Study of chemical changes in pasteurised orange juice during shelf-life: A fingerprinting-kinetics evaluation of the volatile fraction

Food Research International 75 (2015) 295–304

Contents lists available at ScienceDirect

Food Research International journal homepage: www.elsevier.com/locate/foodres

Study of chemical changes in pasteurised orange juice during shelf-life: A fingerprinting-kinetics evaluation of the volatile fraction Scheling Wibowo, Tara Grauwet, Biniam Tamiru Kebede, Marc Hendrickx, Ann Van Loey ⁎ Laboratory of Food Technology, Leuven Food Science and Nutrition Research Center (LFoRCe), Department of Microbial and Molecular Systems (M2S), Katholieke Universiteit Leuven, Kasteelpark Arenberg 22 Box 2457, 3001 Heverlee, Belgium

a r t i c l e

i n f o

Article history: Received 27 March 2015 Received in revised form 12 June 2015 Accepted 17 June 2015 Available online 23 June 2015 Keywords: Orange juice Fingerprinting Kinetics Accelerated shelf-life testing Volatiles

a b s t r a c t The current work used fingerprinting-kinetics for the first time to monitor shelf-life changes in a low-pH, pasteurised, shelf-stable product, more particular in orange juice. Orange juice samples were stored as a function of time at four different storage temperatures (20, 28, 35 and 42 °C). To obtain insight into chemical changes in the volatile food fraction, samples were fingerprinted with headspace GC–MS. The objectives of this work were twofold: (i) to identify major chemical changes of pasteurised orange juice during shelf-life and (ii) to study the kinetics of selected shelf-life compounds in the context of accelerated shelf-life testing (ASLT). At 20 °C, changes in terpenes and a decrease in aldehydes were observed. Oxides and sulphur compounds increased and esters decreased at increased storage temperatures (at 28 °C and above). Concerning ASLT, four volatile compounds had clear temperature and time dependent kinetics within the investigated temperature range. © 2015 Elsevier Ltd. All rights reserved.

1. Introduction The commercial importance of citrus fruits has been recognised for many decades. The global citrus fruit production showed an increasing trend, with the main leading producers being Brazil, China and the United States. Among citrus juices, orange juice is the most widely consumed. According to the United States Department of Agriculture (USDA), its global consumption reached nearly two million metric tons in 2014, which was mainly driven by the United States and the EU (USDA, 2014). Various processing techniques are developed to achieve shelf-stable orange juice: one of the most commonly used methods in food industry being pasteurisation by heat. Nevertheless, orange juice quality changes over time, either during processing or storage. When orange juice is processed with the aim of obtaining a shelfstable product, its product shelf-life is determined by a best-before date, which is limited due to physical and (bio) chemical reactions. The number and type of degradation reactions and their interactions make the quantitative study of quality degradation of orange juice during processing and storage highly complex.

⁎ Corresponding author. E-mail address: [email protected] (A. Van Loey). URL: https://www.biw.kuleuven.be/m2s/clmt/lmt/ (A. Van Loey).

http://dx.doi.org/10.1016/j.foodres.2015.06.020 0963-9969/© 2015 Elsevier Ltd. All rights reserved.

Numerous studies have been aimed at understanding quality changes of orange juice during storage (Bacigalupi et al., 2013; Berlinet, Brat, Brillouet, & Ducruet, 2006; Roig, Bello, Rivera, & Kennedy, 1999; Esteve, Frigola, Rodrigo, & Rodrigo, 2005). In order to obtain insight into colour instability of pasteurised orange juice, our previous works investigated changes in a wide range of a priori selected quality parameters (e.g., acids, sugars, oxygen, vitamin C, furfural and HMF) linked to colour changes as a function of storage time and temperature. It was demonstrated that due to limited contribution of carotenoid degradation reactions, other mechanisms play important roles in browning of orange juice. Furthermore, several quality parameters were selected as potential markers for colour changes: ascorbic acid, sugars and their degradation products, furfural and HMF (Wibowo et al., 2015b; Wibowo et al., 2015a). This strategy is known as a targeted analytical approach, in which specific chemical compounds (e.g., ascorbic acid) or quality attributes (e.g., colour) are focused on (Grauwet, Vervoort, Colle, Van Loey, & Hendrickx, 2014). However, investigating responses related to a particular chemical reaction or characteristic selected at the starting point of the study could result in overlooking unexpected effects. Over the past years, there has been a growing interest in applying untargeted fingerprinting as a methodological approach to gain insight into the reaction complexity of a food system (Picariello, Mamone, Addeo, & Ferranti, 2012). Fingerprinting is defined as a more unbiased and hypothesis-free methodology that considers as many compounds as possible in a particular food fraction (liquid or headspace) compared

296

S. Wibowo et al. / Food Research International 75 (2015) 295–304

to the commonly used targeted approaches. Without fixating on a specifically known compound, it allows an initial fast screening to detect differences among samples. This approach has been suggested to investigate chemical reactions which are influenced by either processing or shelf-life (Grauwet et al., 2014; Vervoort et al., 2012). Advanced analytical methods such as separation techniques by liquid chromatography (LC), gas chromatography (GC) coupled to a mass spectrometry (MS) or nuclear magnetic resonance (NMR)-based detector for compound identification are indispensable analytical tools used in fingerprinting research (Cevallos-Cevallos, Reyes-De-Corcuera, Etxeberria, Danyluk, & Rodrick, 2009; Wishart, 2008). Unfortunately, today, it is not possible to fingerprint the whole chemical composition of a food matrix using one particular analytical method. Therefore, from an analytical point of view, untargeted analysis can be quite challenging because of the diversity of the compounds studied. The use of GC–MS is particularly advantageous, since it is a robust method which allows preliminary compound identification by the availability of mass spectral libraries. Volatile compounds such as aldehydes, esters, terpenes, alcohols, and ketones are often linked to a range of process- and storage-induced chemical reactions (Kebede et al., 2014; Mahattanatawee, Rouseff, Valim, & Naim, 2004; Perez-Cacho & Rouseff, 2008a). Being regularly degradation products of major food components such as sugar, fat, and nutrients, volatiles can be approached not only for understanding chemical changes in the volatile fraction itself, but also as a witness for what is happening in other food fractions (e.g., liquid fraction). In addition, changes in volatile compounds contribute to the perceived aroma and are important for the overall sensorial quality of food products. Applying headspace GC–MS fingerprinting of the volatile fraction as an untargeted method to obtain insight in chemical changes induced by processing has been a subject of study in our research group (Vervoort et al., 2012; Aganovic et al., 2014; Kebede et al., 2014). Recently, our first work was published on the use of fingerprinting integrated with kinetics in the context of the evaluation of quality changes of a sterilised vegetable-based puree as a function of (accelerated) storage (Kebede et al., 2015). The current work will apply this untargeted analytical and engineering toolbox of fingerprinting-kinetics for the first time for monitoring shelf-life changes in a low-pH, pasteurised, shelf-stable product, more particular in orange juice. The objective of this work was to identify major chemical changes of pasteurised orange juice during shelf-life. Orange juice was not only stored at classical ambient storage temperatures but also under increased temperature conditions in order to study the kinetics of selected shelf-life markers in the context of accelerated shelf-life testing (ASLT). 2. Materials and methods 2.1. Sample processing and shelf-life study Single strength orange juice (11.2 corrected °Brix, pH 3.7, titratable acidity 0.8%) was prepared by reconstituting frozen Brazilian orange concentrate (Citrus sinensis (L.) Osbeck) (65 °Brix) with water (1:5, w/w). The 200 L of orange juice mixture was pasteurised at 92 °C for 30 s. Subsequently, hot filling at 85 °C into 500 mL polyethylene terephthalate (PET) bottles, sealing by cap twist inversion and submerging bottles into a water tank to reach ambient temperature were applied. A shelf-life study was conducted at 20 and 28 °C for a total of 32 weeks, at 35 °C for 12 weeks and at 42 °C for 8 weeks. 75 bottles were stored per each incubator (IPP500, Memmert, Schwabach, Germany) protected from light at a specific storage temperature. These orange juice bottles were also used for characterisation in colour, pH, titratable acidity, organic acid profile, °Brix, sugar profile, oxygen, vitamin C, furfural, 5-hydroxymethylfurfural (HMF) and carotenoids (Wibowo et al., 2015b; Wibowo et al., 2015a). At a particular sampling moment, three bottles were randomly taken per storage temperature. Juices were uniformly mixed and divided

into smaller tubes (± 30 mL), frozen in liquid nitrogen and stored at − 80 °C until further analysis. 2.2. Headspace GC–MS analysis Frozen tubes were thawed in a circulating water bath at 25 °C. After thawing, juice inside the tube was homogenised using a vortex mixer. 3 mL juice and 2 mL saturated NaCl solution were pipetted into an amber glass vial (10 mL, VWR International, Radnor, PA, USA). The vials were tightly closed using screw-caps with silicone septum seal (GRACE, Columbia, MD, USA), homogenised and placed in the cooling tray of the autosampler which was maintained at 10 °C. All headspace analyses were conducted on a gas chromatography (GC) system (7890 N, Agilent technologies, Diegem, Belgium) coupled to a mass selective detector (MSD) (5977 N, Agilent Technologies, Diegem, Belgium) and equipped with a CombiPAL autosampler (CTC analytics, Zwingen, Switzerland). Aiming at detection of as many volatiles as possible, an HS–SPME–GC–MS method (e.g., SPME fibre type and GC and MS parameters, etc.) was optimised beforehand. With the aim of selecting a fibre that enables the extraction of a wide range of volatilisable compounds, a comparison between five types of SPME fibre coatings (PDMS, CAR/PDMS, PDMS/DVB, PA, and DVB/CAR/ PDMS) was performed. Based on the observations of the total peak area and the number of peaks (a high area and a high number), a fibre with carboxen/polydimethylsiloxane (CAR/PDMS) coating was chosen for the headspace SPME analysis (data not shown). Samples were equilibrated for 20 min at 30 °C under agitation at 500 rpm. Volatiles were extracted using a SPME fibre coated with 85 μm CAR/PDMS (StableFlex, Supelco, Bellefonte, PA, USA) at 30 °C for 10 min. Each fibre was conditioned according to the manufacturer's guidelines before its first use. After the extraction step, the fibre was inserted into the heated GC injection port (230 °C) for 2 min where the volatile compounds were desorbed. Subsequently, the fibre was thermally cleaned for 5 min at 300 °C in the conditioning station of the autosampler. Injection of the samples to the GC column was performed in split-mode with a splitratio of 1/10. The separation was carried out on a HP-5MS capillary column (30 m × 0.25 mm i.d., 0.25 μm film thickness, Agilent Technologies, Santa Clara, CA, USA) using helium as carrier gas at a constant flow of 1.5 mL min− 1. The oven temperature was programmed at an initial temperature of 40 °C, which was maintained for 2 min, after which it was elevated to 160 °C at a rate of 4 °C min−1, then ramped to 300 °C at 20 °C min− 1 and kept constant for 2 min at 300 °C before cooling back to 40 °C. The mass spectra were obtained by electron ionisation (EI mode) at 70 eV with a scanning range of 35 to 400 m/z. The ion source and quadrupole temperatures were 230 and 150 °C, respectively. A new fibre was used for each storage temperature. During analysis, the samples were randomised as a function of storage time per storage temperature. Possible fibre degradation was monitored by analysis of a reference sample (pasteurised orange juice), every 14 injections. The peak area of the reference samples was investigated as a function of time to monitor the performance of the SPME fibre and the GC–MS. In this work, deviation of the peak area of the reference samples was less than 5% during the analysis, which shows limited fibre degradation and a good performance of the instrument (results not shown). 2.3. Data analysis 2.3.1. Data pre-processing and multivariate analysis GC–MS has long been the method of choice for identifying volatile compounds in complex mixtures. This method can fail, however, when acquired spectra are “contaminated” with extraneous mass spectral peaks, as commonly arise from co-eluting compounds and ionisation chamber contaminants. These extraneous peaks can pose a serious problem for automated identification methods where they can cause identifications to be missed by reducing the spectrum comparison factor below some pre-set identification threshold. Automated Mass

S. Wibowo et al. / Food Research International 75 (2015) 295–304

Spectral Deconvolution and Identification System (AMDIS) is an integrated set of procedures for first extracting pure component spectra and related information from complex chromatograms and then using this information to determine whether the component can be identified as one of the compounds represented in a spectral reference library. The practical purpose is to reduce the effort involved in identifying compounds by GC–MS while maintaining the high level of reliability associated with traditional analysis. The programme first deconvolutes the GC–MS data file to find all of the separate components. Furthermore, the programme can be configured to build a retention index calibration file, to use the retention index data along with the mass spectral data. All chromatograms were analysed with AMDIS (Version 2.66, 2008, National Institute of Standards and Technology, Gaithersburg, MD, USA). The deconvoluted spectra were then analysed with Mass Profiler Professional (MPP) (Version 12.0, 2012, Agilent Technologies, Diegem, Belgium) for filtering and peak alignment. In MPP, there are several filtering options that may be applied: e.g., filter by frequency, abundance. The parameters in which the entities are filtered need to be optimised depending on the data under consideration. Samples are aligned or grouped together if their retention times are within the specified tolerance window and the mass spectral similarity. In cases, where retention time locking is applied, a smaller retention time tolerance window can be used. The potential of the integrated data pre-processing steps, AMDIS and MPP can be described as: (i) improved identification due to the removal of impurities and extraction of pure spectrums; (ii) automated, standardised and much more reproducible data analysis; (iii) time saving and (iv) a possibility to calculate retention index. The MPP yielded a spreadsheet containing peak areas. Before the statistical data analysis, a manual verification of the obtained spreadsheet was performed. The multivariate data were analysed with a MVDA which was carried out in Solo (Version 6.5, 2011, Eigenvector Research, Wenatchee, WA, USA). As pre-processing step for principal component analysis (PCA) and partial least squares (PLS), all data were meancentred and the variables were weighed by their standard deviation to give them equal variance. In a first step, unsupervised classification, PCA, was conducted. Following that, to determine which compound changed the most during storage, a regression based supervised classification technique, namely PLS was implemented. For PLS, the headspace components were considered as X-variables and storage time as continuous Y-variable. In order to visualise these changes during storage, biplots of scores and correlation loadings described by the first two latent variables (LVs) were constructed for each storage temperature using OriginPro 8 (Origin Lab Corporation, Northampton, MA, USA). To quantitatively select headspace components clearly changing during storage, Variable Identification (VID) coefficients were subsequently calculated. These values correspond to the correlation coefficients between each original X-variable and Y-variable (s). Only volatile compounds with a VID coefficient higher than an absolute value of 0.80 were considered of importance. These compounds were plotted individually as a function of time. In these specific compound plots, the mean relative peak areas from the six replicates were depicted. As mentioned earlier (Section 2.2), samples were randomised as a function of storage time per temperature. Taking into account the effect of fibre variation, evaluation was made according to the percentage relative peak area with respect to week 0 for decreasing compounds and to the longest time at each temperature for increasing compounds. Tentative identification of the volatiles was carried out by comparing the deconvoluted mass spectrum with the reference mass spectra from the NIST spectral library (NIST08, version 2.0, National Institute of Standards and Technology, Gaithersburg, MD, USA) and WILEY mass spectra data (WILEY2010 version 9, Hoboken, New York, USA). A threshold match of 90% was taken into account for identification purpose. Moreover, a visual inspection of the spectral matching between the detected compound and the match from the library as well as comparison of the retention index were performed for further confirmation.

297

2.3.2. Kinetic modelling and parameter estimation The degradation rate of a compound at any time can be represented by the general rate law (Eq. (1)), where r is representing the rate of the reaction, A is the concentration of the compound at time t, k is the reaction rate constant at the temperature studied and n is the order of the reaction. Volatiles that were changing during storage at constant temperature can be modelled using a zero-order, n = 0 (Eq. (2)), a firstorder, n = 1 (Eq. (3)), or a first-order fractional conversion kinetic model (Eq. (4)), where A0 is the initial concentration at the start of storage (t = 0 week), A∞ is the concentration at very long storage time. The temperature dependence of the rate constants was expressed by the Arrhenius equation (Eq. (5)), where Ea is the activation energy (kJ mol−1), T is storage temperature (K), krefT is the reaction rate constant at reference temperature (20 °C or 293 K) and R is the universal gas constant (8.3145 J mol−1 K−1). r¼

dA ¼ kAn dt

ð1Þ

A ¼ A0 þ kt

ð2Þ

A ¼ A0 expðkt Þ

ð3Þ

A ¼ A∞ þ ðA0 −A∞ Þ expðkt Þ

ð4Þ

   Ea 1 1 − k ¼ krefT exp R T re f T

ð5Þ

Model evaluation and selection were performed by examining the coefficient of determination or R2adj (Eq. (6)) and by visual inspection of the parity plots (estimated values versus measured values) and the scatter plots (residuals versus measured values).

R2 ad j

  SSQ regression ðDF tot −1Þ 1− SSQ total ¼ 1−½  DF error

ð6Þ

In Eq. (6), DFtot and DFerror are degree of freedom of total and error, respectively and SSQ is the sum of squares. One-step regression analysis was performed by incorporating Eq. (5) into Eqs. (2), (3), or (4). All kinetic parameters were estimated using the SAS software (SAS 9.3, Cary, NC, USA). 3. Results and discussion Results are discussed in three levels. Firstly, using a headspace GC– MS fingerprinting approach, it was qualitatively investigated whether and how the headspace volatile fraction changed during shelf-life (compounds being formed/degraded) (Section 2.3.1). At a second level, three substeps were taken (Section 3.2), (i) quantitative selection of volatile compounds clearly changing during shelf-life through the calculation of Variable Identification (VID) coefficients, (ii) identification of the selected major volatile changes based on available mass spectral libraries and (iii) linkage between selected volatile compounds and possible reaction pathways to occur during storage. In order to obtain insight in what is happening at ambient storage conditions, focus was given first to the evaluation of volatile changes at storage temperatures of 20 and 28 °C. At a third level, in the context of accelerated shelf-life testing (ASLT), changes of selected shelf-life markers were studied over the whole storage temperature range (20, 28, 35 and 42 °C) and their kinetics were modelled as a function of time and temperature. By evaluating the estimated kinetic parameters, the suitability of these volatiles as markers for ASLT was investigated (Section 3.3).

298

S. Wibowo et al. / Food Research International 75 (2015) 295–304

Fig. 1. Total ion chromatogram of the headspace of pasteurised orange juice at the beginning of storage (week 0), obtained by headspace solid-phase microextraction GC–MS (HS-SPME– GC–MS) fingerprinting.

3.1. Headspace GC–MS fingerprinting The volatile fraction of pasteurised orange juice was fingerprinted using a HS-SPME–GC–MS technique. Fig. 1 shows a representative total ion chromatogram of the headspace fingerprint at the beginning of storage, immediately after pasteurisation. In all detected chromatograms over all storage temperatures as a function of time, limonene was clearly depicted as the most abundant among all compounds (retention time 17.1 min). Limonene as the predominant volatile aroma compound of orange juice was also referred by several other researchers (Perez-Cacho & Rouseff, 2008a; Pérez-López & Carbonell-Barrachina, 2006; Bacigalupi et al., 2013; Vervoort et al., 2012). Per storage temperature, the resulting chromatograms were transformed into one data table prior to multivariate data analysis (as described in Section 2.3.1). Each data set was analysed by principal component analysis (PCA), as an exploratory tool to detect potential outliers (results not shown). Subsequently, the changes of volatile compounds during storage per storage temperature were investigated by partial least squares (PLS) regression, in which the volatiles were considered as X-variables and storage time as continuous Y-variable. The changes in the headspace fractions during storage were graphically combined in biplots (Fig. 2). A high variance in the Y-variables explained by the first two LVs was observed for all storage temperatures (96.8%, 95.9%, 95.1% and 97.4% for 20, 28, 35 and 42 °C, respectively). From the biplots, a clear effect of shelf-life on the volatile fraction was observed for all storage temperatures. Additional information on this is given by the vectors, which represent the correlation loading for the Yvariable (time), the longer the vectors, the more Y-variation is explained by the two represented LVs. A qualitative insight about the relationship between volatile compounds and storage time can also be deduced from the biplots. Volatiles which are located far from the centre and situated beyond the inner ellipse are clearly changing during shelf-life (correlation coefficients of 70% and 100% were indicated by inner and outer ellipses in the plot). Compounds which are positioned in the same direction as the Y-vector are positively correlated with increasing storage time which means that those compound concentrations are clearly increasing as a function of storage. On the other hand, a decrease in

concentration was found for compounds located at the opposite direction of the Y-vector projecting the effect of storage time. As indicated by the small open circles in Fig. 2, more volatile compounds are projected to the beginning of shelf-life than to the end of shelf-life. Therefore, it can be concluded that more compound concentrations decreased in concentration during storage than being increased. These trends can be observed for all storage temperatures. 3.2. Identification of major shelf-life volatile changes and linkage to reaction pathways In this section, results are discussed in the following substeps. Firstly, volatile components responsible for the observed shelf-life changes in the volatile fraction were quantitatively selected through the calculation of Variable Identification (VID) coefficients. Secondly, selected volatile compounds were identified based on mass spectral libraries. In a last step, the detected and selected volatile changes were linked to possible reaction pathways. Although it can be clearly observed from Fig. 2 that particular volatiles are clearly changing during storage, a biplot is not a straightforward tool to quantitatively rank volatiles' importance for change during storage. Therefore, in the first substep, the importance of volatile compounds for change was ranked by calculating VID coefficients. In this work, volatiles with a VID coefficient higher than 0.80 in absolute value were found to change significantly during storage. These compounds are marked as open bold circles in Fig. 2 and listed in Table 1, in decreasing order of VID coefficient. A positive VID coefficient represents an increase in concentration as a function of shelf-life while a negative coefficient indicates a decrease. Through this quantitative measure, there were 8 volatiles which reached the VID threshold value and were detected to clearly change during shelf life at ambient storage temperature of 20 °C. A higher number of volatiles was selected at higher storage temperatures, for instance, 10, 12 and 13 volatiles were detected to clearly change during shelf-life at 28, 35 and 42 °C, respectively. In the second substep, these selected volatiles were identified based on both the detected compound mass spectrum and the verification of the obtained mass spectrum with the available mass spectral libraries. As shown in Table 1, selected volatiles can be categorised as aldehydes

S. Wibowo et al. / Food Research International 75 (2015) 295–304

299

Fig. 2. PLS biplots describing the effect of storage time on the orange juice volatile fraction (objects represented by differently shaped symbols) at 20, 28, 35 and 42 °C. The open circles indicate the headspace volatile compounds, of which only the compounds selected through the VID procedure are identified and marked in bold (Table 1). Inner and outer ellipses represent correlation coefficients of 70% and 100%. The vectors represent the correlation loading for the continuous Y-variable (storage time). The percentages of the variances in X and Y explained by each latent variable (LV1 and LV2) are represented on the respective axes.

(decanal, nonanal and octanal), esters (octyl acetate and α-terpinyl acetate), terpene hydrocarbons (α-pinene, α-terpinene, α-terpinolene and α-phellandrene), terpene alcohols (α-terpineol, β-terpineol and linalool), sulphur compounds (dimethyl sulphide) and terpene oxides (linaloyl oxide). When orange juice was stored at 20 and 28 °C, an increase of terpenes (terpene alcohols and terpene hydrocarbons) was observed during shelf-life. Additionally, sulphur compounds (dimethyl

sulphide) and oxides (linaloyl oxide) concentration increased at 28 °C. As indicated by the negative VID coefficients, decreases in terpenes (α-pinene and linalool), aldehydes (octanal) and esters (octyl acetate and α-terpinyl acetate) were detected as well at 28 °C. In the third substep, these volatile changes were linked to possible reaction pathways taken place during storage. These changes will be discussed in detail linked to literature data in the following sections.

300

S. Wibowo et al. / Food Research International 75 (2015) 295–304

Table 1 Volatiles significantly changing as a function of shelf-life, per storage temperature (20, 28, 35 and 42 °C), selected based on the VID procedure, listed in decreasing order of VID coefficient. Positive VID coefficients indicate an increase in concentration during storage while negative coefficients denote a decrease. The retention index (RI) and chemical group are listed. Storage temperature

VID

Identity

RI

Chemical group

20 °C

0.96 0.92 0.89 −0.80 −0.84 −0.84 −0.86 −0.94 0.91 0.91 0.90 0.86 0.85 −0.81 −0.88 −0.90 −0.94 −0.94 0.99 0.96 0.91 0.87 0.85 0.84 0.82 0.80 −0.82 −0.90 −0.97 −0.97 0.98 0.94 0.94 0.91 0.88 0.88 0.86 −0.81 −0.82 −0.86 −0.90 −0.96 −0.97

α-Terpineol α-Terpinolene α-Phellandrene Linalool α-Pinene Decanal Nonanal Octanal α-Terpineol β-Terpineol Linaloyl oxide α-Terpinene Dimethyl sulphide α-Terpinyl acetate Octanal Linalool Octyl acetate α-Pinene α-Terpineol β-Terpineol α-Terpinene 2-Methyl-1,3-butadiene α-Terpinolene α-Phellandrene Linaloyl oxide Dimethyl sulphide Decanal Octanal α-Pinene Linalool α-Terpineol α-Terpinolene β-Terpineol α-Terpinene Dimethyl sulphide 2-Methyl-1,3-butadiene Linaloyl oxide Nonanal Decanal Octyl acetate Octanal Linalool α-Pinene

1185 1083 997 1093 923 1196 1097 994 1185 1139 962 1010 564 1344 994 1093 1201 923 1184 1139 1010 561 1083 997 962 565 1196 994 923 1093 1184 1083 1139 1010 564 561 962 1097 1196 1201 994 1093 922

Terpene (alcohol) Terpene (hydrocarbon) Terpene (hydrocarbon) Terpene (alcohol) Terpene (hydrocarbon) Aldehyde Aldehyde aldehyde Terpene (alcohol) Terpene (alcohol) Terpene (oxide) Terpene (hydrocarbon) Sulphur compound Ester Aldehyde Terpene (alcohol) Ester Terpene (hydrocarbon) Terpene (alcohol) Terpene (alcohol) Terpene (hydrocarbon) Terpene (hydrocarbon) Terpene (hydrocarbon) Terpene (hydrocarbon) Terpene (oxide) Sulphur compound Aldehyde Aldehyde Terpene (hydrocarbon) Terpene (alcohol) Terpene (alcohol) Terpene (hydrocarbon) Terpene (alcohol) Terpene (hydrocarbon) Sulphur compound Terpene (hydrocarbon) Terpene (oxide) Aldehyde Aldehyde Ester Aldehyde Terpene (alcohol) Terpene (hydrocarbon)

28 °C

35 °C

42 °C

3.2.1. Increase in terpene alcohols After 32 weeks of storage, an increase in terpene alcohol concentrations at 20 and 28 °C was observed (Table 1). The increase of αterpineol during storage can be linked to acid-catalysed hydration of limonene and linalool (Haleva-Toledo, Naim, Zehavi, & Rouseff, 1999; Petersen, Tønder, & Poll, 1998). During five months of storage of orange juice at 20 °C, decreases in linalool and limonene concentration at 28% and 49%, respectively, were observed by Berlinet, Ducruet, Brillouet, Reynes, and Brat (2005). Pérez-López and Carbonell-Barrachina (2006) reported that α-terpineol was formed after pasteurisation and increased considerably during storage. Its formation was dependent on the citrus juice pH, for example a fivefold faster formation was observed at pH 2.8 than at pH 3.8 (Haleva-Toledo et al., 1999). Similarly to α-terpineol, β-terpineol is also formed from acid-catalysed reactions of limonene (Berlinet et al., 2006). 3.2.2. Increase in terpene hydrocarbons Three terpene hydrocarbons were increasing in concentration during storage at 20 and 28 °C (Table 1). Increase in terpene hydrocarbons such as α-terpinolene and α-terpinene can be linked to the oxidative

reaction and/or acid-catalysed hydration–dehydration reactions of terpenes as suggested by Perez-Cacho and Rouseff (2008a); Blair, Godar, Masters, and Riester (1952). Some researchers observed a twofold increase of p-cymene in citrus oil at 5 °C during 12 months of storage which could be attributed to rearrangement, hydrogenation and dehydrogenation of α-terpinene, γ-terpinene and limonene (Njoroge, Ukeda, & Sawamura, 1996). Consequently, these chemical reactions of particular terpene hydrocarbons can result in the formation and deformation of other terpene hydrocarbons. Choi and Sawamura (2002) observed increases in p-cymene, α-humulene, α-cedrene and γ-elemene in citrus oil after nine weeks of storage at 30 °C, while others observed a decrease of limonene, α-pinene and β-myrcene during storage of orange juice (Berlinet et al., 2005; van Willige, Linssen, Legger-Huysman, & Voragen, 2003; Bacigalupi et al., 2013; Averbeck & Schieberle, 2011). 3.2.3. Increase in sulphur compounds It was observed that volatile sulphur compounds (dimethyl sulphide) were increasingly formed during storage above 20 °C (no significant change observed at 20 °C). The presence and formation of dimethyl sulphide was possibly due to the reaction involving sulphurcontaining amino acids, such as methionine. Degradation of methionine via Strecker degradation during pasteurisation and storage leads to the formation of methional (Perez-Cacho, Mahattanatawee, Smoot, & Rouseff, 2007). The occurrence of methional in stored pasteurised (11550 ng L− 1) orange juice was reported by Bezman, Rouseff, and Naim (2001). This compound can break down into methanethiol, which can later be oxidised to dimethyl sulphide (Perez-Cacho et al., 2007). Although some compounds such as hydrogen sulphide and methyl sulphide can also be found in the fresh citrus juice (Shaw, Ammons, & Braman, 1980), their formation (e.g., dimethyl sulphide, dimethyl trisulphide, methional, methanethiol, 2-methyl-3-furanthiol and 1-p-menthene-8-thiol) does mainly occur in the thermallyprocessed products and their concentration can be further increased during storage (Perez-Cacho et al., 2007). 3.2.4. Increase in terpene oxides Similar to sulphur compounds, terpene oxides (linaloyl oxide or 2,6,6-trimethoxy-2-vinyltetrahydropyran) displayed an important increase in concentration at a storage temperature of 28 °C. The presence of oxides during storage can be linked to the degradation of monoterpenes (e.g., linalool) induced by pasteurisation and/or storage. In other food matrices, such as passion fruit and grapes, besides the free fraction of volatile terpenes, the bound non-volatiles precursor can be important in the development of aroma if released. It was found that linalool, nerol and geraniol are naturally present in glycosidic bound form. During thermal treatment under acidic conditions, the latter terpene alcohols are being converted to terpenes oxides (e.g., 2,6,6trimethyl-2-vinyltetrahydropyran, linalool oxides and nerol oxides) (Williams, Strauss, & Wilson, 1980; Engel & Tressl, 1983). The effect of storage conditions on the composition of citrus oil was investigated by Choi and Sawamura (2002). After nine weeks of storage, an increase of oxides (limonene oxide, linalool furanoxide, and caryophyllene oxide) was observed, wherein the increase at higher temperatures (20 and 30 °C) was more prominent than lower temperatures (–21 and 5 °C). 3.2.5. Decrease in terpene alcohols From Table 1, volatile compounds with a negative VID coefficient indicate a decrease in concentration of the selected compound as a function of storage. It was observed that the terpene alcohol linalool decreased during storage at 20 and 28 °C. As previously discussed in Section 3.2.1, the loss of linalool during storage was suggested to occur due to acid-catalysed degradation reactions and this decrease may coincide with the observed increase in α-terpineol concentration. According to Haleva-Toledo et al. (1999) linalool was more reactively degraded to α-terpineol than limonene, however, due to the higher

S. Wibowo et al. / Food Research International 75 (2015) 295–304

amount of limonene, the formation of α-terpineol from both compounds occurred to the same extent. A loss up to 40% of linalool after six months of storage at 20 °C in orange juice was reported by Bacigalupi et al. (2013). Besides the formation of α-terpineol, increases in other degradative products of linalool, such as 1,8-cineole, geraniol, nerol and terpinen-4-ol was referred by Perez-Cacho and Rouseff (2008a). 3.2.6. Decrease in terpene hydrocarbons The observed decrease in monoterpene hydrocarbons such as αpinene during storage was in agreement with the observation of Averbeck and Schieberle (2011). They reported a 50% reduction of αpinene in freshly reconstituted orange juice during storage at 20 °C for 16 weeks. It was reported that a considerable decrease of α-pinene occurred through polymerisation and evaporation (Njoroge et al., 1996) and its degradation was accelerated by light exposure (Bacigalupi et al., 2013). Additionally, the absorption of volatiles by packaging materials (also known as scalping) was reported as one of the responsible factors for the decrease of α-pinene and β-pinene in model citrus juices (Lebossé, Ducruet, & Feigenbaum, 1997). Interestingly, the same authors pointed out that scalping was advantageous since this process prohibited pinene compounds from degradation into off-flavours such as α-terpineol, borneol and α-fenchol. In the current work, orange juice was stored in polyethylene terephthalate (PET) bottles. According to Sajilata, Savitha, Singhal, and Kanetkar (2007), plastic polymer characteristics (e.g., polarity and density), flavour compound characteristics (e.g., concentration, carbon chain length and polarity) and external factors (e.g., pH, food composition and storage conditions) can influence this sorption process. Although scalping has been considered as one of the factors contributing to aroma changes during storage, Berlinet et al. (2005) reported that the loss of aroma compounds during storage were mainly due to chemical reactions and not due to the scalping process. 3.2.7. Decrease in aldehydes Among the ones that decreased in concentration, three aldehydes (octanal, nonanal and decanal) were found at 20 °C. A similar decreasing trend was reported by Berlinet et al. (2005) in orange juice. More than 50% decrease in aldehydes (hexanal, octanal, nonanal and decanal) concentration was observed after five months of storage at 20 °C, presumably due to conversion into their corresponding acids. As previously mentioned (Section 3.2.6), related to their polarity characteristics, decreases in aldehydes during storage could also be ascribed to volatiles absorption by PET packaging. Flavour compound characteristics, such as carbon chain length and polarity were reported to be one of the influencing factors for scalping: the longer the chain or the less polar a compound, the higher absorption is expected. Depending on the chain length of their lipophilic portion, the saturated aldehydes (e.g., octanal, nonanal and decanal) are absorbed to a higher extent than the unsaturated aldehydes (e.g., geranial). Moreover, compared to the other volatile groups, esters and aldehydes have higher affinity to LDPE (low-density polyethylene) than alcohols but lower compared to hydrocarbons and ketones (Sajilata et al., 2007). A study was conducted to observe flavour absorption by LDPE, PC (polycarbonate) and PET packaging for 29 days of storage at 20 °C. It was found that limonene, myrcene and decanal were absorbed from orange juice by those packaging materials, with limonene as the highest absorbed compound by LDPE (van Willige et al., 2003). 3.2.8. Decrease in esters In parallel with the observation of Tønder et al. (1998), a decrease in concentration levels of esters during storage was observed. In our work, esters such as octyl acetate and α-terpinyl acetate showed important changes. Hydrolysis of esters producing acids during storage of orange juice could be the possible reason (Blair et al., 1952).

301

Besides the effect of storage, different processing steps are recognised to influence the ester concentration. For example, a higher amount of esters (ethyl acetate, methyl butanoate and ethyl butanoate) was detected in fresh-squeezed juice compared to pasteurised juice (Baldwin et al., 2012). Other researchers reported that ethyl acetate, methyl butyrate and ethyl butyrate were reduced substantially in pasteurised RFC (reconstituted from concentrate) compared to pasteurised NFC (not from concentrate) juice (Nisperos-Carriedo & Shaw, 1990). 3.3. Kinetic modelling of selected shelf-life markers Since shelf-stable products are by definition microbially stable for infinite time, quality deterioration is defining their best-before-date (the time in which the quality characteristics, regulatory issues and consumer demands, are granted). It implies that before a new or modified product can be launched to the market, food companies need to carry out shelf-life studies in order to estimate the appropriate storage time. Shelf-life studies at ambient storage temperature are time- and resource-consuming calling for accelerated shelf-life tests (ASLT), which is typically done by exposing the product to elevated temperatures. Assuming that quality degradation reactions are temperature dependent, the result from ASLT has potential to be converted to actual market conditions (Mizrahi, 2011). In this work, besides evaluating chemical changes in the volatile orange juice fraction during storage at ambient temperatures (20–28 °C), chemical changes in the volatile orange juice fraction were also fingerprinted as a function of time when the product was stored at higher temperatures (35 and 42 °C) with the aim of evaluating the potential of volatile ASLT markers. In this context, we define ‘ASLT markers’ as compounds with a clearly observable time and temperature dependent change. Such compound could have some potential for direct shelf-life prediction if the quality degradation reaction pathway in which the compound takes part, leads to a level of non-acceptable quality of the product. This is also true if the kinetics of the compounds change are the same as the kinetics of a quality degradation reaction leading to a level of non-acceptable quality of the product. In analogy for what is described above, also for the fingerprints evaluated for storage at 35 and 42 °C, VID coefficients were calculated for all detected compounds to relatively rank the volatile compound importance for the change as a function of storage time. Components with an absolute value higher than 0.80 are listed in Table 1 and marked on the biplots in Fig. 2 and were defined as ‘compounds clearly changing during storage’. It can be seen that changes in volatiles were not the same for all storage temperatures and were clearly influenced by the storage temperature. Compared to storage temperature of 20 °C, more volatiles were clearly changing as a function of time at higher temperatures. At higher storage temperatures, increase in concentration was detected for terpene hydrocarbons, terpene alcohols, terpene oxides and sulphur compounds. On the contrary, aldehydes and esters decreasing during storage. In literature, Averbeck and Schieberle (2011) reported nearly ten-fold decrease of aldehydes (octanal and decanal) and terpene alcohols (linalool) during storage of orange juice at 37 °C for 16 weeks, while only twofold decrease of aldehydes and a slight decrease of linalool at 20 °C in the same period of time was observed. In order to determine the suitability of these selected volatiles clearly changing during storage as potential ASLT markers, two criteria need to be fulfilled: (i) the volatile change should be temperature-dependent and (ii) an observable change should be present at reference storage temperature as well (20 °C). In order to evaluate these pre-requisites, all observed compound changes were kinetically modelled. First, an appropriate empirical kinetic model was identified, then reaction rate constants and activation energies were estimated using a non-linear onestep regression analysis. As mentioned in Section 2.3.2, models were evaluated using parity plot, scatter plot and R2 adjusted. In this way, different kinetic models were selected to model the detected compound changes: a zero-order model for describing the changes observed in

302

S. Wibowo et al. / Food Research International 75 (2015) 295–304

Table 2 Estimated kinetic parameters based on one-step non-linear degradation kinetic model (20 °C as reference temperature) describing the changes during storage of headspace compounds in pasteurised orange juice. Samples were stored at 20, 28, 35 and 42 °C. Sulphur compound (dimethyl sulphide) could be adequately modelled by a zero-order empirical kinetic model, and the rest of the compounds could be modelled by a first-order empirical kinetic model. Compounda

Co (%)

C∞ (%)

Zero-order model Dimethyl sulphide

N/Ab

–

First-order model Decanal Nonanal Octanal* Octyl acetate Linalool* α-Pinene* α-Terpinyl acetate

91.03 ± 3.07 92.20 ± 2.87 91.69 ± 2.00 98.52 ± 2.65 96.01 ± 1.12 96.45 ± 1.21 93.57 ± 3.99

– – – – – – –

First-order fractional conversion model α-Terpinolene 41.76 ± 2.89 α-Terpinene 25.76 ± 3.71 α-Phellandrene 17.11 ± 8.22 α-Terpineol* 4.01 ± 1.23 Linaloyl oxide N/A β-Terpineol N/A

98.40 ± 3.71 104.32 ± 6.48 106.02 ± 4.72 110.62 ± 2.51 132.70 ± 24.29 110.34 ± 5.41

kref (week−1) 0.021 ± 0.002c −0.315 ± 0.033 −0.297 ± 0.029 −0.150 ± 0.011 −0.014 ± 0.002 −0.022 ± 0.001 −0.021 ± 0.001 −0.028 ± 0.004 0.113 ± 0.029 0.073 ± 0.018 0.185 ± 0.061 0.057 ± 0.004 0.027 ± 0.009 0.054 ± 0.007

Ea (kJ mol−1)

R2adj

61.66 ± 4.30

0.86

46.98 ± 5.29 56.58 ± 4.89 58.04 ± 3.56 81.41 ± 7.20 85.93 ± 2.45 92.31 ± 2.77 98.37 ± 8.33

0.97 0.97 0.99 0.98 0.99 0.99 0.95

32.51 ± 8.45 39.23 ± 6.59 57.82 ± 14.64 58.35 ± 1.86 64.44 ± 5.76 67.71 ± 4.26

0.83 0.87 0.70 0.99 0.85 0.94

±95% approximate confidence interval. a The selected volatile compounds for ASLT markers are marked with asterisk. b N/A is not applicable. c Unit for dimethyl sulphide % week−1.

dimethyl sulphide; a first-order model for decanal, nonanal, octanal, octyl acetate, linalool, α-pinene and α-terpinyl acetate; and a firstorder fractional conversion model for α-terpineol, β-terpineol, αterpinene, α-terpinolene, α-phellandrene and linaloyl oxide. In total, there were 15 volatiles with clearly observable changes during storage (Table 1). Table 2 provides the estimated apparent kinetic parameters for shelf-life changes listed in increasing order of activation energies (Ea). Based on the two criteria defined above, α-pinene, α-terpineol, linalool and octanal were considered as markers for ASLT. Their changes

were observed at all temperatures and their reaction followed the Arrhenius kinetics wherein higher storage temperatures lead to the acceleration of the rates of reactions. Among the selected compounds, the lower E a was given by α-terpineol and octanal (58 kJ mol− 1), or in other words, a lower temperature dependency than α-pinene and linalool. To have a better visual overview, changes of these compounds are depicted in Fig. 3, in which the kinetic model is indicated by the full lines and the experimental data by the symbols. Evaluation was represented relatively to peak area

Fig. 3. Selected volatile compounds for ASLT markers of pasteurised orange juice stored at 20 (◆), 28 (■), 35 (▲) and 42 °C (×). The full lines represent the fitted values by the kinetic modelling and the different symbols represent the experimental data. The Y-axis indicates the percentage relative peak area with respect to week 0 for decreasing compounds and to the longest time at each temperature for increasing compound.

S. Wibowo et al. / Food Research International 75 (2015) 295–304

detected in week 0 for decreasing compounds and to the longest time at each temperature for increasing compound. Referring to Section 3.2, the linkage between changes of volatile compounds and various reaction pathways was discussed in detail. An insight has been established wherein changes in volatiles were initiated by the processing conditions and then intensified by storage time and temperature. The formation of α-terpineol was due to the acid catalysed reactions of limonene, linalool and α-pinene. The decrease of octanal was due to oxidation into its corresponding acids. It seems that both reactions can be used as indicators for quality determining reactions in orange juice. The selected volatiles represent different chemical groups: α-pinene (terpene hydrocarbons), α-terpineol and linalool (terpene alcohols) and octanal (aldehydes). Combination of these groups are important for determining aroma of orange juice. Within the context of sensorial characteristics, α-terpineol is recognised as one of the most important off-flavour volatile compounds present in stored citrus juices. It imparts stale, musty or piney odour characteristics (Haleva-Toledo et al., 1999). Moreover, a high correlation between α-terpineol and oxidised taste/ odour and bitterness was reported by Petersen et al. (1998). This compound was recommended as an indicator of shelf-life since its increase was found to be linear with storage time (Askar, Bielig, & Terpetow, 1973). In fresh orange juice, linalool is known for a floral, sweet and fruity note, α-pinene is responsible for resin and pine tree odour and octanal contributes to the typical characteristic of citrus note (Perez-Cacho & Rouseff, 2008b; Rega, Fournier, & Guichard, 2003). Reduction in these volatiles concentration due to storage could imply the reduction of the typical orange flavour. Since changes of these compounds can alter the sensorial property of orange juice, monitoring these reactions during storage especially at higher storage temperatures could be important.

4. Conclusion An integrated fingerprinting-kinetics approach was used to obtain more insight into the chemical reactions taking place in pasteurised orange juice during shelf-life (20, 28, 35 and 42 °C). To obtain insight in chemical changes in the volatile food fraction, samples were fingerprinted with headspace GC–MS. The objectives of this work were twofold: (i) to identify major chemical changes of pasteurised orange juice during shelf-life and (ii) to study the kinetics of selected shelf-life compounds in the context of accelerated shelf-life testing (ASLT). At ambient storage temperature (20 and 28 °C), clear changes in volatile compounds were observed. The most significant changes were changes in terpenes (e.g., formation of α-terpineol and degradation of linalool and α-pinene) and increase in oxides (linaloyl oxide) and sulphur compounds (dimethyl sulphide). In addition, a decrease in aldehydes (decanal, nonanal and octanal) and esters (α-terpinyl acetate) was also observed. These volatiles could be linked to the acidcatalysed degradation, Strecker degradation, oxidative and hydrolysis of ester reactions. The volatile absorption into packaging material or scalping process was also mentioned in literature. Concerning objective (ii), relying on a kinetic modelling approach, four volatiles were selected as potential markers for ASLT of pasteurised orange juice: α-pinene, α-terpineol, linalool and octanal. We defined ‘ASLT markers’ as compounds with a clearly observable time and temperature dependent change. In the future, such compounds could have some potential for direct shelf-life prediction if the quality degradation reaction pathway in which the compound takes part, leads to a level of non-acceptable quality of the product. This is also true if the kinetics of this compound change are the same as the kinetics of a quality degradation reaction leading to a level of non-acceptable quality of the product. Consequently, there is a need for quantitative, kinetic studies on the selected markers and further consumer studies (sensory). In addition, the importance of studying orange juice quality changes in other food fractions should not be forgotten.

303

Acknowledgement This research was financially supported by the Seventh Framework Programme (FP7) of the European Union under the Marie Curie Initial Training Network ‘HST FoodTrain’ (Grant agreement 264470). Tara Grauwet is a researcher funded by the Research Foundation Flanders (FWO) as well as the KU Leuven

References Aganovic, K., Grauwet, T., Kebede, B.T., Toepfl, S., Heinz, V., Hendrickx, M., et al. (2014). Impact of different large scale pasteurisation technologies and refrigerated storage on the headspace fingerprint of tomato juice. Innovative Food Science & Emerging Technologies, 26, 431–444. Askar, A., Bielig, H.J., & Terpetow, H. (1973). Aroma changes in orange juice. 2. Mitteilung: Aromaveraderungen bei der herstelllug und eahred der lageerung von rangeensaft in flaschen. Deutsche Lebensmittel-Rundschau, 69, 162–167. Averbeck, M., & Schieberle, P. (2011). Influence of different storage conditions on changes in the key aroma compounds of orange juice reconstituted from concentrate. European Food Research and Technology, 232, 129–142. Bacigalupi, C., Lemaistre, M.H., Boutroy, N., Bunel, C., Peyron, S., Guillard, V., et al. (2013). Changes in nutritional and sensory properties of orange juice packed in PET bottles: An experimental and modelling approach. Food Chemistry, 141, 3827–3836. Baldwin, E.A., Bai, J., Plotto, A., Cameron, R., Luzio, G., Narciso, J., et al. (2012). Effect of extraction method on quality of orange juice: Hand-squeezed, commercial-fresh squeezed and processed. Journal of the Science of Food and Agriculture, 92, 2029–2042. Berlinet, C., Brat, P., Brillouet, J.M., & Ducruet, V. (2006). Ascorbic acid, aroma compounds and browning of orange juices related to PET packaging materials and pH. Journal of the Science of Food and Agriculture, 86, 2206–2212. Berlinet, C., Ducruet, V., Brillouet, J.M., Reynes, M., & Brat, P. (2005). Evolution of aroma compounds from orange juice stored in polyethylene terephthalate (PET). Food Additives and Contaminants, Part A: Chemistry, Analysis, Control, Exposure and Risk Assessment, 22, 185–195. Bezman, Y., Rouseff, R.L., & Naim, M. (2001). 2-methyl-3-furanthiol and methional are possible off-flavors in stored orange juice: Aroma-similarity, NIF/SNIF GC-O, and GC analyses. Journal of Agricultural and Food Chemistry, 49, 5425–5432. Blair, J.S., Godar, E.M., Masters, J.E., & Riester, D.W. (1952). Exploratory experiments to identify chemical reactions causing flavor deterioration during storage of canned orange juice. I. Incompatibility of peel-oil constituents with the acid juice. Journal of Food Science, 17, 235–260. Cevallos-Cevallos, J.M., Reyes-De-Corcuera, J.I., Etxeberria, E., Danyluk, M.D., & Rodrick, G.E. (2009). Metabolomic analysis in food science: A review. Trends in Food Science & Technology, 20, 557–566. Choi, H.S., & Sawamura, M. (2002). Effects of storage conditions on the composition of Citrus tamurana Hort. ex Tanaka (Hyuganatsu) essential oil. Bioscience, Biotechnology, and Biochemistry, 66, 439–443. Engel, K.H., & Tressl, R. (1983). Formation of aroma components from nonvolatile precursors in passion fruit. Journal of Agricultural and Food Chemistry, 31, 998–1002. Esteve, M.J., Frigola, A., Rodrigo, C., & Rodrigo, D. (2005). Effect of storage period under variable conditions on the chemical and physical composition and colour of Spanish refrigerated orange juices. Food and Chemical Toxicology, 43, 1413–1422. Grauwet, T., Vervoort, L., Colle, I., Van Loey, A., & Hendrickx, M. (2014). From fingerprinting to kinetics in evaluating food quality changes. Trends in Biotechnology, 32, 125–131. Haleva-Toledo, E., Naim, M., Zehavi, U., & Rouseff, R.L. (1999). Formation of a-terpineol in citrus juices, model and buffer solutions. Journal of Food Science, 64, 838–841. Kebede, B.T., Grauwet, T., Magpusao, J., Palmers, S., Michiels, C., Hendrickx, M., et al. (2015). Chemical changes of thermally sterilized broccoli puree during shelf-life: Investigation of the volatile fraction by fingerprinting-kinetics. Food Research International, 67, 264–271. Kebede, B.T., Grauwet, T., Mutsokoti, L., Palmers, S., Vervoort, L., Hendrickx, M., et al. (2014). Comparing the impact of high pressure high temperature and thermal sterilization on the volatile fingerprint of onion, potato, pumpkin and red beet. Food Research International, 56, 218–225. Lebossé, R., Ducruet, V., & Feigenbaum, A. (1997). Interactions between reactive aroma compounds from model citrus juice with polypropylene packaging film. Journal of Agricultural and Food Chemistry, 45, 2836–2842. Mahattanatawee, K., Rouseff, R., Valim, M.F., & Naim, M. (2004). Identification and aroma impact of norisoprenoids in orange juice. Journal of Agricultural and Food Chemistry, 53, 393–397. Mizrahi, S. (2011). Accelerated shelf-life tests. In D. Kilcast, & P. Subramaniam (Eds.), Food and beverage stability and shelf life (pp. 482–506). Cambridge: Woodhead Publishing Limited. Nisperos-Carriedo, M.O., & Shaw, P.E. (1990). Comparison of volatile flavor components in fresh and processed orange juices. Journal of Agricultural and Food Chemistry, 38, 1048–1052. Njoroge, S.M., Ukeda, H., & Sawamura, M. (1996). Changes in the volatile composition of yuzu (Citrus junos Tanaka) cold-pressed oil during storage. Journal of Agricultural and Food Chemistry, 44, 550–556. Perez-Cacho, P.R., Mahattanatawee, K., Smoot, J.M., & Rouseff, R. (2007). Identification of sulfur volatiles in canned orange juices lacking orange flavor. Journal of Agricultural and Food Chemistry, 55, 5761–5767.

304

S. Wibowo et al. / Food Research International 75 (2015) 295–304

Perez-Cacho, P.R., & Rouseff, R. (2008a). Processing and storage effects on orange juice aroma: A review. Journal of Agricultural and Food Chemistry, 56, 9785–9796. Perez-Cacho, P.R., & Rouseff, R.L. (2008b). Fresh squeezed orange juice odor: A review. Critical Reviews in Food Science and Nutrition, 48, 681–695. Pérez-López, A.J., & Carbonell-Barrachina, À. A. (2006). Volatile odour components and sensory quality of fresh and processed mandarin juices. Journal of the Science of Food and Agriculture, 86, 2404–2411. Petersen, M.A., Tønder, D., & Poll, L. (1998). Comparison of normal and accelerated storage of commercial orange juice — Changes in flavour and content of volatile compounds. Food Quality and Preference, 9, 43–51. Picariello, G., Mamone, G., Addeo, F., & Ferranti, P. (2012). Novel mass spectrometrybased applications of the ‘omic’ sciences in food technology and biotechnology. Food Technology and Biotechnology, 50, 286–305. Rega, B., Fournier, N., & Guichard, E. (2003). Solid phase microextraction (SPME) of orange juice flavor: Odor representativeness by direct gas chromatography olfactometry (D-GC-O). Journal of Agricultural and Food Chemistry, 51, 7092–7099. Roig, M.G., Bello, J.F., Rivera, Z.S., & Kennedy, J.F. (1999). Studies on the occurrence of nonenzymatic browning during storage of citrus juice. Food Research International, 32, 609–619. Sajilata, M.G., Savitha, K., Singhal, R.S., & Kanetkar, V.R. (2007). Scalping of flavors in packaged foods. Comprehensive Reviews in Food Science and Food Safety, 6, 17–35. Shaw, P.E., Ammons, J.M., & Braman, R.S. (1980). Volatile sulfur compounds in fresh orange and grapefruit juices: Identification, quantitation, and possible importance to juice flavor. Journal of Agricultural and Food Chemistry, 28, 778–781.

Tønder, D., Petersen, M.A., Poll, L., & Olsen, C.E. (1998). Discrimination between freshly made and stored reconstituted orange juice using GC Odour Profiling and aroma values. Food Chemistry, 61, 223–229. USDA (2014). Citrus world market and trade. [online] http://www.fas.usda.gov/data/ citrus-world-markets-and-trade [Last accessed on 8/12/2014] van Willige, R.W.G., Linssen, J.P.H., Legger-Huysman, A., & Voragen, A.G.J. (2003). Influence of flavour absorption by food-packaging materials (low-density polyethylene, polycarbonate and polyethylene terephthalate) on taste perception of a model solution and orange juice. Food Additives & Contaminants, 20, 84–91. Vervoort, L., Grauwet, T., Kebede, B.T., Van der Plancken, I., Timmermans, R.A.H., Hendrickx, M., et al. (2012). Headspace fingerprinting as an untargeted approach to compare novel and traditional processing technologies: A case-study on orange juice pasteurisation. Food Chemistry, 134, 2303–2312. Wibowo, S., Grauwet, T., Santiago, J.S., Tomic, J., Vervoort, L., Hendrickx, M., et al. (2015a). Quality changes of pasteurised orange juice during storage: A kinetic study of specific parameters and their relation to colour instability. Food Chemistry, 187, 140–151. Wibowo, S., Vervoort, L., Tomic, J., Santiago, J.S., Lemmens, L., Panozzo, A., et al. (2015b). Colour and carotenoid changes of pasteurised orange juice during storage. Food Chemistry, 171, 330–340. Williams, P.J., Strauss, C.R., & Wilson, B. (1980). Hydroxylated linalool derivatives as precursors of volatile monoterpenes of muscat grapes. Journal of Agricultural and Food Chemistry, 28, 766–771. Wishart, D.S. (2008). Metabolomics: Applications to food science and nutrition research. Trends in Food Science & Technology, 19, 482–493.