HAL Id: hal-00929554https://hal.archives-ouvertes.fr/hal-00929554

Submitted on 1 Jan 1997

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.

Potentiality of spectroscopic methods for thecharacterisation of dairy products. I. Front-face

fluorescence study of raw, heated and homogenised milksE Dufour, A Riaublanc

To cite this version:E Dufour, A Riaublanc. Potentiality of spectroscopic methods for the characterisation of dairy prod-ucts. I. Front-face fluorescence study of raw, heated and homogenised milks. Le Lait, INRA Editions,1997, 77 (6), pp.657-670. �hal-00929554�

Lait (1997) 77, 657-670© Elsevier/Inra

657

Original article

Potentiality of spectroscopie methodsfor the characterisation of dairy products.1. Front-face fluorescence study of raw,

heated and homogenised milks

E Dufour, A Riaublanc

Laboratoire d'étude des interactions des molécules alimentaires, Inra. BP 71627,44316 Nantes cedex 03, France

(Received 24 December 1996; accepted 13 May 1997)

Summary - The fluorescence spectra of raw (NHa), heated (NHP), homogenised (HOM) andhomogenised + heated (HOP) milks were recorded using a variable angle front-surface accessory. Theemission fluorescence spectra of tryptophans in proteins, vitamin A and anilino-naphthalene sulfo-nie acid and the excitation fluorescence spectra of vitamin A and anilino-naphtha1ene sulfonic acidwere collected. The spectra showed that the treatments applied to milk induced changes in the fluo-rescence characteristics of the probes. Principal component analysis was applied to the normalised fluo-rescence spectral data in order to distinguish between milk samples. It was shown that the map defi-ned by principal components 1 and 2 discriminated NHa, NHP, HOM and HOP samples as a functionof homogenisation and heating, respectively. The potential of fluorescence spectroscopy in combi-nation with a chemometric method to discriminate between heated and homogenised milk sampi eswas demonstrated.

milk / protein / vitamin A / front face fluorescence / multivariate analysis

Résumé - Intérêts des méthodes spectroscopiques pour la caractérisation des produits lai-tiers. I.Utilisation de la fluorescence frontale pour la caractérisation de laits natif, chauffé ethomogénéisé. Les spectres de fluorescence de lait natif (NHa), chauffé (NHP), homogénéisé (HOM)et homogénéisé + chauffé (HOP) ont été enregistrés au moyen d'un accessoire de fluorescence fron-tale. Les spectres d'émission de fluorescence des tryptophanes, de la vitamine A et de l'acide anilino-naphtalène sulfonique, ainsi que les spectres d'excitation de la vitamine A et de l'acide anilino-naphtalène sulfonique, ont été enregistrés. Les traitements appliqués au lait induisent des modificationsdans les spectres de fluorescence. L'analyse en composante principale a été appliquée sur les spectresde fluorescence normés. La carte factorielle 1-2 permet de séparer les échantillons en fonction du trai-tement appliqué au lait.

lait / protéine / vitamine A / fluorescence frontale / analyse multivariée

658 E Dufour, A Riaublanc

INTRODUCTION

Fluorescence spectroscopy, a very sensitivetechnique, has been used for a long time asa powerful analytical tool in many chemical,biochemical and environmental studies. Thepurpose of these studies is to find pairs ofemission-excitation wavelengths with maxi-mum intensity or to record fluorescenceemission and excitation spectra. Fluores-cence spectroscopy provides informationon the presence of fluorescent moleculesand their environment in biological samples.For instance, fluorescence properties of aro-matie ami no acids of proteins (Longworth,1971; Lakowicz, 1983; Dalgalarrondo et al,1992; Dufour et al, 1994), retinol (Dufourand Haertlé, 1990) or extrinsic fluorescentprobes added to the sample can be used tostudy prote in structure and protein-hydro-phobie molecule interactions (Dufour et al,1994).

The aqueous phase of bovine milkcontains six major proteins: ~-lactoglobu-lin, œ-lactalbumin, us1- and us2-caseins,~-casein and x-casein. The amino acid com-positions of aIl these proteins include at leastone tryptophan residue (Fox, 1989). Depen-ding on their structures, each protein exhi-bits, following excitation in the region280-295 nm, a characteristic fluorescenceemission spectrum defined by its maximumemission wavelength and the tryptophanquantum yield (Lakowicz, 1983). Milkretains also fat-soluble vitamins such as vita-mins A, D, E and K. Vitamin A occurs inmore than one form but is generally found asretinol. Because of its alcohol group, reti-nol readily forms esters. In milk, almost aIlthe vitamin occurs in the palmitate or acetateester forms. Vitamin A (about 1 jzmol/l, inbovine milk) is located in the core and inthe membrane of the fat globule (Hartmanand Dryden, 1978). Due to its conjugateddouble bonds, retinol is a good fluorescentprobe with excitation and emission wave-lengths at about 330 and 450 nm, respecti-

vely. The fluorescence properties of retinolchange as a function of the environ ment. Avery weak fluorescence is observed foraqueous solution ofretinol, but its quantumyield is drastically enhanced in an apolarenvironment (Dufour et al, 1994).

Most fluorescence experiments are doneon dilute solutions with absorbance of thesample below 0.1: it is classical right -anglefluorescence spectroscopy. When the absor-bance of the sample is higher than 0.1, thescreening effect (or inner fil ter effect)induces a decrease of fluorescence inten-sity and a distortion of excitation spectra(Genot et al, 1992a). To avoid these pro-blems, an alternative method, frontal illu-mination fluorescence spectroscopy, hasbeen developed (Parker, 1968). Front facefluorescence allows investigation of thefluorescence of powdered, turbid andconcentrated samples. The method has beenused to quantitatively determine hemoglo-bin in undiluted blood (Blumberg et al,.1980), to study hemoglobin R ->T transi-tion kinetics (Hirsch and Nagel, 1989) orproteins in wheat gluten (Genot et al,1992b). However, literature searches encoun-tered very few papers dealing with the appli-cation of front face fluorescence in the cha-racterisation of food products. This couldbe explained by the fact that food productsare complex products containing numerousfluorescent compounds. In such a case thesignaIs of the different chromophores mayoverlap and, for example, it becomes diffi-cult to predict the concentration of one par-ticular compound. However, fluorescencespectroscopy in combination with multiva-riate statistical methods has been used forpredicting the concentrations of two com-ponent synthetic mixtures (Lindberg et al,1983).

Milk is a complex product exhibitingsimultaneously emulsion, colloidal and solu-tion phases. Various chemical and spectro-

Front-face fluorescence study of milk

scopie methods have been used to charac-terise milk samples. Tedious and time-

consuming chemical methods are more andmore often replaced by more rapid and non-invasive spectroscopie methods. In thispaper, the focus is mainly on the study ofmilk intrinsic fluorophores in order to recordfluorescence spectra of 'real' milk samplesand to discriminate between four milksamples (raw, homogenised, heated andhomogenised + heated milks) by applyingprincipal component analysis to the wholefluorescence spectra. The study serves as ageneral investigation of how to enhance thepotential of fluorescence spectroscopy bystatistical methods, as weil as an investiga-tion of how front face fluorescence spec-troscopy in combination with other methodscan be used to characterise food productsand processes.

MATERIALS AND METRODS

l-anilinonaphthalene-8-sulfonic acid (ANS) wasfrom Sigma. ANS stock solution (10 mmol/L)was prepared in methanol.

Milk sampi es

Raw pooled milk (10 L), provided by a dairyplant, was divided in four parts. Two of themwere homogenised in a two-stage laboratoryhomogeniser (25 and 4.5 MPa). Then one rawand one homogenised milk sam pie were heatedin a water bath during 20 min at 70 "C. Sampleswere codedNHO,NHP,HOM and HOP for raw,heated, homogenised and homogenised + hea-ted milks, respectively.

Washed creams were extracted from an ali-quot of each sample by centrifugation in highdensity medium as described by Patton and Hus-ton (1986).

Fat globule size

A Malvern Mastersizer (Mal vern InstrumentsLtd, Malvern, UK) with optical parameters defi-

659

ned by the manufacturer's presentation code 0505was used to determine the fat globule-size dis-tribution, the weight-average diameters (d43), thevolume-surface average diameter (d32) and thetotal interfacial area per unit volume of milk.The measures were made in triplicate. Water wasused to disperse the milk.

Electrophoresis

Proteins adsorbed at the oil/water interface werecharacterised and quantified by electrophoresisfor the four samples. 150 pL of washed creamwas mixed with 50 pL of dissolution buffer (Tris-HCI 150 mmol/L pH 7, SDS 12%, mercapto-ethanol 6%, glycerol 30%, Serva blue G 0.05%)and heated for 5 min at 90 oc. This stage brokeemulsion and desorbed proteins in the waterphase. Proteins were then separated on a 20%homogeneous acrylamide gel according toLaemmli (1970). Protein bands were stained byCoomassie blue and quantified using a Bio Pro-fil densitometer (Vilber Lourmat, Marne la Val-lée, France). Purified n-lactalbumin was usedon the gel as internai standard for quantification.Lipid yields in creams were measured gravime-trically after sol vent extraction and drying.Results were expressed in pg of adsorbed pro-tein by mg of lipids.

Fluorescence spectroscopy

Fluorescence spectra were recorded using a SLM4800C spectrofluorimeter (Bioritech, Chama-rande, France) mounted with a variable anglefront-surface accessory. The incidence angle ofthe excitation radiation was set at 56 ° to ensurethat reflected light, scattered radiation and depo-larisation phenomena are minimised. Emissionand excitation spectra (resolution: 1 nm, avera-ging: 10) were recorded at 22 "C with emissionand excitation slits set at 4 nm. Ail spectra werecorrected for instrumental distortions in excitationusing a rhodamine cell in reference channel. Thespectrum of each sample was recorded six timesusing different aliquots. When ANS fluorescencewas considered, 10 ut: of ANS stock solutionwere added to the milk samples (1 mL) and thefinal concentration of ANS in the cuvette was0.1 mmollL.

The emission spectra of tryptophan (305--400nm), vitamin A (350-500 nm) and ANS (40Q-600

660 E Dufour, A Riaublanc

nm) were recorded with excitation wavelengthsset at 290, 321 and 370 nm, respectively and theexcitation spectra of vitamin A (260-350 nm)and ANS (250-450 nm) were recorded withemission wavelengthsset at 410 and 466 nm,respectively .:

For washed creams, the emission spectra ofprotein tryptophans were recorded for the foursamples (six times for each sample).

Mathematical processingand principal component analysis

In order to reduce scattering effects, the datahave been normalised by reducing the area undereach spectrum to a value of 1 according to theformula (Bertrand and Scotter, 1992):

Ci = Fi/norm (1)

and ŒfLF·2norm = j=l} (2)

where Ci is the corrected value at wavelength i,Fiis the fluorescence intensity at emission wave-length i, Fj is the fluorescence at wavelength jand n is the number of data points for each spec-trum.

Principal component analysis (PCA) wasapplied to the normalised data. PCA is a multi-dimensional statistical method which optimisesthe description of the data with a minimum lossof information (Jolliffe, 1986). From a data set,PCA assesses principal components and theircorresponding eigenvectors. The principal com-ponents are used to draw maps that describe thephysical and chemical variations observed be-tween the samples and make it possible to studythem without any calibration step (Bertrand etal, 1987). Moreover, the eigenvectors are homo-logous to spectra and are called spectral patterns.Both positive and negative peaks of the spectralpattern can be interpreted as characteristic emis-sion or excitation wavelengths of chemical consti-tuents.

The PCA software was written by 0 Bertrand(LTAN, Inra, Nantes, France) and is describedelsewhere (Bertrand et al, 1987).

RESULTS AND DISCUSSION

Fluorescence spectra of milks

A rnilk sarnple was poured into a 1 x 1 cmcuvette and the cuvette was placed in avariable angle front-surface accessory setat 56°. Actually, the pathlength of thecuvette does not matter for front face fluo-rescence since it is the fluorescence of thesurface of the sample which is investigated(Genot et al, 1992a): ail the excitation pho-tons are absorbed within the first few micro-meters. Figure 1 shows the unprocessedexcitation and emission fluorescence spec-tra of raw (NHa), pasteurised (NHP), homo-genised (HOM) and homogenised + pas-teurised (HOP) milks. Considering theintrinsic fluorescence of proteins, trypto-phans were excited at 295 nm and the emis-sion spectra were recorded between 305 and400 nm. The greatest difference was obser-ved between homogenised and non-homo-genised samples: following homogenisa-tion, the emission fluorescence intensityincreased by 6% (fig lA). It appeared alsothat the tryptophan fluorescence intensitiesof heated samples (NHP and HOP) wereslightly lower than those of non-heated milks(NHa and HOM). Ali the maxima of tryp-tophan emission peaks, except that of NHasample (333 nm), were located at 332 nm.The excitation and emission fluorescencespectra of vitamin A for NHa, NHP,HOMand HOP samples are shown in figure 1B.The excitation spectra were characterisedby a maximum located at 322 nm and twoshoulders. The locations of the shouldersdepended on the applied treatment. The firstshoulder was observed at 292 nm for thehomogenised milks and at 293 nm for non-homogenised samples. The locations of thesecond shoulder were 308, 309, 308 and307 nm for NHa, NHP, HOM and HOPmilks, respectively. The shapes of the spec-tra were overall similar, varying mainly inthe maximum/shoulders intensity ratios. lnthe vitamin A emission spectra, the fluo-

Front-face fluorescence study of milk 661

0.2 -

(B) 1.0 (Cl

Ern. Exc. :t': Ern.

0.8 \

o1

0.6

0.4

0.2

0.1 -ë-r---.--r---r--.-..,.J 0.0 L.---.---..----r--...;600250 300 350 400 450 500

Wavelenglh (nm)300 320 340 360 380 400 300 400 500

Fig 1.Fluorescence spectra of NHa, NHP, HOM and HOP milk samples. A. Tryptophan emissionspectra. B. Excitation and emission spectra of vitamin A. C. Excitation and emission spectra ofANS. Exc, excitation spectrum; Em, emission spectrum. See Materials and methods for details.Spectres de fluorescence des échantillons NHO, NHP, HOM et HOP.A. Spectres d'émission des tryp-tophanes. B. Spectres d'excitation et d'émission de la vitamine A. C. Spectres d'excitation et d'émis-sion de l'ANS. Exc: spectre d'excitation; Em: spectre d'émission. Se reporter à « Materials andmethods » pour les détails.

rescence intensities showed, however, largerdifferences. The highest fluorescence inten-sity was observed for NHa milk and treatedmilks were characterised by lower retinolfluorescence: the order was NHa> NHP >HOM> HOP. In addition, the emissionmaxima were at 412 nm and 413 nm fornon-homogenised and homogenised milks,respectively. In addition to these intrinsicprobes, an extrinsic fluorophore (ANS), wasused in this study. ANS is weil known tobind specifically to the hydrophobie poe-kets of proteins (Matarella and Richardson,1983). Excitation and emission spectra ofANS added to the various milks samplesare shown in figure le. In both excitationand emission spectra, the fluorescence yieldincreased in the order: NHa < NHP < HOM

< HOP. The maxima for the emission spec-tra were observed at 468, 467, 466 and 467nm for NHa, NHP, HOM and HOPsamples, respectively. In fluorescence exci-tation, the spectra were characterised by amaximum located at about 388 nm and twoshoulders. The shoulder closest to the maxi-mum was located at 367 nm for all thesamples. The second one was observed at292,294,290 and 291 nm for NHa, NHP,HOM and HOP milks, respectively.

Characterisation of milk fat globulesurface and size

The treatments applied to milk are weilknown to modify protein structure, fat glo-

662 E Dufour, A Riaublanc

bule shape and protein/lipid interactions.Indeed, heating may denature proteins.Denaturation is characterised by the changesof prote in structure modifying tryptophanquantum yield, and by the exposure to thesurface of hydrophobie regions where ANScan bind. Homogenisation breaks up fat glo-bules into smaller ones, increasing drasti-cally the area of the lipid/water interface.The stabilisation of the interface created byhomogenisation results from the adsorptionof proteins at the interface (Walstra, 1995).The adsorption of proteins at the interfacechanges their structure, as weil as their bin-ding and fluorescence properties (Castelainand Genot, 1994). In addition, homogeni-sation should modify retinol quantum yieldsince the structures of the fat globule andof its membrane are altered. The differencesobserved in the reported spectra clearly indi-cate that the physical treatments applied tothe milk modified the characteristics of thefluorescent probes investigated. The oppo-site trends of tryptophan and ANS quantumyields upon heating suggest that the ther-mal treatment of the samples partly dena-tured the milk proteins. Based on the fluo-rescence data only, it appears more difficult,however, to explain, at a molecular level,the effects of homogenisation on milk com-ponents. In this case, the modification offatglobule size may also perturb light scatteringand induce changes in the fluorescence spec-tra which would not be related to the quan-tum yield of the probes. In order to get moreexplanations on the observed fluorescencemodifications of milk samples followingphysical treatments, tryptophan fluorescenceof washed creams, fat globule size and thekind of pro teins adsorbed at the surface ofthe fat globule were studied.

Figure 2 shows tryptophan emissionspectra of the washed creams of the fourmilk samples. It appeared that NHO dis-played the weakest fluorescence intensity.The proteins associated with the nativemembrane of the fat globule are found inlow amounts (Walstra, 1995). After heating

of raw milk, the fluorescence intensity dou-bled, suggesting that heat treatment inducesbinding of proteins to the fat globules. Asshown in figure 2, the highest fluorescencewas observed for the HOM sample. The dra-matie increase of the fluorescence intensityat about 330 nm indicates that the interfacecreated by the homogenisation of milk isstabilised by proteins.

These assumptions were confirmed bythe electrophoresis study of the washedcreams. Electrophoresis of purified fat glo-bules is a convenient method to characte-rise and quantify proteins adsorbed at theoil/water interface (Sharma and Dalgleish,

0.8.,..----------------.

0.6

:;;~€<IlCQl

.5 0.4Ql /cCQl

1o<Il~0 1:0

iL

/ 10.2/ /

// //

/

\\

-,"'-.r-___

<,-,

"- '<, -....._------.'-.

0.0+------,---,---r------,---..,--'300 320 400340 360 380

Wavelength (nm)

Fig 2. Tryptophan fluorescence spectra of thecreams prepared from NHa, NHP, HOM andHOP milk samples. See Materials and methodsfor details.Spectres de fluorescence des tryptophanes descrèmes provenant des échantillons de lait NHO,NHP, HOM et HOP. Se reporter à « Materialsand methods » pour les détails.

Front-face fluorescence study of milk

Table I.Protein yield bound to the fat globule.Quantité de protéine fixée sur le globule gras.

663

total

Prote in (uglmg of lipids)

f3 + x-caseins a-casein a-lactalbumin f3-lactoglobulin

NHa 0 0 0 0NHP 8.4 5.6 4.7 0HOM 25.0 8.8 6.5 0HOP 48.6 19.3 7.1 6.4

o18.7

40.381.4

NHa, raw milk; NHP, heat treated milk; HOM, homogenised milk; HOP, homogenised + heated milk.

NHa, lait cru; NHP, lait chauffé; HOM, lait homogénéisé; HOP, lait homogénéisé et chauffé.

1993). Data reported in table 1 indicate thatno casein, nor whey proteins, were adsor-bed at the surface of raw-milk fat globule.This is in agreement with the literature.Upon homogenisation, caseins adsorbedpreferentially at the lipid/water interface. Inthis case, bound œ-lactalbumin accountedfor 16% of the total interfacial proteins. Heattreatment also induced the interaction ofproteins with the fat globules. The amount ofbound proteins (per mg of lipids) was twiceas low for heated raw milk than for homo-genised milk. Considering HOP samples, asynergistic effect of technological treatments(homogenisation and then heating) wasobserved (table 1).The data reported in table1are somewhat different from the data foundin the literature (McPherson et al, 1984;Houlilan et al, 1992a, b; Kim and Jimenez-Flores, 1995). The main differences affectthe amounts of ~ + x-caseins and œ-caseinswhich should be about the same. The otherconcern is the lack of ~-Iactoglobulin inNHP and HOM cream samples. But oneshould remark that there is no generalagreement in the literature about the amountof protein bound at the fat globule interface.ln fact, the nature of proteins adsorbed atthe interface depends greatly on the heatingtemperature of the milk (see below). Theelectrophoresis results however show thatmore proteins bind to the fat globule during

homogenisation than heating (table 1). Thedifferences in the amount of bound proteinsas a function of the applied treatment suggestthat the fluorescence spectra of the diffe-rent milk samples should exhibit differentshapes. For example, it is weil known thatthe fluorescence properties (maximumwavelength, quantum yield) of proteinsdepend on their environment (Lakowicz,1983).

There are several reasons that can explainthe adsorption of proteins to the fat globulesurface following the heat treatment. Theymay result from the denaturation of pro-teins, the changes in casein micelle struc-ture or from the higher fluidity of the fatglobule phospholipids at 70 oc. Indeed, thesol-to-gel transition of fat globule phos-pholipids occurs at about 30 "C and the bin-ding of pro teins to phospholipids bilayersis enhanced for lipids in gel phase (Subi-rade et al, 1995). Il is generally claimed thatail the milk proteins, except ~-Iactoglobulin,are not denatured at 70 "C. Il has been repor-ted that, at pH 7, ~-Iactoglobulin dilute solu-tion denatures at 73 "C (Kella and Kinsella,1988), a temperature slightly higher thanthe temperature applied to the samples inthis study. In addition, it is weil known thatthe temperature of ~-Iactoglobulin denatu-ration depends on the protein environment.Anema and McKenna (1996) showed that

664 E Dufour, A Riaublanc

~-lactoglobulin in reconstituted whole milkwas not or slightly denatured at 70 and75°C. On the contrary, it was reported thatœ-lactalbumin is partly denatured at 70 "C(Anema and McKenna, 1996). Whateverthe conditions studied, electrophoresis datashowed, however, that ~-lactoglobulinbound weakly to the fat globule. This proteinwas detected at the lipid/water interface onlyfor HOP sample. It suggests that ~-lacto-globulin is not denatured by most of thetreatments investigated in this study. It isgenerally assumed that native ~-lactoglo-bulin does not adsorb to the interface andthat only denatured ~-lactoglobulin inter-acts with the fat globule membrane (McKen-zie et al, 1972; Sharma and Dalgleish, 1993).It has also been reported that ~-lactoglobu-lin interacted with milk fat globule mem-brane after heating of the milk at 87 "C, butonly slightly after heating at 72 "C (Kimand Jimenez-Flores, 1995).The major pro-teins found at the fat globule surface werecaseins. These flexible and hydrophobieproteins move easily to the lipid/water inter-face upon heating and homogenisation ofthe milk.

The results of the study of fat globulesize are shown in table II. The weight-ave-

Table II. Fat globule properties determined withthe granulometer.Caractéristiques des globules gras déterminéeau moyen d'un granulomètre.

NHO 6.02NHP 5.84HOM 1.3HOP 0.85

2.141.880.640.61

2.83.29.39.85

NHa, raw milk; NHP, heat treated milk; HOM, homo-genised milk; HOP, homogenised + heated milk. * Sp,SA, the total interfacial area per unit volume of the fatcontained in the milk.

NHa, lait cru; NHP. lait chauffé; HOM, lait homogé-néisé; HOP, lait homogénéisé et chauffé. * Sp, SA,surface interfaciale totale par unité de volume de lamatière grasse contenue dans le lait.

rage diameters (d43) of raw and pasteurisedmilks were similar. Homogenised samplesshowed lower d43 ranging between 1.3 and0.85 us«. On the one hand, the decrease ofthe fat globule size induced an increase ofthe interfacial area. The specifie surface areawas three times larger for homogenisedsamples than for raw milk. This interfacecreated by the homogenisation of milk wasstabilised by the adsorption of amphipathicproteins, such as caseins. On the other hand,the change of the fat globule size mightmodify the light scattering and could partlybe responsible for the differences observedin the fluorescence spectra. Nevertheless,the scattering effect alone cannot explainail the modifications reported in the fluo-rescence spectra. Indeed, non-heated andheated samples showed different fluores-cence spectra but similar sizes for the fatglobules.

The characterisation of the creams byfluorescence, electrophoresis and granulo-metry indicated that the size and the surfaceof fat globule, as weil as the structure ofproteins, was modified by homogenisationand heating of milk samples.

Multivariate analysisof rnilk fluorescence spectra

The data show that the fluorescence spectraofproteins, fat-globule vitamin A and ANSare different for NHa, NHP, HOM andHOP samples. The applied treatmentsmodify the environments of the fluorophoresand, consequently, their fluorescence pro-perties (quantum yield, anisotropy and life-time) (Lakowicz, 1983). Despite most ofthe fluorescence studies only focus on emis-sion spectra, it can be interesting to considerexcitation fluorescence spectra too. Princi-pal component analysis, a multivariate tech-nique, was applied to the full emission orexcitation fluorescence spectra data in orderto discriminate between the milk samples.This method is weil suited to optimise the

5

4

3

N 2

'"10<-e

0

·1

·2

·3

·1 -0.5 0 0.5Axe 1

0.5

N

'"10< 0-e

-0.5

-1

·1.5

-0.01

1.5

Front-face 665

(A)

1.5 Fig 3. PCA similaritymap defined by theprincipal components1 and 2 for (A) trypto-phan emission and (H)vitamin A excitationfluorescence spectraldata. Each label cor-responds to a spec-trum. See Materialsand methods fordetails.Carte factorielle /-2de l'analyse en com-posantes principalesréalisée sur lesspectres d' émissiondes tryptophanes (A)et d'excitation de lavitamine A (R). Chaqueétiquette correspond àun spectre. Se reporterà « Materials andmethods » pour lesdétails.

o

Axe 1

description of the data collection with aminimum loss of information.

As suggested earlier, the distinct fat-glo-bule sizes of the sampI es may induce diffe-rences in the fluorescence intensities. Inorder to reduce the scattering effect, thespectra were normalised using equation 1(Bertrand and Scotter, 1992): the area under

-0.005 0.010.005

each spectrum was reduced to a value of 1.In this way, only the shapes of the spectrawere considered in the analyses. PCA wasapplied separately on the three collections of24 normalised spectra corresponding to tryp-tophan emission fluorescence, vitam in Aexcitation fluorescence and ANS excitationfluorescence. For the PCA results, the maps

666 E Dufour, A Riaublanc

defined by principal components 1 and 2for tryptophan emission and vitamin A exci-tation fluorescence data are shown in figure3A and 3B, respectively. The first two prin-cipal components took into account 96.9%(tryptophan data) and 99.3% (vitamin Adata) of the total variability. For the twodata collections, a discrimination of thesamples as a function of homogenisationwas observed according to the principalcomponent 1: non-homogenised milks hadnegative scores, whereas homogenised milksshowed positive scores. For principal com-ponent 2, a discrimination of the samplesas a function of heat treatment was observedfor both data collections: negative scoresare observed for heated samples, whereaspositive scores characterised non-heatedsamples. These results indicate that the treat-

0.25 ;--------------,

0.20

(A)

0.15

0.10

~ 0.05ci

0.00

-0.05

-0.10

-0.15320 340 360 380

Wavelength (nm)

ments applied to the milk induce specifiemodifications in the shape of the fluores-cence spectra and that tryptophan and vita-min A fluorescence spectra allow to discri-minate between NHa, NHP, HOM andHOP samples.

Moreover, the spectral patterns corres-ponding to the principal components pro-vide information about the characteristicpeaks which are the most discriminating forthe samples observed on the maps. The spec-tral patterns corresponding to the principalcomponents 1 and 2 for tryptophan emis-sion and vitamin A excitation spectral dataare given in figures 4 and 5, respectively.In figure 4, the spectral pattern l, associatedwith principal component 1 discriminatingthe tryptophan fluorescence spectra accor-ding to homogenisation, showed an oppo-

0.2 -r-------------.,

0.1

0.0

-0.1

-0.2

-0.3

-0.4 .1.....--.---.-----.----r---.J400 340 380 400360320

Wavelength (nm)

Fig 4. Spectral patterns corresponding to the principal compone nt 1 (A) and 2 (B). Fluorescenceemission spectra of tryptophans.Vecteurs propres 1 (A) et 2 (R) des analyses en composantes principales réalisées sur les donnéesspectrales correspondant aux spectres d'émission de fluorescence des tryptophanes.

Front-face fluorescence study of milk

0.15

0.10

0.05

0.00

:::) -0.05...:

-0.10

-0.15

-0.20

(A)

-0.25 .L.,.__ ....-_-.. __ ..--_-.-_...J320 340260 280 300

667

0.2..---------------,

(B)

0.1

0.0

-0.1

-0.2

-0.3 l...r__ ..--_-.- __ ..--_-.--_...J260 280 300 320 340

Wavelenglh(nm) Wavelength(nm)

Fig 5. Spectral patterns corresponding to the principal component 1 (A) and 2 (B). Fluorescenceexcitation spectra of vitamin A.Vecteurs propres 1 (A) et 2 (B) des analyses en composantes principales réalisées sur les données spec-trales correspondant aux spectres d'excitation de fluorescence de la vitamine A.

sition between a negative peak at 360 nmand a positive one at 320 nm: a blue shiftof the maximum emission of proteins wasinduced by homogenisation. It confirms thatthe environment of the tryptophans of pro-teins adsorbed at the surface of the fat glo-bule becomes more apolar and that the quan-tum yield of tryptophans is increased (fig 1A).The spectral pattern 2 is noisy and more dif-ficult to analyse. ft however exhibits a posi-tive maximum at 333 nm suggesting thatthe spectra of the heated sampl es are some-what larger than the non-heated one. Forthe vitamin A data shown in figure SA, thespectral pattern of the first principal com-po nent presented an opposition between alarge positive band at about 290 nm and anegative peak (and a shoulder) at 322 nm.The positions of these three peaks corres-

pond to the maximum and the two shoul-ders described above for the excitation spec-tra of vitamin A (fig 2B). Spectral pattern 1also indicated that 322 nm/290 nm inten-sity ratios were modified by homogenisa-tion. Spectral pattern 2 was noisy, but it sho-wed two weil defined peaks: a maximum at310 nm and a minimum at 335 nm (fig SB).This opposition between the two bands sug-gests that a red shift of the maximum emis-sion located at about 322 nm occurs in thespectra upon heating of the milk samples.

The examination of the map defined byprincipal component 1 and 2 correspondingto the principal component analysis of ANSnormalised spectra showed no discerniblepattern according to the applied treatments(data not shown). Despite the differencesobserved in ANS fluorescence intensity for

668 E Dufour, A Riaublanc

the four milk samples (fig l C), the shapes ofthe spectra are overall similar since, afternormalisation, it was impossible to discri-minate between the samples according tohomogenisation and heating.

In a second step, the three spectral col-lections (tryptophan emission, retinol exci-tation and ANS excitation fluorescence spec-tra) were gathered together in one matrixand this new table was analysed by PCA.The aim of this approach was to improvethe discrimination of the samples using anumber of different fluorescence spectra.The PCA similarity map defined by princi-pal components 1 and 2 is shown in figure 6.The first two principal components took intoaccount 98.45% of the total variability. Thismap and the maps described above for tryp-tophan and vitamin A fluorescence data sho-wed similar trends: a discrimination of the

0.5

M.-~ 0

-0.5

-1

-1.5

1.5

samples as a function of homogenisationand heating was observed according to theprincipal component 1 and 2, respectively.In addition, the comparison of the maps sho-wed that the four populations, NHa, NHP,HOM and HOP, are better discriminatedusing the large data matrix than consideringtryptophan or vitamin A data alone. The useof ANS data in this study also appeared use-ful (data not shown). Indeed, the additionof ANS data to the matrix made up of tryp-top han and vitamin A data improved thediscrimination of the samples, although PCAanalysis of ANS excitation spectra aloneshowed no discernible pattern according toheating and homogenisation. The resultsdemonstrate that the fluorescence spectrarecorded reflect the physico-chemical cha-racteristics of the milk sample. The analysisof combined fluorescence spectra (trypto-

oAxe 1

0.005 0.01-0.01 -0.005

Fig 6. PCA similarity map defined by the principal components 1 and 2 for the data table includingtryptophan emission, retinol excitation and ANS excitation spectra. Each label corresponds to aspectrum. See Materials and methods for details.Carte factorielle 1-2 de l'analyse en composantes principales réalisée sur le tableau de données conte-nant les spectres d'émission des tryptophanes, d'excitation de la vitamine A et d'excitation de l'ANS.Chaque étiquette correspond à un spectre. Se reporter à « Materials and methods » pour les détails.

Front-face fluorescence study of milk

phan, vitamin A and ANS), recorded on thesame sample, with chemometric methodsshould allow the determination of the phy-sical treatments (homogenisation and/or hea-ting) applied to milk.

Front face fluorescence is a very usefultechnique to record the excitation and emis-sion spectra of powdered, turbid and concen-trated samples. The potential of fluorescencespectroscopy in combination with chemo-metric methods to discriminate betweenmilk samples has been demonstrated. Asearch of the literature showed that thereare very few papers reporting the use offluorescence to characterise food composi-tion or process (Norgaard, 1996; Novaleset al, 1996). Fluorescence, a sensitive andrapid technique, could, however, be usedfor the development of fast at-Iine or on-line analyse methods in the food industry(Marangoni, 1992). The development offluorescence analytical tools implies, firstly,the characterisation of the fluorescence pro-perties of the probes, ie, excitation and emis-sion maximum wavelengths and, secondly,the recording of excitation and emissionspectra. Multivariate statistical methodsapplied to spectra are useful to describevariations between samples and to derivefrom the spectral patterns the characteristicwavelengths of the samples. In general,spectrometrie methods and, in particular,front face fluorescence, in combination withmultivariate statistical analyses, have a hugepotential in the development of fast ana-lyses and in the quality control applicationsto food systems.

ACKNOWLEDGMENTS

We are grateful to Dr Bertrand (LTAN-Inra,Nantes) for the PCA program and to Dr Ollivon(CNRS, Chatenay-Malabry) for valu able dis-cussions.

669

REFERENCES

Anema SG, McKenna AB (1996) Reaction kinetics ofthermal denaturation of whey proteins in heatedreconstituted whole milk. J Agric Food Chem 44,422-428

Bertrand D, Scotter CNG (1992) Application of mul-tivariate analyses to NIR spectra of gelatinizedstarch. Appl Spectrosc 46,1420-1425

Bertrand D, Robert P, Launay D, Devaux MF (1987)Application of principal component analysis to thedetermination of fat and protein content of milk bynear-infrared spectroscopy. Rapid Analysis in FoodProcessing and Food Control. Proc Euro FoodChem IV. 1-4 June, Loen, Norway

Blumberg WE, Doleiden FH, Lamola AA (1980)Hemoglobin determined in 15 il L of whole blood by'front face' fluorometry. Clin Chem 26,409-413

Castelain C, Genot C (1994) Conformational changesof bovine serum albumin upon adsorption in dode-cane-in-water emulsions as revealed by front-facesteady-state fluorescence. Biochim Biophys Acta1199,59-64

Dalgalarrondo M, Dufour E, Harb C, Chobert JM,Haertlé T (1992) Purification and characterizationof two porcine ~-Iactoglobulin variants by NaCIsalting-out and reversed phase-HPLC. Lait 72,35-42

Dufour E, Haertlé T (1990) Alcohol-induced changesof ~-lactoglobulin/retinol binding stoichiometry.Protein Eng 4, 185-190

Dufour E, Genot C, Haertlé T (1994) ~-Lactoglobu-lin binding properties during its folding changesstudied by fluorescence spectroscopy. Biochim Bio-phys Acta 1205, 105-1 12

Fox PF (1989) The milk protein system. In: Develop-ment in dairy chemistry. 4. Functional milk pro-teins (Fox PF, ed), Elsevier, New York, 1-54

Genot C, Tonetti F, Montenay-Garestier T, Drapron R(l992a) Front-face fluorescence applied to struc-turai studies of proteins and lipid-protein interac-tions of visco-elastic food products. 1. Designing offront-face adaptor and validity of front-face fluo-rescence measurements. Sei Alim 12, 199-212

Genot C, Tonetti F, Montenay-Garestier T, Marion D,Drapron R (1992b) Front-face fluorescence appliedto structural studies of proteins and lipid-proteininteractions of visco-elastic food products. 2. Appli-cation to wheat gluten. Sei Alim 12,687-704

Hartman AM, Dryden LP (1978) The vitamins in milkand milk products. In: Fundamentals of dairy che-mistry (Webb BH, Johnson AH, Alford JA, eds),The Avi Publishing Company Inc, Westport, CT,USA,325-401

Hirsch RE, Nagel RL (1989) Stopped flow front-facefluorometer: a prototype design to measure hemo-globin R -> T transition kinetics. Anal Biochem176,19-21

670 E Dufour, A Riaublanc

Houlilan AV, Goddard PA, Nottingham SM, KitchenBJ, Masters CJ (l992a) Interactions between thebovine milk fat globule membrane and skim milkcomponents on heating whole milk. J Dairy Res59, 187-195

Houlilan AV, Goddard PA, Kitchen BJ, Masters CJ(l992b) Changes in structure of the bovine milk fatglobule membrane on heating whole milk. J DairyRes 59, 321-329

Kella NKD, Kinsella JE (1988) Enhanced thermody-namic stability of ~-Iactoglobulin at low pH: a pos-sible mechanism. Biochem J 255, 113-118

Kim HHY, Jimenez-Flores R (1995) Heat-inducedinteractions between the proteins of milk fat globulemembrane and skim milk. J Dairy Sei 78, 24-35

Jolliffe IT (1986) Principal component analysis. Sprin-ger, New York

Laemmli UK (1970) Cleavage of structural proteinsduring the assembly of the head of bacteriophageT4. Nature 227,680-685

Lakowicz JR (1983) Protein fluorescence. In: Prin-ciples of fluorescence spectroscopy (Lakowicz JR,ed), Plenum Press, New York, 341-389

Lindberg W, Persson JA, Wold S (1983) Partial least-squares method for spectrofluorometric analysis ofmixtures of humic acid and lignin sulfonate. AnalChem 55,643-646

Longworth JW (1971) Luminescence of polypeptidesand proteins. In: Excited states of proteins andnucleic acids (Steiner RF, Weinryb l, eds) Mac-Millan Press Ltd, London, 319-483

Marangoni AG (1992) Steady-state fluorescence pola-rization spectroscopy as a tool to determine micro-viscosity and structural order in food systems. FoodRes [nt 25,67-80

Matarella NL, Richarson T (1983) Physico-chemicaland functional properties of positively charged deri-vatives of ~-lactoglobulin. J Agric Food Chem 31,972-978

McKenzie GH, Ralston GB, Shaw DC (1972) Loca-tion of sulfhydryl and disulfide groups in bovine~-Iactoglobulin and the effects of urea. Bioche-mistry 11,4349-4357

McPherson AV, Dash MC, Kitchen BJ (1984) Isolationand composition of milk fat globule membranematerial. II. From homogenised and ultra heat trea-ted milks. J Dairy Sei 51,289-297

Norgaard L (1996) Classification and prediction ofquality and process parameters of beet sugar andthick juice by fluorescence spectroscopy and che-mometrics. Zuckerindustrie (in press)

Novales B, Bertrand D, Devaux MF, Robert P, Sire A(1996) Multispectral fluorescence imaging for theidentification of food products. J Sei Food Agric71,376-382

Parker CA (1968) Apparatus and experimentalmethods. In: Photoluminescence of solutions withapplications to photochemistry and analytical che-mistry (Parker CA, ed), Elsevier, Amsterdam,128-302

Patton S, Huston GE (1986) A method for isolation ofmilk fat globules. Lipids 21,170-174

Sharma SK, Dalgleish DG (1993) Interactions betweenmilk serum proteins and synthetic fat globule mem-brane during heating of homogenized whole milk.J Agric Food Chem 41,1407-1412

Subirade M, Salesse C, Marion D, Pézolet M (1995)Interaction of a nonspecific wheat lipid transferprotein with phospholipid monolayers imaged byfluorescence microscopy and studied by infraredspectroscopy. Biophys J 69,974-988

Walstra, P (1995) Physical chemistry of milk fat glo-bule. In: Advanced dairy chemistry, 2. Lipids (FoxPF, ed), Elsevier, New York, 131-178