University of Birmingham Preparation and rheological properties of whey protein emulsion fluid gels

A shear-gel approach was used to coat an O/W emulsion with a whey protein shell. Mechanical shear was applied to an aqueous solution of whey protein isolate (WPI) and oil then heated through the sol – gel transition. The formation of a continuous WPI network was prevented through the shear ﬂ ow, resulting in discreet spherical capsules, with sizes in the micron scale ( (cid:1) 25 m m). Through careful control over both the oil and WPI components, encapsulation e ﬃ ciencies of up to 99% were obtained. Subsequent rheological properties highlighted elastic behaviour ( G 0 ) dependent on oil content; where higher oil fractions increased the e ﬀ ective phase volume of the particles. The result was packing fractions that exceeded those for hard spheres (>0.64), leading to pseudo-solid characteristics at rest, apparent yield stresses and thixotropic behaviour under shear. As such, emulsion ﬂ uid gel (EmFG) rheology was closely comparable to those of soft microgel suspensions ( ﬂ uid gels) and soft-coated particles.


Introduction
The use of particulates as rheological modiers spans across multiple industries: paint, cosmetic, food etc. For this reason, systems such as colloidal suspensions and emulsions have been well documented. However, more recently microparticulate gel suspensions have received increasing interest for their ability to create weakly structured uids, with rheological properties characterised between those for colloidal particles and polymeric gels; pseudo-solid behaviour at rest, but ow above a critical stress. [1][2][3][4] Enhanced ow behaviours have been observed in polysaccharide uid gels where, at volume fractions as low as f 0.2 suspensions showed high viscosities and marked shear thinning behaviour, typical of highly aggregated systems. 5,6 The observed changes in ow properties were assigned to the particle microstructures, where, as a result of incomplete gelation at the particle surface, interactions between microgel spheres resulted in a degree of structuring. Additionally, the deformability of such so hydrogel particles led to volume fractions that exceeded those typical for hard spheres. At such high volume fractions particles become sterically conned, thus rheology became more closely governed by the particle moduli. [7][8][9] Particle intrinsic properties such as strength and deformability can be better understood by looking at a comparable quiescently formed gel. Again, mechanical properties are closely linked to the gel microstructure. [10][11][12][13][14] Complex microstructures therefore provide another means of controlling a gel's physicomechanical properties. Besides polymer mixes where two or more polymers undergo a sol-gel transition, ller particles can be used to prepare gel composites. 15 These composites display viscoelastic properties as a function of the gel matrix, volume fraction and rigidity of the ller. Chemical affinity between the ller and surrounding substrate is also very important. 16 Typically, where the ller interface is included into the gelled network strengthening of the matrix occurs. 17 It was demonstrated that increasing the time scale between emulsication and gelation resulted in a shi from stronger to weaker gels. Here, polymerisation of the emulsier lowered the affinity between oil ller and substrate matrix, causing inherent weak spots within the gel. [17][18][19] Processing is therefore a major consideration in the preparation of gels, on both a macro and micro-scale. Microgels are typically formed by conning the polymer during the sol-gel transition. This can be achieved by means of chemical 20 or mechanical 1 separation. Controlled microstructures and further rheological properties can be achieved by a microstructural design approach; whereby changing the formulation, i.e. composition, degree of cross-linking and system pH, leads to careful control over the gelled network. 4,21 Additionally, in the case of shear gels, modulating the two processing parameters shear and thermal history can lead to controlled particle size and morphology. When the separation applied to the system is such that it becomes comparable to the timescale for polymer ordering, random coil to helix transition and subsequent gelation through the formation of junction sites, thermodynamically favourable spherical particles ($1 mm) are formed: where the enthalpies of melting are much smaller to comparable quiescently prepared gels. 22 The rate of particle growth therefore becomes key. Large anisotropic ($100 mm) particles form when structuring is so rapid that initial gels are able to form, becoming subsequently broken down in the shear ow. However, in whey protein systems an inverse relationship between aggregation kinetics and system elasticity was observed, showing faster rates to yield weaker suspensions, with lower yield stresses and viscosities. 23 Supramolecular chemistry has also shown a viable route to the preparation of micro-composites. [24][25][26] These particles are formed through electrostatic bonding of the polymers. The particulates are formed by carefully controlling the pH so that the zeta-potential of both polymers is of an opposite charge, with a potential great enough to drive complexation. Such micro-particulates, based on an O/W 1 /W 2 system, exhibited a several fold increase in viscosity when compared to a simple O/W emulsion. 24 These viscosity changes were attributed to a change in the effective phase volume of the hydrogel spheres, which were further increased through the inclusion of an oil ller. However, interactions between the particles were not observed.
We report here the preparation of emulsion uid gel (EmFG) particles, a micro-composite of whey protein gel and oil. The research builds on work presenting O/W 1 /W 2 lled hydrogel systems by applying a "shear gel approach" to promote interactions between resulting particles. As such, it takes an additional step to surfactant or Pickering stabilised emulsions by gelling a continuous WPI layer around an oil core. The technique applied results in an elastic suspension, whereby the particles become trapped, suspended in an aqueous phase. Unlike similar studies 27 the nal systems act in a pseudo-solid fashion until a great enough stress is applied to induce ow, thinning the suspension into a liquid-like state. The microparticles thus offer the capacity to act as a multifunctional composite, for both controlled rheological applications and pose the potential to encapsulate poorly soluble molecules.
The work focuses on whey protein isolate (WPI) as the coating material owing to its thermo-denaturation and subsequent hydrophobic aggregation, to form a gel layer on the surface of an oil substrate. Shear separation will then be applied to prevent complete gelation, promoting particle-particle interactions. Therefore the production of emulsion uid gels has been investigated, with particular attention to the resulting rheological properties.

Preparation of oil lled uid gels
Preparation of the oil lled uid gels involved a three-step process. Primary solutions of WPI were rstly prepared and used to form O/W emulsions. The emulsions were then heat treated under shear conditions resulting in emulsion uid gels (EmFG).
2.2.1. Preparation of WPI stock solutions. Whey protein stock solutions (5 to 30 wt% on a protein basis) were prepared by dispersing WPI in deionised water. An anti-microbial, sodium azide (0.02 wt%) was added to all solutions to enhance storage times. Solutions were stirred overnight at ambient conditions until completely hydrated and stored at 5 C until further usage. For the preparation of stained samples Rhodamine B (0.015 mM) was added to stain the protein.
Stained samples were kept covered to prevent photo bleaching.
2.2.2. Preparation of emulsions. Oil in water emulsions were prepared by the addition of high oleic oil to WPI primary solution (protein concentration ranging from 5 to 30 wt%) so that total volumes resulted in oil fractions ranging from 5 to 20 vol%. The mixtures were subsequently mixed in a high shear mixer (Silverson, SL2T) at 4000 rpm for 10 minutes. System pH was adjusted to pH 4.6 with concentrated hydrochloric acid (12 M) and aged for 72 hours before gelling under shear. Stained EmFG were prepared with the addition of Nile Red (0.015 mM) to the oil phase and kept covered to prevent photo bleaching.
2.2.3. Preparation of oil lled uid gels. A jacketed vessel and overhead stirrer equipped with pitched blade impeller was used to prepare all WPI emulsion uid gels (EmFG). Aliquots of emulsion were added to a jacketed vessel set to 50 C, controlled through a circulating water bath. Shear was applied through the stirrer and impeller at 450 rpm. Once thermal equilibrium was obtained (ca. 10 minutes), emulsions were heated at a rate of 0.5 C min À1 to 80 C. Suspensions were subsequently decanted and le to cool quiescently under ambient conditions. Cooled EmFG were further stored at 5 C for seven days until rheologically tested. In all experiments a cover was applied to minimise water loss. When staining, the system was covered in aluminium foil to avoid uorescence quenching.

Static light scattering (SLS)
A MS2000 Mastersizer with attached Hydro SM manual small volume dispersion unit (Malvern Instruments Ltd, UK) was used to obtain size distributions for EmFG particles. Distributions are the average of three repeats. Particle size calculations were based upon the Mie theory, thus particles were assumed to be monodisperse, homogenous spheres. Additionally, once coated, particles were a binary system of protein and oil; as such the refractive index of the shell was used.

Microscopy
2.4.1. Optical light microscopy. Samples were prepared by rst diluting the EmFG in deionised water (7.5 vol%). A Brunel SP300- (Brunel Microscopes Ltd, UK) optical light microscope tted with an SLR camera (Cannon EOS Rebel XS, DS126 191) at 20 and 40Â optical magnication was used to image the particles. Slides were prepared by addition of 50 mL of sample to a microscope slide (VWR, UK) and covered with a coverslip (thickness no.1, VWR, UK).

Confocal laser scanning microscopy (CLSM).
Stained samples prepared with both Rhodamine B and Nile Red were imaged using concave microscope slides (60 mL), with a coverslip sealed using super glue. Fluorescence free, UV transparent immersion oil (Sigma-Aldrich, UK) was used to bridge the gap between coverslip and objective lens (40Â magnication). A Leica TCS-SPE confocal microscope (Leica Microsystems Ltd, UK) tted with an argon laser was used for all CLSM analysis. Rhodamine B stained protein was excited at 532 nm and detected at 560-600 nm, and Nile Red was excited at 488 nm and detected at 680-700 nm respectfully. 3-Dimensional images were obtained using 0.5 mm slices throughout. Further image processing was undertaken using an image soware package (ImageJ).

Rheological analysis
Rheometry was conducted using a Bohlin Gemini HR Nano stress-controlled Rheometer (Malvern Instruments Ltd, UK) equipped with serrated parallel plate (25 mm diameter) at 1 mm gap height. Experiments were undertaken at 25 C using a silicon oil moisture trap. An equilibrium was achieved for 15 minutes prior to testing, allowing for consistent loading conditions. Particle phase volumes were obtained using a method outlined by Moakes, et al.; 23 whereby aliquots of a given volume were centrifuged, water phase separated and volumes obtained. Phase volumes were calculated as followed (eqn (1)): Particle phase volume ¼ 1 À volume of water removed total initial volume (1) 2.5.1. Yield stress determination. Yield stresses were determined as the stress at which the transition from storage modulus (G 0 ) to loss modulus (G 00 ) dominance occurred; the G 0 /G 00 cross over point on a stress controlled amplitude sweep. Amplitude sweeps were conducted at 1 Hz from 0.1 to 400 Pa.
2.5.2. Dynamic oscillatory measurements. Frequency sweeps were obtained between 0.1 and 10 Hz at controlled stress. The stress was determined by amplitude sweeps as a value within the linear viscoelastic region (LVR) for all EmFG tested. For samples containing no oil a, couette, double gap geometry was used: as the large surface area allowed lower values of G 0 to be probed in the lower viscosity uids.
2.5.3. Viscosity measurements. Dynamic viscosity measurements were undertaken between 0.1 and 400 s À1 . Shear sweep time was set to 10 minutes (ramp ascending and descending). Two consecutive sweeps were run with the second commencing immediately aer the rst having been completed.
2.5.4. Recovery analysis. EmFG recovery was probed by primarily rejuvenating the system. A pre-shear (10 s À1 for 10 s) was applied to the system and subsequent change in storage modulus (G 0 ) recorded over the following 30 min. G 0 was obtained at 1 Hz and 1 Pa.

Encapsulation efficiency
Optical microscopy was used to analyse oil droplet entrapment for each system. Micrographs of the raw emulsion were obtained and emulsion droplets in each image manually counted to give an average droplet count (N em ) determined over 12 images. Values for uncoated emulsion droplets (N fo ) in the nal EmFG systems were then obtained in the same manner. An example of both a coated and uncoated droplet has been shown in Fig. 1. The ratio of the two was used to calculate the percentage of encapsulated oil, as shown in eqn (2): Encapsulation was averaged over 12 micrographs with error calculated as the 95% condence interval.

Preparation of emulsion uid gels (EmFG)
An O/W system where the excess WPI emulsier exceeded the critical gelling concentration 28 (>1 wt%), was subjected to heat treatment under shear conditions. The shear ow exerted on the system during the sol-gel transition prevented the formation of a continuous gel network, resulting in single discreet particles/encapsulates.
3.1.1. Encapsulation efficiency. Fig. 2 shows the encapsulation efficiencies for systems with both increasing WPI concentration (a) and oil phase volume, f oil , (b). It was observed that high levels of encapsulation, $99%, were achieved in systems with lower WPI concentrations, however as the WPI exceeded 20 wt%, entrapment decreased. It is argued that the degree of encapsulation is closely correlated to the ow behaviour of the system during the sol-gel transition. It has previously been reported that around the isoelectric point of WPI (ca. pH 5) a transition from Newtonian to pseudoplastic ow occurs in systems containing 20 wt% whey protein, with yield stresses observed on further concentrated solutions. 29,30 Such changes in the system viscosity retard or prevent the diffusion of denatured protein to the oil/water interface due to a reduction in mobility and increased collisions with other denatured polymer chains. In turn, both increased formation of WPI aggregates without included oil and free emulsion droplets were observed in the nal suspensions.
Increasing the phase volume of the oil from 5 to 20% whilst retaining a standard WPI concentration (15 wt%) however, had little effect on the emulsion entrapment, yielding droplet entrapment in excess of 95%. Here, the increase in oil was not sufficient to raise the viscosity of the system, and hinder polymer diffusion. However, at low oil fractions increasing the WPI concentration from 5 to 15 wt% caused an observed transition from suspension creaming to sedimentation, indicating a change in particle density, probably as a result of a change in shell thickness.
3.1.2. Particle morphology. Particle morphology was studied using confocal laser scanning microscopy (CLSM). Previous reports regarding the formation of WPI microgel particles through the application of shear, 23 show irregular shaped particles, characterised by a larger length to width ratio. However, the incorporation of oil in to the system resulted in particles with spherical morphology, whereby a gel layer surrounded an oil substrate, as shown in Fig. 3.
Imaging the EmFG particles at 0.5 mm intervals gave enhanced topographical detail (Fig. 3a), allowing the coating to be observed. A non-uniform shell with much greater thickness than expected for emulsied droplets was shown, inferring the presence of a gel layer. Additionally, cross-sections were obtained using CLSM, which again show shell thickness and nonuniformity, but also through negative staining and dying, oil reservoirs in the centre of the particles (Fig. 3b and c respectfully).
The mechanism for particle formation is thought to be based upon the oil acting as a substrate for shell growth. Growth occurs through enrichment from the surrounding un-gelled biopolymer. Primarily led by the b-lactoglobulin, heat induced denaturation of the native structure causes hydrophobic regions to become exposed. 31,32 Hydrophobic interactions then dominate the gelation causing diffusion of the denatured polymer to the oil/water interface. The oil droplet thus acts as a point for nucleation and growth. Shear imposed upon the system then restricts particle-particle aggregation, preventing a continuous network from forming. As a result, particles grow to an extent permitted by the shear ow, however are primarily dictated by the size of the emulsion droplets.
3.1.3. EmFG particle size distributions. Particle size distributions for all EmFG systems are shown in Fig. 4. It is clear from Fig. 4a that increasing WPI concentration results in a shi towards smaller particle sizes. Particles are primarily a function of the emulsion droplet size, thus such observations would be expected as increasing the emulsier concentration causes the formation of smaller droplets. 33,34 However, by further increasing the protein concentration up to 30 wt%, the formation of a bimodal system centred at $9 mm, with a second peak at much higher particle sizes ($200 mm) was observed. The shi from monomodal to bimodal is due to the formation of gelled particles where no oil has been entrapped as observed through microscopy. As previously described, the change in system ow results in aggregated protein, as diffusion of the denatured polymer becomes restricted, causing a transition from proteinsubstrate to protein-protein interactions. The extent of such a transition is thus reected by the change in peak intensity observed in systems with 20 to 30 wt% WPI. Fig. 4b shows little effect on the resulting particle size as a function of oil fraction, as a result of consistent emulsier concentrations and unrestricted diffusion of biopolymer to the substrate interface. Thus, all distributions are centred around 25-30 mm complimenting sizes observed via microscopy ( Fig. 1 and 3).

EmFG material properties
3.2.1. Small deformation oscillation testing. To further understand and characterise the EmFG physico-mechanical properties, small deformation rheology was carried out. Fig. 5a shows frequency sweeps obtained for EmFG systems prepared with varying oil fractions from 0 to 20 vol%. Systems containing no oil displayed typical liquid-like behaviour where both moduli were dependent on frequency and G 00 is higher than G 0 throughout the measured frequency range. 35 The addition of oil caused a transition in rheological behaviour to pseudo-solid where G 0 is higher than G 00 , with both moduli becoming further independent to frequency as a function of the oil. Here particle proximity is such that inter-particle interactions arise as observed for WPI microgel particles, 23 however, as the shear and thermal history exerted was consistent across all systems, it is argued that system elasticity becomes a function of the oil content; where increasing oil fractions increase the effective phase volumes of the particles, as shown in Fig. 5b.
To better understand the mechanism through which system elasticity arises, the results obtained have been compared to models already proposed for particulate suspensions. The Krieger-Dougherty (KD) model is used to describe the relationship between relative viscosity (h rel ) and particle phase volume for hard sphere suspensions, 36 eqn (3).
The equation relates the ratio between the phase volume of the suspension, f, to the maximum packing fraction for monodisperse hard spheres (0.64), f max , as a function of the intrinsic viscosity, [h]. What is clear from eqn (3) is that as the maximum packing fraction is approached the relative viscosity will asymptote and eventually the equation fail as the suspension reaches/surpasses the maximum packing. As such, the KD equation cannot be used to describe the correlation observed in Fig. 5b, where the maximum packing fraction for hard spheres has been exceeded.
Similar observations have been reported for agar microgel suspensions, 9 where above a critical volume fraction, f c , elastic response was observed, becoming insensitive to phase volumes above f max . It was explained that above f c , elasticity was a function of the particle modulus; hence particles were typically acting as so spheres, however the plateauing effect was le unexplained, suggested as an artefact of the phase volume calculation. EmFG are therefore assumed to act as so spheres. It is suggested that the so oil core and elastic whey protein shell allows for particle deformation when highly concentrated, reaching phase volumes that exceed those expected for rigid spheres. At such high phase volumes a jamming phenomenon is observed, thus particle rheology mainly represents a function of the shell. 37 Hence above a volume fraction of 0.64 particles are packed to an extent that system elasticity is close to those shown for lled quiescent gels, 38 hence frequency sweeps show gel-like behaviour (Fig. 5a).
EmFG prepared using 5 vol% oil showed marked storage moduli even though f max had not been reached. Here, elastic response is observed not through the jamming of particles, but assumed to arise from particle trapping as a result of steric connement through inter-particle interactions. Between f c and f max it is argued that EmFG act as a glass where particles become trapped allowing localised motion but not long range diffusion, similarly to results published by Koumakis, et al. 37 and le Grand. 39 Therefore, the EmFG can be categorised into three regimes; suspended particles below f c , glassy between f c and f max , and jammed above f max .
3.2.2. Material yielding behaviour. The effect of oil fraction on material response was further probed through the use of stress sweeps. Fig. 6 shows the data obtained for EmFG with increasing oil content (from (a) to (d)). The stress sweeps indicate that at a critical stress the network started to break down. Further increasing the stress led to a transition at which point the loss modulus dominated the storage. At this point a change in material response occurs, where the system no longer resembles a pseudo-solid, but is much more uid-like. For systems in a glassy state, where particle have become trapped through inter-particle interactions, the linear viscoelastic region (LVR) was observed to be much shorter than those that are jammed. Again, such observations can be interpreted through the deformability of the particles; where particles are in close proximity they become compressed and deform, a larger stress is needed to induce ow as observed for k-carrageenan uid gels. 6 However, for systems where f does not surpass f max ,  particles appear to act as hard spheres, as such deformation does not occur which is reected by the lower stresses required to induce ow, 39 as shown in Fig. 7. Here it is possible to see a similar correlation as observed for the frequency sweeps, where a plateau is reached for oil fractions above 10 vol%. This yielding insensitivity towards increasing oil fractions again infers a network of deformed particles closely packed together, as previously suggested. 39 3.2.3. Suspension ow behaviour. System ow behaviours for all WPI EmFG were studied and presented in Fig. 8. EmFG showed marked shear thinning behaviours typical of highly occulated suspensions. 40 At very low shear rates ($0.01 s À1 ) an apparent shear thickening can be observed as a resultant effect of so jamming. Continued increase in the applied strain resulted in thinning of the suspensions. The observed thinning is due to the inhomogeneous ow across the shear prole applied, as a result of the break down to the weakly occulated network. 40 Thus a degree of inter-particle interactions is suggested, where initially clusters of mesostructures are broken down to form single mesostructures (smaller ocs) and eventually microstructures (single particles). To analyse this further, data has been presented for the ramp up, down and additionally a second sweep taken immediately aer the rst.
The presence of hysteresis highlighted a thixotropic nature arising through the breakdown of occules and inter-particle interactions, as seen in Fig. 9. The plot shows a similar correlation as previously observed for yield stresses and frequency sweeps whereby the highly packed systems (10 to 20 vol% oil) have a greater hysteresis. This indicates that jammed systems have a much higher degree of inter-particle interactions as a result of greater packing arising through particle deformation. As expected, where the particles remain unjammed a lower degree of thixotropy is observed, as non-deformed particles  present a smaller surface area for inter-particle interactions to occur. However, data obtained for the second sweep showed a similar thixotropic behaviour for all systems, irrespective of oil content. Additionally, a combination of microscopy and light scattering techniques showed that the applied shear was insufficient to break the single particles, with the same particle size distributions observed pre and post shear sweeps (data not shown). As such it is argued that the shear applied throughout the rst sweep disrupts the network to an extent that all particles behave as independent spheres. This is followed by restructuring between the particles, but on a scale that is much slower than the break down.
3.2.4. Material recovery. The recovery was further probed using oscillatory rheology. Primarily the system underwent rejuvenation, whereby the occulated structure is broken down at a shear rate found within the shear-thinning region for all systems (10 s À1 for 10 s). The structuring was then observed through the storage modulus (G 0 ) over the subsequent 30 min, Fig. 10. The recovery curves show a two-step process, initially rapid, followed by a more gradual increase in elastic modulus, observed as a power function (0.45 AE 0.7). The same power law dependency observed across all systems can be argued as the same recovery mechanism being observed i.e. initially there is rapid formation of a large number of small ocs, as the ocs grow in size the change in G 0 slows as fewer larger particles are available to interact. The extent and rate of recovery however was dependent on the volume fraction of the particles. Volume fractions greater than 0.64 could not be signicantly differentiated, with the extent and rate of recovery depending on the self-similarity of the  shells. For EmFG < 0.64, G 0 values were lower by a factor of 10; as is observed from the frequency sweep data, again due to much less interacting where particles are further apart in space. As such, the hysteresis observed through dynamic shear experiments is suggested to arise not through the rupturing of a weakly gelled structure, but from the break down of reversible interactions between the particles.

Conclusions
This study has shown that oil droplets can be incorporated into a WPI gel layer by applying a shear-gel technique. Entrapment efficiency was observed to be dependant on the protein concentration, as a function of the viscosity and ow behaviours, reaching up to 99% encapsulation. By applying shear during the WPI sol-gel transition discreet micron sized spherical capsules were obtained, with enhanced structuring properties. Small deformation rheology was used to characterise the suspensions, which showed pseudosolid like behaviour at rest, however, by applying a shear force to the system that is greater than the yield stress, the suspensions could be made to ow. Suspension rheology highlighted a significant dependence on the oil fraction, with the addition of oil increasing the effective phase volume of the particles, resulting in an increase in particle proximity. Increasing the oil content to around 10 vol% led to packing fractions that exceeded those for hard spheres. As such, particle properties have been expressed as so and deformable. Flow behaviours of the suspensions were indicative of highly occulated systems, where marked shear thinning was observed through the break up of weakly aggregated ocs and mesostructures. Furthermore, when le under quiescent conditions the particulate suspensions showed signicant recovery, displaying the occurrence of reversible interactions.