pubs.acs.org/Langmuir © 2010 American Chemical Society
Revealing the Size, Conformation, and Shape of Casein Micelles and Aggregates with Asymmetrical Flow Field-Flow Fractionation and Multiangle Light Scattering Maria Glantz,*,† Andreas Ha˚kansson,† Helena Lindmark Ma˚nsson,†,‡ Marie Paulsson,† and Lars Nilsson† †
Department of Food Technology, Engineering and Nutrition, Lund University, P.O. Box 124, SE-221 00 Lund, Sweden, and ‡Swedish Dairy Association, Ideon Science Park, SE-223 70 Lund, Sweden Received May 12, 2010. Revised Manuscript Received June 23, 2010
Casein (CN) micelles are naturally occurring colloidal protein aggregates present in a dispersed state in milk. In this paper we aim to obtain a detailed description of physicochemical properties of CN micelles over the entire size distribution using asymmetrical flow field-flow fractionation (AsFlFFF) connected to multiangle light scattering (MALS) and refractive index (RI) detection. Conclusions are drawn on the colloidal level regarding shape and conformation by comparison with models of colloidal particles. By using AsFlFFF-MALS-RI, it is concluded that the CN micelles are highly polydisperse with an average rms radius and hydrodynamic radius of 177 and 116 nm, respectively. The results show that the majority of CN micelles have a spherical shape, whereas a low concentration exists of larger and elongated aggregates. By comparison with models of aggregates of colloidal particles, the aggregates are shown to be anisotropic, e.g., aggregating linearly (threadlike) or in a sheet, rather than forming randomly spherical clusters. The results show that the characterization of colloidal dispersions with AsFlFFF-MALS-RI and the comparison with theoretical models are of a general character and, thus, of fundamental importance for colloidal dispersions.
1. Introduction Casein (CN) micelles are naturally occurring colloidal protein aggregates present in a dispersed state in milk. The detailed structure of CN micelles remains a subject of debate, and several different models have been proposed.1-4 CN micelles contain four different proteins: Rs1-, Rs2-, β-, and κ-CN. The latter resides mainly at the CN micelle surface providing colloidal stabilization with steric and electrostatic contributions.5 Experimental studies of CN micelle size have shown that the physical diameter can range from 50 to 600 nm, with an average of around 200 nm,3,6,7 showing that the population is highly polydisperse. The CN micelle size distribution and their structure have been shown to be of great importance for technological applications, such as aggregation with relevance for cheese and yoghurt production.1,5,8,9 The complex nature and polydispersity of CN micelles put high demands on the characterization methods. Asymmetrical flow field-flow fractionation (AsFlFFF or sometimes AF4) is a versatile chromatography-like separation technique well suited for the characterization of colloidal particles.10 The method is part of larger FFF family of techniques *Corresponding author: e-mail
[email protected]; tel þ46 46 222 96 53; fax þ46 46 222 46 22. (1) Horne, D. S. Int. Dairy J. 1998, 8, 171–177. (2) Walstra, P. Int. Dairy J. 1999, 9, 189–192. (3) Holt, C.; de Kruif, C. G.; Tuinier, R.; Timmins, P. A. Colloids Surf., A 2003, 213, 275–284. (4) Dalgleish, D. G.; Spagnuolo, P. A.; Goff, H. D. Int. Dairy J. 2004, 14, 1025– 1031. (5) Tuinier, R.; de Kruif, C. G. J. Chem. Phys. 2002, 117, 1290–1295. (6) Fox, P. F.; Brodkorb, A. Int. Dairy J. 2008, 18, 677–684. (7) de Kruif, C. G. J. Dairy Sci. 1998, 81, 3019–3028. (8) Glantz, M.; Devold, T. G.; Vegarud, G. E.; Lindmark Ma˚nsson, H.; Sta˚lhammar, H.; Paulsson, M. J. Dairy Sci. 2010, 93, 1444–1451. (9) Amenu, B.; Deeth, H. C. Aust. J. Dairy Technol. 2007, 62, 171–184. (10) Wahlund, K. G.; Giddings, J. C. Anal. Chem. 1987, 59, 1332–1339.
Langmuir 2010, 26(15), 12585–12591
which includes symmetrical flow FFF (FlFFF) and sedimentation FFF (SdFFF).11 When connected to suitable detectors, such as multiangle light scattering (MALS) and refractive index (RI), mass and root-mean-square radius (rrms) can be obtained throughout the size distribution. Furthermore, calculations from AsFlFFF retention times allow determination of hydrodynamic radii (rh) over the entire size distribution.12 Together these physical parameters make available additional structural/conformational information, such as apparent densities and the ratio between rrms and rh. CN micelles have previously been studied with FlFFF13 and SdFFF.7,14,15 However, the most common characterization methods are typically various types of electron microscopy4,16,17 or batch light scattering without preceding fractionation.8,18,19 In this paper, the purpose is to obtain a detailed description of physicochemical properties of CN micelles, such as mass, rrms and rh, over the entire size distribution. By determining these properties, important information regarding conformation and shape of CN micelles can be acquired. Furthermore, the aim is to draw (11) Giddings, J. C.; Yang, F. J. F.; Mayers, M. N. Anal. Chem. 1974, 46, 1917– 1924. (12) Nilsson, L.; Leeman, M.; Wahlund, K. G.; Bergensta˚hl, B. Biomacromolecules 2006, 7, 2671–2679. (13) Jussila, M. A.; Yohannes, G.; Riekkola, M. L. J. Microcolumn Sep. 1997, 9, 601–609. (14) Udabage, P.; McKinnon, I. R.; Augustin, M. A. J. Dairy Res. 2003, 70, 453–459. (15) Udabage, P.; Sharma, R.; Murphy, D.; McKinnon, I.; Beckett, R. J. Microcolumn Sep. 1997, 9, 557–563. (16) Waninge, R.; Kalda, E.; Paulsson, M.; Nylander, T.; Bergensta˚hl, B. Phys. Chem. Chem. Phys. 2004, 6, 1518–1523. (17) McMahon, D. J.; Oommen, B. S. J. Dairy Sci. 2008, 91, 1709–1721. € (18) Bauer, R.; Hansen, M.; Hansen, S.; Ogendal, L.; Lomholt, S.; Qvist, K.; Horne, D. J. Chem. Phys. 1995, 103, 2725–2737. (19) Devold, T. G.; Brovold, M. J.; Langsrud, T.; Vegarud, G. E. Int. Dairy J. 2000, 10, 313–323.
Published on Web 07/12/2010
DOI: 10.1021/la101892x
12585
Article
Glantz et al.
conclusions on the colloidal level by comparison with models of colloidal particles. Such conclusions are highly relevant for technological applications and of fundamental importance for colloidal dispersions. In order to fulfill the aims, powerful characterization techniques are needed, and hence, AsFlFFFMALS-RI is the method of choice. To our knowledge, this is the first time CN micelles are investigated with FFF with in-line MALS detection.
2. Experimental Section 2.1. Milk Samples. Individual morning milk samples were collected in September 2008 from 10 Swedish Holstein (SH) and 10 Swedish Red (SR) cows, all being healthy and fed the same diet. The samples were defatted by centrifugation at 2000g for 30 min and stored at -40 °C. The samples were thawed at 4 °C overnight and diluted to 3.3% (v/v) in the carrier liquid, described below, prior to FFF analysis. CN was precipitated by acid coagulation to identify CN micelles in the AsFlFFF elution profile. 50 mL of pure water prepared with a Milli-Q system (Millipore Corp., Bedford, MA) and 0.5 mL of 10% (w/v) acetic acid (Merck KGaA, Darmstadt, Germany) were added to 15 g of skim milk. After 10 min, 0.5 mL of 1 M sodium acetate (Fluka Chemie GmbH, Buchs, Switzerland) was added, and the mixture was filtered through a 5 μm filter (Munktell & Filtrak GmbH, B€arenstein, Germany). The serum was collected and diluted to 15% (v/v) in the carrier liquid, described below, prior to analysis. To exclude that results occur due to the carrier liquid used, commercial, nonhomogenized milk (3.5% protein) was defatted by centrifugation at 2000g for 30 min and diluted to 3.3% (v/v) in milk ultrafiltrate, described below, before analysis. 2.2. FFF Analysis Equipment. The AsFlFFF instrument was an Eclipse 3þ separation system (Wyatt Technology Europe, Dernbach, Germany). It was connected to a Dawn Heleos II multiangle light scattering (MALS) detector (Wyatt Technology) operating at a wavelength of 658 nm, an Optilab DSP differential refractive index (RI) detector (Wyatt Technology) operating at a wavelength of 632.8 nm, and a Jasco UV-970 ultraviolet light detector (Jasco Corporation, Tokyo, Japan) operating at wavelength of 280 nm. An Agilent 1200 series isocratic pump (Agilent Technologies, Waldbronn, Germany) with an in-line vacuum degasser and an Agilent 1200 series autosampler delivered the carrier flow and handled sample injection onto the AsFlFFF channel. Between the pump and the channel was placed a filter holder with a 100 nm pore size poly(vinylidene fluoride) membrane (Millipore Corp.) to ensure that particle-free carrier entered the channel. The AsFlFFF channel was a Wyatt short channel (Wyatt Technology) having a tip-to-tip length of 17.4 cm and a nominal thickness of 250 μm. The actual thickness was determined to be 197 μm by calibration against ferritin according to the procedure described in the literature.20 The ultrafiltration membrane forming the accumulation wall was made of regenerated cellulose with a cutoff of 10 kDa (Microdyn-Nadir GmbH, Wiesbaden, Germany).
2.3. FFF Separation Parameters and Data Processing. The sample injection onto the channel was commenced at a flow rate of 0.20 mL min-1 for 1.0 min. The sample volume injected onto the channel was 20-50 μL for an injected sample mass of approximately 20-50 μg. The injected amount was optimized in order to ensure no overloading in the channel by confirming that retention times were independent of the injected amount. A 0.5 min focusing/relaxation step was performed prior to elution with the focusing flow rate being identical to the initial cross-flow rate during elution (1.0 mL min-1). Injection was followed by 0.5 min of focusing/relaxation. In order to avoid excessive retention and long elution times, a cross-flow rate which decays linearly (20) Litzen, A. Anal. Chem. 1993, 65, 461–470.
12586 DOI: 10.1021/la101892x
Figure 1. Cross-flow (Vc) vs elution time (t) as used in the AsFlFFF experiments. with time was used as shown in Figure 1. After elution the channel was flushed without any cross-flow for 5 min before the next analysis. Detector flow rate (Vout) was constant at 2.0 mL min-1 throughout the separation. The carrier liquid was 50 mM imidazole (Merck KGaA) dissolved in pure water prepared with a Milli-Q system (Millipore Corp.) with addition of 11 mM CaCl2 (Merck KGaA) in order to resemble the serum content of Ca in milk and 0.02% (w/v) NaN3 (Sigma-Aldrich, St. Louis, MO) to prevent microbial growth in the carrier liquid. The pH was adjusted to 6.7. To exclude that results occur due to the imidazole buffer used, milk ultrafiltrate was used as carrier liquid. The milk ultrafiltrate was prepared from bulk milk (Ska˚nemejerier, Malmoe, Sweden) defatted by centrifugation at 1500g for 20 min. The milk was filtered using Pellicon XL (Millipore Corp.) with a cutoff of 10 kDa and subsequently dialyzed using Float-a-lyzer (Spectrum Laboratories Inc., Rancho Dominguez, CA) with a cutoff of 500 Da to remove lactose, all according to the manufacturer’s instructions. The milk ultrafiltrate was dried in a freeze-dryer (HETOSICC, Heto, Birker€ od, Denmark) and stored at -20 °C. The filtrate powder was dissolved in pure water prepared with a Milli-Q system (Millipore Corp.) to its volume before freezedrying prior to use. FFF separation was performed in the same manner as described above with the exception that Vout was reduced to 1.0 mL min-1, and in order to keep the initial Vc/Vout constant, the initial Vc was reduced to 0.5 mL min-1. Processing of light scattering data was made by the Astra software, version 5.3.4.14 (Wyatt Technology). The molar mass and the rrms were obtained by the Berry method21 fitting a straight line to data obtained at 25.9°-100.3° scattering angle. The lowest scattering angle, 14.4°, was not included, as the data obtained was imprecise. A dn/dc value of 0.166 mL g-1 was used,22 and the second virial coefficient was assumed to be negligible. The rh was obtained from the Stokes-Einstein equation23 r h, i ¼
kb T 6πηDi
ð1Þ
where kb is the Boltzmann constant, T is the temperature, η is the dynamic viscosity of the solvent, and Di is the diffusion coefficient. In turn, the diffusion coefficient was obtained from dzi ¼ kVc ðt, zÞDi dt
ð2Þ
where zi is the position of sample component i along the channel, t is the time, k is a constant including flow conditions and geometrical parameters, and Vc is the cross-flow rate (which is a function of both t and z). Equation 2 is a highly condensed version (21) Berry, G. C. J. Chem. Phys. 1966, 44, 4550–4564. (22) Alexander, M.; Rojas-Ochoa, L. F.; Leser, M.; Schurtenberger, P. J. Colloid Interface Sci. 2002, 253, 35–46. (23) Einstein, A. Ann. Phys. 1905, 17, 549–560.
Langmuir 2010, 26(15), 12585–12591
Glantz et al.
Article
of the full differential equation of which the derivation of a solution procedure was reported in an earlier paper.12 The apparent densities were obtained from the molar mass and rrms or rh distributions assuming homogeneous distribution of mass and a spherical shape. As the two radii give only an approximate description of the volume of possible shapes, the density obtained should be considered as an apparent property. The apparent density, Fi, for component i of the sample is calculated from Frms, i ¼
Mi Mi q or Fh, i ¼ Vðrrms Þi Vðrh Þi
ð3Þ
where M is the molar mass, V(r) is the volume of a sphere with radius r, and q is given by eq 4. pffiffiffiffiffiffiffiffi 3=2 ð 3=5rÞ3 Vsphere ðrrms Þ rrms 3 3 q ¼ ¼ 3 ¼ ¼ Vsphere ðrÞ 5 r r3
ð4Þ
The mass-weighted average apparent density was obtained from P mi F Fi ¼ P i mi
ð5Þ
where mi is the mass flow in each class i.
2.4. Dynamic Light Scattering. The z-average hydrodynamic diameter of native CN micelles was determined by dynamic light scattering (DLS) at a scattering angle of 90° at 25 °C on skim milks diluted in simulated milk ultrafiltrate24 according to Glantz et al.8 Each sample was analyzed in triplicate. 2.5. Modeling Hydrodynamic Radius. In order to aid in interpreting the experimental results, these were compared with simplified models of aggregates. The rh of rigid particles can only be calculated analytically for a few simple geometries such as spheres and ellipsoids.25 Kirkwood et al.26 and Kirkwood27 derived an expression for the diffusion coefficient or rh of particles consisting of identical subunits, and this approach has later been utilized in order to estimate the hydrodynamic properties of an arbitrary particle as approximated by a number of spheres; see ref 28 for an overview. In this study, rh of linear and polygonal aggregates have been calculated by means of Monte Carlo simulation of the Kirkwood formula applied to shell bead models as described by Hansen.29 Planar aggregates are described as cylindrical slices and can, thus, be determined using the Perrin formula.25 The results were also compared to simulations using the computer programmes HYDRO30 and HYDROSUB.31 The rrms can be calculated directly from the definition for a large number of geometries, e.g., ellipsoids and cylinders. For geometries where no analytical expression was available, the rrms was calculated by combination of subunits; see for example ref 31 and references therein.
Figure 2. (a) Mass (0) and RI signal (-) vs elution time and (b) root-mean-square radius (rrms; 2) and Rayleigh ratio (90° scattering angle; O) vs elution time for casein micelles in milk with a high concentration of large micelles.
3. Results CN micelles in bovine milk were investigated with AsFlFFFMALS-RI. Through this method we were able to obtain the distributions of molar mass and rrms of the samples. In addition, rh, apparent densities and the ratio rrms/rh of the samples were calculated.
Figure 3. (a) Mass (0) and RI signal (-) vs elution time and (b) root-mean-square radius (rrms; 2) and Rayleigh ratio (90° scattering angle; O) vs elution time for casein micelles in milk with a low concentration of large micelles.
(24) Jenness, R.; Koops, J. Neth. Milk Dairy J. 1962, 16, 153–164. (25) Perrin, F. J. Phys. Radium 1936, 7, 1–11. (26) Kirkwood, J. G.; Riseman, J. J. Chem. Phys. 1948, 16, 565–573. (27) Kirkwood, J. G. J. Polym. Sci. 1954, 12, 1–14. (28) de la Torre, J. G.; Bloomfield, V. A. Q. Rev. Biophys. 1981, 14, 81–139. (29) Hansen, S. J. Chem. Phys. 2004, 121, 9111–9115. (30) de la Torre, J. G.; Navarro, S.; Lopez Martinez, M. C.; Diaz, F. G.; Lopez Cascales, J. J. Biophys. J. 1994, 67, 530–531. (31) de la Torre, J. G.; Carrasco, B. Biopolymers 2002, 63, 163–167.
The results from the FFF analyses are shown in Figures 2 and 3. The graphs show three peaks that were registered with the RI signal (Figures 2a and 3a). The largest peak, with a maximum after 8 min elution, is the CN micelles. This was confirmed by precipitating CN by acid coagulation, after which the peak was undetectable (results not shown). Since AsFlFFF separates molecules based on differences in diffusion coefficient,
Langmuir 2010, 26(15), 12585–12591
DOI: 10.1021/la101892x
12587
Article
Glantz et al.
Table 1. Mass, Root-Mean-Square Radius (rrms), and Hydrodynamic Radius (rh) of Casein Micelles in Groups of Samples with High and Low Concentration of Large Micellesa group of samples high concentration of large micelles
low concentration of large micelles
mass (Da) (1.7 ( 0.8) 109 (4.4 ( 1.7) 108 rrms (nm) 217 ( 53 121 ( 34 132 ( 19 95 ( 17 rh (nm) a Results are presented as average value ( standard deviation (P < 0.01).
i.e., hydrodynamic size, the molecules eluted earlier than the CN micelles are expected to be smaller than the latter. Hence, the first peak, eluted close to the void (t0 = 0.15 min) during the first 2 min, is most likely the whey proteins present in milk. The second peak, eluted after 3 min, is not identifiable. However, it can be concluded that the peak does not represent proteinaceous matter as no UV absorption can be observed (results not shown). A possible explanation may be that it originates from small lipid aggregates that may be retained in the milk after centrifugation. No difference is found between the two studied breeds, SH and SR, regarding molar mass and rrms (P > 0.05, two-sample t test). However, regardless of cow breed, two distinctive groups of samples are found, one group of samples having a high concentration of larger micelles (Figure 2) and one group having a low concentration of larger micelles (Figure 3). Yet, for both groups of samples it can be seen that the CN micelles are polydisperse. The mass for the group of samples with a high concentration of larger micelles ranges between 5.9 107 and 1.2 1010 Da, whereas the rrms ranges between 40 and 360 nm (Figure 2). On the other hand, for the group of samples with a low concentration of larger micelles, the mass ranges between 6.5 107 and 7.5 109 Da and the rrms between 35 and 220 nm (Figure 3). The average values of mass (Mw), rrms (z-average), and rh (z-average) for the two groups of samples are shown in Table 1. Significant differences are found for both mass (P = 0.002) and radii (P = 0.001), with a higher average mass and larger average radii for the group of samples with a high concentration of larger micelles. The average values of mass, radii, and Frms obtained with AsFlFFF-MALS-RI have a high Pearson correlation (P = 0.000-0.007) with the z-average hydrodynamic diameter of CN micelles determined with DLS (173 ( 18 nm), which is expected. The average values of CN micelle radii obtained in this study are in the range found previously for bulk milk and milk of individual cows determined with DLS8,19 and bulk milk studied with cryotransmission electron microscopy16 as well as obtained from elution times from FlFFF.13 Similar results have also been obtained for milk powder studied with SdFFF7 and field-emission scanning electron microscopy.4 Figures 4a and 5a show the apparent density, calculated from rrms and rh, respectively, versus rrms for both groups of samples. In these plots it can be seen that Frms decreases with increasing rrms, whereas there is an increase in Fh as the micelles increase in size. The same trend as in the latter case is observed when calculating the ratio rrms/rh (Figures 4b and 5b), which is a qualitative measure on conformational shapes that are present in a solution. A rrms/rh ratio of 0.775 corresponds to hard, homogeneous spheres, whereas ratios above 1.7 represents anisotropic shapes.32 The differential mass distributions plotted in Figures 4b and 5b show that most of the CN micelles present in milk have a rather (32) Wittgren, B.; Borgstr€om, J.; Piculell, L.; Wahlund, K. G. Biopolymers 1998, 45, 85–96.
12588 DOI: 10.1021/la101892x
Figure 4. (a) Apparent density (F) calculated from the root-meansquare radius (rrms; 2) and the hydrodynamic radius (rh; 0) vs rrms and (b) rrms/rh (b) and differential mass distribution (-) vs rrms for casein micelles in milk with a high concentration of large micelles.
Figure 5. (a) Apparent density (F) calculated from the root-meansquare radius (rrms; 2) and the hydrodynamic radius (rh; 0) vs rrms and (b) rrms/rh (b) and differential mass distribution (-) vs rrms for casein micelles in milk with a low concentration of large micelles.
spherical shape as expected4,17 and that a smaller fraction appears more anisotropic. To exclude that results occur due to the imidazole buffer used, milk ultrafiltrate was prepared and used as carrier liquid in the AsFlFFF. The results from these analyses are shown in Figure 6. The same trend of rrms/rh is found for commercial, Langmuir 2010, 26(15), 12585–12591
Glantz et al.
Article
Figure 6. (a) Root-mean-square radius (rrms; 2) and Rayleigh ratio (90° scattering angle; O) vs elution time and (b) rrms/ rh vs rrms for casein micelles in commercial, nonhomogenized milk.
nonhomogenized milk diluted in milk ultrafiltrate as for the individual milk samples diluted in imidazole buffer. Hence, the buffer does not influence the results obtained, and the results represent a general behavior, not only in individual milk samples but also in bulk milk. Furthermore, the imidazole buffer with added CaCl2 constitutes an adequate carrier liquid that avoids inherent problems of the milk ultrafiltrate, i.e., high and noisy baselines. The utilization of the latter may also increase the risk of microbial growth in the instrumentation.
4. Discussion As the results show, AsFlFFF-MALS is an efficient way to separate and characterize CN micelles. This enables the acquisition of detailed information on size, mass, and conformational properties over the entire size distribution. To our knowledge, it is the first time that CN micelles have been characterized using AsFlFFF-MALS. An interesting observation from the results is that the scaling between radii and mass varies over the size distribution. This means that the relationship between mass and radii are not constant over the size distribution, which is reflected in the density plots (Figures 4a and 5a). Frms has a maximum for rather small micelles, approximately 300-400 kg/m3, and then decreases with increasing micellar rrms. This suggests either that Frms is lower for larger CN micelles or that the decrease in Frms corresponds to aggregated clusters as Frms for an aggregate will be lower than for the individual micelle. Fh, on the other hand, displays somewhat different behavior as a function of rrms. For low rrms, the Fh is rather constant, which shows that the relationship between mass and rh is constant for the smaller micelles. In turn, this can be interpreted as that rh depends mainly on the external solventsurrounded surface of the CN micelles. In Figure 5a, Fh decreases for an intermediate rrms range, approximately 80-120 nm, which is somewhat difficult to interpret as this is not observed in Figure 4a. A possible explanation, however, could be a difference in the surface composition and/or surface morphology of the CN micelles. Figure 5a shows a decrease in Fh, i.e. an increase in rh, which may be due to an uneven surface on the micelles and, thus, giving rise to a higher rh. Somewhat similar behavior has been observed for macromolecules, such as dissolved and mechanically degraded starch.12,33 κ-CN, which is located primarily at the CN (33) Rojas, C. C.; Wahlund, K. G.; Bergensta˚hl, B.; Nilsson, L. Biomacromolecules 2008, 9, 1684–1690.
Langmuir 2010, 26(15), 12585–12591
micelle surface, is known to protrude into the surrounding solution.1-4 It has been shown that κ-CN may not evenly cover all of the micellar surface, and this uneven distribution may have an influence on the results observed.4 For larger micelles, Fh increases, and several possible explanations exist. CN micelles contain calcium phosphate nanoclusters, which are distributed throughout the micelle.34,35 Thus, one explanation may be that large CN micelles contain a higher relative amount of calcium phosphate nanoclusters than small micelles. However, one would expect that this would also be reflected in an increase in Frms, which is not observed (Figures 4a and 5a). Another possible explanation is, as already discussed above, that these larger size objects are aggregates of CN micelles with a different scaling behavior. As briefly mentioned in the Results section, large conformational differences can be seen over the size distribution (Figures 4b and 5b). The major part of the distribution has a conformation which corresponds to spherical-like objects, as has been shown in the literature.4,16,17 As micellar size increases, rrms/rh also increases which shows that objects appear increasingly anisotropic (Figures 4b and 5b). For instance, a rod with axial ratio 20 has a ratio of rrms/rh of 1.73 and a rod with axial ratio 100 has a ratio of 2.66.32 Taking into account existing knowledge about structure of CN micelles,1-4 it is highly unlikely that the anisotropy observed would be single CN micelles. Rather, it could correspond to aggregated clusters of CN micelles, which is further supported by the density plots (Figures 4a and 5a) as discussed above. Thus, it appears that a small fraction of CN micelles in milk can aggregate to form elongated clusters, and this is confirmed for both individual milk and bulk milk (Figures 4b, 5b, and 6b), hence making this a general observation. We have also shown that this behavior is not dependent on the carrier liquid used in the AsFlFFF. Furthermore, it is striking that a similar aggregation behavior has been reported by Bauer et al.18 In this paper, the authors studied chymosin-induced aggregation of CN micelles with time-resolved batch, static, and dynamic light scattering and found during the aggregation that anisotropic clusters were formed (rrms/rh = 1.0-2.5). As the light scattering experiments were allowed to continue, the authors found that the aggregates ceased to increase in size, whereas the ratio rrms/rh decreased, (34) Holt, C.; S€orensen, E. S.; Clegg, R. A. FEBS J. 2009, 276, 2308–2323. (35) de Kruif, C. G.; Holt, C. In Advanced Dairy Chemistry - 1: Proteins; Fox, P. F.,McSweeney, P. L. H., Eds.; Kluwer Academic/Plenum Publishers: New York, 2003; Chapter 5.
DOI: 10.1021/la101892x
12589
Article
Figure 7. Schematic illustration of four aggregate structures used in the theoretical models: (a) linear (threadlike); (b) planar (cylindrical with height equal to the diameter of a single spherical micelle; (c) polygonal; and (d) spherical aggregates.
Figure 8. Ratio of root-mean-square radius (rrms) and hydrodynamic radius (rh) vs size for experimentally obtained data from groups of samples with a high (A) and low (B) concentration of larger micelles and theoretical models of linear (O), planar (0), polygonal (]), and spherical (-) aggregates. rrms has been normalized with rrms in the minima of rrms/rh.
indicating a structural rearrangement of the aggregated clusters. The formation of elongated clusters has also been observed by Green et al.36 with electron microscopy. However, to interpret the results obtained in this study simply as extended linear aggregates is most likely an oversimplification. This as other structures of clusters, which may give rise to a similar behavior, can be considered, and thus, it is interesting to compare with theoretical models for aggregated clusters of colloidal particles. The theoretical models for the ratio rrms/rh and Frms as a function of size are obtained by hydrodynamical calculations as described in the Experimental Section. Four general types of aggregate structures are investigated (linear, planar, polygonal, and spherical) and are shown in Figure 7. The results of the calculations are shown in Figures 8 and 9. The rh of the linear and polygonal models are calculated using Monte Carlo simulations, as described in the Experimental Section. The obtained rh is also compared with simulations using HYDROSUB and HYDRO, which show qualitatively similar results. In Figure 8, rrms has been normalized with rrms in the minima of rrms/rh (Figures 4b and 5b) assumed to correspond to individual spherical CN micelles. As can be seen, none of the models offer a complete explanation to the experimentally observed behavior. Spherical clusters cannot describe the experimental results, as expected (Figure 8). However, the linear and planar aggregates, which are increasingly (36) Green, M. L.; Hobbs, D. G.; Morant, S. V.; Hill, V. A. J. Dairy Res. 1978, 45, 413–422.
12590 DOI: 10.1021/la101892x
Glantz et al.
Figure 9. Apparent density (F) calculated from the root-meansquare radius (rrms) vs size for experimentally obtained data from groups of samples with a high (A) and low (B) concentration of larger micelles and theoretical models of linear (O), planar (0), polygonal (]), and spherical (-) aggregates. Frms has been normalized with the value in the maxima of Frms, and rrms have been normalized with rrms in the minima of the ratio between rrms and hydrodynamic radius.
anisotropic, offer a considerably better description, which depends somewhat on the individual sample and on the different intervals over the distribution. For instance, the group of samples with a low concentration of larger micelles (Figure 8B) show better agreement with a planar model at smaller clusters, whereas larger clusters appear to be more threadlike (linear). The properties of the group of samples with a high concentration of larger micelles (Figure 8A), on the other hand, are better described by a threadlike cluster structure. Also included in Figure 8 is a model of polygonal aggregates. Inclusion of such aggregates is made based on the common view of the structure of cheese curd, i.e., a threadlike network in which the aqueous phase is retained.37 The polygonal aggregates display an increase in rrms/rh as does the linear and planar aggregates, however, to a lower extent. In Figure 9, Frms is normalized with the value in the maxima of Frms (Figures 4a and 5a) assumed to correspond to individual spherical CN micelles. As discussed above, the decrease in Frms could be due to formation of elongated clusters. In Figure 9, the linear, planar, and polygonal aggregates show a similar behavior, i.e., decrease in Frms, which is in contrast to the spherical aggregates. Once again, the agreement depends somewhat on the individual sample and on the different intervals over the distribution. An interesting observation is that the group of samples with a low concentration of larger micelles (Figure 9B) shows a better agreement with polygonal aggregates, whereas the group of samples with a high concentration of larger micelles (Figure 9A) is closer to planar aggregates. Thus, the model showing best agreement with experimentally obtained rrms/rh does not necessarily offer the best fit for experimentally obtained Frms. This shows that the CN micelles display interesting characteristics of the models but are not identical to either. Naturally, the models are simplifications, and the experimental systems can be expected to be more complex. Other factors may influence the cluster structure, such as polydispersity and combinations of cluster structures. However, the combined results show the strength of combining theoretical models and data obtained experimentally from AsFlFFF-MALS and that detailed information regarding size, structure, and conformation can be obtained in colloidal systems. (37) Walstra, P.; Geurts, T. J.; Noomen, A.; Jellema, A.; van Boekel, M. A. J. S. Dairy Technology: Principles of Milk Properties and Processes; Marcel Dekker: New York, 1999; Chapter 22.
Langmuir 2010, 26(15), 12585–12591
Glantz et al.
Article
5. Conclusions By using AsFlFFF-MALS-RI, it is concluded that the CN micelles are highly polydisperse with an average rrms and rh of 177 and 116 nm, respectively, based on 20 individual milk samples. By combining the results obtained from MALS detection with calculation of rh from elution times, we obtain information on the shape of the CN micelles. These results show that the majority of CN micelles present in milk have a spherical shape, whereas a low concentration exists of larger and elongated objects. From rrms/rh and Frms, it is concluded that this population most likely consists of aggregates, rather than individual CN micelles. By comparison with models of aggregates of colloidal particles, we obtain information on possible morphologies of aggregates. In order to explain the experimental results, the aggregates must be anisotropic and elongated, e.g., having a threadlike, planar, or polygonal shape, rather than being randomly aggregated in a spherelike cluster. The occurrence of aggregates is somewhat surprising, and the underlying reason for their presence cannot be concluded at this stage. It should, however, be remembered that the aggregates represent a rather small fraction of the entire
Langmuir 2010, 26(15), 12585–12591
distribution. The morphology of CN micelle aggregates can be expected to have a large impact on technological properties during renneting and acidification of milk. Furthermore, the characterization of colloidal dispersions with AsFlFFF-MALS-RI and the comparison with theoretical models are of a general character and, thus, of fundamental importance for colloidal dispersions. Acknowledgment. Catja Freiburghaus at the Department of Food Technology, Engineering and Nutrition, Lund University, Sweden, is greatly acknowledged for her help with the preparation of milk ultrafiltrate. The Swedish Farmer’s Foundation for Agricultural Research (SLF), Stockholm, Sweden, and the Swedish Research Council (VR), Stockholm, Sweden, are acknowledged for financial support. Funding for instrumentation is gratefully acknowledged from The Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS), Stockholm, Sweden, The Royal Physiographic Society of Lund, Sweden, and The Crafoord Foundation, Lund, Sweden.
DOI: 10.1021/la101892x
12591
