Studies of polyphosphate composition and their interaction with dairy matrices by ion chromatography and 31P NMR spectroscopy

International Dairy Journal 28 (2013) 102e108

Contents lists available at SciVerse ScienceDirect

International Dairy Journal journal homepage: www.elsevier.com/locate/idairyj

Studies of polyphosphate composition and their interaction with dairy matrices by ion chromatography and 31P NMR spectroscopy Célie Rulliere a, Corinne Rondeau-Mouro b, c, *, Sana Raouche a, Marie Dufrechou a, Sylvie Marchesseau a a

Université Montpellier 2, UMR IATE, Place E. Bataillon, 34095 Montpellier Cedex 05, France IRSTEA, UR TERE, CS 64427, 17 avenue de Cucillé, 35044 Rennes Cedex, France c Université Européenne de Bretagne, France b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 6 April 2012 Received in revised form 6 September 2012 Accepted 15 September 2012

The use of ion-exchange chromatography and 31P nuclear magnetic resonance (31P NMR) to analyse the composition and the chain length of phosphate emulsifying salts were studied, as well as the impact of these salts in dairy products. Ion chromatography was more appropriate than 31P NMR to study polyphosphate composition in complex environments, whereas interactions between phosphate species and dairy components were elucidated by 31P NMR. Phosphate species interacting with calcium, as well as the percentage of chelated calcium, were identified using 31P NMR. Thus, ion chromatography and solidsate 31P NMR could be used as complementary methods to study compositions of polyphosphate blends and their interactions with dairy matrices. Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction Polyphosphate salts are food additives used in dairy products such as processed cheese or cheese analogues for their mineral sequestering and buffering properties. These salts also play a key role in protein hydration (Ennis, O’Sullivan, & Mulvihill, 1998; Gaucher, Piot, Beaucher, & Gaucheron, 2007) and improve the shelflife of products due to their bacteriostatic effect (Obritsch, Ryu, Lampila, & Bullerman, 2008). During the manufacture of processed cheese, polyphosphates disrupt the insoluble calciumparacaseinateephosphate network present in cheese by removing calcium from caseins, leading to the formation of a homogeneous lipido-protein network after heat treatment. In the food industry, blends of linear and sometimes cyclic polyphosphate salts are used, which contain few or numerous phosphorus atoms linked together by energy-rich phosphoanhydride bonds. The textural properties of the processed cheese network greatly depend on the nature of the emulsifying salts used and especially on their composition in phosphate chain lengths (Caric, Gantar, & Kalab, 1985). Recently, Sadlikova et al. (2010) and Weiserova et al. (2011) confirmed these results and showed that increasing the proportion of polyphosphates in binary mixtures of long- and short-chain phosphates led to an increase in hardness of processed cheese. During food manufacture, polyphosphate composition

* Corresponding author. Tel.: þ33 0 2 23 48 21 43. E-mail address: [email protected] (C. Rondeau-Mouro). 0958-6946/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.idairyj.2012.09.005

might change due to hydrolysis reactions, as demonstrated by a recent study showing the susceptibility of such molecules to hydrolysis into short chains at temperatures above 100
C (Rulliere, Perenes, Senocq, Dodi, & Marchesseau, 2012). Thus, the precise composition of emulsifying salts may be monitored carefully during processing by consistent unvarying methods. In addition, further investigations are needed to understand the interaction of these phosphate-chelating agents with dairy matrices and their relation with processed cheese functional properties. Ion chromatography is an interesting method for evaluating the composition in phosphate chain lengths of such additives (Baluyot & Hartford, 1996; McBeath, Lombi, McLaughlin, & Bumann, 2007), but does not give information on potential interactions between phosphates and other food components. In addition, several studies in environmental applications demonstrated the potential of solidstate 31P nuclear magnetic resonance (31P NMR) spectroscopy to evaluate the average chain length (n) of polyphosphates (Ahlgren et al., 2007; Carman, Edlund, & Damberg, 2000; Hupfer, Gloss, Schmieder, & Grossart, 2008; Turner, Mahieu, & Condron, 2003) using the well-established equation formulated by Fluck (1963). Moreover, based on specific chemical shifts, 31P NMR spectroscopy has been applied to study the local structure and composition of native casein micelles (Rasmussen, Sorensen, Petersen, Nielsen, & Thomsen, 1997; Thomsen, Jakobsen, Nielsen, Petersen, & Rasmussen, 1995) and the phosphorylated compounds present in milk (Belloque, De la Fuente, & Ramos, 2000). As a recent study demonstrated the feasibility of magic angle spinning (MAS) solidstate 31P NMR to analyse the phosphorylated compounds of

C. Rulliere et al. / International Dairy Journal 28 (2013) 102e108

a home-made semi-hard cheese at a molecular scale (RondeauMouro, Gobet, Mietton, Buchin, & Moreau, 2009), it appeared interesting in this study to use 31P NMR spectroscopy and ion chromatography as complementary methods to study the composition of polyphosphate salts and their interaction with dairy matrices. The objective of this research was thus to investigate the composition of a commercial sodium polyphosphate blend used in some dairy foods by ion chromatography and 31P NMR methods, as well as its potential interactions with proteins and minerals. Firstly, the composition of the polyphosphate blend in different chain lengths was studied in aqueous solutions. Then, 2-dimensional MAS 31P NMR was used to investigate polyphosphate interactions within two dairy matrices; (i) milk and (ii) processed cheese, at two stages of manufacture, i.e., before and after heat treatment. 2. Materials and methods 2.1. Sample preparation A commercial polyphosphate blend (JohaÒ, BK Giulini Chemie GmbH & Co., Ladenburg, Germany) was analysed in aqueous solutions as well as in milk and processed cheese samples. According to the supplier, this blend contained a mix of sodium polyphosphate, citrate, pyrophosphate and orthophosphate. However, the exact composition and the proportion of the different chain lengths were not stated. Aqueous solutions of polyphosphates were prepared by dissolving the commercial polyphosphate blend in deionised water at 0.5% (w/w). Solutions had a pH of 6.95  0.01. To investigate the impact of pH on the polyphosphate blend composition, polyphosphate aqueous solutions were analysed at pH 5.60 and 6.95; pH adjustments were made with 5 M HCl (Sigma Aldrich, Steinheim, Germany) using a pH-meter (Mettler-Toledo, Schwerzenbach, Switzerland). Milk solutions were prepared from skimmed milk powder (Lactalis Ingredients, Vern sur Seiche, France), dissolved in deionised water at 10% (w/w) and stirred for 2 h at 20
C. Afterwards, the commercial polyphosphate blend was added at a final concentration of 0.5% (w/w) corresponding to a polyphosphate/calcium ratio of 4 (w/w). The pH was adjusted to 5.60  0.01 to reproduce processed cheese pH conditions. Spreadable processed cheese samples containing 40% dry matter and 50% fat in dry matter, with or without the polyphosphate blend, were purchased from a French processed cheese factory. The polyphosphate blend concentration was added at a concentration of 2% (w/w) to conserve the same polyphosphate/ calcium ratio as in milk samples. Hard cheese, butter and skimmed milk powder from the same batch of ingredients were used as raw materials for their production and mixed together at 37
C before heat treatment. The mix was then sterilised for 5 s at 140
C, kept for 10 min at 80
C and cooled down to 4
C. Samples containing the polyphosphate blend were taken before and after heat treatment and samples without polyphosphates were taken after heat treatment. All samples had a pH of 5.60  0.05 and were stored for 3 days at 4
C before analyses.

103

The anion self-regenerating suppressor ASRSÒ 300-4 mm was operated in the auto-suppression recycle mode. Dionex Chromeleon software 6.8 was used for data acquisition and instrument control. The 4  50 mm guard and 4  250 mm analytical columns used were the Dionex IonPacÒ AGll HC and IonPacÒ ASll, respectively. A Dionex anion trap column (ATC-3 IonPac) was placed between the eluent pump and the injection valve. Aqueous solutions of polyphosphate were diluted by a factor of 1:1000 and then filtered through 0.2-mm filters (PALL GHP Acrodisc, New York, USA) before ion chromatography measurements. Disodium hydrogen phosphate (Na2HPO4), tetrasodium pyrophosphate decahydrate (Na4P2O7, 10H2O), pentasodium tripolyphosphate (Na5P3O10), trisodium trimetaphosphate (Na3P3O9) and trisodium citrate dehydrate (Na3C6H5O7) were purchased from Sigma Aldrich (Steinheim, Germany) and used as standard solutions at 0.001%, 0.002% and 0.005% (w/w) of phosphorus oxide. For all chromatography analyses, a 60-min gradient program was used according to the Baluyot & Hartford method (1996). Briefly, a sample volume of 50 mL was injected at a constant flow rate of 1 mL min1 in a NaOH gradient increasing from 20 to 140 mM over the initial 47-min period. Each measurement was made in triplicate. 2.3.

31

P NMR measurements

Measurements with 31P NMR were made in duplicate in aqueous solutions of polyphosphate at 0.5% (w/w) at pH 5.60 and 6.95, as well as in milk and processed cheese at pH 5.60. 31 P NMR spectra were recorded on a Bruker DMX 400 spectrometer (Bruker, Karlsruhe, Germany) operating at a phosphorus frequency of 161.98 MHz. A double resonance H/X CPMAS 4 mm probe was used for single pulse excitation magic angle spinning experiments (SPEMAS). The MAS rate was fixed at 9 kHz for analyses in water and 5 kHz in milk and processed cheese samples. Each acquisition was recorded at a temperature of 298 K (1). The SPEMAS experiments used a 90
phosphorus pulse of 4 ms, a 30 s recycling delay and 230 ms acquisition time. The 31Pe31P 2D-COSY and 31Pe31P 2D-TOCSY spectra were acquired with a spectral window of 6493 Hz in both dimensions, 4 K data points and increments of 192 and 256, respectively. Chemical shifts were referenced to 85% (w/w) phosphoric acid solution resonating at 0 ppm. The calculation of length of the longest polyphosphate chain was based on the following equation (Fluck, 1963):

n ¼

ðPP1 þ PP2 þ PP3 þ PPnÞ  2 PP1

(1)

where PP1 corresponded to phosphates at the terminal position, PP2 and PP3 to phosphates in the second and third position, and PPn to the inner phosphates. Due to the complexity of the matrices investigated in the present study, a new version of this equation was developed. Details of the modifications applied to equation (1) are presented in Section 3 (Results). 3. Results

2.2. Ion exchange chromatography 3.1. Ion chromatography measurements in aqueous solutions The composition of polyphosphate aqueous solutions in different phosphate chain lengths at pH 5.60 and 6.95 was analysed with a Dionex ICS 2500 chromatography system (Thermo Scientific, Illkirch Cedex, France). The system was equipped with a GP50 Gradient pump, an ED50 Pulsed Electrochemical Conductivity detector, a DS3 Conductivity Cell and an AS 50 automated sampler.

Fig. 1 displays the chromatogram obtained for the polyphosphate blend at pH 5.60 in an aqueous solution. Similar results were obtained for each repetition and for samples at pH 6.95 (not shown). As expected, a decrease in pH from 6.95 to 5.6 at room temperature did not influence the polyphosphate composition. In

104

C. Rulliere et al. / International Dairy Journal 28 (2013) 102e108

3.2.

31

P NMR measurements in aqueous and dairy matrices

3.2.1. Measurements in aqueous solutions containing the polyphosphate blend at pH 6.95 3.2.1.1. Signal assignment. The 31P NMR, 31Pe31P 2D-COSY and Pe31P 2D-TOCSY NMR spectra of the polyphosphate blend in aqueous solutions at pH 6.95 are shown in Figs. 2a, 3a and 3b, respectively. The two-dimensional sequences, 2D-COSY and 2D-TOCSY, were used to detect coupled pairs of phosphorus atoms and correlations between phosphorus nuclei along the chain. Complex series of signals were identified in three chemical shift zones: (1) around 2 ppm, (2) between 5 and 8 ppm, and (3) around 21 ppm. The signal around 2 ppm was due to the presence of P1 (Roberts, Ray, Wade-Jardetzky, & Jardetzky, 1980), which showed no connectivity on the 2D spectra (not shown in Fig. 3a and b). Assignment of the singlet around 7.1 ppm was coherent with the presence of P2 (Turner et al., 2003), which also displayed no cross-pick on 2D spectra due to its conformational and structural symmetry. Signals between 5 and 8 ppm corresponded to phosphates at the terminal position, also named PP1. Indeed, according to Turner et al. (2003), who determined the polyphosphate composition of temperate pasture soils by 31P NMR, PP1 resonates at between 4 and 10 ppm depending on the pH. As shown in Fig. 3a, these PP1 were associated with different signals between 19.5 and 20.6 ppm, corresponding to phosphates in the second position or penultimate phosphates (PP2). PP2 signals were connected or not to a PP3 triplet resonating in the 21 ppm zone, revealing the presence or absence of polyphosphates with more than three phosphates. The signal around 21.4 ppm was assigned to inner phosphates (PPn), as shown by several authors for polyphosphates in soil and aquatic sediments (Hupfer et al., 2008; Turner et al., 2003). The high intensity observed did not permit detection of correlations with PP3 signals in the 2D-TOCSY spectrum. Fig. bb showed that the terminal phosphates PP1 for long polyphosphate chains (more than three phosphates) showed cross peaks with PP2, PP3 and PPn. The PP1 doublets centred at 5.14, 5.18 and 5.25 ppm (Fig. 2a) were coupled in the TOCSY spectrum (Fig. 3b) with PP2 signals resonating between 19.8 and 20.3 ppm, PP3 peaks around 20.93 ppm and two PPn at 21.33 and 21.40 ppm. The TOCSY spectrum was also characterised by PP1 signals coupled with a PP2 triplet without correlation with PP3 nor PPn. This result is characteristic of triphosphate structures. As shown in the 2D-COSY spectrum (Fig. 3a), the PP2 signals of these triphosphates were characterised by special hyperfine structures, related to their fine structure and apparently depending on the nature of the cation in interaction with phosphates. The PP1 doublet centred at 6.47 ppm was coupled with the PP2 triplet centred at 20.33 ppm (2JPP ¼ 19.6 Hz). Another triphosphate was characterised by two terminal phosphates exhibiting two overlapping doublets centred at 5.85 ppm and 5.87 ppm (2JPP ¼ 13.5 Hz), coupled with the same PP2 centred at 20.19 ppm. These two triphosphates were distinguished by the nature of the terminal phosphates, depending on the cations: PP1ePP2ePP1 and PP1ePP2ePP10. This last structure, which was composed of two different terminal phosphates, displayed a smaller J-coupling constant compared to the symmetrical triphosphate. The J coupling constants measured on the PP1 multiplets ranged from 13 to 19 Hz. These results were in agreement with values determined in adenosine triphosphate from human brain (Jung et al., 1997). The signal at 21.44 ppm may be assigned to P3c (Lack, Dulong, Picton, Le Cerf, & Condamine, 2007). This trimetaphosphate form has been 31

Fig. 1. Ion chromatogram of a commercial polyphosphate blend in aqueous solution at pH 5.60.

fact, hydrolytic degradation of phosphates only happens at very acidic pH (McBeath et al., 2007). Standard solutions of orthophosphate (P1), pyrophosphate (P2), triphosphate (P3), trimetaphosphate (P3c) and citrate allowed peak assignment. As no polyphosphates with chains longer than three (>P3) were commercially available, it was then considered that each peak corresponded to a phosphate chain length and that this length increased according to its order of appearance. This peak assignment was confirmed by fitting a logarithmic curve based on the retention time of P1, P2 and P3 to our results, according to the work of Ohtomo, Sekiguchi, Mimura, Saito, and Ezawa (2004). Phosphate chains containing up to 18 phosphorus atoms were then detected (Fig. 1). A linear relationship between peak area and concentration of phosphorus was established for the short phosphate chains (SPC). The peak area for each chain length longer than P3 was assumed to be proportional to their concentration, in a similar manner to SPC. The proportion of each phosphate species was calculated by dividing the peak area of the specific species by the total area detected in the polyphosphate blend. Table 1 shows the polyphosphate average composition determined by ion chromatography for the commercial salt. The same composition was obtained at both pH 6.95 and 5.60.

Table 1 Percentage of the different phosphate chain lengths contained in the commercial polyphosphate blend dissolved in aqueous solution and determined by ion chromatography and 31P NMR.a Chain lengthb

Percentage of each phosphate form Ion chromatography pH 6.95 or 5.60

P1 P2 P3 P3c P4 P5 P6 SPC

6.9 31.9 13.4 2.7 13.5 10.5 21.1 52.2

       

1.0 1.2 1.0 0.3 0.9 0.9 1.8 1.2

31

P NMR

pH 6.95

pH 5.60

5.4 26.5 18.1 n.d.c n.d. 4.7 45.3 50  1

1.1 16.6 18 n.d. n.d. 5 59.2 35.7  1

a All results obtained by ion chromatography measurements are mean values (n ¼ 3)  standard deviation. Each result obtained by 31P NMR represents a 2% error on the signal integration. b P1, orthophosphate; P2, pyrophosphate; P3, triphosphate; P3c, trimetaphosphate; P4, tetraphosphate; P5, pentaphosphate; P6, polyphosphates with six or more phosphorus atoms; SPC, short phosphate chains from P1 to P3. c not determined by 31P NMR; the percentage of these species are included in polyphosphates P6.

C. Rulliere et al. / International Dairy Journal 28 (2013) 102e108

Fig. 2. 1-D

31

P NMR spectrum of the commercial polyphosphate blend in water at (a) pH 6.95 and (b) pH 5.60.

detected by ion chromatography, especially after heat treatment of polyphosphates in presence of calcium (Rulliere et al., 2012). The sample was also characterised by the presence of a low content of P5. 2D spectra and in particular TOCSY spectra (Fig. 3b) displayed a PP3 triplet centred at 20.73 ppm, with a 2J constant of about 15 Hz correlated to a PP2 signal centred at 20.1 ppm with PP1 chemical shifts, at 5.38 ppm. It should not be excluded that signals from phosphorus pentoxide (P4O10) appeared in these spectra. However, if the ionic distribution of this species was not symmetrical, correlations between two PP2 should have appeared on spectra. Due to the overlapping signals in the PP2 region, it was difficult to observe these correlations. Moreover, in the case of symmetrical P4O10, the phosphate chemical shift should have been similar to inner phosphate signals of linear polyphosphates (PPn) around 21.4 ppm. It is also known that phosphorus pentoxide is a cyclic molecule which is easily converted to phosphoric acid in the presence of water. For all these reasons, the content of phosphorus pentoxide was deduced from the ion chromatography studies.

Fig. 3. (a)

105

31

Pe31P COSY and (b)

31

3.2.1.2. Determination of the polyphosphate chain lengths The detailed integration of the 1D-31P spectrum led to the determination of the polyphosphate composition (% of the different chain lengths) presented in Table 1. The polyphosphate compositions determined by ion chromatography and by 31P NMR spectroscopy at pH 6.95 were similar, with regard to SPC, representing 52.2 and 50.0% respectively of the total phosphate blend. For long chains (>P3), the proportion of inner PPn phosphates was overestimated by NMR, due to the overlapping of signals from phosphorus pentoxide and P3c (cyclic triphosphate in which each phosphate shares two oxyanions with another phosphate). The longest chain of polyphosphate was estimated in this work according to the following equation based on equation (1) (see 2.3 31 P NMR measurements):

n ¼

ðR1*fPP1g þ R2*fPP2; PP3; PPngÞ  2 R1*fPP1g

(2)

This new equation took into account the total integration of signals resonating between 5 and 8 ppm (integral noted {PP1} in equation (2)), which represented the zone of the terminal PP1 of

Pe31P TOCSY NMR spectra of the commercial polyphosphate blend at pH 6.95.

106

C. Rulliere et al. / International Dairy Journal 28 (2013) 102e108

polyphosphates but also signals from P3 and P5 determination. The contribution of P3 and P5 to the total signal integral {PP1} can easily be determined in the zone between 5 and 8 ppm, where the terminal phosphates of P3 and P5 are well resolved and assigned using 2D spectroscopy. This contribution corresponds to 1  R1, which is used to determine 1  R2, the contribution of P3 and P5 on PP2 and PP3 integrals in the spectral zone between 19 and 23 ppm showing a total integral noted {PP2, PP3, PPn}. For this polyphosphate blend analysis, R1 is equal to 0.2889, as it corresponds to 28.89% of the signal integral between 5 and 8 ppm, while R2 (¼ 0.7808) accounts for 78.08% of the total integral noted {PP2, PP3, PPn} between 19 and 23 ppm. Based on this new equation, the longest chain of polyphosphate at pH 6.95 was estimated at 18 (Table 2), which was in agreement with the result obtained by ion chromatography (Fig. 1). 3.2.2. Measurements in aqueous solutions of the polyphosphate blend at pH 5.60 At pH 5.60, the 31P NMR spectrum indicated a loosening of the signal resolution (signal broadening), as shown in Fig. 2b. As a consequence, it was difficult to assign and to integrate the proportion of each phosphate form, leading to an incorrect estimation of the polyphosphate composition by 31P NMR at this pH (Table 1). Using a phosphate buffer solution at pH 5.60 as an external chemical shift reference, the signal detected at 0 ppm on Fig. 2b was easily assigned to orthophosphate (P1). The intensity of this signal was significantly reduced. In fact, the decrease in pH favoured the formation of hydrogenophosphates ðHPO2 4 Þ and dihydrogenophosphate (H2PO4), due to the release of sodium ions. Chemical exchanges between these two forms may have led to an increase of the P1 signal width and so to a decrease of its intensity. In addition, a significant shielding of P3, P5, and PP1 signals was observed at pH 5.60. These results indicated that the recording of polyphosphate spectra at neutral pH was important to help the assignment of phosphates and the estimation of ratios R1 and R2. However, calculation of the polyphosphate longest chain length at pH 5.60

using equation (2) gave a value of 16, consistent with the length estimated by NMR at pH 6.95 (Table 2). 3.2.3. Measurements in dairy matrices: milk and processed cheese The impact of polyphosphate addition to complex matrices such as milk and processed cheese at different stages of its manufacture was followed by 31P NMR (Fig. 4aec). A spectrum of processed cheese without polyphosphates, after heat treatment, is shown in Fig. 4d. As already shown by Andreotti, Trivellone, and Motta (2006) for buffalo milk samples, phosphates from caseins were observed at 0.64 ppm, while small signals characteristic of phospholipids resonated between 1 and 0 ppm, (precisely at 0.21 ppm, 0.51 ppm and 0.86 ppm). A large signal for caseins was observed for processed cheese without polyphosphates, compared to processed cheese or milk containing these additives. This result might be due to the reduced molecular mobility of caseins in absence of polyphosphates. Moreover, a large variation in the polyphosphate chemical shifts (PP1, PP2, PP3 and PPn), as well as signal width, was observed in these dairy matrices, compared to aqueous solutions. This indicated a modification of the chemical environment of the polyphosphate. This change was taken into account by the Delta parameter (the missing percentage) which represents the difference between the total integral (in %) for polyphosphate signals (P  2) measured in water at pH 5.60 (98%) and the same integral in milk and processed cheese (23 and 26%, respectively) (Table 2). This difference between water and dairy matrices represented a loss of signal resulting from a reduction of the phosphate mobility due to calciumephosphate complex formation. According to this finding, around 38% of polyphosphates were associated with calcium in milk, whereas 75% and 72% interacted in processed cheese before and after cooking respectively.

Table 2 Percentage of each phosphate form determined by 31P NMR in water, milk and processed cheese; total percentage of polyphosphates (with chain length  2); polyphosphate chain length calculated using the proposed equation 2 (in the text) corrected or not with Delta (the missing percentage, corresponding to polyphosphates in interaction with ions and/or the matrix). Phosphate form

Water

Milk

Processed cheese Before cooking

Phospholipids Phosphoproteins Orthophosphates (P1) Pyrophosphates (P2) P3, P4 PPn (5) Unknown Total polyphosphates  P2 (%) Polyphosphate (PPn) chain length Deltaa (%) Corrected PPn chain lengthb

After cooking

pH 6.95

pH 5.6

pH 5.6

pH 5.6

e e 5.4 26.5 18.1 48.3 1.7 93

e e 1.1 16.6 18 63.3 0.9 98

5.4 18.4 29.7 7.4 13 23.7 2.4 60

e 17.1 61.4 0 3.3 15.2 2.9 23

3.4 16.6 58 8.5 2.7 9.2 1.6 26

18

16

12

27

20

e

e

38 16

75 47

72 35

a Difference between the percentage of the total integral of the polyphosphate signals measured in water at pH 5.6 (98%) and the total integral at the same pH but in milk (60%) or processed cheese before (23%) or after (26%) cooking. b Corrections taking into account Delta, the missing percentage.

Fig. 4. 1-D 31P NMR spectra of the commercial polyphosphate blend in (a) milk at pH 5.60, (b) processed cheese at pH 5.60 before cooking, (c) processed cheese at pH 5.60 after cooking and (d) processed cheese without polyphosphates at pH 5.60 after cooking.

C. Rulliere et al. / International Dairy Journal 28 (2013) 102e108

The introduction of the Delta parameter in equation (2) confirmed that the longest chain of polyphosphates in milk contained around 16 phosphorus atoms (Table 2). Two spectral changes appeared for polyphosphates in the processed cheese matrix before cooking, comparing to its spectrum in milk (Fig. 4a and b). Firstly, P2 signals at 9.18 ppm disappeared, the association of these species with calcium ions could explain this result, as discussed in Section 4. Secondly, the chemical environment of the polyphosphate chains changed with the emergence of a higher signal for the polyphosphate inner phosphates at 21.44 ppm. After introduction of the Delta parameter into equation (2), the longest chain of polyphosphate was estimated to contain around 47 phosphates (Table 2). As the condensation of polyphosphates does not seem possible at ambient temperature, the apparent increase of the polyphosphate longest chain represents an artefact. In fact, the calculation of the chain length from the NMR integrals may not be applicable in this system, due to the complexity of interactions between components, a phenomenon that could result in larger signals with weak signal-to-noise ratio not quantifiable by NMR. After heat treatment, the P2 signal at 9.18 ppm reappeared (Fig. 4c), whereas polyphosphates completely disappeared. The hydrolysis of the long-chain phosphates into shorter ones may explain this, as hydrolysis of polyphosphates occurred in aqueous solutions heated above 100
C (Rulliere et al., 2012). However, no increase of P1 was observed in this present study. It is possible that hydrolytic degradation of long-chain phosphates stopped at the “pyrophosphate step”, due to the formation of complexes between proteins, calcium and pyrophosphate. This phenomenon was observed in the presence of calcium during hydrolysis of polyphosphates in aqueous solutions at pH 5.60 (Rulliere et al., 2012). Also, complex matrices such as processed cheese could protect the polyphosphate species towards total hydrolysis during heat treatment. The estimation by NMR of the polyphosphate chain length gave once again, after correction, a surprising value of 35. This result confirmed that the calculation of the chain length using equation (2) in complex matrices was wrong, due to an overestimation of the percentage of PPn. 4. Discussion Ion chromatography and 31P NMR gave comparable results concerning the polyphosphate composition of aqueous solutions at neutral pH. The amount of SPC in the commercial blend was estimated at 52.2% and 50.0%, respectively, by both techniques, with comparable ratios of each form (Table 1). Moreover, chain lengths up to 16 phosphates were successfully identified by both methods at neutral pH (Fig. 1 and Table 2). Thus, the new equation proposed (equation (2)), derived from the Fluck equation, was adapted for the determination of polyphosphate chain length by 31P NMR in aqueous solutions at neutral pH. This new equation took into account the specific contribution of the terminal phosphate signals (PP1) and the inner phosphates of polyphosphates (PP2, PP3, PPn) through ratios R1 and R2. At lower pH (5.60), the same results were obtained by ion chromatography for the polyphosphate blend composition. Indeed, when using ion chromatography, all phosphate species were converted into their negatively charged ionic forms by the addition of increasing amounts of sodium hydroxide. Thus, increasing the amount of protonated forms by decreasing the pH from 6.95 to 5.60 did not influence the results obtained with this method. This finding is promising for further analysis of dairy matrices like processed cheese, as this product has an average pH value of 5.60. In contrast, a lower pH value significantly impacted the determination of the polyphosphate composition by the 31P NMR

107

technique. The modification of the ionic environment due to the pH decrease strongly modified the 31P NMR spectrum. The chemical shifts observed, associated with a decrease in the signal resolution, led to an incorrect determination of the polyphosphate composition (SPC: 35.7%). However, these changes in the 31P NMR chemical signals as a function of the environment were useful in evaluating the interaction of polyphosphates with minerals and proteins in dairy matrices. Indeed, a loss of signal was observed for some phosphate forms in the 31P NMR spectra of skimmed milk and processed cheese when compared to the polyphosphate spectrum obtained in aqueous solution at pH of 5.60 (Delta, Table 2). This signal loss was due to interactions between added phosphates and dairy components such as calcium and permitted estimation of the amount of chelated forms. It was calculated that around 38% of polyphosphates were linked to calcium in milk at pH 5.60, whereas 75% and 72% interacted with this mineral in processed cheese before and after cooking respectively. Two main conclusions can be reached from these results. Firstly, as milk and processed cheese had the same ratio of added phosphate and endogenous calcium, polyphosphate chains seemed to present different chelating properties in these products. In milk, with 10e12% dry matter, the long phosphate chains seemed to chelate the calcium with a non-stoichiometric ratio (numerous calcium ions per molecule) as 60% phosphate remained unchelated. In processed cheese (40% dry matter) before cooking, as only 23% of polyphosphate remained unchelated, SPC seemed more involved in the chelating phase. The other interesting conclusion concerns the calcium chelation step during processed cheese manufacture. This reaction, also named “peptisation”, seemed to take place during the mixing step of ingredients before cooking, as the temperature increase didn’t change the percentage of phosphate chelated forms. The comparison of 31P NMR spectra of dairy matrices with and without polyphosphates in Fig. 4 showed differences concerning the mobility of organic phosphate, linked to casein molecules (signal at 0.64 ppm). In processed cheese without emulsifying salts, caseins are trapped in a three-dimensional calcified paracasein network which negatively impacts their emulsifying properties (Cernikova et al., 2010). In samples with polyphosphates, calcium is chelated by these complexing agents (De Kort, Minor, Snoeren, Van Hooijdonk, & Van Der Linden, 2009; Mekmene & Gaucheron, 2011; Udabage, McKinnon, & Augustin, 2001), increasing the degree of dispersion of the insoluble casein matrix, and thus casein mobility. After heat treatment, a reassociation of CaePhosphates complexes with the dispersed caseins could occur and may participate in the restructuring of the casein network. Indeed, Mizuno and Lucey (2005, 2007) showed that pyrophosphates, which are present in polyphosphate blends, could form complexes with caseins and calcium, leading to gelation of milk at pH 5.80. The formation of such complexes could explain the disappearance of the pyrophosphate signal in processed cheese before cooking. However, the pyrophosphate signal reappeared after heat treatment. A simple interpretation of this result would be that dissociation of the pyrophosphateecalcium complexes occurs after heat treatment. Indeed, thinner signals in the 31P NMR spectrum at 9.18 ppm (Fig. 4c) might be linked to the loosening of these interactions with bridging mineral ions. However, this hypothesis would mean that (i) dissociation could have occurred during the process without any change of pH and (ii) phosphates and calcium would not be involved in the protein network, as proposed by Mizuno and Lucey (2007). The most probable hypothesis for the presence of this thin signal after heat treatment would be the presence of an excess of pyrophosphate, due to the release of high amounts of this species after the hydrolysis of the long chains. This hypothesis would be in

108

C. Rulliere et al. / International Dairy Journal 28 (2013) 102e108

agreement with results obtained in aqueous solutions of polyphosphate in the presence of calcium, where pyrophosphate was the most abundant phosphate species after heating at 120
C at pH 5.60 (Rulliere et al., 2012). Thus, the signal detected may be linked to this excess of pyrophosphate, while pyrophosphate interacting with proteins through calcium ions would not be visible due to the signal broadening. The evaluation of the amount of pyrophosphate present in processed cheese samples could support this hypothesis. However, this quantification was not possible by 31P NMR, due to the failure to properly integrate the phosphate signals in this matrix. 5. Conclusion Ion chromatography appeared as an interesting method to analyse polyphosphate composition in aqueous solutions even at low pH. In contrast, the influence of environmental conditions on 31 P NMR measurements led to incorrect polyphosphate compositions at acidic pH values. However, 31P NMR, as a non-invasive method, permitted analysis of samples in their native state and was used to study interactions of phosphates with milk and processed cheese matrices. Thus, the complementarity of ion chromatography and 31P NMR has been demonstrated in this work. The comparison of processed cheese with and without polyphosphates by 31P NMR confirmed the role of these ingredients on solubilisation of casein. Moreover, the percentage of calcium chelated to phosphates was evaluated in milk and processed cheese at different steps of its manufacture and demonstrated that calcium chelation occurred in the first mixing step of the manufacture. Finally, 31P NMR results emphasised the role of pyrophosphate in processed cheese, as this species seemed to be trapped in complexes with proteins and calcium. Its concentration increased during the process through the hydrolysis of polyphosphate during heat treatment. However, further investigation is required to elucidate the precise role of this species after heat treatment and its potential interaction with caseins and calcium. The quantification of the amount of pyrophosphate, chelated with calcium or not, and interacting with protein or not, in complex dairy matrices would be useful in confirming these results. Acknowledgement The authors would like to thank Ms Terri Andon for editing the manuscript. They also acknowledge INRA of Nantes for the access to the NMR facilities of the BIBS platform. References Ahlgren, J., De Brabandere, H., Reitzel, K., Rydin, E., Gogoll, A., & Waldeback, M. (2007). Sediment phosphorus extractants for phosphorus-31 nuclear magnetic resonance analysis: a quantitative evaluation. Journal of Environmental Quality, 36, 892e898. Andreotti, G., Trivellone, E., & Motta, A. (2006). Characterization of buffalo milk by 31 P-nuclear magnetic resonance spectroscopy. Journal of Food Composition and Analysis, 19, 843e849. Baluyot, E. S., & Hartford, C. G. (1996). Comparison of polyphosphate analysis by ion chromatography and by modified end-group titration. Journal of Chromatography A, 739, 217e222. Belloque, J., De la Fuente, M. A., & Ramos, A. M. (2000). Qualitative and quantitative analysis of phosphorylated compounds in milk by means of 31P-NMR. Journal of Dairy Research, 67, 529e539. Caric, M., Gantar, M., & Kalab, M. (1985). Effects of emulsifying agents on the microstructure and other characteristics of process cheese e a review. Food Microstructure, 4, 297e312.

Carman, R., Edlund, G., & Damberg, C. (2000). Distribution of organic and inorganic phosphorus compounds in marine and lacustrine sediments: a 31P NMR study. Chemical Geology, 163, 101e114. Cernikova, M., Bunka, F., Pospiech, M., Tremlova, B., Hladka, K., Pavlinek, V., et al. (2010). Replacement of traditional emulsifying salts by selected hydrocolloids in processed cheese production. International Dairy Journal, 20, 336e343. De Kort, E., Minor, M., Snoeren, T., Van Hooijdonk, T., & Van Der Linden, E. (2009). Calcium-binding capacity of organic and inorganic ortho- and polyphosphates. Dairy Science and Technology, 89, 283e299. Ennis, M. P., O’Sullivan, M. M., & Mulvihill, D. M. (1998). The hydration behaviour of rennet caseins in calcium chelating salt solution as determined using a rheological approach. Food Hydrocolloids, 12, 451e457. Fluck, E. (1963). Die kernmagnetische Resonanz und ihre anwendung in der anorganischen chemie. Berlin, Germany: Springer-Verlag. Gaucher, I., Piot, M., Beaucher, E., & Gaucheron, F. (2007). Physico-chemical characterization of phosphate-added skim milk. International Dairy Journal, 17, 1375e1383. Hupfer, M., Gloss, S., Schmieder, P., & Grossart, H. P. (2008). Methods for detection and quantification of polyphosphate and polyphosphate accumulating microorganisms in aquatic sediments. International Review of Hydrobiology, 93, 1e30. Jung, W. I., Staubert, A., Widmaier, S., Hoess, T., Bunse, M., vanErckelens, F., et al. (1997). Phosphorus J-coupling constants of ATP in human brain. Magnetic Resonance in Medicine, 37, 802e804. Lack, S., Dulong, V., Picton, L., Le Cerf, D., & Condamine, E. (2007). High-resolution nuclear magnetic resonance spectroscopy studies of polysaccharides crosslinked by sodium trimetaphosphate: a proposal for the reaction mechanism. Carbohydrate Research, 342, 943e953. McBeath, T. M., Lombi, E., McLaughlin, M. J., & Bunemann, E. K. (2007). Polyphosphate-fertilizer solution stability with time, temperature, and pH. Journal of Plant Nutrition and Soil Science [Zeitschrift Fur Pflanzenernahrung Und Bodenkunde], 170, 387e391. Mekmene, O., & Gaucheron, F. (2011). Determination of calcium-binding constants of caseins, phosphoserine, citrate and pyrophosphate: a modelling approach using free calcium measurement. Food Chemistry, 127, 676e682. Mizuno, R., & Lucey, J. A. (2005). Effects of emulsifying salts on the turbidity and calciumephosphateeprotein interactions in casein micelles. Journal of Dairy Science, 88, 3070e3078. Mizuno, R., & Lucey, J. A. (2007). Properties of milk protein gels formed by phosphates. Journal of Dairy Science, 90, 4524e4531. Obritsch, J. A., Ryu, D., Lampila, L. E., & Bullerman, L. B. (2008). Antibacterial effects of long-chain polyphosphates on selected spoilage and pathogenic bacteria. Journal of Food Protection, 71, 1401e1405. Ohtomo, R., Sekiguchi, Y., Mimura, T., Saito, M., & Ezawa, T. (2004). Quantification of polyphosphate: different sensitivities to short-chain polyphosphate using enzymatic and colorimetric methods as revealed by ion chromatography. Analytical Biochemistry, 328, 139e146. Rasmussen, L. K., Sorensen, E. S., Petersen, T. E., Nielsen, N. C., & Thomsen, J. K. (1997). Characterization of phosphate sites in native ovine, caprine, and bovine casein micelles and their caseinomacropeptides: a solid-state phosphorus-31 nuclear magnetic resonance and sequence and mass spectrometric study. Journal of Dairy Science, 80, 607e614. Roberts, J. K. M., Ray, P. M., Wade-Jardetzky, N., & Jardetzky, O. (1980). Estimation of cytoplasmic and vacuolar pH in higher plant cells by 31P NMR. Nature, 283, 870e872. Rondeau-Mouro, C., Gobet, M., Mietton, B., Buchin, S., & Moreau, C. (2009). Identification and quantification of phosphorus in cheeses e methodological investigations by solid-state 31P NMR spectroscopy. In M. Guðjónsdóttir, P. Belton, & G. Webb (Eds.), Magnetic resonance in food science e Challenges in a changing world (pp. 126e135). Cambridge, UK: Royal Society of Chemistry. Rulliere, C., Perenes, L., Senocq, D., Dodi, A., & Marchesseau, S. (2012). Heat treatment effect on polyphosphate chain length in aqueous and calcium solutions. Food Chemistry, 134, 712e716. Sadlikova, I., Bunka, F., Budinsky, P., Barbora, V., Pavlinek, V., & Hoza, I. (2010). The effect of selected phosphate emulsifying salts on viscoelastic properties of processed cheese. LWT e Food Science and Technology, 43, 1220e1225. Thomsen, J. K., Jakobsen, H. J., Nielsen, N. C., Petersen, T. E., & Rasmussen, L. K. (1995). Solid-state magic-angle spinning 31P-NMR studies of native casein micelles. European Journal of Biochemistry, 230, 454e459. Turner, B. L., Mahieu, N., & Condron, L. M. (2003). The phosphorus composition of temperate pasture soils determined by NaOHeEDTA extraction and solution 31P NMR spectroscopy. Organic Geochemistry, 34, 1199e1210. Udabage, P., McKinnon, I. R., & Augustin, M. A. (2001). Effects of mineral salts and calcium chelating agents on the gelation of renneted skim milk. Journal of Dairy Science, 84, 1569e1575. Weiserova, E., Doudova, L., Galiova, L., Zak, L., Michalek, J., Janis, R., et al. (2011). The effect of combinations of sodium phosphates in binary mixtures on selected texture parameters of processed cheese spreads. International Dairy Journal, 21, 979e986.