at SciVerse ScienceDirect

Food Hydrocolloids 34 (2014) 46e53

Contents lists available

Food Hydrocolloids

journal homepage: www.elsevier .com/locate/ foodhyd

Physico-chemical properties of casein micelles in unheated skimmilk concentrated by osmotic stressing: Interactions and changesin the composition of the serum phase

Pulari Krishnankutty Nair, Marcela Alexander, Douglas Dalgleish, Milena Corredig*

Department of Food Science, University of Guelph, Guelph, Ontario, Canada N1G 2W1

a r t i c l e i n f o

Article history:Received 10 August 2012Accepted 3 January 2013

Keywords:Casein micellesConcentrated milkOsmotic stressingRheology

* Corresponding author. Tel.: þ1 519 824 4120; faxE-mail addresses: milena.corredig@uoguelph

(M. Corredig).

0268-005X/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.foodhyd.2013.01.001

a b s t r a c t

The changes in processing functionality of concentrated milk are caused by a number of factors, amongstthe most important, the ionic equilibrium and the increase in the interactions between the casein mi-celles because of their increased volume fraction. The objective of this work was to characterize thephysico-chemical properties of casein micelles as a function of their volume fraction, by using osmoticstressing as a non-invasive method to obtain concentrated milk, in the attempt to preserve the ionicbalance during concentration. Osmotic concentration was carried out for 18 h at 4 �C, using differentconcentrations of polyethylene glycol dissolved in permeate as the stressing polymer. The viscosity of theconcentrated milk could be predicted using established rheological models, when the changes occurringto the viscosity of the serum phase were taken into account. Both Eilers and Mendoza equations pre-dicted a maximum packing volume fraction of 0.8 for the casein micelles. After concentration up to 20%protein, the casein micelles did not show a change in their size upon redilution. Light scattering mea-surements carried out using diffusing wave spectroscopy without dilution suggested that casein micellesbehave as hard spheres with the characteristic of free diffusing Brownian particles up to a volumefraction of 0.3, and restricted motion at higher concentrations. Results of total and soluble calciumsuggested release of colloidal calcium phosphate from the micelles at volume fractions >0.35. Thisresearch brings new insights on the changes occurring in skim milk during concentration.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

The caseins are a family of calcium binding phosphoproteinsthat comprise about 80% of the protein present in milk with overallconcentration of approximately 25 g L�1and the majority (Fox,2003) exist as particles of colloidal dimensions generally referredto as casein micelles. The remainder of the milk protein is wheyproteins, which are composed mostly of b-lactoglobulin, a-lactal-bumin and bovine serum albumin (Fox, 2003). The casein micellesare of great interest for colloid chemists as they represent theresponse of nature to the need to deliver a high level of calcium tothe neonate. These micelles are highly hydrated colloids (>4 g ofwater per g of protein), composed of a core of highly phosphory-lated caseins (as and b caseins) interacting with calcium phosphate,with a stabilizing layer of k-casein on the surface (Dalgleish, 2011).

: þ1 519 824 6631..ca, mcorredi@uoguelph.ca

All rights reserved.

The casein micelles contain calcium phosphate nanoclusters boundto the phosphoserine groups of the aS1, aS2, and b-caseins. Thiscolloidal component is in equilibrium with the calcium and phos-phate present in the soluble phase (Holt, 2002). The casein micellesare polydisperse in size (between 60 and 500 nm in diameter). Inbovine milk, the content of k-casein decreases as the micellar sizeincreases, balanced by an increase in the content of b-casein, whilethe proportions of aS1 and aS2-casein are independent of micellesize (Dalgleish, Horne, & Law, 1989). The micellar calcium phos-phate is distributed uniformly inside the micellar particles, inclusters of about 2.5 nm in diameter (Marchin, Putaux, Pignon, &Léonil, 2007).

The understanding of the supramolecular structure of nativecasein micelles and their changes during processing still holdsmany challenges. This aggregated structure is very dynamic as itresponds in various ways to environmental changes as well as tothe presence of the other components present in milk (Horne,2009). In the past years there has been an increased interest inthe study of the effects of concentration on the structure andprocessing functionality of casein micelles. With concentration, the

P. Krishnankutty Nair et al. / Food Hydrocolloids 34 (2014) 46e53 47

serum compositionmay change as well as the interactions betweenthe colloidal particles, perhaps affecting the structure and functionof the casein micelles.

The viscosity of milk changes with concentration in a non-linearfashion (Snoeren, Damman, & Klok, 1982); as concentration in-creases there is a change from Newtonian to non-Newtonianbehaviour (Walstra & Jenness, 1984). The rheological properties ofconcentrated milk prepared by either heat evaporation (Ruiz &Barbosa-Cánovas, 1997; 1998), ultrafiltration (Karlsson, Ipsen,Schrader, & Ardö, 2005; Pignon et al., 2004) or powder reconsti-tution (Anema, 2009; Dahbi, Alexander, Trappe, Dhont, &Schurtenberger, 2010) have been investigated. In addition to this,information is available on age gelation or renneting of ultra-filtered, concentrated milk (Bienvenue, Jimenez-Flores, & Singh,2003; Karlsson, Ipsen, & Ardö, 2007) or evaporated milk,(Harwalkar, Beckett, McKellar, Emmons, & Doyle, 1983; Hwang, Lee,Park, Min, & Kwak, 2007) because of the technological implicationsin dairy processing. However, little research has been carried out onthe fundamentals of the interactions of casein micelles in concen-trated milk. When membrane filtration is employed as a means toconcentrate milk, shear effects and membrane fouling may occur,while when concentrating by evaporation, heating is applied andthe serum composition may change. These differences in process-ing history make a study on the properties of the casein micellesduring concentration quite challenging, and for this reason,knowledge of the fundamental aspects of how the physico-chemical properties of caseins evolve during concentration ofmilk is very limited.

Osmotic stress is an alternative process for concentrating milkwhich may be considered non-invasive. This technique has alreadybeen successfully employed to characterize concentrated milksystems by other groups (Bouchoux, Cayemitte, Jardin, Gésan-Guiziou, & Cabane, 2009; Bouchoux, Debbou, et al., 2009). By us-ing milk permeate (the serum phase of milk) it is possible tomaintain the original serum environment for the casein micelles.This has important implications for the equilibrium between thecolloidal and the soluble calcium, which may otherwise change,with consequences to proteineprotein interactions within thecasein micelle. In short, this allows the investigation of the funda-mental aspects of concentrated systems as a function of volumefraction of the casein micelles, without applying shear (as inmembrane filtration) or affecting the serum composition. Adetailed study of the dependence of the osmotic pressure of sodiumcaseinate solutions as a function of concentration has been pub-lished (Farrer & Lips, 1999), encompassing the dilute, semi-diluteand highly concentrated regimes. The relative viscosity of sodiumcaseinatewas shown to increase gradually with concentration up toabout 10% (w/w) and then steeply after (Farrer & Lips, 1999). Recentwork (Bouchoux, Cayemitte, et al., 2009) explored model disper-sions of phosphocaseinate over a wide range of casein concentra-tions (from 20 to 500 g L�1). The results were described in terms ofthree compression regimes: dilute, transition and concentratedregimes. The same authors also studied in detail the rheologicalbehaviour of the system (Bouchoux, Debbou, et al., 2009),concluding that when casein micelles are below close-packingconditions, these protein particles behave like polydisperse hard-spheres. At concentrations close to close-packing (178 g L�1), theelastic modulus increases rapidly and the system progressivelyshows a frequency independent elastic modulus.

The objective of this work was to extend the knowledge of thecolloidal properties of casein micelles in untreated skim milk as afunction of volume fraction. In addition to the determination of therheological properties, a detailed composition analysis of the serumphase was carried out as well as a study of the colloidal propertiesof the casein micelles using diffusing wave spectroscopy, a light

scattering technique which does not require dilution of the sample.This work will allow for a better understanding of the effects ofconcentration on the physico-chemical properties of the caseinmicelles, and may strengthen our current understanding of thestructure of these colloidal particles.

2. Materials and methods

2.1. Skim milk and permeate preparation

Sodium azide (0.2 g L�1) was added to fresh raw milk (Universityof Guelph Dairy Research Station, Ponsonby, Ontario, Canada) topreventmicrobial growth. Skimmilkwas prepared by centrifuging at4000� g for 25minat 4 �C (J2-21 centrifuge, BeckmanCoulter CanadaInc, Mississauga, Canada) and filtering four times through Whatmanfibreglass filter (Fisher Scientific, Mississagua, Ontario, Canada). Ul-trafiltrationpermeatewasprepared byultrafiltration of reconstitutedskim milk powder (100 g L�1 solids) (Gay Lea Foods Cooperative,Guelph, Ontario, Canada) by passing it through an OPTISEP� Filtermodule (Smartflow Technologies, Apex, NC, USA) with 10 kDa mo-lecular mass cutoff at ambient temperature. The average ioniccomposition ofUFpermeate is:w20mMNaþ,w40mMKþ,w10mMCa2þ, w30 mM Cl�, w10 mM phosphate, w10 mM citrate, in agree-ment with previous reports (Jenness & Koops, 1962).

2.2. Milk concentration

Polyethylene glycol (PEG) with a molar mass of 35,000 Da(Fluka, Oakville, Ontario, Canada) was used as the stressing poly-mer. PEG is a flexible, water-soluble polymer, and preliminary ex-periments showed that it has no specific interactions with calcium.This polymer is widely used to obtain high osmotic pressures andthe systems are well characterized (Koning, van Eendenburg & DeBruijne, 1993). All experiments were carried out by dispersingPEG in permeate (prepared as described above) containing 0.2 g L�1

sodium azide as a bacteriostatic. The use of permeate ensured thatthe ionic composition remained similar across the dialysis mem-brane, which was a standard regenerated cellulose Spectra/Por 1(Fisher Scientific, Whitby, Ontario, Canada) with a molecular masscut-off of 6e8 kDa. This pore size ensured the exchange of water,ions, and lactose but not caseins or PEG. Before experiments, thedialysis membranes were washed in MilliQ water and conditionedin milk permeate. Milk samples (40 mL) were inserted in dialysistubing, and immersed in a 1 L permeate solution containingdifferent PEG concentrations. The dialysis was conducted for 18 h at4 �C, to minimize sample degradation. Significant degradation mayoccur conducting the dialysis of unheated milk at 20 �C (Bouchoux,Cayemitte, et al., 2009), and in the present experiments, milk wasuntreated. The pH of all dispersions remained unchanged duringthe experiment.

The volume fraction was calculated by assuming the volumi-nosity of the micelles to be constant (at 4.4 mL/g (Holt, 1992)),throughout the concentration range of our experiment.

2.3. Separation of the serum phase

The concentrated milk samples were equilibrated at roomtemperature for 1 h before serum separation. Preliminary experi-ments were used to determine a suitable centrifugation speed. Thecentrifugation speed chosen was the minimum required to effec-tively deposit the casein micelles as a firm pellet. During pre-liminary trials, the serum was also measured by dynamic lightscattering with no further dilution, and very little scattering wasdetected, suggesting that a serum devoid of casein micelles wasobtained at this centrifugation speed. Soluble whey proteins for the

PEG (g L-1)

0 20 40 60 80 100 120 140

Prot

ein

(g L

-1)

0

50

100

150

200

250

300Vo

lum

e Fr

actio

n

0.0

0.2

0.4

0.6

0.8

Fig. 1. Amount of protein in milk, after 18 h of dialysis in permeate, as a function ofdifferent PEG concentrations. Corresponding casein micelles volume fractions are alsoindicated. Error bars indicate standard error of three independent trials.

P. Krishnankutty Nair et al. / Food Hydrocolloids 34 (2014) 46e5348

present experiment were therefore defined as those that did notsediment from the milk during ultracentrifugation at 100,000� gfor 1 h at 20 �C in a Beckman Coulter Optima� LE-80K ultracen-trifuge with rotor type 70.1Ti (Beckman Coulter Canada Inc., Mis-sissauga, Canada). The clear supernatant was carefully removedfrom the pellets using a pasteur pipette, and was given twosequential filtrations using 0.45 mm and then 0.22 mm (syringedriven filters, Fisher Sci.) and then analysed for protein, calciumcontent, viscosity, refractive index and particle size. Protein analysiswas carried out using a Dumas nitrogen analyzer (FP-528, Leco Inc.Lakeview Avenue, St. Joseph, MI) and the protein concentrationwasdetermined using 6.38 as conversion factor.

2.4. Diffusing wave spectroscopy (DWS)

DWS allows the investigation of the static and dynamic behav-iour of colloidal particles in fairly concentrated suspensions (Weitz,Zhu, Durian, Gang, & Pine, 1993). Static properties of the sampleswere measured via the value of the photon transport mean freepath, l*, which represents the length scale over which the directionof the light passing through a sample has been fully randomized.Turbidity was measured as the inverse of the l*. In addition, theapparent diffusion coefficient (D) was obtained by probing thecolloidal mobility over a very short time scale. TheD is derived fromthe characteristic decay time of the intensity auto-correlationfunction and can be used to calculate the apparent particleradius, via StokeseEinstein relation, when the colloidal particlesare freely diffusing.

Once the particle dynamics are changed to a sub-diffusive mo-tion (e.g., the point after which a liquid-like colloidal suspension isconverted to a more gel-like state), mean squared displacement(MSD) values can be calculated to probe the system dynamics(Weitz & Pine, 1993). At very short times, when the time of themeasurement is much smaller than the characteristic decay time ofthe system, the MSD can be written as:

< Dr2ðtÞ > ftp (1)

where the exponent p has a value of 1 for a freely diffusing particle.In an arrested system the value of p is always less than 1 (Krall &Weitz, 1998).

The sample (w1.5 mL) was poured into an optical glass cuvette(Hellma Canada Ltd., Concord, Ontario, Canada) with a 5 mm pathlength. The cuvette was placed in a thermostatted water bath at atemperature of 25 �C. The sample was illuminated by a solid-statelaser light with a wavelength of 532 nm and a power of 350 mW(Coherent, Santa Clara, CA, USA). Scattered light intensity wascollected in transmissionmodeaspreviouslydescribed (Gaygadzhiev,Corredig, & Alexander, 2008). When measuring the various concen-trated milk samples as a function of time using DWS, values weretaken after 40min from transfer of the sample in the cuvette, to allowfor equilibration.

2.5. Particle size determination by dynamic light scattering

The particle size of the casein micelles was measured by dy-namic light scattering (Zetasizer Nano-ZS). After concentration, themilk samples were diluted w2000 times in filtered (0.2 mm nylonfilters, Fisher scientific) milk permeate and analysed.

2.6. Mineral determination

Determination of insoluble and soluble calcium was carried outusing non-suppressed ion chromatography (Rahimi-Yazdi, Ferrer, &Corredig, 2010). The amount of soluble calcium was defined as the

total calcium in the serum phase after centrifugation at 100,000� g(see above).Determinationof total and solublephosphatewas carriedout using Inductively Coupled PlasmaOptical Emission Spectrometryat the Laboratory Services facilities of the University of Guelph.

The levels of colloidal calcium and phosphate in the sampleswere calculated as the difference between the total amount ofcalcium/phosphate and that measured in the centrifugal superna-tant fraction.

2.7. Rheology measurements

A controlled stress rheometer (Paar Physica MC 301, Anton Paar,Graz, Austria), was used tomeasure the viscosity of the concentrate.The concentrated milk samples as well as the centrifugal super-natants were subjected to a shear sweep test from 0.1 to 300 s�1,using a cone and plate geometry, with a set gap of 0.51 mm. Thetemperature of the platewas controlledwithwater circulating froma Julabo F25-HP refrigerated and heated water bath (JulaboLabortechnik, GbmH, Germany). All measurements were made at25 �C. The values of viscosity measured at 300 s�1 were employedfor the calculations of the rheological behaviour.

To model the rheological behaviour of concentrated milk, asemi-empirical Eilers equation is usually applied (Karlsson et al.,2005; Snoeren et al., 1982). In this work, the Mendoza model forsolid sphere suspension was applied to the experimental data. Thismodel takes into account the hydrodynamic interactions betweenthe colloidal particles (Mendoza & Santamaría-Holek, 2009).

h ¼ ho

�1�

�f

1� cf

���5=2(2)

Where ho is the viscosity of the continuous phase and c representsthe ratio between [(1�4max)/4max. The 4max is the critical packingvolume fraction of the dispersed particles; for. milk, the 4max hasbeen found to be 0.79 (Snoeren et al., 1982). It is important to notethat this model could be applied to particles of different shape bychanges to the value of the exponent (Mendoza & Santamaría-Holek,2009). However, themodel does not take into account polydispersity.

3. Results and discussion

Fig. 1 depicts the changes in protein concentration in the milksample as a function of PEG concentration in the permeate sideafter 18 h of dialysis. It is very clear that by increasing the osmotic

P. Krishnankutty Nair et al. / Food Hydrocolloids 34 (2014) 46e53 49

stress level it was possible to achieve high protein concentrationswhile maintaining the serum composition close to that of skimmilk. Using 120 g L�1 PEG, milk was concentrated from its original32 g L�1 protein to about 250 g L�1 protein with a correspondingreduction in the volume of the sample.

The increase in the concentration of protein in milk willdecrease the spacing between the micelles. It has been reportedthat at the native concentration in milk (25 g L�1), the casein mi-celles, with a mass of approximately 2.8 � 108 Da and an averageradius of 78 nm (Morris, Foster, & Harding, 2000) are about 121 nmapart. At higher concentrations their decreased interparticle dis-tance will hence result in strong interactions. It has been previouslyreported that with concentration, the viscoelastic properties ofmilk change (Bouchoux, Debbou, et al., 2009; Dahbi et al., 2010;Snoeren et al., 1982). At protein concentrations of 125 g L�1, cor-responding to an osmotic pressure of 4.5 kPa, phase transition ofnative phosphocaseinate from a sol to a gel has been observed(Bouchoux, Cayemitte, et al., 2009). Jamming concentrations of180 g L�1 have also been reported for a lactose free micellar sus-pension (Dahbi et al., 2010).

3.1. Characterisation of the serum phase of milk

To fully understand the behaviour of the different milk samplesas a function of concentration, the soluble fraction was also char-acterized. In particular, the protein concentration, the refractiveindex and viscosity of the serumwere measured, as they are criticalto the interpretation of the light scattering and rheological prop-erties of the concentrates. It is important to note that as the con-centration of protein increased, an efficient centrifugal separationof the serum phase was increasingly challenging (because theactual amount of serum decreased dramatically). Therefore onlythe serum fraction of samples up to 150 g L�1 protein were sepa-rated for further analysis.

The amount of protein recovered in the serum fraction increasedgradually with concentration (Fig. 2); however, when the serumfraction was analysed by SDS-PAGE, loading the same amount ofprotein in each lane, the ratio betweenwhey protein and caseins inthe serum phase did not appear to change with concentrationwithin the experimental error (data not shown). These results were

Total Protein (g L-1)

0 20 40 60 80 100 120 140 160 180

Prot

ein

in th

e su

pern

atan

t (g

L-1)

0

10

20

30

40

50

60

Volume Fraction

0.0 0.1 0.2 0.3 0.4 0.5

Fig. 2. The amount of soluble protein present in the serum phase as a function of totalprotein concentration or casein micelles volume fraction.

not surprising, as both caseins and whey proteins were retained bythe dialysis membrane. At the highest concentrations it appearsthat a higher amount of proteinmay be present in the serum phase:this could be protein originally occluded in the serum contained inthe hairy layer around the micelles being squeezed out at highconcentration factors because of the close approach of the micelles.Indeed, it has been previously reported that sedimentation maycause compression of the hairy layer (Walstra, 1979) and recentwork showed that during the evaporation process water is removedpreferentially from the serum phase (Liu, Dunstan, & Martin, 2012).

Fig. 3 summarizes the values of refractive index and viscosity ofthe centrifugal serum as a function of protein concentration in themilk. Both the index of refraction and viscosity increased, and thiswas caused by the increase in the concentration of soluble protein.The serum showed a Newtonian fluid behaviour in the whole rangeof concentration investigated.

3.2. Changes in calcium during concentration

Determination of insoluble and soluble calcium, although ofcritical importance, has not beenwell reported in the literature as afunction of concentration of casein micelles. Ca2þ is a structuralcomponent of the casein micelles and its equilibrium can affect theprocessing functionality of the casein micelles (Lucey & Horne,2009). The colloidal calcium phosphate (CCP) is in dynamic equi-libriumwith the mineral components in the soluble phase, and it isnot yet clear to which extent this equilibrium is affected duringconcentration of the micelles (Holt, 2002). It has been previouslyhypothesized that as the milk is already saturated with calciumphosphate, a considerable proportion of soluble calcium andphosphate may be transferred into the colloidal state during con-centration by evaporation (Hardy, Donald Muir, Maurice Sweetsur,& West, 1984; Liu et al., 2012; Nieuwenhuijse, Timmermans, &Walstra, 1988). This would lead to a greater amount of colloidalcalcium phosphate per weight of casein in the concentrated milk ascompared with normal milk. Nevertheless, as a consequence ofconcentration, the pH decreases and ionic strength increases(Anema, 2009), both of these reducing the amount of calciumassociated with themicelles (Snoeren, Brinkhuis, Dammam, & Klok,1984). However, during membrane concentration of milk, calciumdeposition on the membrane may occur, while during osmoticstressing using permeate the ionic concentration in the serumphase is maintained as close as possible to that of the original milk.

Protein (g L-1)

20 40 60 80 100 120 140 160

Ref

ract

ive

inde

x

1.330

1.335

1.340

1.345

1.350

1.355

1.360Se

rum

Vis

cosi

ty (P

a s)

0.001

0.002

0.003

0.004

Volume Fraction

0.1 0.2 0.3 0.4 0.5

Fig. 3. Refractive index (- and right-hand scale) and viscosity (: and left-hand scale)of the serum phase as a function of concentration of protein in the original milk.

P. Krishnankutty Nair et al. / Food Hydrocolloids 34 (2014) 46e5350

The amount of soluble calcium present in the centrifugal su-pernatant increased linearly with concentration. The initial con-centration of soluble Ca2þ in the skim milk was 9.8 � 0.48 mM.10 mM Ca2þ was present in permeate in the outer phase of thedialysis membrane as the dialysis was conducted against milkpermeate.

Fig. 4 summarizes the amount of total and soluble calcium as afunction of the protein concentration. As expected, the totalamount of calcium increased with concentration, as the calciumassociated with the micelles continued to be retained. However,Fig. 4A clearly shows that at >100 g L�1 protein (at a volumefraction >0.3) the amount of total calcium may be reaching aplateau. The amount of soluble Ca2þ present in the serum phase(Fig. 4B) also increased with concentration (in spite of the dialysisequilibration) in a linear fashion, and this increase can be attributedto the calcium associated with the soluble proteins present in thesoluble phase (whey proteins and non sedimentable caseins)(Rahimi-Yazdi et al., 2010).

An analogous trend was observed in the case of total phosphate,the control milk had a total phosphate content of 28.1 � 0.2 mMand increased to 64.5� 0.5 when the protein content of the samplewas 100.24 � 0.2 g L�1. Similarly the ratio of soluble phosphate tosoluble protein showed a steady decline.

Because of the experimental design (the use of constant serumcomposition during osmotic stressing), this study could not confirmprevious reports that at high protein concentrations an amount ofcolloidal calcium is released to the serum phase (Anema, 2009;

Tota

l Cal

cium

(mg

L-1)

0

1000

2000

3000

4000

5000

Volume Fraction

0.0 0.1 0.2 0.3 0.4 0.5

B

Protein (g L-1)

0 20 40 60 80 100 120 140 160 180

Solu

ble

Cal

cium

(mg

L-1)

300

400

500

600

700

A

B

Fig. 4. Total (A) and soluble (B) calcium content present in milk as a function of proteinconcentration. The mean and standard error of three independent trials are shown.

Snoeren et al., 1984). At least up to 90 g L�1 there was a steadyincrease in the amount of total Ca2þ (Fig. 4A) and only at thatconcentration the data may indicate a critical concentration wherethe micelles are trying to resist structural changes caused by theextraction of the water from the dispersion.

The release of colloidal calcium phosphate will help balance theosmotic gradient between the inner core of the micelles and theserum phase, and will result in an increase in negative chargeswithin the micelles. Therefore the reaching of a plateau value at90 g L�1 (Fig. 4) may signal the beginning of changes in compositionof the calcium phosphate nanoclusters of the micelles. Indeed it hasbeen hypothesized using X ray scattering (Bouchoux, Gésan-Guiziou, Pérez, & Cabane, 2010.). That there are hard regions inthe casein micelle of about of about 25 nm of diameter and con-taining about 7 calcium phosphate nanoclusters, and these regionsresist compression even at much higher concentrations. These re-sults would lead to the conclusion that calcium is released withoutdisrupting the internal supramolecular structure of the caseinmicelles.

3.3. Rheological properties of concentrated milk

Fig. 5 illustrates the changes in viscosity of the samplesconcentrated by osmotic stressing, as a function of volume fraction.The inset shows the entire volume fraction measured, using a logscale for relative viscosity. A conversion from protein concentrationto volume fraction (4) was necessary to be able to compare theexperimental data with theoretical models.

Concentrated milk up to a volume fraction of 0.4 exhibitedNewtonian behaviour and shear thinning behaviour was observedat higher concentrations (results not shown). Previous researchersassumed constant serum viscosity to predict the rheologicalbehaviour of concentrated milk (Dahbi et al., 2010; Karlsson et al.,2005); however, the Eilers equation could not fit the experi-mental values when a constant value of viscosity (the viscosity ofthe initial serum or permeate) was employed in the calculations.The experimental data shown in Fig. 5 were predicted by the Eilersequation only when the value of ho was varied, using experimentalvalues (Fig. 3). A critical packing volume fraction (4max) of 0.79 wasused in these calculations in accordance with previous literature(Karlsson et al., 2005; Snoeren et al., 1982). For each volume

Volume fraction

0.0 0.2 0.4 0.6 0.8

seru

m

0

10

20

30

40

Volume fraction

0.0 0.2 0.4 0.6 0.8 1.0

seru

m

0.1

1

10

100

1000

η/η

η/η

Fig. 5. Changes in viscosity measured at 300 s�1 of the concentrated milk as a functionof volume fraction. Symbols are experimental values (mean and standard error of threeindependent trials). The solid line corresponds to theoretical predictions of viscosityusing Mendoza model (Eq. (2)) for interacting colloidal hard spheres.

P. Krishnankutty Nair et al. / Food Hydrocolloids 34 (2014) 46e53 51

fraction the corresponding value of serum viscosity determinedexperimentally was used. It is important to note that in Fig. 5 it wasassumed that the voluminosity of the micelles did not change withconcentration. It has been recently demonstrated that duringconcentration by evaporation the water is preferentially removedfrom the serum than from the micelles (Liu et al., 2012).

The experimental data were fit to a model which takes in-teractions into account (Eq. (2)) (Mendoza & Santamaría-Holek,2009) with a r2 of 0.999. The viscosity of the concentratedmicellar suspensions could be predicted by knowing the back-ground viscosity (Fig. 3). The predicted values for hard spherebehaviour showed a 4max of 0.81. It may also be worth pointing outthat at the very low concentrations a slight deviation from the fitmay be noted, and this could be due to an increase in the volumi-nosity of the casein micelles at the very low concentrations. Thiswould be consistent with previous studies showing both for heatedand unheated milk a higher voluminosity of the casein micelles indiluted milk (Hallström & Dejmek, 1988).

Nonetheless, it was possible to conclude that assuming nochanges in voluminosity, the casein micelles behave as hardspheres with a critical packing volume in untreated skim milk of0.8. This is important, as it has been previously suggested that adecrease in the voluminosity of the casein micelles with concen-tration occurs (Hallström & Dejmek, 1988). Rheological data onlactose free micellar caseins showed a deviation from hard spherebehaviour at a volume fraction of 0.69 (Dahbi et al., 2010). Ourresults are in full agreement with those of Bouchoux, Debbou, et al.(2009) who also described the casein micelles as incompressiblepolydisperse hard spheres, however their critical volume fractionwas 0.65. This divergence in the critical packing volumemay derivefrom their use of native phosphocaseinate material compared tothe present work, where untreated skim milk was employed.

3.4. Light scattering properties of concentrated milk

The turbidity parameter (1/l*) as a function of the volumefraction of the various concentrated milk samples is illustrated inFig. 6. Increasing the concentration of casein micelles led to a cor-responding gradual growth of 1/l*. The values of 1/l* increasedmonotonically up to a 4 of w0.4. At higher volume fractions, thevalue of 1/l* no longer increased. Fig. 6 also illustrates the

Volume fraction0.0 0.1 0.2 0.3 0.4 0.5 0.6

1/l*

(mm

-1)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Fig. 6. Changes in 1/l* as a function of casein micelle volume fraction. The symbolscorrespond to three independent experiments. The solid line indicates the theoreticalbehaviour of the turbidity parameter 1/l* fitted to the theory for a hard-sphere of216 nm of diameter with a refractive index of 1.38 and 1.34 as refractive index of thesolvent.

theoretical behaviour for a hard sphere system (line), calculatedaccording to literature (Rojas-Ochoa, Méndez-Alcaraz, Sáenz,Schurtenberger, & Scheffold, 2004): as the protein content in thesample increases, more scattering events happen in a given spaceand the light becomes randomized faster, decreasing the value of l*.At higher volume fractions there is a decrease in the 1/l*.

At low volume fractions (<0.32) the values obtained experi-mentally (Fig. 6) fit well to the theoretically predicted values for ahard-sphere system (solid line, Fig. 6) with the experimentallyobtained values of size by DWS and refractive index parameterscorresponding to those used before for casein micelles (Alexander,Rojas-Ochoa, Leser, & Schurtenberger, 2002); however, at highervolume fractions, the experimental values deviated from thosecalculated from theory, as only hydrodynamic interactions can nolonger take into account the changes in the turbidity of the system.At higher concentrations, there will be interparticle interactionsbesides those of hydrodynamic drag which will cause strong po-sitional correlations between the particles and further decrease thevalue of l*. However, it is important to point out that refractiveindex contrast is also an important factor affecting the 1/l*parameter. It cannot be excluded that the micelles may undergosome internal rearrangements (due to the release of colloidal cal-cium phosphate, see Fig. 4) but these changes will decrease therefractive index of the micelles. In addition, the rheological datashown in Fig. 5 suggest that the voluminosity does not vary withconcentration.

To determine if aggregation occurred in the milk during osmoticstressing concentration experiments, the size of the casein micellewas measured by dynamic light scattering after diluting the sam-ples in permeate (data not shown). After redilution, for milkconcentrated up to 20% protein (corresponding to approximately0.65 volume fraction) the casein micellar size was constant afterdilution, confirming that no aggregation occurred in these samples.

The self-diffusion coefficient of the particles was also derivedfrom the DWS experiments, for the milk concentrated by osmoticstressing. At the low volume fractions the normalized diffusioncoefficient decreased with increasing concentration (data notshown) similarly to the behaviour of a hard-sphere suspension, inagreement with previous work (Alexander et al., 2002). At the lowvolume fractions (<0.4) the diffusion coefficient was used tocalculate the radius of the particle, with the assumption that thesystem was free diffusing. The values of the radius, calculated byDWS were in full agreement with those measured by DLS(82.43 � 1.38).

To better understand the behaviour of the casein micelles as afunction of concentration, the dynamics were also evaluated, bycalculating the mean squared displacement (MSD). In other words,for free diffusing particles, there is a linear relationship betweentheir displacement (MSD) and the time. Fig. 7 describes the MSDobtained from the correlation functions of milk with differentvolume fractions. At the low volume fractions <0.32%, the particlesshowed a behaviour characteristic of free diffusing Brownian par-ticles as it is clear that there is a linear relation between time andMSD.

The milk with a volume fraction of 0.39 started to deviate fromthis behaviour, and at higher protein content, the particles are nolonger free diffusing and experience restrictions in their move-ment. These results show for the first time that restricted motion ofcasein micelles starts to occur at such low volume fractions. Similarlight scattering experiments, carried out on micellar dispersionsshowed a hard sphere behaviour up to volume fractions of 0.54(Dahbi et al., 2010). The decrease in the mobility of the casein mi-celles shown in Fig. 7 may be related to their polydiperse nature.Indeed at a volume fraction of 0.3, the interparticle distance be-tween the micelles (see above) is much smaller than their size,

Time (s)0 1e-5 2e-5 3e-5 4e-5 5e-5 6e-5

MSD

(m

2 )

0

2e-5

4e-5

6e-5

8e-5

μ

Fig. 7. Time dependence of MSD of milk samples with different volume fraction. Fromleft to right: 0.1 (control milk), 0.29, 0.39 and 0.5. To show the deviation of MSD withtime, straight lines are drawn from the origin to each set of experimental results.

P. Krishnankutty Nair et al. / Food Hydrocolloids 34 (2014) 46e5352

causing the small casein micelles to be caged in, surrounded by thelarger ones.

4. Conclusions

The present work further characterized the physico-chemicalproperties of casein micelles in concentrated milk. Osmoticstressing is a convenient and non-destructive method to study theconcentrated milk as a function of volume fraction. Up to a volumefraction of 0.6 casein micelles do not irreversibly aggregate, aswhen measured under diluted conditions, their size is constant.

At volume fractions <0.32, the behaviour of casein micelles canbe fullymodelled to that of hard spheres of similar size and refractiveindex. In addition, rheological properties were well described by theEilers’ and Mendoza equations, with a critical packing volume frac-tion of about 0.8. This is a much higher critical volume fraction thanthat reported by recent literature (Bouchoux et al., 2010; Dahbi et al.,2010). The polydispersity of the casein micelles allows such highpacking values, as shown in rheological experiments, and the dis-crepancies from previous literature derive from differences in thesample preparation (this study used fresh milk while others usedreconstituted milk or phosphocaseinate powders). The rheologicalmeasurements would also suggest that the voluminosity of thecasein micelles does not vary with concentration, once again con-firming that the casein micelles behave as uncompressible hardspheres up to high critical packing volume fractions.

However, hydrodynamic interactions between the micellesoccur at much lower volume fractions. Diffusing wave spectroscopydata clearly indicated that when milk is concentrated to a volumefraction higher than 0.32, the behaviour of the casein micelles canno longer be predicted using only hydrodynamic effects, and thatstronger interactions occur between the particles. The casein mi-celles are no longer free diffusing, as the interparticle distancecreates caging of the small protein particles, which are surroundedby the larger casein micelles. The combination of serum composi-tion analysis, light scattering and rheological measurements wouldhence suggest that in the range of concentration studied, caseinmicelles behave as uncompressible hard spheres.

Acknowledgements

This work was funded by the Natural Sciences and EngineeringResearch Council of Canada.

References

Alexander, M., Rojas-Ochoa, L. F., Leser, M., & Schurtenberger, P. (2002). Structure,dynamics, and optical properties of concentrated milk suspensions: an analogyto hard-sphere liquids. Journal of Colloid and Interface Science, 253, 35e46.

Anema, S. (2009). Effect of milk solids concentration on the pH, soluble calcium andsoluble phosphate levels of milk during heating. Dairy Science & Technology, 89,501e510.

Bienvenue, A., Jimenez-Flores, R., & Singh, H. (2003). Rheological properties ofconcentrated skim milk: importance of soluble minerals in the changes inviscosity during storage. Journal of Dairy Science, 86, 3813e3821.

Bouchoux, A., Cayemitte, P., Jardin, J., Gésan-Guiziou, G., & Cabane, B. (2009a). Caseinmicelle dispersions under osmotic stress. Biophysical Journal, 96, 693e706.

Bouchoux, A., Debbou, B., Gesan-Guiziou, G., Famelart, M., Doublier, J., & Cabane, B.(2009b). Rheology and phase behavior of dense casein micelle dispersions.Journal of Chemical Physics, 131, 165106e165111.

Bouchoux, A., Gésan-Guiziou, G., Pérez, J., & Cabane, B. (2010). How to squeeze asponge: casein micelles under osmotic stress, a SAXS study. Biophysical Journal,99, 3754e3762.

Dahbi, L., Alexander, M., Trappe, V., Dhont, J. K. G., & Schurtenberger, P. (2010).Rheology and structural arrest of casein suspensions. Journal of Colloid andInterface Science, 342, 564e570.

Dalgleish, D. G. (2011). On the structural models of bovine casein micelles-reviewand possible improvements. Soft Matter, 7, 2265e2272.

Dalgleish, D. G., Horne, D. S., & Law, A. J. R. (1989). Size-related differences inbovine casein micelles. Biochimica Et Biophysica Acta (BBA)/General Subjects,991, 383e387.

Farrer, D., & Lips, A. (1999). On the self-assembly of sodium caseinate. InternationalDairy Journal, 9, 281e286.

Fox, P. F. (2003). Milk proteins: general and historical review. In P.F. Fox &P.L.H. McSweeney (Eds.), Advanced dairy chemistry (pp. 1e48). New York; Lon-don: Kluwer Academic/Plenum.

Gaygadzhiev, Z., Corredig, M., & Alexander, M. (2008). Diffusing wave spectros-copy study of the colloidal interactions occurring between casein micellesand emulsion droplets: comparison to hard-sphere behavior. Langmuir, 24,3794e3800.

Hallström, M., & Dejmek, P. (1988). Rheological properties of ultrafiltrated skimmilk. II. Protein voluminosity. Milchwissenschaft, 43, 95e97.

Hardy, E. E., Donald Muir, D., Maurice Sweetsur, A. W., & West, I. G. (1984). Changesof calcium phosphate partition and heat stability during manufacture of ster-ilized concentrated milk. Journal of Dairy Science, 67, 1666e1673.

Harwalkar, V. R., Beckett, D. C., McKellar, R. C., Emmons, D. B., & Doyle, G. E. (1983).Age-thickening and gelation of sterilized evaporated milk. Journal of DairyScience, 66, 735e742.

Holt, C. (1992). Structure and stability of bovine casein micelles. Advances in ProteinChemistry, 43, 63e143.

Holt, C. (2002). Milk salts and their interaction with caseins. In Hubert Roginski(Ed.), Encyclopedia of dairy sciences (pp. 2007e2015). Oxford: Elsevier.

Horne, D. S. (2009). Casein micelle structure and stability. In A. Thompson &M. Boland (Eds.), Milk proteins from expression to food (pp. 133e162). Amster-dam; Boston: Academic Press/Elsevier.

Hwang, J., Lee, S., Park, H., Min, S., & Kwak, H. (2007). Comparison of physico-chemical and sensory properties of freeze-concentrated milk with evaporatedmilk during storage. Asian - Australasian Journal of Animal Sciences, 20, 273e282.

Jenness, R., & Koops, J. (1962). Preparation and properties of a salt solution whichsimulates milk ultrafiltrate. Netherlands Milk and Dairy Journal, 1(16), 153e164.

Karlsson, A. O., Ipsen, R., & Ardö, Y. (2007). Rheological properties and micro-structure during rennet induced coagulation of UF concentrated skim milk.International Dairy Journal, 17, 674e682.

Karlsson, A. O., Ipsen, R., Schrader, K., & Ardö, Y. (2005). Relationship betweenphysical properties of casein micelles and rheology of skim milk concentrate.Journal of Dairy Science, 88, 3784e3797.

Koning, M. M. G., van Eendenburg, J., & De Bruijne, D. W. (1993). Mixed biopolymersin food systems: determination of osmotic pressure. In E. Dickinson & P. Walstra(Eds.), Food colloids and polymers: Stability and mechanical properties (pp. 103e112). Cambridge: Royal Society of Chemistry.

Krall, A. H., & Weitz, D. A. (1998). Internal dynamics and elasticity of fractal colloidalgels. Physical Review Letters, 80, 778e781.

Liu, D. Z., Dunstan, D. E., & Martin, G. J. O. (2012). Evaporative concentration ofskimmed milk: effect on casein micelle hydration, composition, and size. FoodChemistry, 134, 1446e1452.

Lucey, J. A., & Horne, D. S. (2009). Milk salts: technological significance. InP.L.H. McSweeney & P.F. Fox (Eds.), Advanced dairy chemistry, Vol. 3 (pp. 351e380).New York: Springer.

Marchin, S., Putaux, J., Pignon, F., & Léonil, J. (2007). Effects of the environmentalfactors on the casein micelle structure studied by cryo transmission electronmicroscopy and small-angle x-ray scattering/ultrasmall-angle X-ray scattering.The Journal of Chemical Physics, 126, 045101e045110.

Mendoza, C. I., & Santamaría-Holek, I. (2009). The rheology of hard sphere sus-pensions at arbitrary volume fractions: an improved differential viscositymodel. Journal of Chemical Physics, 130, 044904e044911.

Morris, G. A., Foster, T. J., & Harding, S. E. (2000). Further observations on the size,shape, and hydration of casein micelles from novel analytical ultracentrifugeand capillary viscometry approaches. Biomacromolecules, 1, 764e767.

P. Krishnankutty Nair et al. / Food Hydrocolloids 34 (2014) 46e53 53

Nieuwenhuijse, J. A. W., Timmermans, W., & Walstra, P. (1988). Calcium andphosphate partitions during the manufacture of sterilized concentrated milkand their relation to the heat stability. Netherlands Milk and Dairy Journal, 42,387e421.

Pignon, F., Belina, G., Narayanan, T., Paubel, X., Magnin, A., & Gèsan-Guiziou, G.(2004). Structure and rheological behavior of casein micelle suspensions duringultrafiltration process. Journal of Chemical Physics, 121, 8138e8146.

Rahimi-Yazdi, S., Ferrer, M. A., & Corredig, M. (2010). Nonsuppressed ion chro-matographic determination of total calcium in milk. Journal of Dairy Science, 93,1788e1793.

Rojas-Ochoa, L. F., Méndez-Alcaraz, J. M., Sáenz, J. J., Schurtenberger, P., &Scheffold, F. (2004). Photonic properties of strongly correlated colloidal liquids.Physical Review Letters, 93(7), 073903.

Ruiz, J. F., & Barbosa-Cánovas, G. V. (1997). Effects of concentration and temperatureon the rheology of concentrated milk. Transactions of the American Society ofAgricultural Engineers, 40, 1113e1118.

Ruiz, J. F., & Barbosa-Canovas, G. V. (1998). Rheological properties of concentratedmilk as a function of concentration, temperature and storage time. Journal ofFood Engineering, 35, 177e190.

Snoeren, T. H. M., Brinkhuis, J. A., Dammam, A. J., & Klok, H. J. (1984). Viscosity and agethickeningof skimmilk concentrate.NetherlandsMilk andDairy Journal, 38, 43e53.

Snoeren, T. H. M., Damman, A. J., & Klok, H. J. (1982). The viscosity of skim milkconcentrates. Netherlands Milk and Dairy Journal, 36, 305e316.

Walstra, P. (1979). The voluminosity of bovine casein micelles and some of itsimplications. Journal of Dairy Research, 46, 317e323.

Walstra, P., & Jenness, R. (1984). Dairy chemistry and physics. New York: Wiley.Weitz, D. A., & Pine, D. J. (1993). Diffusing-wave spectroscopy. In W. Brown (Ed.),

Dynamic light scattering: The method and some applications (pp. 653e719).Oxford: Clarendon Press.

Weitz, D. A., Zhu, J. X., Durian, D. J., Gang, H., & Pine, D. J. (1993). Diffusing-wavespectroscopy: the technique and some applications. Physica Scripta, T49B,610e621.