J.Dairy Sci.86:1101–1107
©American Dairy Science Association,2003.
紧逼防守Application of Fluorescence Spectroscopy and Chemometrics in the Evaluation of Procesd Chee During Storage
J.Christenn*,V.T.Povln*,and J.Sørenn†
*Food Technology,Department of Dairy and Food Science,The Royal Veterinary and Agricultural University,
Rolighedsvej 30,DK-1958Frederiksberg C,Denmark
†Arla Foods Innovation,Søderupvej 26,DK-6920Videbæk,Denmark
ABSTRACT
Front face fluorescence spectroscopy is applied for an evaluation of the stability of procesd chee during storage.Fluorescence landscapes with excitation from 240to 360nm and emission in the range o
f 275to 475nm were obtained from chee samples stored in darkness and light in up to 259d,at 5,20and 37°C,respectively.Parallel factor (PARAFAC)analysis of the fluorescence landscapes exhibits four fluorophores pres-ent in the chee,all related to the storage conditions.The chemometric analysis resolves the fluorescence sig-nal into excitation and emission profiles of the pure fluorescent compounds,which are suggested to be tryp-tophan,vitamin A and a compound derived from oxida-tion.Thus,it is concluded that fluorescence spectro-scopy in combination with chemometrics has a potential as a fast method for monitoring the stability of pro-cesd chee.
(Key words:chee,chemometrics,fluorescence spec-troscopy,PARAFAC)
Abbreviation key:GC-MS =gas chromatography-mass spectrometry,PARAFAC =parallel factor analysis.
INTRODUCTION
The development of undesirable flavor caud by lipid oxidation and nonenzymatic browning are critical qual-ity factors during storage of procesd chee.The dete-rioration of the chee product is dependent on the han-dling in the post manufacturing process.Since chee mainly consists of protein,fat,minerals and water,oxi-dation is reflected in the composition of the constit-uents.Monito
ring the changes in structure and compo-sition of the chee constituents,especially protein and fat,will help understand the effect of stress factors
Received September 2,2002.Accepted October 25,2003.
Corresponding author:Jakob Christenn;e-mail:jach@kvl.dk.
1101
during storage.Common stress factors in the distribu-tion retails and production are light exposure and vary-ing temperature,which can result in reduced shelf life partly due to incread formation of free radicals.Therefore,procesd chee samples stored under dif-ferent light and heat conditions are investigated in the prent study.
Many methods have been developed to shed light on the degree of oxidation of dairy products,a process that consists of veral stages.The early stage of lipid oxida-tion can form hydroperoxides,which normally are mea-sured by HPLC or by evaluation of the peroxide value (Emmons et al.,1986).Secondary oxidation products can be analyzed by static or dynamic headspace GC-MS (Sunen et al.,2002)or methods using thiobarbi-turic acid (Kristenn et al.,200
1).Methods bad on electron spin resonance spectrometry were recently suggested for monitoring the formation of radicals dur-ing the oxidation of procesd chee (Kristenn and Skibsted,1999).All the methods for evaluation of the oxidative levels of dairy products have in common,that they are destructive and time consuming.In this study,the potential of front face fluorescence,measured di-rectly on the chee surface were investigated,as an alternative,fast and nondestructive method.Theoreti-cally the potential of fluorescence ems sound,since the chee product contains well known fluorescent compounds in form of aromatic amino acids,vitamin A and riboflavin (Duggan et al.,1957),which all have been reported to be affected during structural changes in chee (Dufour et al.,2001)or during light and heat exposure (Kristenn et al.,2001;Whited et al.,2002;Wold et al.,2002).孤儿英语
Fluorescence spectroscopy is a nsitive,rapid and noninvasive analytical technique that can provide in-formation on the prence of fluorescent molecules and their environment in all sorts of biological samples.The development and improvement of chemometric meth-ods (Bro,1996;Bro,1997;Andersson and Bro,2000)combined with the technical and optical development of spectrofluorometers have in recent years incread the possibilities for the u of fluorescence spectroscopy.
CHRISTENSEN ET AL. 1102
Thus,online monitoring nsors that enable measure-ments of complete excitation emission spectra(fluores-cence landscapes)are now commercially available.
In the last years,a few studies have focud on the potential of using front facefluorescence of dairy prod-ucts without any pretreatment of the samples.Pre-viously heat treatment and structural changes during coagulation have successfully been investigated in milk usingfluorescence spectroscopy(Dufour and Riaublanc, 1997;Birlouez-Aragon et al.,1998;Herbert et al.,1999). Changes in fat and protein composition and structure have been characterized by the means of measuring the tryptophan and vitamin Afluorescence of chees during ripening(Dufour et al.,2000;Mazerolles et al., 2001)and for identification of different chees at a molecular level(Dufour et al.,2000;Herbert et al., 2000).Wold et al.(2002)demonstrated the potential offluorescence spectroscopy for measuring the light-induced oxidation,ascribed to the photodegradation of riboflavin.
Common to all the studies is that basic chemome-tric tools like Principal Component Analysis and Partial Least Squares Regression are applied for the evaluation of single excitation or emissionfluorescence spectra. The multivariate approach increas the extracted in-formation and is
very uful when handling thefluo-rescence signal of complex food products.Even more information can be obtained,if thefluorescence mea-surements are not limited to single emission or excita-tion spectra.The possibilities when measuring whole fluorescence landscapes(excitation emission matrices) will be investigated here.New chemometric methods (Andersson and Bro,2000)make it possible to handle fluorescence landscapes keeping the2-dimensional data structure of each measurement.The techniques are known as N-way or multiway chemometrics,and in the ca offluorescence signals,a3-way(samples×excitation×emission)data analysis is an obvious choice.The advantage of the multiway analysis is that one can utilize the original and true structure in data, which can stabilize the decomposition of the data,and potentially increa the interpretability(Bro,1996; Bro,1997).
In the prent study Parallel Factor analysis(PARA-FAC)(Bro,1997)is applied on thefluorescence land-scapes of procesd chee expod to light and varying temperature during storage.PARAFAC analysis of fluorescence data is previously ud with success on model system of mixtures offluorophores and in other food applications like sugar andfish(Bro,1999;Bauns-gaard et al.,2000a;Baunsgaard et al.,2000b;Pedern et al.,2002)to investigate the prentfluorescent com-pounds in complex matrices.PARAFAC is bad on the decomposition of thefluorescence data repre
nted in a Journal of Dairy Science Vol.86,No.4,2003three-way array,into a few spectral loadings expressing the common structure of the data.The feature of PARA-FAC is that the retrieved loading spectra can be directly related to the originalfluorescence characteristics of the prentfluorophores,which means that the emis-sion and excitation maximum of the loadings can be ud in the interpretation and identification of theflu-orophores(Bro,1997).
Thus,the overall objective of the prent investiga-tion is to u multivariate analysis onfluorescence spec-tra keeping the3-dimensional structure and extract information about the product at hand regarding age and storage conditions.This is pursued by using a non-destructive and rapid high-nsitivefluorescence method,which is simple to perform,and does not in-volve sample preparation.
MATERIALS AND METHODS
Procesd Chee:Product and Storage Conditions The product and storage conditions are identical to the experimental plan ud by Kristenn et al.,2001.
A batch of procesd chee spread samples(density approximately1.1g/mL)with65%fat in dry matter was obtained from Arla Foods amba,Denmark.The procesd chee was produced according to standard production of procesd chee and was constituted of bovine milk,starter cult
ure,salt and emulsifier.After production the product wasfilled without any head-space(140g)in transparent glass containers and aled with a metal lid.The samples were stored for10months at three temperatures5,20,and37°C and were expod by placing the samples at a distance of approx.55cm from afluorescent lamp or protected from light by wrap-ping the glass container in tin foil.The light source was fluorescent tubes(Phillips TLD18/83W)with a light intensity of2000lx as measured by a Topcon IM-1 illumination meter(Tokyo Kogaku Kikai K.K.).Sam-ples were taken out at the beginning of the experiment and then after14,28,56,84,112and256d.Only the 1cm outer layer which had been in contact with the wall of the containers were ud and each of the samples were taken from the glass jars by breaking the original al prior to freezing at−80°C.The samples were frozen for a year before being thawed.Two chee samples from each treatment were withdrawn for each analy-sis time.
Fluorescence Spectroscopy and Sampling
All samples were measured on a Perkin-Elmer LS 50B spectrometer equipped with a Front Surface Acces-sory and controlled with FLDM software.The stored chee samples were mixed thoroughly before spread-
FLUORESCENCE AND CHEMOMETRICS
1103
你知道不知道
Figure 1.Illustration of the decomposition scheme into f number of components of the PARAFAC model for the data array X .The cube E reprents the residual.
澎湃的拼音怎么写
ing directly onto the quartz window of a powder cell,which was then asmbled and placed in the light path in an angle of around 60°.The spectral range of the experiment was lected upon an exploratory basis.A preliminary investigation measuring excitation wave-lengths from 200to 600nm,and emission wavelengths from 220to 800nm on different chees were per-formed,and resulted in focusing on excitation wave-lengths in the UV region.Strong fluorescence signals were obtained from the chee samples in this area,leaving no signal from higher excitation and emission wavelengths when using this technique and t-up.The lected spectral range of the excitation wav火龙果的营养价值
elength was 240to 360nm with 20nm intervals.Emission was obtained for every nm from 275to 475nm.The slit width was 6nm for excitation and 5nm for the emission and a 1%attenuation filter was ud.
It should be noted that the lected spectral range does not cover riboflavin fluorescence,which exhibit emission around 520nm (Duggan et al.,1957),despite it would be an obvious compound to monitor throughout storage.However,the preliminary studies on chee samples showed that no detectable signal was obtained in this spectral area when using the described measur-ing t-up.
Data Analysis—PARAFAC
PARAFAC decompos the fluorescence spectra,into tri-linear components according to the number of fluor-ophores prent the chee samples (objects).The num-ber of fluorophores prent in the samples is equal to the minimal number of factors (f =1,…,F )needed to describe the fluorescence matrix X .
A graphical illustration of the decomposition of the data array X is given in Figure 1.The object mode is expresd by the A-scores (a 1,…,a f )and the two spec-tral loadings excitation and emission are expresd as
B loadings (b 1,…,b f )and
C loadings (c 1,…,c f ),respec-tively.The loadings in a spectral bilinear decomposition reflect the pure spectra of the fluorophores and the true underlying spectra can be recovered in the single com-ponents.
Journal of Dairy Science Vol.86,No.4,2003
The principle behind the PARAFAC decomposition is to minimize the sum of squares of the residual e ijk ,e Equation 1.
x ijk =
∑F
f =1
a if
b jf
c kf +e ijk
[1]
(i =1,…,I ;j =1,…,J ;k =1,…,K ;f =1,…,F )The element x ijk reprents the raw fluorescence excita-tion/emission spectra (X )of the stored chee,where i is the number of measured samples,j is the number of excitation wavelengths,k is the number of emission wavelengths and f is the number of factors.a ؒf is the object score (magnitude of the fluorophore)for factor f (first mode),b ؒf is the excitation loading for factor f (cond mode),and loading c ؒf express the emission spectra (third mode).e ijk is the residual (E )and contains the variation not captured by the PARAFAC model (Bro,1997).Split half analysis is suggested for valida-tion of PARAFAC models by Bro (1997).The idea of this strategy is to divide the data t into two halves and make a PARAFAC model on both halves.Due to the uniqueness of the PARAFAC model one will obtain the same result—same loadings in the nonsplitted ,excitation and emission mode—on both data-ts,if the correct number of components is chon.Calculating the PARAFAC Model
The following sampling was performed:45samples ×2replicates ×2repetitions =190samples.Seven samples were removed,as they were considered to be spectral outliers bad on a preliminary data i
nspection and resulted in a total of 183samples.The preliminary PARAFAC modelling indicated that nonnegativity con-straints on all three modes (samples,excitation,and emission)were necessary.Validation of the PARAFAC modelling was performed with split half test,bad on replicated splitting of the repetitions.In addition to the split-half experiment,the residuals were inspected,and the results were judged,interpre-ted and compared with external knowledge.
All calculations were performed in Matlab version 6.1(MathWorks,Inc.)with the N-way Toolbox (Andersson and Bro,2000)and the PLS Toolbox ().
RESULTS AND DISCUSSIONS
The fluorescence landscapes of two chee samples are shown in Figure 2.The two samples reprent the extremes in the experimental ,a fresh chee sample (a)and a chee sample stored under the most vere conditions (b).The highest fluorescence peak for
自兹
CHRISTENSEN ET AL.
1104
Figure 2.Three dimensional plot of fluorescence landscapes of procesd chee samples.a)fresh chee sample,and b)chee sample stored in 259d at 37°C expod to light.
both samples is en with excitation around 280nm and emission around 350nm,with a significant higher and apparently broader signal from the fresh chee.The excitation and emission characteristics indicate that the fluorescence peak corresponds to tryptophan fluorescence,which is reported to have excitation/emis-sion wavelength maximum at 285/365nm in pure solu-tions (Dugga
n et al.,1957),and previously measured in chee products with excitation 290nm and emission from 305–400nm (Herbert et al.,2000;Dufour et al.,2001;Mazerolles et al.,2001).Apart from this major peak,a vague peak is obrved in the higher wavelength region with excitation around 320to 360nm and emis-sion round 400to 460nm,especially for the chee sample stored for 259d.
The aforementioned patterns in the fluorescence landscapes were investigated further by the u of PAR-AFAC analysis with the objective to resolve the fluores-cence signal into the contributions of each of the fluo-rescent compounds prent in the t of stimate the excitation and emission profiles of fluoro-phores directly from the three-dimensional fluorescence landscapes.PARAFAC models of the fluorescence data were estimated with one to five components,but the four-component model was chon bad on split half analysis (Bro,1997).A high explained variation of 99.76%is captured by the PARAFAC model,and the resulting PARAFAC components are shown in Figure 3.The model indicates that four different fluorophores are prent in the chee samples with the excitation and emission profiles shown in the figure.The excita-Journal of Dairy Science Vol.86,No.4,2003
tion/emission maximum for the two compounds are 300/347nm and 280/339nm,respectively,as listed in Table 1.The loading profiles of the cond PARAFAC compo-nent corresponds quite well with the
characteristics of tryptophane,whereas the excitation maximum of the first component ems a little too high for tryptophan.Having the rather low resolution of 20nm in the excita-tion mode in mind,and knowing that the fluorescence properties of protein-bound amino-acids are known to be affected by the structure of protein (Lakowicz,1999),we dare to suggest that the first PARAFAC components is also due to tryptophan fluorescence,but simply shifted due to inclusion to different protein structures.The score values in the first column of Figure 3repre-nt the concentration mode for each of the fluoro-phores,and since the excitation and emission loadings are normalized when calculating the PARAFAC model,the contribution for each of the components can be com-pared to the overall variation bad on the level of the scores.The score values are arranged so the develop-ment of the fluorophores easily can be caught through-out the storage time.Looking at the two propod tryp-tophan components,a significant decrea is obrved throughout the storage period for the samples stored at 37°C.This shows that alterations in the protein structure,monitored by the decrea in tryptophan fluorescence,somehow can reflect the conditions of the chee samples during storage.The samples expod to light during storage show a systematically higher tendency to be degraded throughout the storage than the samples stored in the dark.Compared with the
FLUORESCENCE AND CHEMOMETRICS
1105
物业项目经理岗位职责Figure 3.A-(scores),B-(excitation),and C-(emission)loadings of a four component PARAFAC model,bad on the fluorescence landscapes of 183procesd chee samples.Samples stored at 5°C are indicated with triangles,and connected with dotted line (......).Samples stored at 20and 37°C are shown with squares and dashed line (–––)and circles connected with a full line (——),respectively.Open signs reprent samples stored in light,and filled sign illustrates storage in darkness.温室甲鱼养殖
effect of different temperatures,the light exposure ems negligible for the two tryptophan components,though.The same storage experiment showed a similar tendency of light exposure having little,if any influence on the browning of chee (Kristenn et al.,2001),and thereby indicate that the obrved differences in the protein structure are somehow related to the browning he formation of Maillard products from the protein and lipid oxidation products in chee,even though tryptophan itlf may not be part of the brown-ing reaction scheme.As indicated by the first visual inspection of the fluorescence landscapes,the level of the score values for the two first components are much Journal of Dairy Science Vol.86,No.4,2003
higher than the third and fourth component,simply showing that the development in the tryptophan signal reprents the major variation in the fluorescence data.The development of the third estimate
d fluorophore (score values of the third PARAFAC component)shows a similar pattern as the decrea in the tryptophan signal.Thus,the chee samples stored at 37°C contain less of this component throughout the storage,espe-cially the samples expod to light during storage.Com-paring the fluorescence profiles en in Figure 3and the excitation/emission wavelength maximum of 320/411nm (Table 1)with obrved maximum of 325/470nm in pure solution reported (Duggan et al.,1957)and
