Journal of Colloid and Interface Science 296 (2006) 332341

www.elsevier.com/locate/jcis

Microstructure of -lactoglobulin-stabilized emulsions containing non-ionic

surfactant and excess free protein: Influence of heating
Sven Kerstens, Brent S. Murray, Eric Dickinson
Procter Department of Food Science, University of Leeds, Leeds LS2 9JT, UK
Received 7 June 2005; accepted 21 August 2005
Available online 15 September 2005

Abstract
The influence of the non-ionic surfactant Tween 20 on the microstructure of -lactoglobulin-stabilized emulsions with substantial excess free
protein present was investigated via confocal microscopy. The separate distributions of oil droplets and protein were determined using two different
fluorescent dyes. In the emulsion at ambient temperature the excess protein and protein-coated oil droplets were associated together in a reversibly
flocculated state. The pore-size distribution of the initial flocculated emulsion was found to depend on the surfactant/protein ratio R, and at higher
values of R the system became more inhomogeneous due to areas of local phase separation. Evidence for competitive displacement of protein
from the oilwater interface by surfactant was obtained only on heating (from 25 to 85 C) during the process of formation of a heat-set emulsion
gel. By measuring fluorescence intensities of the protein dye inside and outside of the oil-droplet-rich areas, we have been able to quantify the
evolving protein distribution during the thermal processing. The results are discussed in relation to previous work on the competitive adsorption
of proteins and surfactants in emulsions and the effect of emulsion droplets on the rheology of heat-set protein gels.
2005 Elsevier Inc. All rights reserved.
Keywords: Confocal laser scanning microscopy; Competitive displacement; Emulsion gel; Phase separation; -Lactoglobulin; Non-ionic surfactant; Whey protein;
Tween 20

1. Introduction
Obtaining stability in complex food systems remains a major challenge for the food industry. Whereas two ingredients
may individually enhance the desired properties of a product,
the interplay of the mixed components is sometimes a destabilizing factor. Whey proteins are widely used as structural
building blocks in food products [13]; their ability to denature
and aggregate into a three-dimensional network makes them especially suitable as gelling agents and texture modifiers. Most
frequently the gelation of whey proteins is induced by heating,
although cold-set gels can also be obtained by the addition of
salts or by using high pressure or enzymatic treatments [47].
The main molecular interactions involved in the association
and cross-linking of denatured whey proteins are hydrophobic
forces and covalent bonding via intermolecular disulfide linkages [812]. The final gel microstructure is determined by a
* Corresponding author. Fax: +44 (0) 113 343 2982.

E-mail address: e.dickinson@leeds.ac.uk (E. Dickinson).

0021-9797/$ see front matter 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcis.2005.08.046

balance of local attractive and repulsive forces involving both

whey protein molecules and their aggregates, and it is strongly
influenced by pH and ionic strength [1316].
The presence of small-molecule surfactants (emulsifiers)
can affect the functional properties of proteins. For instance,
the addition of the water-soluble non-ionic surfactant, polyoxyethylene sorbitan monolaurate (Tween 20), influences the
thermal gelation behavior of the major whey protein, -lactoglobulin (-lg), both in solution and in -lg-stabilized emulsions [17,18]. Recently, we reported [19] on the use of confocal
microscopy to follow the microstructure of heated -lg solutions at pH 6.8 with and without added surfactant. While the
strands making up the final heat-set gel were too small to be resolved by the technique, and the homogeneous final microstructure was visibly unaffected by the presence of Tween 20, we
observed that during the process of heating from 25 to 85 C
there appeared micrometer-sized aggregates which disappeared
again upon further heating towards the gelation point. Moreover, we found that the temperature-dependent evolution of the
appearance and sizes of these aggregates was dependent on the

S. Kerstens et al. / Journal of Colloid and Interface Science 296 (2006) 332341

surfactant/protein molar ratio R. While the exact nature of the

aggregates is yet has to be established, we do know from the
literature [2022] that Tween 20 binds specifically to -lg in a
1:1 fashion; so it is possible that some kind of reversible selfassembly involving -lg/Tween 20 complexes lies at the origin
of these pseudo-micellar aggregates.
Due to their much higher adsorption energy per molecule,
milk proteins saturate fluid interfaces at much lower concentrations than do small-molecule surfactants [23]. However, at
higher concentrations, due to their more effective interfacial
packing, the interfacial tension is lowered more by smallmolecule surfactants than by proteins. Hence, milk proteins are
competitively displaced by surfactants from oilwater and air
water interfaces, as has been extensively reported [2028] for
systems containing a mixture of Tween 20 + whey proteins (including pure -lg). Typically, in oil-in-water emulsions, there
is complete displacement of -lg from the oilwater interface
by non-ionic surfactants at high surfactant/protein ratios (i.e.,
R  15) [29].
From a material science point of view, the rigidity of a heatset protein gel can be strongly affected by the presence of socalled filler particles like emulsion droplets. The oil volume
fraction, as well the size, shape and deformability of the inclusions, influences the overall mechanical behavior of the filled
gel. The composition of the adsorbed monolayer around the
droplets also has important implications for the bulk rheology
of the system because it influences the nature of the interaction
between the filler particles and the protein gel network [30].
This role of the fillermatrix interaction has been demonstrated
using scanning electron microscopy [31,32] and confirmed by
rheological measurements on heat-set whey protein emulsion
gels [33,34]. Droplets coated with protein are incorporated into
the protein gel network via hydrophobic and covalent bonding,
and hence they act as active fillers; but droplets stabilized by
small-molecule surfactants have little affinity for the surrounding protein matrix, and so they behave as inactive fillers. Previous work in our laboratory on heat-set gels made from whey
protein-stabilized (or -lg-stabilized) emulsions has shown [17,
18,3439] that the elastic shear modulus is indeed greatly enhanced by the presence of the active filler particles. On the
other hand, the presence of emulsion droplets stabilized by a
low-molecular-weight surfactant (such as Tween 20) leads to a
lower elastic modulus.
Most of this existing knowledge about the effect of surfactant/protein competitive adsorption on emulsion properties has
come from a combination of protein surface coverage measurements made at ambient temperature [29,3843] and bulk
rheological measurements [17,18,3337,4449]. There is very
little information available on the influence of added surfactants
on the emulsion microstructure. Visualizing the microstructure
in situ seems essential, however, for a proper understanding of
the factors affecting the bulk mechanical behavior and stability of emulsions before, during and after thermal processing.
In a preliminary confocal microscopy study of this type, using an oil-soluble dye to identify the oil phase, Chen and coworkers [35] revealed substantial qualitative effects of added
emulsifier on the final microstructure of a heat-set emulsion gel.

333

But this earlier study [35] provided no quantitative information

about the distribution of the protein in the system; nor was the
microstructure followed in time during heating.
Here, for the first time, we report on the use of confocal microscopy to follow time-dependent changes in microstructure
during the formation of a heat-set emulsion gel. Spatial distributions of oil droplets and protein are separately monitored using
two different fluorescent dyes. Specifically, we examine the effect of varying amounts of Tween 20 on the initial and evolving
microstructure of heated -lg-stabilized emulsions containing
a high content of excess unadsorbed protein.
2. Materials and methods
2.1. Materials
Bovine -lactoglobulin (L-3908, lot number 024K7051),
Tween 20 (polyoxyethylene sorbitan monolaurate), polyethylene glycol, Nile Blue and Nile Red were purchased from Sigma
Chemicals (Poole, UK). The protein sample was a mixture of
-lactoglobulin variants A and B, and no further purification
was applied. In order to calculate the surfactant/protein molar
ratio R, we assumed average molecular weights of 1228 Da and
18.3 kDa respectively for Tween 20 and -lg [50]. 1-Bromohexadecane (99%, Acros Organics, Belgium) was used as the
emulsion oil phase and distilled water was used for the preparation of the solutions.
2.2. Preparation of protein solutions
Concentrated protein solutions were used to investigate the
effect of temperature and environment on the fluorescence behavior of the protein dye Nile Blue. Solutions were prepared
containing 13.2 wt% -lg with Tween 20 present at surfactant/protein ratios of R = 0, 0.5 and 2. The solutions were
adjusted to pH 6.8 with 0.1 M HCl.
2.3. Preparation of emulsions
A starting oil-in-water emulsion (45 vol% oil, 2.7 wt% -lg
in the aqueous phase) was made using a jet homogenizer operating at a constant pressure of 300 bar. This stock emulsion was
divided into several aliquots. During gentle stirring each aliquot
was topped up with an aqueous solution containing known concentrations of -lg and Tween 20 to obtain a set of final oil-inwater emulsions (20 vol% oil) having a total protein content of
13.2% (w/v) and values of the surfactant/protein ratio R varying between 0 and 2. The pH was adjusted dropwise with 1 M
HCl to pH 6.8.
The particle-size distribution and mean diameter d32 of the
emulsions were determined by static light scattering using a
Malvern Mastersizer 2000 with an assumed value of 1.1 for the
relative refractive index and 0.0005 for the particle absorption.
2.4. Gelation time test
Test tubes containing ca. 4 ml of emulsion were placed in a
water bath at 85 C. The tubes were taken out for observation at

334

S. Kerstens et al. / Journal of Colloid and Interface Science 296 (2006) 332341

regular intervals, and the gel point was defined empirically as

the time at which a tube could be turned upside-down without
observable downwards flow of its contents, i.e., indicating the
formation of a self-supporting gel.
2.5. Confocal microscopy
A Leica TCS SP2 confocal system mounted on a Leica DMRXE upright microscope was used to examine the microstructure of the samples. Emulsions and protein solutions were put
in small cylindrical containers (80 l), covered with a cover
glass (size 1.5), and sealed with silicone oil to prevent evaporation. The containers were placed on a Peltier microscope
heating stage. The temperature was increased from 25 to 85 C
at 2 C min1 . Following this heating ramp, the temperature
was kept constant at 85 C for 30 min, after which time the
sample was allowed to cool down to 25 C.
The proteins were stained with Nile Blue (12.5 l of dye solution added to 2.5 ml emulsion to obtain a dye/protein ratio
1:100) and excited at 633 nm. Nile Red was used for staining of the emulsion oil phase (25 l of 0.01% (w/v) dye in
polyethylene glycol added to 2.5 ml emulsion) and the sample was excited at 488 nm. Fluorescence intensity data for Nile
Blue and Nile Red were collected in two separate channels corresponding to 503613 and 653750 nm, respectively. Only
one collection channel was required for the protein solutions.
A bleed-through of the signal from the first channel (Nile Red)
could be measured in the second channel, where Nile Blue was
detected. Therefore, fluorescence data for the two dyes were
not collected simultaneously, but instead the sequential linescanning mode was used. Every line of pixels in the image was
sequentially scanned for the Nile Red and Nile Blue fluorescence, thereby completely avoiding interference due to crossfluorescence. Before and after heating both a dry 20 lens (NA
0.7) and a 63 water-immersion lens (NA 1.2) were used to
study the microstructure of the sample, while a dry 20 lens
(NA 0.5) of longer working distance (1.15 mm) was used during the heating process to prevent thermal damage to the other
lenses.
2.6. Image analysis
Fluorescence intensity distributions in the micrographs were
analyzed using the confocal microscope operating software.
The aqueous protein solutions gave a uniform picture with
no structure visible, except when Tween 20 was present and
the system was heated, in which case small aggregates appeared and then disappeared again during heating, as reported
elsewhere [19]. Here we are solely interested in the behavior of the Nile Blue fluorescence, since this parameter will be
used to quantify the distribution of the protein component in
the emulsions. At set time intervals we measured the average
grey value intensity of the image, corresponding to the fluorescence intensity of the dye. We ignored the small protein
aggregates that appeared and disappeared in the sample, on the
basis that they would not affect the average fluorescence intensity.

Micrographs of emulsion samples were analyzed using the

Leica image analysis software QWin. The protein distribution,
in images taken with the 63 lens, was determined by measuring the fluorescence intensity of Nile Blue inside and outside
the oil-droplet-rich areas (using the Nile Red fluorescence of
the oil phase as a mask to define the different regions). To determine the change in fluorescence intensity distribution during
heating, the analysis was based on micrographs taken with the
20 lens (NA 0.5).
Pore sizes were measured on the micrographs showing Nile
Red fluorescence (indicating the oil phase), taken with the lower
magnification 20 lens (NA 0.7) in order to have a bigger field
of view than with the 63 lens. A crucial step in image analysis
is thresholdingin order to differentiate the background from
the objects of interest within the image. The type of images
we obtained here did not have a clear bimodal distribution of
grey levels, whereby the two sub-populations of pixels clearly
belong to the background and the foreground. So, in principle, some deconvolution technique has to be used to define the
threshold between the background and the objects of interest
[51,52]. Here we implemented thresholding in a series of consecutive steps using the Leica image analysis software Qwin.
The typical steps involved Gaussian smoothing to reduce noise,
after which the bright detail was enhanced by adding a whitetop-hat filtered image to the original. This enhanced image was
transformed using a point function (LUT-transform) whereby
lower intensities were suppressed. After this step the image was
thresholded, and the dark enclosed regions were assigned to
pores and then quantified. As these dark regions were generally
non-circular (non-spherical in 3-D), the pore size was expressed
in terms of area (m2 ).
3. Results and discussion
3.1. Effect of environmental conditions on fluorescence
behavior
We aim to quantify the evolving protein distribution in regions of the emulsion sample images by analyzing the fluorescence intensity of Nile Blue. Therefore it was necessary first to
determine the sensitivity of the dye fluorescence to the different
conditions.
Fluorescence intensities measured for the protein solutions
during heating are given in Fig. 1 for three surfactant/protein ratios (R = 0, 0.5 and 2). The electronic gain setting was adjusted
so that each sample had the same initial fluorescence intensity
value (170 10 units) and that the changing fluorescence intensity values remained throughout well within the full 0255
grey scale. Off-set values (signal cut-off values) were all set to
zero. These settings were not changed during the course of each
heating experiment.
The general tendency shown in Fig. 1 is that, in the presence
or absence of Tween 20, the fluorescence intensity declines as
temperature increases. On holding the temperature at 85 C for
30 min, the fluorescence intensity increases slightly again. On
cooling back down to 25 C, the average fluorescence intensity returns to an even higher value than before the sample was

S. Kerstens et al. / Journal of Colloid and Interface Science 296 (2006) 332341

335

3.2. Emulsion systems

Fig. 1. Effect of thermal processing on Nile Blue fluorescence of protein solutions (13.2 wt% -lg, dye/protein ratio 1:100) in the presence and absence
of Tween 20. The fluorescence intensity is plotted against time: (!) R = 0,
gain setting = 425 V; (") R = 0.5, gain setting = 400 V; (Q) R = 2, gain
setting = 385 V. The solid line represents the heating profile.

heated. This latter behavior suggests to us that no substantial

photo-bleaching is involved. Heating of fluorochromes can often induce reduction in quantum efficiency due to the increased
vibration and torsion of their flexible side-arms, resulting in
an internal radiationless transition, as reported previously for
Rhodamine B [19,53]. In our system, when the temperature
reaches 85 C, gelation takes place [19]. Together with the fact
that complexation of Nile Blue with the protein hinders rotational relaxation processes and increases the quantum yield of
the dye [54], gelation might explain the increase in fluorescence
intensity on holding at 85 C, arising from some immobilization of the dye within the aggregated protein network. A further rigidity-enhancing effect in heat-set globular protein gels
takes place during the cooling down stage, due to strengthening of hydrogen bonding and electrostatic forces [17,18]. This
could explain the increase in the fluorescence intensity after
cooling to a value even higher than for the original unheated
sample.
The second main feature of the protein solution data is the
increase in fluorescence intensity of Nile Blue on addition of
Tween 20. A lower gain setting has to be used when increasing the surfactant concentration in order to obtain similar average grey scale values in the image at room temperature. One
plausible explanation [54] is that the presence of surfactant micelles prevents radiationless deactivation of the excited dye to
its ground state, resulting in a higher overall fluorescence intensity. In the solution of low surfactant content (R = 0.5) all of
the Tween 20 is expected to be bound to -lg [1922]. No surfactant micelles would be expected to exist, therefore, until the
protein denatures on heating and releases the bound Tween 20
back into the solution, further complicating the interpretation of
the fluorescence intensity.
Others factors that could also potentially affect the measured
fluorescence intensity are the relative concentrations of dye and
protein and the pH. But each of these is kept constant. So in our
systems the fluorescence intensity is predominantly dependent
on the surfactant (micelle) concentration, the temperature, and
the aggregation state of the protein (or proteindye complexes).

In the low-protein-content stock emulsion, straight after homogenization, there was no evidence of flocculation. The most
probable droplet diameter was around 0.5 m and the d32 value
given by the Mastersizer was 0.38 0.02 m. While flocculation was induced on subsequently adjusting the protein concentration to 13.2% (see below), the measured particle-size distribution was unaffected by the subsequent operations of diluting
the aliquots or adjusting the pH.
Fig. 2 shows a set of images for the 20 vol% oil-in-water
emulsion with 13.2% -lg in the aqueous phase (no surfactant
present). Image (a) represents the fluorescence collected from
the oil phase stained with Nile Red, and image (b) represents
the fluorescence from the protein stained with Nile Blue. The
overlay image (c) permits location of the two dyes with respect to each other. Individual oil droplets could not normally
be distinguished, except for some exceptionally large ones. For
instance, in Fig. 2, we can see how one big oil droplet towards
the left side of the picture gives a dark round hole (no discernible intensity) in the image recording the Nile Blue/protein
fluorescence.
For the system represented in Fig. 2, it is clear that the
protein-coated oil droplets and the excess protein present in the
aqueous continuous phase are in a closely associated state
hence the predominantly yellow appearance of the overlay
image (green + red yellow). The original stock emulsion
(45 vol% oil, 2.7 wt% protein) gave a set of images (not shown)
with a homogeneous distribution of the oil droplets. The flocculation evident in Fig. 2 had therefore been induced by topping
up the high-volume-fraction stock emulsion with the concentrated -lg solution as described in Section 2.3. As no discernible change in mean droplet size d32 was detectable on
dilution in the Mastersizer, the heterogeneity apparent in Fig. 2
must be the result of a process of weak, reversible flocculation. It would therefore seem that, at the low ionic strength of
this system, the protein added after homogenization has greater
affinity for the weakly aggregated protein-coated oil droplets
than for the aqueous serum, and that this manifests itself in the
micrographs as local phase separation.
Fig. 3 shows the effect of the added surfactant on the microstructure of the emulsion at 25 C. Overlay micrographs,
including fluorescence from both the protein (Nile Blue) and
the oil (Nile Red) are given for R = 0, 0.75, 1.25 and 2. By
eye we can see that, depending on the amount of Tween 20
present in the system, the pore-size distribution varies substantially. Fig. 4 shows the effect of surfactant content on the mean
pore size (white bars, left axis) and the maximum pore size
(dashed bars, right axis) as derived from the image analysis.
With addition of Tween 20, the mean pore size decreases initially, and then it increases strongly. The quoted error bars refer
to the standard error. For large surfactant additions (R > 1.5),
the error in mean pore size diverges as the structure becomes
quite inhomogeneous. For instance, we can see in Fig. 3d that,
despite some large open (dark) areas in the image, there are still
some areas with much smaller pores. For this reason we also
include data in Fig. 4 for the maximum pore size. Whereas the

336

S. Kerstens et al. / Journal of Colloid and Interface Science 296 (2006) 332341

(a)

(b)

(c)
Fig. 2. Confocal micrograph of a sample of an oil-in-water emulsion (20 vol%
oil, 13.2 wt% -lg in the aqueous phase, 25 C) with no Tween 20 added:
(a) image from Nile Red fluorescence (highlighting the oil phase); (b) image
from Nile Blue/protein fluorescence; (c) overlay image combining fluorescence from both dyes. The arrow identifies a circular dark hole in the Nile
Blue/protein image corresponding to a single large oil droplet. Scale bar represents 30 m.

average pore size only changes by a factor of 5 over the whole

range of surfactant contents, we can see that there is a nearly
30-fold difference between the lowest (R = 0.5) and the highest (R = 1.75) values of the maximum pore size. So it would
seem that, while the average pore size is a reasonable indicator
of structural homogeneity, it should properly be supplemented
by information on the whole range of pore sizes.

It is noteworthy that, at the lower Tween 20 concentrations (R  1), the pores are smaller than in the absence of
surfactant. This is possibly attributable to the known 1:1 complexation of Tween 20 with -lg [2022]. From our previous calculations [19], all the surfactant present is expected
to be bound to the protein for R  1. The protruding polyoxyethylene chains might thereby affect the solubility of the
aggregated proteins. At higher surfactant/protein ratios, however, free surfactant (monomer + micelles) would be expected
to be present in the system, inducing local phase separation, as
seen in Figs. 3c and 3d. Such reversible depletion flocculation
of emulsion droplets due to the presence of surfactant micelles
is well recognized in the literature [55,56]. The R values quoted
above are, of course, calculated with respect to all the protein
present in the system, but it is likely that the protein adsorbed
at the oilwater interface interacts differently with the Tween
20. Furthermore, small amounts of surfactant probably already
absorb at the interface for R  1. These combined factors add
some uncertainty as to the precise value of the free surfactant
concentration in the system.
Fig. 3 shows that the surfactant content strongly affects
the way that the protein is distributed within the microstructure. In micrographs (a), (b) and (d), corresponding to R = 0,
0.75 and 2, respectively, we can see that the protein-coated oil
droplets and the excess protein are roughly located within the
same local regions. At higher magnification, the images corresponding to the separately collected fluorescence data have
the same general appearance as images (a) and (b) in Fig. 2
for R = 0. In strong contrast, however, Fig. 3c (R = 1.25) has
a red colored background instead of a black one, indicating
that at this composition the non-adsorbed protein is spread out
evenly throughout the aqueous continuous phase, instead of being closely associated with the aggregated protein-coated oil
droplets. Corresponding to Fig. 3c, the primary image indicating just the Nile Blue/protein fluorescence has a rather uniform
appearance (not shown), in contrast to the clustered regions of
aggregated protein as seen in Fig. 2b.
In order to quantify how the protein is distributed within
these samples, we have separately estimated the average fluorescence intensities of Nile Blue/protein complex inside and
outside of the oil-droplet-rich areas. Absolute quantification
would require a single calibration curve under strictly defined
conditions [54]. In our case this is not feasible, because the environment experiences continuously changing conditions that
affect the fluorescence behavior in a complex manner (see Section 3.1). Furthermore, an increase in the Tween 20 concentration in our emulsion systems was found to lead to a decrease
in fluorescence intensity. This is the opposite effect to that observed for the pure protein solutions (Section 3.1), illustrating
the further complexity of this notionally relatively simple system of -lg + Tween 20. In practice, the electronic gain and
off-set settings on the confocal microscope had to be changed
from sample to sample, and adjusted during heating in order
to keep the measured fluorescence level within the 0255 grey
scale range. In order to standardize this process, we took the
occasional voids of big oil droplets in the micrographs of the
protein fluorescence (e.g., see Fig. 2b) as points of zero fluores-

S. Kerstens et al. / Journal of Colloid and Interface Science 296 (2006) 332341

(a)

(b)

(c)

(d)

337

Fig. 3. Effect of the addition of surfactant on the microstructure of an oil-in-water emulsion (20 vol% oil, 13.2 wt% -lg in the aqueous phase, 25 C). Each of the
four confocal micrographs is an overlay image incorporating both Nile Red and Nile Blue fluorescence: (a) R = 0; (b) R = 0.75; (c) R = 1.25; (d) R = 2. Scale bar
represents 100 m.

Fig. 4. Effect of the surfactant/protein ratio R on the mean pore size (white bars, left scale) and maximum pore size (shaded bars, right scale). Standard errors are
given for the mean pore size.

cence intensity. Setting this reference level to a zero intensity

value, then any other region in the micrograph showing a higher
intensity value could be properly regarded as a region with a
true positive fluorescence contribution, as opposed to one with
just amplified background noise. On this basis we introduce a

protein distribution ratio (PDR) defined as the protein fluorescence within the oil-droplet-rich areas divided by the protein
fluorescence outside the oil-droplet-rich areas. So, a PDR substantially higher than unity implies that the excess unadsorbed
protein is mainly associated with the oil droplets; a value near

338

S. Kerstens et al. / Journal of Colloid and Interface Science 296 (2006) 332341

Table 1
Mean values of absolute grey levels and their standard deviations (SD) for Nile
Blue/protein fluorescence before and after heating measured inside and outside
of oil-droplet-rich areas as a function of the surfactant/protein ratio R
R

0
0.5
0.75
1
1.25
1.5
1.75
2

Before heating

After heating

Oil-rich areas

Oil-free areas

Oil-rich areas

Oil-free areas

Mean

SD

Mean

SD

Mean

SD

Mean

SD

164.3
176.7
178.4
191.2
175.8
177.3
166.3
152.7

2.7
8.0
11.2
9.5
16.1
8.9
2.5
10.1

36.8
28.6
38.0
51.1
150.0
135.5
101.4
64.6

2.6
1.8
7.5
8.6
19.5
7.4
2.5
12.0

188.8
161.3
83.7
95.7
98.6
128.3
82.6
81.4

8.0
6.2
3.2
9.7
6.2
19.1
6.9
6.4

183.1
193.1
157.4
184.0
179.4
204.3
174.9
177.5

8.9
9.3
2.7
16.6
5.8
13.3
11.2
9.0

Fig. 5. Effect of the surfactant/protein ratio R on the protein distribution ratio

PDR as calculated from the grey values listed in Table 1: (") before heating;
(!) after heating.

unity corresponds to a fairly homogeneous distribution of protein within the system; and a value considerably below unity
means that the protein is mainly distributed outside the oildroplet-rich areas.
Table 1 lists the sets of measured grey level intensities inside and outside the oil-droplet-rich areas, both before and after
heating. The calculated PDR values are plotted as a function of
R in Fig. 5. Before heating, for R  1, we see that the PDR
value is high (5 1). This means that five times more fluorescence is measured in the oil-droplet-rich areas than outside.
At R = 1.25 and R = 1.5, we have PDR 1, indicating that
the protein is fairly evenly distributed throughout the system,
as already qualitatively observed in Fig. 3c. For higher surfactant/protein ratios, Fig. 5 shows that the protein again becomes
more closely associated with the oil droplets, although substantially more protein is still present outside the oil-droplet-rich
areas as compared to the systems of low surfactant content.
Hence, for the case of no surfactant present or only a small
amount, the excess (non-adsorbed) protein has a high affinity
for the oil-coated droplets. On the other hand, when the surfactant/protein ratio exceeds unity, it appears that the excess
Tween 20 makes the protein more soluble (dispersible) within
the continuous phase, possibly by a molecular mechanism involving the shielding of hydrophobic regions of the protein

from the solvent [57]. At even higher surfactant concentrations, there is a local phase separation between a surfactant-rich
phase and a phase containing non-adsorbed protein + proteincoated oil droplets. Comparison of the data in Figs. 4 and 5
shows no obvious direct correlation between the effects of
the added surfactant on the pore size and the protein distribution.
The systems before thermal treatment were stable towards
changes in visible appearance and microstructure for at least 3
days on storage at refrigerator temperature. Under the confocal
microscope there was no obvious evidence of displacement of
protein from the oilwater interface by Tween 20. This is not
really surprising, as the thickness of an adsorbed monolayer
at the oilwater interface is no more than a few nanometers,
as compared with a resolution of not less than a few hundred
nanometers for the confocal microscope. Previous research has
demonstrated [2029] that addition of Tween 20 or some other
non-ionic surfactant after emulsion formation leads to complete
spontaneous displacement of protein from the oilwater interface at high surfactant/protein ratios (R  15). But typically in
this earlier work the emulsion contained a low concentration of
protein (e.g., 0.5 wt% [29]), as compared to 13.2 wt% for the
emulsions studied here. In the present study we are only considering relatively low surfactant/protein ratios (up to R = 2),
even though in absolute terms the concentration of Tween 20
in our emulsion at R = 0.75 is already higher than for R = 15
in the earlier study [29]. It has been recognized recently [58]
that the stability properties of -lg-based emulsions at ambient
temperature are affected considerably by the presence of free
protein. Furthermore, in our emulsions the large amount of free
protein may neutralize the competitive adsorption effect of the
surfactant present in the system, due to the known 1:1 complexation between Tween 20 and -lg [1922]. This means that, for
R  1, most of the Tween 20 may be unavailable for disrupting
the protein adsorbed layer or for displacing protein from the interface [17,18]. What is more likely, however, is that some limited displacement did occur in our unheated emulsions, but that
it was not detectable under the resolution of the confocal microscope. Indeed, we have confirmed separately that an emulsion
of much higher surfactant/protein ratio (20 vol% oil, 1 wt% protein, R 16) possessed a completely uniform microstructure,
with no evidence of any discernible protein depletion from the
surface of the largest oil droplets (and, incidentally, no change
in microstructure on heating).
In our emulsions containing 20 vol% oil and 13.2 wt% protein, it was only during the heat treatment that confocal microscopy gave evidence of protein becoming displaced from
the oilwater interface. All the systems exhibited gelation when
held at 85 C for several minutes in tubes in a water bath. The
heating profile used for the emulsions was exactly the same
as that used for the protein solutions (see Fig. 1). The PDR
values for the different surfactant/protein ratios after heating
are plotted in Fig. 5. It is clear that, in the final gel state of
the surfactant-free emulsion after heating and cooling down,
the protein is evenly distributed within the system (PDR = 1).
However, with Tween 20 added, the PDR value decreases until
it reaches the value of PDR 0.5 for R  0.75, which means

S. Kerstens et al. / Journal of Colloid and Interface Science 296 (2006) 332341

(a)

(b)

(c)
Fig. 6. Confocal micrograph of the final gel state, after cooling down, of a
heated emulsion with surfactant/protein ratio R = 2: (a) Nile Red fluorescence
(oil phase); (b) Nile Blue/protein fluorescence; (c) overlay image. Scale bar
represents 30 m.

that a significant fraction of the protein has moved out of the

oil-droplet-rich areas.
Fig. 6 shows a set of images for the final gel emulsion system with R = 2 after heating and cooling down. The oil droplets
still appear in a flocculated state (Fig. 6a), but clearly these
areas are now depleted of protein (Fig. 6b). Much of the protein is concentrated in the gaps between the oil-droplet-rich
areas. In the corresponding overlay image (Fig. 6c) we can
see clear separation of the green and red colors, representing
the distinct regions of Nile Red and Nile Blue fluorescence,

339

respectively. (Contrast this with Fig. 2c, where fluorescence

from the protein overlaps directly the fluorescence from the oil
phase, resulting in the composite yellow color.) What has happened, we infer, is that the Tween 20 has displaced the -lg
from the oilwater interface. Before heating, the free surfactant
and proteinsurfactant complex are in equilibrium, the relative amounts depending on the value of R. During heating, the
gradually unfolding protein releases the bound Tween 20 back
into the aqueous phase, where the increasing concentration of
free surfactant starts to displace the protein from the oilwater
interface. That is, during heating, the system is transformed
from a protein-stabilized emulsion into a surfactant-stabilized
emulsion. The newly formed surfactant-coated oil droplets are
incompatible with the high content of protein in the aqueous
phase, and consequently they exclude themselves from the developing protein network. That is, the oil droplets start to act as
inactive fillers, reducing the overall strength of the final emulsion gel [17,18].
Heat treatment in the presence of surfactant leads to conversion of the initial fine emulsion into a coarse emulsion. We
can observe in the images of Fig. 6 that there is a considerable fraction of large droplets of several micrometers diameter in the system with R = 2 after heating. This suggests that
the heat denaturation of -lg, and its associated displacement
from the oilwater interface, is accompanied by significant oil
droplet coalescence. Evidence for considerable droplet coalescence during emulsion gel formation was also reported previously by Chen and co-workers [35]. Nevertheless, despite this
coalescence, there is no visible loss of stability due to creaming or oiling off in the final heat-set emulsion system because
the coarse aggregated droplets eventually become immobilized
in the cross-linked protein gel matrix.
The PDR value of the final gel state for R = 0.75 is the
same as that for R = 2 (see Fig. 5). Probably this implies that
there is already enough surfactant present at R = 0.75 in the
heated emulsion to completely remove the protein from the
oilwater interface. Thus adding more surfactant does not necessarily mean that any more protein becomes excluded from the
oil-droplet-rich areas. For lower Tween 20 contents (R < 0.75),
the protein and oil droplets were observed to remain evenly dispersed together during the heating process (not shown).
Table 2 presents the absolute fluorescence values from the
Nile Blue/protein complex recorded during heating within the
two different regions of the emulsion sample with R = 2. The
corresponding PDR values are plotted in Fig. 7. Clearly we
can see that there is more protein initially located in the oildroplet-rich areas, but during heating the protein distribution
becomes shifted towards the areas devoid of oil droplets. The
distribution value changes steadily from PDR 1.4 at 25 C to
PDR 0.5 at 85 C, and it stays around this latter value until
the temperature is lowered again to 25 C. The sequence of micrographs corresponding to the data points in Table 2 and Fig. 7
is presented in an accompanying movie. The separate timedependent images from Nile Red and Nile Blue fluorescence
are displayed in this movie, together with the overlay image.
Initially more (excess) protein is located within the oil-dropletrich-areas. Gradually protein moves away from these regions

340

S. Kerstens et al. / Journal of Colloid and Interface Science 296 (2006) 332341

Table 2
Mean values of absolute grey levels and their standard deviations (SD) for Nile
Blue/protein fluorescence measured inside and outside of oil-droplet-rich areas
for a system of surfactant/protein ratio R = 2 as a function of temperature/time
during thermal processing
Temperature ( C)
(time)

Oil-rich areas
Mean

SD

Mean

SD

25
30
35
40
45
50
55
60
65
70
75
80
85
85 (5 min)
85 (10 min)
85 (15 min)
85 (20 min)
85 (25 min)
85 (30 min)
25

177
150
154
138
123
118
119
120
128
93
87
63
63
72
72
93
70
73
87
88

30
29
27
29
26
29
31
27
34
33
36
36
35
38
36
40
36
38
42
40

124
115
116
112
102
117
133
134
149
150
161
141
156
170
165
189
163
161
170
177

26
25
23
24
23
26
31
28
38
39
43
52
44
43
41
39
43
47
48
47

Oil-free areas

ing the changing microstructure of fine emulsions with droplets

of mean size of the order of the resolution of the microscope. But, despite the differences, both these studies have
demonstrated the power of the technique for exploring timedependent changes induced by surfactants in food colloid systems.
4. Conclusions
Model whey protein emulsions containing a high concentration of non-adsorbed protein (-lg) and variable amounts of
non-ionic surfactant Tween 20 have been examined by confocal microscopy before, during and after heat-induced gelation. In the emulsion system before heat treatment, we have
observed that the excess protein has more affinity for the
protein-coated droplets than for the aqueous medium. However, the protein distribution is substantially affected by the
concentration of added surfactant, as is the emulsion morphology as measured by the pore-size distribution, although these
two structural characteristics appear not to be directly correlated.
Not unexpectedly, we found no evidence for displacement
of -lg from the oilwater interfacial layer (thickness: a few
nanometers) based on observations at ambient temperature with
the confocal microscope (resolution: hundreds of nanometers).
However, the manifestation of such competitive displacement
was very apparent on heating the samples (up to 85 C). During this process the protein component was seen to move
away from the aggregated oil droplets into the aqueous continuous phase regions of the emulsion microstructure, so that
finally the surfactant-coated droplets became phase-separated
inactive filler particles within the heat-set protein gel network.
These findings may have potential implications for using the
combination of added emulsifiers and thermal processing to
control the texture and rheology of protein-based emulsion
gels.
Acknowledgments

Fig. 7. Time-dependent change in protein distribution ratio PDR during thermal

processing for the emulsion system of surfactant/protein ratio R = 2, as calculated from the data in Table 2. A best-fit polynomial has been added as a guide
to the eye. Also shown is the temperature/time profile.

until the Nile Blue and Nile Red images become completely
complementary, indicating that the aggregated oil droplets and
the gelling protein are mutually excluded from each other.
Finally, it seems appropriate to make mention of the pioneering confocal microscopy investigation of Heertje and coworkers [59,60] demonstrating the displacement of protein
from the oilwater interface by emulsifier. There are two main
differences between Heertjes experiments and those reported
here. Firstly, their experiments were carried out in a more uniform physico-chemical environment, with time as the only variable factor, whereas here we describe changes in protein distribution with time, temperature and emulsifier concentration. Secondly, Heertjes experiments were done at a semi-macroscopic
oilwater interface, whereas here we are concerned with study-

The authors thank EPSRC (Grant No. GR/S01566/01) and

the IMPACT Faraday Partnership (UK) for financial support of
this project.
Supplementary material
The online version of this article contains additional supplementary material (the movie).
Please visit DOI: 10.1016/j.jcis.2005.08.046.

References
[1] J.E. Kinsella, D.M. Whitehead, Adv. Food Nutr. Res. 33 (1989) 343.
[2] C.V. Morr, E.A. Foegeding, Food Technol. 44 (4) (1990) 100.
[3] J.E. Kinsella, in: P.F. Fox, J.E. Condon (Eds.), Food Proteins, Applied
Science, London, 1982, p. 51.
[4] C.M. Bryant, D.J. McClements, Trends Food Sci. Technol. 9 (1998) 143.
[5] M. McSwiney, H. Singh, O.H. Campanello, Food Hydrocoll. 8 (1994) 441.

S. Kerstens et al. / Journal of Colloid and Interface Science 296 (2006) 332341

[6] V.B. Galazka, E. Dickinson, D.A. Ledward, Curr. Opin. Colloid Interface
Sci. 5 (2000) 182.
[7] N. Kitabatake, E. Doi, Food Res. Int. 9 (1993) 445.
[8] A.C. Alting, R.J. Hamer, C.G. de Kruif, R.W. Visschers, J. Agric. Food
Chem. 48 (2000) 5001.
[9] D. Galani, R.K. Owusu Apenten, Int. J. Food Sci. Technol. 34 (1999) 467.
[10] M.A.M. Hoffmann, P.J.J.M. van Mil, J. Agric. Food Chem. 45 (1997)
2942.
[11] G.A. Manderson, M.J. Hardman, L.K. Creamer, J. Agric. Food Chem. 46
(1998) 5052.
[12] P. Relkin, Int. J. Biol. Macromol. 22 (1998) 59.
[13] M. Langton, A.-M. Hermansson, Food Hydrocoll. 5 (1992) 523.
[14] T. Lefevre, M. Subirade, Biopolymers 54 (2000) 578.
[15] M. Stading, M. Langton, A.-M. Hermansson, Food Hydrocoll. 6 (1992)
455.
[16] M. Stading, M. Langton, A.-M. Hermansson, Food Hydrocoll. 7 (1993)
195.
[17] S.-T. Hong, E. Dickinson, in: G.O. Phillips, P.A. Williams, D.J. Wedlock
(Eds.), Gums and Stabilizers for the Food Industry, vol. 8, Oxford Univ.
Press, Oxford, 1996, p. 320.
[18] E. Dickinson, S.-T. Hong, J. Agric. Food Chem. 43 (1995) 2560.
[19] S. Kerstens, B.S. Murray, E. Dickinson, Food Hydrocoll. 19 (2005) 625.
[20] A.R. Mackie, P.A. Gunning, P.J. Wilde, V.J. Morris, J. Colloid Interface
Sci. 210 (1999) 157.
[21] P.J. Wilde, D.C. Clark, J. Colloid Interface Sci. 155 (1993) 48.
[22] D.C. Clark, P.J. Wilde, D.J.M. Bergink-Martens, A.J.J. Kokelaar, A. Prins,
in: E. Dickinson, P. Walstra (Eds.), Food Colloids and Polymers: Stability
and Mechanical Properties, Royal Society of Chemistry, Cambridge, UK,
1993, p. 354.
[23] E. Dickinson, in: E.D. Goddard, K.P. Ananthapadmanabhan (Eds.), Interactions of Surfactants with Polymers and Proteins, CRC Press, Boca
Raton, FL, 1993, p. 295.
[24] P.J. Wilde, A.R. Mackie, F. Husband, P.A. Gunning, V.J. Morris, Adv. Colloid Interface Sci. 108109 (2004) 63.
[25] A.R. Mackie, Curr. Opin. Colloid Interface Sci. 9 (2004) 357.
[26] L.A. Pugnaloni, E. Dickinson, R. Ettelaie, A.R. Mackie, P.J. Wilde, Adv.
Colloid Interface Sci. 107 (2004) 27.
[27] J. Krgel, D.C. Clark, P.J. Wilde, R. Miller, Prog. Colloid Polym. Sci. 98
(1995) 239.
[28] J.-L. Courthaudon, E. Dickinson, Y. Matsumura, D.C. Clark, Colloids
Surf. 56 (1991) 293.
[29] J. Chen, E. Dickinson, J. Sci. Food Agric. 62 (1993) 283.
[30] T. van Vliet, Colloid Polym. Sci. 266 (1988) 518.

341

[31] R.A. Yost, J.E. Kinsella, J. Food Sci. 57 (1992) 892.

[32] K.R. Langley, M.L. Green, J. Texture Stud. 20 (1989) 191.
[33] D.J. McClements, F. Monahan, J.E. Kinsella, J. Texture Stud. 24 (1993)
411.
[34] J. Chen, E. Dickinson, Colloids Surf. B 12 (1999) 373.
[35] J. Chen, E. Dickinson, M. Langton, A.-M. Hermansson, Lebens. Wiss.
Technol. 33 (2000) 299.
[36] J. Chen, E. Dickinson, J. Agric. Food Chem. 46 (1998) 91.
[37] E. Dickinson, S.-T. Hong, Y. Yamamoto, Neth. Milk Dairy J. 50 (1996)
199.
[38] E. Dickinson, S. Tanai, J. Agric. Food Chem. 40 (1992) 179.
[39] J. Chen, E. Dickinson, Colloids Surf. A 101 (1995) 77.
[40] J.-L. Courthaudon, E. Dickinson, Y. Matsumura, A. Williams, Food
Struct. 10 (1991) 109.
[41] J.-L. Courthaudon, E. Dickinson, D.G. Dalgleish, J. Colloid Interface
Sci. 145 (1991) 390.
[42] J.-L. Courthaudon, E. Dickinson, W.W. Christie, J. Agric. Food Chem. 39
(1991) 1365.
[43] E. Dickinson, S.R. Euston, C.M. Woskett, Prog. Colloid Polym. Sci. 82
(1990) 65.
[44] J. Chen, E. Dickinson, J. Texture Stud. 29 (1998) 285.
[45] J. Chen, E. Dickinson, Int. J. Food Sci. Technol. 34 (1999) 493.
[46] E. Dickinson, J. Chen, J. Dispers. Sci. Technol. 20 (1999) 197.
[47] E. Dickinson, Y. Yamamoto, J. Food Sci. 61 (1996) 811.
[48] E. Dickinson, Y. Yamamoto, Food Hydrocoll. 10 (1996) 301.
[49] E. Dickinson, S.-T. Hong, Colloids Surf. A 127 (1997) 1.
[50] C.V. Morr, E.Y.W. Ha, Crit. Rev. Food Sci. Nutr. 33 (1993) 431.
[51] M. Mellema, J.W.M. Heesakkers, J.H.J. van Opheusden, T. van Vliet,
Langmuir 16 (2000) 6847.
[52] L.A. Pugnanoli, L. Matia-Merino, E. Dickinson, Colloids Surf. B 42
(2005) 211.
[53] T. Karstens, K. Kobs, J. Phys. Chem. 84 (1980) 1871.
[54] S.H. Lee, J.K. Suh, M. Li, Bull. Korean Chem. Soc. 24 (2003) 45.
[55] D.J. McClements, Colloids Surf. A 90 (1994) 25.
[56] E. Dickinson, C. Ritzoulis, M.J.W. Povey, J. Colloid Interface Sci. 212
(1999) 466.
[57] A. Malhotra, J.N. Coupland, Food Hydrocoll. 18 (2004) 101.
[58] H.-J. Kim, E.A. Decker, D.J. McClements, Langmuir 20 (2004) 10394.
[59] I. Heertje, J. Nederlof, H.A.C.M. Hendrickx, E.H. Lucassen-Reynders,
Food Struct. 9 (1990) 305.
[60] I. Heertje, H. van Aalst, J.C.G. Blonk, A. Don, J. Nederlof, E.H. LucassenReynders, Lebens. Wiss. Technol. 29 (1996) 217.