB (when comparing 4-ethylguaiacol with4-ethylphenol), while the introduction of analdehyde function led to a slight increase inKB (when comparing vanillin with guaia-col). The observed effect was only smallbut significant, as was the difference in thelog P values.

That hydrophobicity cannot be the onlyfactor influencing BLG-flavour compoundinteraction is also evident when regardingthe results obtained for the terpenic alco-hols (Tab. I). Linalool and nerol, with a sim-ilar polarity, differ in KB by a factor of 2.The higher KB of nerol can be explained bythe position of the OH function at the end ofthe carbon chain, allowing the rest of themolecule to enter into the hydrophobicpocket of BLG. In linalool, the hydroxylgroup is positioned within the hydrophobicchain. Therefore, only a part of the moleculeinteracts with the protein. The cyclic ter-penic alcohols are supposed to enter by theapolar end of the molecule opposite the OHfunction into the hydrophobic pocket ofBLG. Other parameters taking into accountthe topological characteristics of themolecule may be investigated using QSAR(quantitative structure-activity-relationship).

Apart from the described steric effectsof different origin, the influence of the stere-ochemistry of flavour compounds on theirinteractions with BLG was shown by theanalysis of the 4 3-oxo-p-menthane-8-thiolenantiomers. As the only difference betweenthe molecules was the arrangement of thesubstituents in positions 1 and 4, and as eachsubstituent is once in the axial and once inthe equatorial position for cis- and trans-iso-mers respectively, the difference in KB forcis- and trans-3-oxo-p-menthane-thiol mustbe due to the relative position of the 2 sub-stituents. Lamiot et al. [9] also found a greateraffinity for trans isomers of dodecadienols

nature, binding constants KB were relatedto log P values of the substances (Fig. 1).Regarding especially the pyrazines, anincrease in the chain length of the substituentcaused an increase in the hydrophobicity ofthe molecules and therefore of their inter-actions with BLG. Pelletier et al. [19] didnot observe any significant interactionbetween different alkylpyrazines and BLG.On the other hand, Boudaud and Dumont[2] observed quenching of the fluorescenceemission of BLG by 3-isobutyl-2-methoxy-pyrazine and also by 2,3,5-trimethy-lpyrazine, indicating complex formation.They concluded that the ring structure andnot the methoxy group was important forthe interactions between BLG and thepyrazines. In the present study, 3-isobutyl-2-methoxypyrazine and 3-sec-butyl-2-methoxypyrazine with the same theoreticalhydrophobicity differ in their binding con-stants (Tab. I), indicating a steric influence.We suppose that the 3-isobutyl-2-methoxy-pyrazine enters less easily into the hydropho-bic pocket of BLG because of the ramifica-tion at the end of the chain resulting inweaker interactions with the protein. Thissteric effect was less important than thehydrophobicity effect, because the differ-ence in KB values of the 2 stereoisomerswas only 13%, while the difference betweenKB of 3-isopropyl- and of 3-isobutylmetho-xypyrazine was 43%.

The interactions of an alkyl chain withBLG can also be reduced by a polar group,as in the case of 2-phenylethanol. The OHfunction at the end of the hydrophobic sub-stituent is responsible for weaker interac-tions when compared with the KB of4-ethylphenol. The KB of 4-vinylguaiacolis higher than that of 4-ethylguaiacol despiteits lower hydrophobicity. This may beexplained by a stabilization of the 4-vinyl-guaiacol–BLG-complex resulting from astructural analogy between the planarflavour compound and the planar side-chainof an amino acid in the hydrophobic pocketof BLG. Monaco et al. [12] suggested thatthe planar ring of β-ionone was fixed by

357

J. Reiners et al.

in comparison to the cis isomers. However,we observed no stereoselectivity of BLGfor these compounds, which confirms theresults obtained by Massolini et al. [11].The importance of the geometry of themolecule has also been pointed out byLübke et al. [10], who showed, by infraredspectroscopy, that the fixation of retinol andβ-ionone into the hydrophobic cavityinduced no significant conformationalchange in BLG, whereas the fixation ofα-ionone did. Both isomers only differ bythe position of the cyclohexene double bond,which is in the ‘retinol-like’ position forβ-ionone.

The weak interactions between BLG andthe furanones are supposed to be due to thepolarity of the furanone ring combined withonly a few short hydrophobic side-chains.The higher KB values of the 2(5H)furanones,sotolone and abhexone, compared to the KBvalues of the corresponding 3(2H)furanones,furaneol and ethylfuraneol, may be due tothe proximity of the apolar substituents.

Looking at Figure 1, a general tendencytoward increased interactions indicated byincreasing KB with increasing hydropho-bicity (expressed by log P values) wasobserved. The KB rose when the number ofC-atoms in the side-chain of the ligand(pyrazines) was increased. For less homo-geneous classes, for example when func-tional groups change (phenols and relatedcompounds like 2-phenylethanol, vanillinand ethylvanillin; terpenes), the relationbetween KB and log P is not evident. Obvi-ously, the topological characteristics offlavour molecules play an important role intheir interactions with BLG, which arenot taken into account by calculated logP-values.

4.2. Sensory analysis

The assessor effect was significant forthe sample sets of vanillin and eugenol,meaning that the assessors estimated theintensity of the odorants in a different way.A different use of the scale may be due to

the interindividual variations in the percep-tion of BLG odour. The odour of BLG couldalso be the reason for the BLG–assessorinteraction observed for the vanillin sam-ple set. The significance of this factor resultsfrom the different tendencies of the assessorsin evaluating the samples containing 0, 0.3and 1% BLG.

The vanillin and the eugenol effects weresignificant, as samples were grouped accord-ing to their flavour compound concentra-tion. The F-value for the concentrations ofvanillin and eugenol was much higher thanthe F-value for the assessor effect, reflectingthe greater importance of the flavour com-pounds. The higher F-value for eugenolcompared with that for vanillin indicated agreater difference between the 9 sampleswhen they contained eugenol as the odor-ous substance.

There was a significant effect for theeugenol-BLG interaction, corresponding tothe fact that a significant influence of BLGon the perception of the odour intensity ofeugenol was observed for only a certain con-centration of eugenol. This means that theinfluence of BLG on the odour intensity ofeugenol is dependent on the eugenol con-centration. Headspace analyses performedon the same solutions showed a retentionof around 15% eugenol by BLG, which is inagreement with the binding constants cal-culated and the concentrations used.

No significant effect was found forvanillin, in agreement with the relativelylow binding constant, indicating weakerinteractions with BLG. It was not possible todetect vanillin by headspace analysis at theconcentrations studied.

5. CONCLUSION

Affinity chromatography is a rapid, repeat-able method to measure global affinitybetween BLG and flavour compounds. Evenif the reversible interactions measured areclosed related to the hydrophobicity of the

358

β-Lactoglobulin-flavour compound interactions

[9] Lamiot E., Dufour E., Haertlé T., Insect sexpheromone binding by bovine β-lactoglobulin,J. Agric. Food Chem. 42 (1994) 695–699.

[10] Lübke, M., Guichard, E., Le Quéré, J.L., Infraredspectroscopic study of β-lactoglobulin interac-tions with flavour compounds, in: Roberts D.,Taylor A. (Eds.), Flavour Release, ACS Symp.Ser., Am. Chem. Soc., Washington, DC, 2000(in press).

[11] Massolini G., De Lorenzi E., Lloyd D.K.,McGann A.M., Caccialanza G., Evaluation ofβ-lactoglobulin as a stationary phase in high-performance liquid chromatography and asbuffer additive in capillary electrophoresis:observation of a surprising lack of stereoselec-tivity, J. Chromatogr. 712 (1998) 83–94.

[12] Monaco H.L., Zanotti G., Spandon P., BolognesiM., Sawyer L., Eliopoulos E.E., Crystal structureof the trigonal form of bovine β-lactoglobulinand of its complex with retinol at 2.5 Å resolu-tion, J. Mol. Biol. 197 (1987) 695–706.

[13] Narayan M., Berliner L.J., Fatty acids andretinoids bind independently and simultaneouslyto β-lactoglobulin, Biochemistry 36 (1997)1906–1911.

[14] Nilsson K., Larsson P.-O., High performanceliquid affinity chromatography on silica-boundalcohol dehydrogenase, Anal. Biochem. 134(1983) 60–72.

[15] O’Neil T.E., Kinsella E., Binding of alkanoneflavors to β-lactoglobulin: effects of conforma-tional and chemical modification, J. Agric. FoodChem. 35 (1987) 770–774.

[16] Ohmes R.L., Marshall R.T., Heymann H., Sen-sory and physical properties of ice creams con-taining milk fat or fat replacers, J. Dairy Sci. 81(1998) 1222–1228.

[17] Overbosch P., Afterof W.G.M., Haring P.G.M.,Flavor release in the mouth, Food Rev. Int. 7(1991) 137–184.

[18] Papiz M.Z., Sawyer L., Eliopoulos E.E., NorthA.C.T., Findlay J.B.C., Sivaprasadarao R., JonesT.A., Newcomer M.E., Kraulis P.J., The struc-ture of β-lactoglobulin and its similarity toplasma retinol–binding protein, Nature 324(1986) 383–385.

[19] Pelletier E., Sostmann K., Guichard E., Mea-surement of interactions between β-lactoglobu-lin and flavor compounds (esters, acids, andpyrazines) by affinity and exclusion size chro-matography, J. Agric. Food Chem. 46 (1998)1506–1509.

[20] Plug H., Haring P., The influence of flavour-ingredient interactions on flavour perception,Food Qual. Pref. 5 (1994) 95–102.

[21] Rekker R. F., The Hydrophobic FragmentalConstant: Pharmacochemistry Library, in: NautaW.T., Rekker R.F. (Eds.), Elsevier Sci. Publ.Co., Amsterdam, Vol. 1, 1977.

molecules, topological characteristics haveto be taken into account for a better inter-pretation. The influence of binding on odourintensity was demonstrated for 2 phenoliccompounds in water solution, showing asignificant effect only for the compoundwith the higher binding constant. Furtherexperiments are in progress to study the inci-dence of these interactions on flavour per-ception in real food systems.

ACKNOWLEDGEMENTS

The authors thank I. Andriot and N. Fournierfor technical assistance and the Conseil Régionalde Bourgogne for financial support. This pro-gramme was part of a collaborative study withENSBANA-Dijon, ENSIA-Massy, GBSA-Mont-pellier, INRA-Dijon, INRA-Nantes and waspartly financed by the French Ministry of Agri-cultural and Fisheries.

REFERENCES

[1] Aubel M., Rogers L.B., Effects of pretreatmenton the enantioselectivity of silica-bound bovineserum albumin used as high-performance liq-uid chromatographic stationary phases, J. Chro-matogr. 392 (1987) 415–420.

[2] Boudaud N., Dumont J.-P., Interaction betweenflavor components and β-lactoglobulin, in:McGorrin R.J., Leland J.V. (Eds.), Flavor-FoodInteractions, ACS Symp. Ser. 633, Am. Chem.Soc., Washington, DC, 1996, pp. 90–97.

[3] Delahunty C.M., Piggott J.R., Current methodsto evaluate contribution and interactions of com-ponents to flavour of solid foods using hardcheese as an example, Int. J. Food Sci. Tech-nol. 30 (1995) 555–570.

[4] Dufour E., Marden M.C., Haertlé T., β-Lac-toglobulin binds retinol and protoporphyrin IXat two different binding sites, FEBS Lett. 277(1990) 223–226.

[5] Felix G., Liu M., New method for grafting pro-teins on silica gel, Biol. Sci. 8 (1989) 2–6.

[6] Fischer N., Widder S., How proteins influencefood flavor, Food Technol. 51 (1997) 68–70.

[7] Hansch C., Leo A., Substituent Constants forCorrelation Analysis in Chemistry and Biology,John Wiley & Sons, New York, USA, 1979.

[8] Hansen A.P., Heinis J.J., Decrease of vanillinflavor perception in the presence of casein andwhey proteins, J. Dairy Sci. 74 (1991) 2936–2940.

359

J. Reiners et al.

[22] Rousseau F., Castelain C., Dumont J.P., Oil-water partition of odorant: discrepancy betweensensory and instrumental data, Food Qual. Pref.7 (1996) 299–303.

[23] Schieberle P., Hofmann T., Evaluation of thecharacter impact odorants in fresh strawberryjuice by quantitative measurement and sensorystudies on model mixtures, J. Agric. Food Chem.45 (1997) 227–232.

[24] Schnermann P., Schieberle P., Evaluation ofkey odorants in milk chocolate and cocoa massby aroma extract dilution analyses, J. Agric.Food Chem. 45 (1997) 867–872.

[25] Sostmann K., Guichard E., Immobilized β-lac-toglobulin on a HPLC-column: a rapid way todetermine protein–flavour interactions. FoodChem. 62 (1998) 509–513.

[26] Vidal-Madjar C., Jaulmes A., Racine M., SebilleB., Determination of binding equilibrium con-stants by numerical simulation in zonal high-performance affinity chromatography, J. Chro-matogr. 458 (1988) 13–25.

[27] Wu S.Y., Perez M.D., Puyols P., Sawyer L.,β-Lactoglobulin binds palmitate within its cen-tral cavity, J. Biol. Chem. 274 (1999) 170–174.

360