Interfacial rheology and stability of air bubbles stabilized by mixtures of hydrophobin and β-casein

Food Hydrocolloids 34 (2014) 119e127

Contents lists available at SciVerse ScienceDirect

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

Interfacial rheology and stability of air bubbles stabilized by mixtures of hydrophobin and b-casein Joanne Burke a, Andrew Cox b, Jordan Petkov c, Brent S. Murray a, * a

Food Colloids Group, School of Food Science and Nutrition, University of Leeds, Leeds LS2 9JT, UK Unilever Research & Development, Colworth Science Park, Sharnbrook MK44 1LQ, UK c Unilever Research & Development, Port Sunlight, Wirral, Merseyside CH63 3JW, UK b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 17 October 2012 Accepted 27 November 2012

Class II hydrophobin (HFBII) is a highly surface active molecule and in the context of aeration can be considered to be an air structuring protein conferring exceptional stability to foams for periods far in excess of that obtained with any other commonly used protein. This is of interest to the food industry, since producing shelf stable foams in food formulations is very difficult. Although HFBII has proven to be very promising in terms of foam stability when used alone, it is still unknown whether HFBII will be able to maintain its functionality when other surface active agents are present, such as in real food systems. The surface rheology: surface shear viscosity (hs) and surface dilatational elasticity (ε), for HFBII and b-casein mixes at various ratios is described in this paper and how this relates to bubble stability. The addition of b-casein up to a certain ratio seems to increase hs significantly, whilst ε is less affected. This is accompanied by improved stability of air bubbles to coalescence and allows the formation of very small air bubbles that remain extremely stable to disproportionation. Overall, there is the suggestion of some kind of synergy between the two proteins. The exact nature of this interaction is unknown. Measurements of the z-potential of the proteins suggest that electrostatic interactions are probably not important at the pH investigated (pH 7). Confocal microscopy of individual bubbles over prolonged periods of time, stabilized by HFBII and a fluorescently labeled b-casein, suggests that the enhanced stability is due to highly unusual and complicated interfacial packing phenomena plus local bubble curvature effects that require further investigation. Ó 2012 Elsevier Ltd. All rights reserved.

Keywords: Hydrophobin b-casein Protein Bubbles Stability

1. Introduction Over the past ten years, hydrophobins have attracted a lot of attention especially in the context of aeration as they are considered to be an air-structuring protein, conferring exceptional stability to foams. Hydrophobins are small surface active proteins produced by filamentous fungi, found to cover the hyphal cell walls of fungal aerial structures (Hakanoaa, 2006; Wosten, 2001). Based on their solubility properties, two types of hydrophobins may be distinguished. Class I hydrophobins are highly insoluble and form amyloid-like rodlet membranes that can only be dissolved in certain strong acids. Class II hydrophobins are more water-soluble and can be readily dissolved in SDS or ethanol (Linder, Szilvay, Nakari-Setälä, Soderlund, & Penttilä, 2002; Scholtmeijer, Rink, Hektor, Wösten, & Dilip, 2005). In this study a class II hydrophobin (HFBII) was used. This same HFBII has recently been used by

* Corresponding author. Tel.: þ44 (0)113 343 2962; fax: þ44 (0)113 343 2982. E-mail address: [email protected] (B.S. Murray). 0268-005X/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.foodhyd.2012.11.026

Cox, Cagnol, Russell, and Izzard (2007), Cox, Aldred, and Russel (2009) to produce liquid foams that do not coarsen and that are stable to coalescence for periods of several months. This period of stability is far in excess of that obtained with any other commonly used proteins. This is of interest to the food industry, since producing shelf stable foams in food formulations is very difficult. The HFBII used in this study was produced by Trichoderma reesei and in common with other Class II hydrophobins it is a small globular protein (molecular weight z7 kDa) possessing eight conserved cysteine residues that form four intermolecular disulfide bonds. These bonds give HFBII an extremely rigid and stable structure (Hakanpaa, Linder, Popov, Schmidt, & Rouvinen, 2006). A hydrophobic patch is also located on the protein surface, estimated to be 12e19% of the total surface area of the protein (Hakanpaa et al., 2004; Kallio, Linder, & Rouvinen, 2007; Linder et al., 2002) and this motif is key in conferring high surface activity to the molecule. Indeed, to date hydrophobins are considered to be the most surface-active proteins known. They are able to spontaneously self-assemble into a robust amphiphilic membrane at a hydrophilicehydrophobic interface and can change the nature of

120

J. Burke et al. / Food Hydrocolloids 34 (2014) 119e127

the surface from hydrophobic to hydrophilic or vice versa. The driving force for this self-assembly would seem to be the concealment of the hydrophobic patch. Based on the multimerization model proposed by Hakanpaa, Szilvay, et al. (2006), HFBII is found in a dimeric state, concealing 34% of the hydrophobic patch compared to monomers, which is considered to be energetically more favorable. HFBII monomers are able to form monolayers at an oilewater (OeW) or airewater (AeW) interface by exposing their hydrophobic patch to the hydrophobic phase, creating an even more stable energetic state, as the entire patch can be concealed (Burtko et al., 2001; Linder et al., 2002; Lumsdon, Green, Stieglits, 2005; Szilvay, Nakari-Setälä & Linder, 2006). Furthermore, unlike other surface active globular proteins, HFBII does not seem to unfold significantly once it is adsorbed. The strong adsorption and the close and ordered packing of the molecules result in mechanically strong films that are excellent at resisting bubble shrinkage via disproportionation (Cox et al., 2009; Murray, Dickinson, Lau, Nelson, & Schmidt, 2005). Consequently, hydrophobin seems unique amongst proteins in behaving more like a rigid Janus-type particle (Cox et al., 2007). Linder et al. (2002) and Linder (2009) have provided several good reviews on hydrophobins and their functionality. Foams and bubbles play a very important role in many food products in terms of their structure and texture. However, foams are renowned for being unstable, even in frozen solidified products such as ice cream. Environmental changes, such as changes in pressure and/or temperature can lead to changes in the foam microstructure that affect both the physical and sensory properties of the product (Heuer, Cox, Singleton, Scott, Barigou & van Ginkel, 2007). The ability to produce liquid stable foams without relying on a gelled or solidified continuous phase is very much desired by food manufacturers since this would allow the development of new aerated products with better functionality, new textures and lower calorie content. One of the ways to create extremely stable bubbles is to adsorb a layer of surface active solids at the surface of the gas bubbles (Dickinson, Ettelaie, Kostakis & Murray, 2004; Du et al., 2003; Horozov, 2008; Hunter, Pugh, Franks, & Jameson, 2008; Murray & Ettelaie, 2004). In practice, finding surface-active particles in the correct size range that are acceptable for consumption is very difficult. Hydrophobins may be able to fulfill these criteria (Murray, Dickinson, & Wang, 2008). Although HFBII may be more surface active than other proteins when compared at equal bulk concentrations, real food products may contain a wide range of concentrations of different proteins and other surface active agents. At present, it is unknown whether HFBII will be able to maintain its functionality in such systems or if synergistic effects will occur with other proteins. It is therefore of great practical importance to test this point and in this work bcasein (BC) was chosen as a second surface active protein for the study of competitive effects with HFBII. This is a significant choice, since b-casein is the most surface active milk protein and milk protein powders are widely exploited for their surface active properties in foods and non-food products. The effects of pH and different concentration ratios of HFBII to b-casein on the stability of individual bubbles were studied in detail and compared with the corresponding changes in interfacial film rheology. 2. Materials and methods 2.1. Materials and sample preparation Class II Hydrophobin (HFBII) was supplied by Unilever (Colworth, UK), provided in an ammonium acetate buffer solution. It was then freeze-dried and stored in a vacuum oven (Gallenkamp) at 40  C for 18 h, in order to remove the water and buffer, to enable

solutions to be prepared at different pH values. The HFBII was then reconstituted in pure water at a concentration of 1.44 wt.% and stored frozen. Before conducting any measurements on HFBII samples, 1 min of sonication (via a Kerry sonicator, Kerry Ultrasonics, Hitchin, Herts, UK) at 45 kHz was applied separately to the original 1.44 wt.% solution and also to the diluted samples. This step is necessary to remove any small bubbles and to dissociate any protein aggregates. b-casein (BC) from bovine milk (98%, PAGE), potassium dihydrogen phosphate (K2H2PO4) disodium phosphate (Na2HPO4), sodium chloride (NaCl, 7647-14-5), sodium azide (NaN3), D-gluconic acid d-lactone (GDL) and xanthan gum were supplied by SigmaeAldrich (Poole, UK). Fluorescently labeled b-casein was obtained by conjugating FITC fluorescent probe to the protein and was provided by Ecole Ltd. (Qingdao, China), with a ratio of 5.5 fluorophores per BC molecule. Separate measurements (not shown) of the surface tension of 102 wt.% FITC-labeled BC versus time were not significantly different from the corresponding measurements on non-labeled BC, so that the labeling of the BC itself was not thought to significantly affect its surface active properties. All solutions used in this study were prepared in phosphate buffered saline using 0.02 mol dm3 KH2PO4 þ Na2HPO4 þ 0.05 mol dm3 NaCl buffer with the pH adjusted to pH 7  0.05 via addition of drops of 1 mol dm3 NaOH as necessary. Water from a Milli-Q apparatus (Millipore, Watford, UK), free from surface-active impurities and with a conductivity of less than 107 S cm1, was used throughout. To all solutions sodium azide was added (0.02 wt.%) to inhibit microbial growth during measurements. Mixed solutions of b-casein þ HFBII were prepared with 104 wt.% HFBII and concentrations of b-casein ranging from 103 wt.% to 101 wt.%. Each solution was stored overnight at 4  C before use and each experiment was repeated at least three times. To investigate the effect of pH on the surface activity of HFBII, GDL was used to acidify the solutions. The GDL granules were slowly added to the protein solution and were vigorously stirred for 5 min to allow thorough dissolution and even distribution of the GDL. The sample was then split into two parts: one was used for experimental purposes whilst the pH of the other portion was simultaneously monitored using a pH meter (Jenwau 3310, Essex) in the same laboratory at the same temperature (22  3  C). At this concentration and temperature, GDL slowly hydrolyzes and lowers the pH uniformly throughout the sample. After 24 h the desired pH was obtained and remained stable. This method was preferable to attempting to dissolve the proteins in buffers of different pH at the start, which led to problems of reproducibility, presumably due to difficult and slow solubilization of the proteins at some pH values. 2.2. Surface shear rheology The surface shear viscosity was measured using a twodimensional Couette-type viscometer which has been described in detail previously (Murray, 2002; Borbás, Murray, & Kiss, 2003; Martin, Bos, Stuart, & Van Vliet, 2002) and only a brief explanation will be given here. A biconical disk, hanging from a thin wire of known torsion constant, is positioned with its edge touching the AeW interface of the solution contained in a concentric circular dish. The deflection of the disk is measured by reflection of a 1 mW laser off a mirror on the spindle of the disk onto a scale at a fixed distance from the axis of the spindle. In most of the experiments described here the rheometer was operated in a constant shear-rate mode, although oscillatory measurements of low amplitude and frequency can also be performed (Borbás, Murray, & Kiss, 2003; Jourdain, Christophe Schmitt, Leser, Murray, & Dickinson, 2009). The motion of the beam on the scale was recorded digitally via a CCD camera for subsequent analysis of the disk deflection versus

J. Burke et al. / Food Hydrocolloids 34 (2014) 119e127

time. From the beam deflection and torsion constant of the wire the corresponding surface shear stress versus time can be calculated. Where a second protein was introduced after initial adsorption of a different protein, a peristaltic pump was used to exchange the sub-phase in the rheometer dish with the contents of a sealed flask external to the rheometer, without affecting the final volume (and therefore the height) of the sub-phase in contact with the disk edge. In this way the concentration of protein in the sub-phase can be diluted down to negligible levels and/or a second protein can be introduced at increasing concentration, through repeated replacement of the flask with fresh buffer and/or a solution of the second protein. Previous experience with this technique (Jourdain et al., 2009) has shown that if the adsorbed protein film is sufficiently robust, the exchange process itself does not disrupt the adsorbed layer. The surface shear viscosity is calculated from:

hs ¼ gf K qi =u

(1)

2 where gf is the geometric factor of the equipment ¼ ðR2 i  Ro Þ, ð4pÞ1 ; where Ri ¼ the radius of the disk (¼1.5 cm) and Ro ¼ the radius of the dish (¼7.3 cm); u is the angular velocity of the dish; K is the torsion constant and Өi is the angle of rotation of the disk.

2.3. Surface dilatational rheology A Langmuir trough with a flexible square rubber barrier, as described in detail elsewhere (Xu, Dickinson, & Murray, 2007) was used to measure the dilatational rheology of adsorbed films (Murray & Nelson, 1996). The surface tension was measured by the Wilhelmy plate method, using a thoroughly cleaned, roughened mica plate (3e5 cm in length), which was suspended in the middle of the trough from a force transducer. The inside dimensions of the trough were 9.5 cm and the rubber barrier was able to expand from 4 cm square to 7.2 cm square. After filling the trough to the required level with the protein sample, the surface was sucked away using a clean Pasteur pipette and vacuum pump. This defined ‘zero’ adsorption time. The interface was then left for 3 h, to allow protein to adsorb to the interface, before the barrier and therefore the film was expanded at different speeds. The advantage of the apparatus is that it can be used to expand the interface at fairly fast rates that also match the rate of surface expansion of bubbles in various pressure drop test of foam stability (see later). The change in interfacial tension Dg ¼ gg0, where g0 is the tension (g) just before the expansion, increases in magnitude as a film expands and then starts to fall back to zero when the expansion stops. In terms of this work, the more interesting part of the Dg versus time (t) plot is the initial rising part of Dg because previous work has shown that bubbles subjected to similar surface expansion rates are more likely to coalesce during the early stages of the expansion (Murray, Campbell, et al., 2002; Murray, Cattin, Schüler, & Sonmez, 2002; Soderberg, Dickinson, & Murray, 2003). One convenient way to analyze the whole rising part of the Dgt plot is to fit the curve to a Maxwell model (Murray, Campbell, et al., 2002; Murray, Cattin, et al., 2002):

Dg ¼ ½kds=dt þ ½kðds=dtÞexpðεt=kÞ

(2)

where s is the area strain DA/A, ε is an elastic (surface dilatational) modulus, and k is a viscous (surface dilatational) modulus, assuming the strain rate is approximately constant. In all cases the fit of the Maxwell model was always good enough to give a regression coefficient >0.95. Other, more complex models may be more justifiable for such a complex process of simultaneous interfacial protein expansion and protein adsorption. However, in

121

view of the experimental error in the measurements plus the assumptions that probably have to made in more advanced models, the Maxwell model is adequate to characterize the differences in behavior of the adsorbed films as a function of bulk protein composition. 2.4. Zeta potential measurements The z-potential of the solutions was measured by a Malvern Zetasizer Nano-ZS (Malvern Instruments Ltd, UK) in a DTS 1060 experimental cell. Each sample was filtered through a Whatman puradisc filter (25 mm diameter) with a 1.0 mm PTFE membrane and polypropylene housing (6784-2510) directly into the cell before measurements commenced. 2.5. Disproportionation measurements Bubble disproportionation behavior was observed as described previously (Dickinson, Ettelaie, Murray, & Du, 2002; Du et al., 2003; Murray, Campbell, et al., 2002; Murray, Cattin, et al., 2002). Briefly, air bubbles are injected beneath the AeW interface of the test solution in a sealed chamber and their behavior is recorded through a glass window via a microscope connected to a CCD camera and digital image capture system. Bubbles were created by vigorously hand-shaking some of the test solution in a clean stoppered glass tube for approximately 60 s. A sample of the lower part of the solution, which contained the smaller bubbles, was then injected into the chamber via a special port. Once injected, the bubbles rose quickly (in less than 10 s) to the AeW interface where they were prevented from moving out of the field of view by a thin mica barrier floating at the interface. Bubbles touching the walls of the barrier and bubbles less than 2 bubble diameters apart in the interface were discarded from the analysis, since this complicates their dissolution kinetics considerably (Ettelaie, Dickinson, Du, & Murray, 2003). From the images recorded at various time intervals after injection bubble sizes were measured via ImageTool v.3.0. 2.6. Confocal laser scanning microscopy of shrinking bubbles A Leica TCS SP2 confocal laser scanning microscopy (CLSM) was used in fluorescent mode in order to obtain images of the surface of bubbles stabilized by 104 wt.% HFBII þ 102 wt.% FITC-BC. After mixing the two proteins in a sample tube (10 mm diameter, 50 mm length), xanthan gum solution (0.9 wt.%) was added and the mixture stirred for 30 min. Bubbles were then formed by handshaking the tube in an up and down motion for 1 min. The tube was then rotated slowly end over end for a further 1 h at 15 rpm. The end of rotation was considered to be ‘zero’ bubble age. Xanthan was added to increase the bulk viscosity and therefore inhibit movement of bubbles once they had formed, so that they did not coalesce with each other or the AeW interface before sufficient protein had adsorbed to their surface to prevent their coalescence or shrinkage. Nevertheless, at these fairly low protein concentrations, it was difficult to maintain a sufficient number of stable bubbles and there was still substantial bubble creaming. It was not possible to increase the viscosity further otherwise it would have been too difficult to obtain small bubbles via shaking. For this reason, the tube rotation was also necessary to cancel out the effects of gravity and prevent significant bubble movement until aging produced some stable bubbles. Samples from the tubes were then carefully transferred via a Pasteur pipette to a confocal sample cell (diameter of 19 mm and 3 mm depth). A cover slip (0.17 mm thickness) was placed on top of the cell, ensuring that there was no air gap trapped between the sample and cover slip. The sample was scanned at 22  C (3  C) using a 10 objective lens, approximately

122

J. Burke et al. / Food Hydrocolloids 34 (2014) 119e127

30e50 mm below the bottom of the cover slip, in order to minimize hydrodynamic (and other) interactions with the cover slip. The samples were observed intermittently in the same cell for up to 4 days. Fluorescence from the sample was excited by the 488 nm Ar laser line and/or the 633 nm HeNe laser line. CLSM images were recorded at a resolution of 1024  1024 pixels and analyzed using Image-Pro Plus software. 3. Results and discussion 3.1. Surface rheology Interfacial rheology, and especially interfacial shear rheology, is a useful way of monitoring the formation and structuring of adsorbed protein layers. If any changes occur in the structure, composition or interactions within adsorbed protein layers, this is generally reflected in a change in the interfacial rheological properties, which in turn may be related to the corresponding foam (or emulsion) stability (Murray, 2002, 2007; Murray & Dickinson, 1996; Martin et al., 2002). This is especially useful when studying mixed systems, since the shear rheology can give insight into competitive and/or co-operative interactions at the interface. The surface shear viscosity (hs) measured after 24 h for the mixed HFBII þ BC systems, at pH 7 and 22  C, is shown in Fig. 1, where the HFBII concentration was fixed at 104 wt.%. Even at this low concentration, hs of HFBII alone was always very high (>1 N s m1 after 2e3 h adsorption). Unfortunately, there was also significant variation in hs between different batches of protein, for unknown reasons. Although all solutions were sonicated for a set time before use (see Methods section) the natural strong tendency for HFBII to aggregate may mean that some aggregation in the bulk persisted and altered the kinetics of evolution of hs for different preparations. It was partly for this reason that the initial test concentration was kept so low, to reduce bulk aggregation effects, but also because of the need to repeat experiments several times in order to obtain reproducible results and the scarcity of pure HFBII. The variations between batches were likely to be due to the fact that they were purified samples and therefore more sensitive to slight changes in the true protein concentration or denaturation or aggregation differences as a result of their preparation. In terms of real applications, less pure samples would probably be used anyway. Simple mass balance calculations suggest that there should be insignificant depletion of the bulk HFBII concentration for the usual adsorbed surface loads of protein (ca. 2 mg m2). It was also thought wise to fix the HFBII concentration at a low value for mixtures with other less surface active proteins, otherwise these proteins might have to be present at significantly higher concentrations to exert any effect at all. These higher

ηs /N s m-1 20

concentrations might be prohibitively expensive and irrelevant to real systems, as well as unrealistic because of limitations on protein solubility. Indeed, this is demonstrated in Fig. 1 by the result for 104 wt.% HFBII þ 0.1 wt.% BC, since even though the BC concentration is 1000 higher, hs for the mixture is still much closer to that of HFBII alone, since at 0.1 wt.% BC alone hs was still too low (i.e., <1 mN s m1) to be measured accurately using the set-up currently available. (Previous measurements by Murray & Dickinson, 1996; suggested hs values of around 0.5 mN s m1 for 103 wt.% BC, close to the maximum sensitivity of the biconical bob technique, due to the bulk drag on the disk due to water alone). However, it is interesting that the addition of BC up to a certain ratio consistently increased hs, by an amount of approximately 6 N s m1 at around 0.005 wt.% BC (i.e., at a bulk concentration ratio of BC: HFBII z 50:1). At BC concentrations above 102 wt.% hs decreased, to 4 N s m1 at 0.1 wt.% BC (BC: HFBII ¼ 1000:1), lower than for HFBII alone, but still very much higher than BC alone at 0.1 wt.%, as noted above. The results therefore corroborate the greater surface activity and higher film strength of HFBII compared to BC, but also indicate that some kind of interaction still occurs between the two proteins at the interface. The decrease in hs at higher concentrations of BC might be due to BC interfering in some way with the close-packing of HFBII, whilst the increase in hs at intermediate BC concentrations suggests strengthening via complex formation. Fig. 2 shows the time taken in minutes of hs for the mixtures to reach the arbitrarily chosen value of 0.1 N s m1. Note again that hs ¼ 0.1 N s m1 is significantly higher than for BC alone at any concentration, so that reaching this value should indicate significant HFBII adsorption. The lag times show that, by adding BC between 103 and 0.1 wt.%, it took progressively longer for hs to reach this value. Therefore, these results provide additional evidence that BC exerts some influence on film formation and structure, particularly at short adsorption times, even if the final interfacial film may be dominated by HFBII. The possibilities include: (a) significant BC adsorption takes place before HFBII starts to adsorb and displace BC from the interface and form its own monolayer; (b) interactions occur between BC and HFBII in the bulk that significantly curtail HFBII adsorption. Wang (2009) has shown that sodium caseinate adsorption can proceed more quickly than HFBII under similar conditions, although HFBII always results in a lower surface tension at long enough adsorption times, supporting the first possibility. To test further the competitive adsorption effects of BC and HFBII, the effect of sequentially exposing the proteins to the AeW interface was investigated. The results are shown in Fig. 3. Data for three different situations are shown: (i) both proteins present in the bulk solution at the start, i.e., 0.01 wt.% BC þ 104 wt.% HFBII; (ii) where 104 wt.% HFBII is present at the start and then after 24 h the sub-phase exchanged with BC solution to reach a final composition of 0.01 wt.% BC þ 106 wt.% HFBII; (iii) where 0.01 wt.% BC is present 200

15

t /min

10

100

5 0 0.00

0.05 0.10 BC concentration/ wt%

Fig. 1. Surface shear viscosity (hs) versus the bulk concentration of BC after 24 h adsorption for: 104 wt.% HFBII þ BC (-); BC alone (:).

0 0.00

0.05

0.10

BC concentration/ wt% Fig. 2. Time (t) taken for the surface shear viscosity (hs) to reach 0.1 N s m1 for each mixture of 104 wt.% HFBII þ BC (-).

J. Burke et al. / Food Hydrocolloids 34 (2014) 119e127 s/

Nsm

123

-1

10

5

0

20

30

t/h

40

50

Fig. 3. Effect of sequentially adding proteins on the surface shear viscosity (hs) for: BC added to 104 wt.% HFBII resulting in a final concentration of 106 wt.% HFBII þ 102 wt.% BC (-); HFBII added to 102 wt.% BC resulting in a final concentration of 104 wt.% BC þ 104 wt.% HFBII (,). The dashed line at hs ¼ 9.7 N s m1 indicates the value obtained for 24 h co-adsorption of 104 wt.% HFBII þ 102 wt.% BC from the start, for comparison.

at the start and then after 24 h the sub-phase exchanged with HFBII solution to reach a final composition of 104 wt.% BC þ 104 wt.% HFBII. Case (i) is therefore the same as already discussed above, represented in Fig. 1, and is the mixed composition that tends to give high hs values. For both 106 wt.% HFBII in case (ii) and 104 wt.% BC in case (iii), these concentrations of protein may be considered negligible, i.e., over the adsorption times considered, no measurable hs could be recorded for the proteins on their own at these concentrations: the other protein is therefore the dominant species in the bulk. In case (ii) the effect of adding BC to the previously adsorbed HFBII film is to decrease hs fairly quickly (in approximately 4 h) from ca. 11.2 N s m1 to ca. 5.5 N s m1, followed by a slight recovery to ca. 7.4 N s m1 in the following 24 h, when the experiment was terminated. This ‘final’ value is just slightly lower than the value (9.5 N s m1) obtained when both proteins were left to co-adsorb for 24 h or 48 h at this composition from the start. Thus, even though the creation of a huge excess of BC over HFBII in the bulk initially seems to weaken the HFBII layer already adsorbed, it seems that after this initial period the HFBII layer is able to recover its initial state. This recovery must be through replacement of any transient adsorption of BC by HFBII, or through some sort of association between the BC and HFBII at the interface. The case (iii) results show that the addition of HFBII to the pre-adsorbed BC film causes a slow increase in hs from almost zero to 5.5 N s m1 over ca. 24 h. This value is lower than the values for HFBII alone at this concentration (104 wt.%) but very much greater than for BC alone, i.e., much more characteristic of HFBII than BC. The results in cases (ii) and (iii) again reflect the higher surface activity of HFBII over BC, which is now well known (Cox et al., 2007, 2009; Hakanpaa, 2006; Radulova et al., 2012; Wang, 2009). Thus, even after saturation coverage of an interface by BC, subsequent exposure to low concentrations of HFBII gives adsorbed layer properties dominated by HFBII. All the time-scales studied here are rather long and the adsorption occurs under quiescent conditions, so that they probably represent the long-term, equilibrium adsorption position. How far these findings relate to the much shorter time scales involved during foam formation remains to be seen but we now turn to surface rheological measurements made under such time-scales. Fig. 4 shows the surface dilatational elastic (ε) and viscous (k) components obtained by fitting to the Maxwell model of Dg versus t for mixtures of 104 wt.% HFBII þ the same range of concentrations (103 to 0.1 wt.%) of BC as studied in the surface shear viscosity measurements. The adsorbed layer was expanded at a rate of dlnA/dt ¼ 0.012 s1. Across the full range of compositions, the

Fig. 4. Effect of concentration of BC on the fitted dilatational viscous modulus (,) for: 104 wt.% HFBII þ BC (-); 0.1 wt.% BC alone (,). Also the fitted elastic modulus (ε) for: 104 wt.% HFBII þ BC (C); 0.1 wt.% BC alone (B) (The inset illustrates the lower values of ε more clearly).

higher values of k over the ε are noted. However, k decreases rapidly as the BC concentration is increased, whilst the elastic modulus (ε) is hardly affected. This is similar to the results found by Radulova et al. (2012) for 0.005 wt.% HFBII þ BC varying from 0 to 0.14wt.%. The white square shows the value of k for 0.1 wt.% BC alone. Only a small increase in ε is noted (highlighted in the Fig. 4 inset), in a similar region (103102 wt.% BC) to where the much more pronounced maximum in hs occurred. It should be remembered that the dilatational elasticity and viscosity are affected by the dynamics of protein re-arrangement within the interface and how protein adsorption/desorption kinetics affect the interfacial tension, whilst hs is essentially only sensitive to in-film structure. The results in all the surface rheological experiments indicate that an adsorbed layer of HFBII persists at the AeW interface at all compositions studied. Possibly some small amount of HFBII desorption occurs and we can hypothesize that BC might fill these voids, affecting the overall rheological behavior of the film to a limited extent. If BC was able to displace HFBII significantly, or break up the adsorbed HFBII monolayers, a much more dramatic change in the surface rheological parameters would be expected, but this is not observed. An additional possibility, however, is that BC forms an adsorbed layer, or patches of adsorbed layer, beneath the largely coherent HFBII film. 3.2. Zeta potential measurements The most obvious way in which HFBII and BC might interact together in bulk solution and/or at the interface is via electrostatic interactions. It is therefore essential to know how the charge properties on the proteins vary with pH. There seems to be some debate about the charge on HFBII as a function of pH (Alexandrov et al., 2012; Basheva, Kralchevsky, Christov, et al., 2011; Basheva, Kralchevsky, Danov, et al., 2011; Cox et al., 2007; Kisko, Szilvay, Vainio, Linder, & Serimaa, 2008) and therefore it was decided to measure the zeta potential of our sample ourselves. Fig. 5 shows zpotential versus pH for 0.1 wt.% BC, 103 wt.% HFBII and 104 wt.% HFBII. The isoelectric pH values for BC and HFBII seem to be around pH 5 and pH 3.5, respectively. The value of pH 3.5 for HFBII is in agreement with the work of Cox et al.. (2007). Thus at pH 7, the pH at which the surface rheological measurements were made, both proteins should have an overall negative charge, which would tend to inhibit electrostatic attraction between the two proteins. Previous work carried out by Wang (2009) has shown that there does not seem to be any hydrophobic interaction between HFBII and the caseins either. Radulova et al. (2012) also investigated

124

J. Burke et al. / Food Hydrocolloids 34 (2014) 119e127

10 mV 0

pH

2

4

6

-10

Fig. 5. Effect of pH on the zeta potential (z) of: 104 wt.% HFBII (-); 103 wt.% HFBII (,); 101 wt.% BC (:).

mixtures of BC and HFBII and concluded that any interactions between the two proteins must be of a weak nature because hardly any changes were noted in the surface shear rheological properties of HFBII adsorbed layer properties when adding BC to HFBII. Of course it is possible that oppositely charged regions of either protein may interact with each other even if the overall charge on the proteins is the same. However, if the interaction was significant, the large excess of BC over HFBII in the mixtures might be expected to complex most of the HFBII and make it unavailable for adsorption. The results do not indicate that this is the case. Overall then, the z-potential measurements seem to confirm that strong interactions between BC and HFBII in the bulk solution at pH 7 should not be expected. We now examine how these ideas about the structuring of mixed HFBII þ BC films relate to the corresponding bubble stability. 3.3. Bubble disproportionation Previous studies (Dickinson et al., 2002) have shown that milk proteins do not confer good stability against disproportionation of air bubbles. Studies carried out on HFBII, on the other hand, have shown that hydrophobins can form a film around air bubbles that provides them with an extremely strong resistance to bubble shrinkage (Cox et al., 2007). Therefore, it was of interest to investigate the disproportionation behavior of mixtures of BC þ HFBII. Fig. 6 illustrates the main types of behavior that were observed for air bubbles with initial diameters ranging between 15 and 130 mm (There are no results for any bubbles smaller than this because any that were observed disappeared after only a few seconds and this was too rapid to quantify). The results are summarized in terms of the % of bubbles that were still visible (i.e., diameter > 5 mm) after 24 h. Firstly, it is important to note that any air bubbles in this initial size range stabilized by 0.1 wt.% BC alone dissolved away completely only a few seconds after they had reached the AeW interface. Also, it was impossible to create any significant number of bubbles using 104 wt.% HFBII alone. In contrast, mixtures of the two proteins enabled the formation of an adequate number of

60 % stable 40 20 0 0.00

0.05 0.10 BC concentration/ wt%

Fig. 6. Fraction (%) of bubbles still stable after 24 h versus concentration of BC for: 104 wt.% HFBII þ BC (-); BC alone (:).

bubbles for analysis and some of these bubbles were able to resist disproportionation for very long periods of time, i.e., several days, although some bubble shrinkage was still noted. In agreement with the surface rheological results, these general observations for the mixtures suggest that both BC and HFBII are present at the bubble surface and affect their stability. It would be expected that BC might provide foamability at the expense of reduced film strength and therefore increase susceptibility to shrinkage, whilst HFBII would bolster film strength and therefore increase stability to bubble dissolution. Thus, there seemed to be an optimum ratio of BC to HFBII for maximum bubble stability and it is interesting that this seemed to occur at around 102 wt.% BC, where higher values of hs were observed. For all bubbles, bubble diameter versus time (data not shown) was obtained via image analysis. It was noted that, for a fair proportion (approximately 70%) of bubbles that remained stable, the diameter initially decreased then leveled off at certain size, thereafter the bubbles remained indefinitely stable. We could find no correlation between this tendency for shrinkage followed by stability and the initial size of the bubbles, nor between this behavior and the composition of the systems. However, it is tempting to suggest that the initial shrinkage represented BC desorption from the interface, until the bubble surface was more completely covered by the less desorbable HFBII, which then formed a coherent film that resisted shrinkage effectively. In other words, the surface composition changed with time until a ‘pure’ hydrophobin-stabilized bubble was formed, more resistant to shrinkage. In order to confirm this hypothesis, the surface composition of the AeW interface would have to be monitored as a function of interface compression and time and we hope to report in the future the results of such experiments currently underway. For now, we describe some interesting observations of the surface of shrinking bubbles that goes part way to understanding how the surface composition and structure might actually change, as follows. 3.4. CLSM observations of shrinking bubbles CLSM images of air bubbles stabilized by 104 wt.% HFBII þ 102 wt.% FITC-labeled BC are shown in Figs. 7 and 8, at different aging times. This composition was chosen because it was the one which seemed to give optimum foamability and foam stability, as described above. In the images the lighter regions represent higher fluorescence intensity, i.e., BC-rich areas. Darker regions are therefore assumed to be more rich in HFBII. The appearance of the bubbles is very striking and unexpected, in particular the long thin bright striations across the bubble surfaces, which on closer examination appear to be bordered by many small, dark roundish zones. When analyzed more closely, in fact most of the small dark regions appear to contain a small bright region near their center (e.g., see Figs. 7(d) and 8(f)). The same sort of structures have been observed many times in different preparations of the bubbles, so that we do not think the features observed are an artifact of the measurement method. Overall, the clearest statement that can be made is that both HFBII and BC appear to be present in patches all over the bubble surfaces, i.e., the adsorbed film is a mixture of the two proteins. As referred to in the previous section, bubbles tended to shrink with time and then stabilize, or disappeared completely. Both Figs. 7 and 8 show images of bubbles at different aging times over several days. The difference between the conditions of these measurements compared to the measurements of disproportionation described in Section 3.3 is that, in the former the bubbles are immobilized within the bulk aqueous phase, whereas in the latter the bubbles are trapped beneath an AeW interface. As such, the

J. Burke et al. / Food Hydrocolloids 34 (2014) 119e127

125

Fig. 7. CLSM images of a smaller sized air bubbles (initial diameter 170 mm) stabilized by 104 wt.% HFBII þ 102 wt.% FITC-BC at various aging times: (a) 0.5 h; (b) 24 h and (c) 48 h. The size bar ¼ 40 mm. The region highlighted by the white dashed box in (a) is the same region highlighted by the black dashed box in (b). Image (d) is an enlargement of the indentation region in (b).

CLSM measurements are much more difficult to make, in terms of following the same bubble over prolonged periods of time, since some bubble movement still occurred even at the relatively high bulk viscosity used. Despite these difficulties, some good examples were recorded of the general effects observed on prolonged storage (over 7 days), as in Fig. 8. A common qualitative observation was that on aging the proportion of black regions seemed to increase with time. This was measured quantitatively for selected bubbles where good clear images could be obtained over the time course of the shrinkage. Fig. 9 gives examples, where the fraction of pixels

below a threshold gray level of 50 is shown, as a % of the total area of the bubble surface visible at that particular time. This therefore takes into account the decrease in the total bubble surface area with time as the bubbles slowly dissolve. It is seen that the % of dark pixels does not change significantly and this does not substantiate the previously stated hypothesis of BC preferentially desorbing with time, leaving the interface richer in HFBII. The driving force for the dissolution of bubbles is higher for smaller bubbles due the higher Laplace pressure and indeed initially larger bubbles (>100 mm diameter) were much more likely

Fig. 8. CLSM images of a larger sized air bubble (initial diameter 620 mm) stabilized by 104 wt.% HFBII þ 102 wt.% FITC-BC at various aging times: (a) 0.5 h; (b) 24 h; (c) 48 h; (d) 120 h; (e) 168 h. The size bar ¼ 135 mm. Image (f) is an enlargement of the indentation region in (e).

126

J. Burke et al. / Food Hydrocolloids 34 (2014) 119e127

1. 5

1. 5

(a)

Br 1. 0

1. 0

0. 5

0. 5

0. 0

0. 0

0

t/h

100

(b)

Ar

200

0

t/h

100

200

Fig. 9. Effect of shrinkage on the fraction of dark (HFBII-rich) regions of several air bubbles stabilized by 104 wt.% HFBII þ 102 wt.% BC. In (a) the ratio (Br) of the % of black pixels/% of black pixels at the start of shrinkage is shown. In (b), the corresponding ratio (Ar) of the bubble area (in pixels)/the initial bubble area at the start of shrinkage is shown.

to persist and be visible after 7 days than bubbles that initially started out smaller. However, the above observations of light and dark regions suggest that the fraction of HFBII in the interface does not increase significantly. Since HFBII is the component that confers resistance to dissolution and shrinkage, one might expect all stable bubbles to have the same surface composition after a certain time. On the other hand, the more rapid shrinkage of the smaller bubbles will mean that they have less time for an enriched HFBII interface to form a coherent film around the bubbles that is able to arrests shrinkage completely. At the same time, the resolution of CLSM simply might be not good enough to distinguish the true % of light (BC) and dark (HFBII) domains at the interface. Finally, another common and striking aspect observed was the formation of crinkled edges of the bubbles. A good example is highlighted in Fig. 7(d) and 8(f). In Fig. 7(a), a larger dark region is highlighted by the dashed line box. In Fig. 7(b), this dark region appears to correspond to the indentation at the edge of the bubble also highlighted by a dashed box. Note the bubble itself has shrunk considerably in the time between the 2 images. Fig. 7(d) is a closeup of this indented region. Thus the dark, supposedly HFBII-rich regions appear to have opposite curvature to that of the overall bubble surface. On closer observation of all bubbles that had shrunk considerably, the bubble surface appeared to be covered in large numbers of these puckered regions, which gave the overall surface a rough or crinkled appearance. As far as we are aware, this has not been observed before, but this goes a long way to explaining the stabilizing effect of HFBII. If HFBII regions are so rigid that they flip the curvature in this way then the overall average curvature of the bubble surface will tend to zero. The evolution of separate BC-rich and/or HFBII-rich domains as a function of film compression could obviously be more easily observed in flat, planar films in Langmuir trough type experiments, but this would not reveal these curious local curvature effects. The crinkled surface is reminiscent of the puckered surface of so-called armored bubbles stabilized by particles (Subramaniam, Mejean, Abkarian, & Stone, 2006) or crystallized domains of sucrose esters (Dressaire, Bee, Bell, Lips, & Stone, 2008), although the local curvature of these surface features seemed to remain positive. 4. Conclusions The results of this study suggest that the addition of BC up to a certain ratio increases the surface viscosity (hs) of HFBII films and improves bubble stability by decreasing the shrinkage rate and allowing the formation of small bubbles that can survive indefinitely. There is the suggestion of some kind of synergy between the proteins at the interface at these ratio. This interaction between BC and HFBII is probably not electrostatic in nature; possibly it is

related to the way the proteins pack together at the interface. The order of addition of the proteins has no significant effect on hs, confirming the greater surface activity of HFBII. In theory, this suggests that during bubble shrinkage BC would be more likely to desorb from the airewater interface, but no direct evidence for this was obtained experimentally. Confocal microscopy of individual shrinking bubbles suggested that the interface structure and local curvature became highly inhomogeneous and many of these features require further investigation in order to be fully explained. Acknowledgments The authors gratefully acknowledge the support from Unilever R&D in both Colworth and Port Sunlight as well as funding from EPSRC (Industrial CASE award 09002749). References Alexandrov, N. S., Marinova, K. G., Gurkov, T. D., Danov, K. D., Kralchevsky, P. A., Stoyanov, S. D., et al. (2012). Interfacial layers from the protein HFBII hydrophobin: dynamic surface tension, dilatational elasticity & relaxation time. Journal of Colloid and Interface Science, 376, 296e306. Basheva, E. S., Kralchevsky, P. A., Christov, N. C., Danov, K. P., Stoyanov, S. D., Blijdenstein, T. B. J., et al. (2011). Uniques properties of bubbles and foam films stabilized by HFBII hydrophobin. Langmuir, 27, 2382e2392. Basheva, E. S., Kralchevsky, P. A., Danov, K. P., Stoyanov, S. D., Blijdenstein, T. B. J., Pelan, E. G., et al. (2011). Self-assembled bilayers from the protein HFBII hydrophobin: nature of the adhesion energy. Langmuir, 27, 4481e4488. Borbás, R., Murray, B. S., & Kiss, E. (2003). Interfacial rheological behaviour of proteins in two liquid systems with low interfacial tension. Colloids and Surfaces A: Physicochemical and Engineering Aspects, 213, 93e103. Butko, P., Buford, J. P., Goodwin, J. S., Stroud, P. A., Mccormick, C. L., & Cannon, G. C. (2001). Spectroscopic evidence for amyloid-like interfacial self-assembly of hydrophobin Sc3. Biochemical and Biophysical Research Communications, 208, 212e215. Cox, A., Aldred, D. L., & Russel, A. B. (2009). Exceptional stability of food foams using class II hydrophobin HFBII. Food Hydrocolloids, 23, 366e376. Cox, A., Cagnol, F., Russell, A. B., & Izzard, M. J. (2007). Surface properties of class II hydrophobins from Trichoderma reesei and influence on bubble stability. Langmuir, 23, 7995e8002. Dickinson, E., Ettelaire, R., Kostakis, T., & Murray, B. S. (2004). Factors controlling the formation and stability of air bubbles stabilized by partially hydrophobic silica nanoparticles. Langmuir, 20, 8517e8525. Dickinson, E., Ettelaie, R., Murray, B. S., & Du, Z. P. (2002). Kinetics of disproportionation of air bubbles beneath a planar air-water interface stabilized by food proteins. Journal of Colloid and Interface Science, 16, 321e331. Dressaire, E., Bee, R., Bell, D. C., Lips, A., & Stone, H. A. (2008). Interfacial polygonal nanopatterning of stable microbubbles. Science, 320, 1198e1201. Du, Z. P., Bilbao-Montoya, M. P., Binks, B. P., Dickinson, E., Ettelaie, R., & Murray, B. S. (2003). Outstanding stability of particle-stabilized bubbles. Langmuir, 19, 3106e 3108. Ettelaie, R., Dickinson, E., Du, Z., & Murray, B. S. (2003). Disproportionation of clustered protein-stabilized bubbles at planar air-water interfaces. Journal of Colloid and Interface Science, 263, 47e58. Hakanpaa, J. (2006). Structural studies of T. reesei hydrophobins HFBI and HFBII e the molecular basis for the function of fungal amphiphiles. Dissertation, University of Joensuu, Finland.

J. Burke et al. / Food Hydrocolloids 34 (2014) 119e127 Hakanpaa, J., Linder, M., Popov, A., Schmidt, A., & Rouvinen, J. (2006). Hydrophobin HFBII in detail: ultrahigh-resolution structure at 0.75 angstrong. Acta Crystallographica. Section D, Biological Crystallography, 62, 356e367. Hakanpaa, J., Paananen, A., Askolin, S., Nakari-Setälä, T., Pakkinen, T., Penttilä, M. E., et al. (2004). Atomic resolution structure of the HFBII hydrophobin, a selfassembling amphiphile. The Journal of Biological Chemistry, 279, 534e539. Hakanpaa, J., Szilvay, G. R., Kaljunen, H., Maksimainen, M., Linder, M., & Rouvinen, J. (2006). Two crystal structures of Trichoderma reesei hydrophobin HFBIIdthe structure of a protein amphiphile with and without detergent interactions. Protein Science, 15, 2129e2140. Heuer, A., Cox, A., Singleton, S., Barigou, M., & van Ginkel, M. (2007). Visualisation of foam microstructure when subject to pressure change. Colloids and Surfaces A: Physicochemical and Engineering Aspects, 311, 112e123. Horozov, T. S. (2008). Foams and films stabilised by solid particles. Current Opinion in Colloid & Interface Science, 13, 134e140. Hunter, T. N., Pugh, R. J., Franks, G. V., & Jameson, G. J. (2008). The role of particles in stabilising foams and emulsions. Advances in Colloid and Interface Science, 137, 57e81. Jourdain, L. S., Christophe Schmitt, C., Leser, M. E., Murray, B. S., & Dickinson, E. (2009). Mixed layers of sodium caseinate þ dextran sulfate: influence of order of addition to oilewater interface. Langmuir, 25, 10026e10037. Kallio, J. M., Linder, M. B., & Rouvinen, J. (2007). Crystal structures of hydrophobin HFBII in the presence of detergent implicate the formation of fibrils and monolayer films. Journal of Biological Chemistry, 282, 28733e28739. Kisko, K., Szilvay, G. R., Vainio, V., Linder, M. B., & Serimaa, R. (2008). Interaction of hydrophobin protin in solution studied by small angle X-ray scattering. Biophysical Journal, 94, 198e206. Linder, M. B. (2009). Hydrophobins: proteins that self-assemble at interfaces. Current Opinion in Colloid & Interface Science, 14, 356e363. Linder, M. B., Szilvay, G. R., Nakari-Setälä, T., Soderlund, H., & Penttila, M. (2002). Surface adhesion of fusion proteins containing the hydrophobins HFBI and HFBII from Trichoderma reesei. Protein Science, 2257e2266. Lumsdon, S. O., Green, J., & Stieglitz, B. (2005). Adsorption of hydrophobin proteins at hydrophobic and hydrophilic interfaces. Colloids and Surfaces B: Biointerfaces, 44, 172e178. Martin, A., Bos, M., Stuart, M. C., & Van Vliet, T. (2002). Stress-strain curves of adsorbed protein layers at the A-W interface measured with surface shear rheology. Langmuir, 18, 1238e1243. Murray, B. S. (2002). Interfacial rheology of food emulsifiers and proteins. Current Opinion in Colloid & Interface Science, 7, 426e431. Murray, B. S. (2007). Stabilization of bubbles and foams. Current Opinion in Colloid & Interface Science, 12, 232e241.

127

Murray, B. S., Campbell, L., Dickinson, E., Maisonneuve, K., Nelson, P. V., & Soderberg, I. (2002). Technique for studying the effects of rapid surface expansion on bubble stability. Langmuir, 15, 5007e5014. Murray, B. S., Cattin, B., Schüler, E., & Sonmez, Z. O. (2002). The response of adsorbed protein films to rapid expansion. Langmuir, 18, 9476e9484. Murray, B. S., & Dickinson, E. (1996). Interfacial rheology and the dynamic properties of adsorbed films of food proteins and surfactants. Food Science and Technology International, 2, 131e145. Murray, B. S., Dickinson, E., Lau, C. K., Nelson, P. V., & Schmidt, E. (2005). Coalescence of protein-stabilized bubbles undergoing expansion at a simultaneously expanding planar air-water interface. Langmuir, 21, 4622e4630. Murray, B. S., Dickinson, E., & Wang, Y. (2008). Bubble stability in the presence of oil-in-water emulsion droplets: influence of surface shear versus dilatational rheology. Food Hydrocolloids, 23, 1198e1208. Murray, B. S., & Ettelaie, R. (2004). Foam stability: proteins and nanoparticles. Current Opinion in Colloid & Interface Science, 9, 314e320. Murray, B. S., & Nelson, P. V. (1996). A novel Langmuir trough for equilibrium and dynamic measurements oil air-water and oil-water monolayers. Langmuir, 12, 5973e5979. Radulova, G. M., Golemanov, K., Danov, K.,D., Kralchevsky, P.,A., Stoyanov, S. D., Arnaudov, L. N., et al. (2012). Surface shear rheology of adsorption layers from the protein HFBII hydrophobin: effect of added b-casein. Langmuir, 28, 4168e 4177. Scholtmeijer, K., Rink, R., Hektor, H. J., Wösten, H. A. B., & Dilip, K. A. A. R. M. B. (2005). Expression and engineering of fungal hydrophobins. In D. K. Arora, & R. M. Berka (Eds.), Applied Mycology and Biotechnology (pp. 239e255). Amsterdam: Elsevier. Soderberg, I., Dickinson, E., & Murray, B. S. (2003). Coalescence stability of gas bubbles subjected to rapid pressure change at a planar air/water interface. Colloids and Surface B: Biointerfaces, 30, 237e248. Subramaniam, A. B., Mejean, C., Abkarian, M., & Stone, H. A. (2006). Microstructure, morphology, and lifetime of armored bubbles exposed to surfactants. Langmuir, 22, 5986e5990. Szilvay, G. R., Nakari- Setälä, T., & Linder, M. B. (2006). Behavior of Trichoderma reesei hydrophobins in solution: interactions, dynamics and multimer formation. Biochemistry, 45, 8590e8598. Wang, Y. (2009). Coalescence and disproportionation of air bubbles stabilized by proteins and emulsion droplets. PhD thesis, University of Leeds, UK. Wosten, A. B. H. (2001). Hydrophobins: multipurpose proteins. Annual Review of Microbiology, 55, 625e646. Xu, R., Dickinson, E., & Murray, B. S. (2007). Morphological changes in adsorbed protein films at the air-water interface subjected to large area variations, as observed by Brewster angle microscopy. Langmuir, 23, 5005e5013.