Multiple Steps during the Formation of β-Lactoglobulin Fibrils

Luben N. Arnaudov,†,‡ Renko de Vries,*,†,‡ Hans Ippel,§ and Carlo P. M. van Mierlo§. Laboratory of Physical Chemistry and Colloid Science, W...
0 downloads 0 Views 286KB Size
Biomacromolecules 2003, 4, 1614-1622

1614

Multiple Steps during the Formation of β-Lactoglobulin Fibrils Luben N. Arnaudov,†,‡ Renko de Vries,*,†,‡ Hans Ippel,§ and Carlo P. M. van Mierlo§ Laboratory of Physical Chemistry and Colloid Science, Wageningen University, Dreijenplein 6, 6700 EK Wageningen, The Netherlands, Food Physics Group, Wageningen University, Bomenweg 2, 6703 HD Wageningen, The Netherlands, and Laboratory of Biochemistry, Wageningen University, Dreijenlaan 3, 6703 HA Wageningen, The Netherlands Received March 27, 2003; Revised Manuscript Received August 8, 2003

In this study, the heat induced fibrilar aggregation of the whey protein β-lactoglobulin is investigated at low pH and at low ionic strength. Under these circumstances, tapping mode atomic force microscopy results indicate that the fibrils formed have a periodic structure with a period of about 25 nm and a thickness of one or two protein monomers. Fibril formation is followed in situ using light scattering and proton NMR techniques. The dynamic light scattering results show that the fibrils that form after short heating periods (up to a few hours) disintegrate upon slow cooling, whereas fibrils that form during long heating periods do not disintegrate upon subsequent slow cooling. The NMR results show that even after prolonged heating an appreciable fraction of the protein molecules is incorporated into fibrils only when the β-lactoglobulin concentration is above approximately 2.5 wt %. The data imply multiple steps during the heat induced formation of β-lactoglobulin fibrils at low pH and at low ionic strength: (partly) denatured protein monomers are either incorporated into fibrils or form instead a low molecular weight complex that is incapable of forming fibrils. Fibril formation itself also involves (at least) two steps: the reversible formation of linear aggregates, followed by a slow process of “consolidation” after which the fibrils no longer disintegrate upon slow cooling. Introduction Many food proteins are known to form gels when subjected to heating.1-13 Knowledge about the relation between the conditions under which the gels form and the accompanying gel characteristics is of major importance for optimizing industrial food processing. Mainly two types of gels are formed from heated protein solutions. Opaque, fractal-like gels form at pH values close to the isoelectric point and/or at high ionic strength.7-9,16-20 Instead, finestranded, transparent gels form at pH values far from the isoelectric point and/or at low ionic strength.5,6,8-10,14-20 It seems that under the latter conditions the strong, nonspecific electrostatic repulsion between the (partly denatured) proteins, causes the relatively open network structure of these transparent gels.14-20 The formation of both types of gels has been extensively investigated for β-lactoglobulin. Bovine β-lactoglobulin (β-lg) is a globular whey protein with a molecular mass of 18 400 Da and a radius of about 2 nm.21 The protein is of significant importance for the food industry because of its high occurrence in whey (it comprises more than 50% of the total whey protein concentration). Especially, the opaque, fractal-like gels of β-lg that form at pH values around the isoelectric point and/or at high ionic strength have been extensively studied.12,13,17-19 * To whom correspondence should be addressed. Fax: +31 317 483777. E-mail: [email protected]. † Laboratory of Physical Chemistry and Colloid Science. ‡ Food Physics Group. § Laboratory of Biochemistry.

Recently, studies of the fine-stranded, transparent gels that form at low pH and at low ionic strength after prolonged heating of β-lg solutions at relatively low protein concentration were reported.4-6,8,10,17 Aymard et al.11 investigated the aggregates that form in dilute β-lg solutions (of the order of 1 wt %) at pH ) 2.0 after prolonged heating at 80 °C. Using static (SLS) and dynamic (DLS) light scattering, small angle neutron scattering (SANS), and small-angle X-ray scattering (SAXS), they conclude that long linear aggregates form, with persistence lengths ranging from 38 nm at 0.1 M ionic strength up to 600 nm at 0.013 M ionic strength. The presence of these long linear aggregates is confirmed by electron microscopy. The kinetics of β-lg aggregate formation is followed using a selective precipitation technique.11 At higher β-lg concentrations (of the order of 10 wt %), the fibrils induced at pH ) 2.0 and at low ionic strength form transparent gels. The gelation kinetics and the rheology of these gels has been studied by Kavanagh et al.6 and by Veerman et al.10 Changes in the protein secondary structure upon β-lg fibril formation have been studied by Kavanagh et al.5 and by Lefevre and Subirade8 using both wide-angle X-ray diffraction (WAXD) and Fourier transform infrared spectroscopy (FTIR). An increased amount of intermolecular hydrogen bonds and intermolecular beta-sheets are observed upon fibril formation. Finally, atomic force microscopy (AFM) was used by Ikeda and Morris20 to visualize fine-stranded (at pH ) 2.0) and particulate (at pH ) 7.0) β-lg and whey protein isolate (WPI) aggregates. AFM was also used in a recent study by Gosal et al.22 to investigate fibrilar networks derived from

10.1021/bm034096b CCC: $25.00 © 2003 American Chemical Society Published on Web 09/16/2003

Formation of β-Lactoglobulin Fibrils

β-lg in water solutions or in trifluoroethanol (TFE)-water solutions. A “beaded” appearance of the latter fibrils is observed. Aggregates formed by β-lg and whey protein isolates (WPI) at pH ) 2 were studied by Ikeda23 by using AFM and Raman scattering spectroscopy. The fine stranded aggregates of β-lg appeared to be strings of monomers, whereas WPI aggregates were found to consist of granular aggregates. For the opaque, fractal-like, gels that form close to the isoelectric point and/or at high ionic strength, the dependence of the structure of the protein aggregates (and eventually, protein gels) on the heating time has been modeled as a diffusion-limited aggregation27 process, using both analytical theory and computer simulations.24-26 In the case of β-lg at a pH close to its isoelectric point, the dependence of the fraction of aggregated protein molecules on the heating time is modeled in detail by Roefs and De Kruif24 under the assumption that the protein molecules aggregate in a manner similar to an ordinary radical addition polymerization reaction. The model involves an initiation, propagation, and a termination step and shows that the decrease of the concentration of the native β-lg molecules follows the experimentally observed reaction order of 3/2. In the work of Clark et al.,28 a mean field approach is used to model the kinetics of heat-set fine-stranded globular protein gels, in general, and for β-lg at acidic pH, in particular. The model recognizes three main stages in the process of heat-set globular protein gelation: an initial unfolding step or dimerization, a step of linear fibrilar aggregation via nucleation and growth, and a step of random association of the fibrils. Nevertheless, a model that describes the dependence of the fraction of aggregated protein molecules on the heating time at pH values far from the isoelectric point and/or at low ionic strength has not been reported yet. The aggregation kinetics of β-lg at pH ) 2.0 and at low ionic strength, at 80 °C, is studied by both Veerman et al.10 and Aymard et al.11 using a selective precipitation technique. Apparently, under these circumstances, not all β-lg monomers are converted into aggregates, even after prolonged heating. The higher the initial β-lg concentration is, the higher the final relative fractional conversion factor to aggregates becomes. This fits with the nucleation and growth model of Oosawa29 for linear protein polymerization: only a protein amount exceeding a “critical concentration” is incorporated into fibrils, the rest remains monomeric. However, the Oosawa model pertains to equilibrium polymerization circumstances, under which continuous depolymerization cannot be neglected. Although it has been shown that β-lg aggregation at pH ) 2.0 and 80 °C is irreversible with respect to rapid temperature quenches, its reversibility with respect to slow temperature changes, dilution at 80 °C, etceteras, has not been investigated. In this paper, the fibrilar aggregation of β-lg at pH ) 2.0 and at low ionic strength, is studied at 80 °C. AFM is used to probe the resulting β-lg fibrilar structures in great detail. Insight is obtained in the various steps that lead to β-lg fibrils. The question whether β-lg aggregation is reversible under slow temperature changes is addressed by following the aggregation process in situ by using DLS. Previous reports

Biomacromolecules, Vol. 4, No. 6, 2003 1615

about the kinetics of β-lg aggregation used a selective precipitation method10,11 which is rather indirect and cannot distinguish between denatured protein molecules that are monomeric and those ones incorporated into fibrils. We instead use proton NMR spectroscopy, to follow the β-lg aggregation process in situ. Aggregates larger than ∼200 kDa do not contribute to the observed NMR intensities as they have large rotational correlation times and thus very short NMR relaxation times.30,31 Consequently, by using proton NMR spectroscopy, we are able to measure the concentration of low molecular weight β-lg material in solution that is not incorporated into fibrils. The interpretation of the experimental results obtained sheds light on the processes occurring during β-lg aggregation and fibril formation. Materials and Methods Protein Solutions. β-lg from bovine milk was obtained from Sigma (ref. L0130, Lot 20K7043) and is a mixture of genetic variants A and B and is used throughout all experiments. For comparison, we also used β-lg from bovine milk obtained from ICN Biomedicals (Cat. No 100363, Lot 4941C) with a similar mixture of genetic variants A and B. Both β-lg solutions gave rise to indistinguishable results in various AFM and DLS experiments. β-lg solutions were obtained by dissolving the protein in deionized water and by directly adjusting the pH through the addition of small amounts of 1 M HCl to the solution. The β-lg solutions were subsequently extensively dialyzed against a 10 mM HCl solution. All solutions contained 200 ppm NaN3 to prevent bacterial growth. After dialysis, the β-lg solutions were centrifuged for 3 h at 45 000 g using a Beckman Avanti J-25I High performance centrifuge and subsequently filtered through 0.45 µm low-protein adsorbing filters (Sterile Acrodisk, Gelman Sciences). Protein concentrations in the final solutions were determined by UV spectrophotometry at 278 nm using an extinction coefficient of 0.83 L cm-1 g-1. Tapping Mode Atomic Force Microscopy (AFM). AFM observations were performed using a Digital Instruments NanoScope III Multimode Scanning Probe Microscope. Protein solutions of different concentrations were put into tightly closed glass tubes and introduced in a water bath preheated at 80 ( 0.1 °C. At certain time intervals ranging from 1 to 24 h, the tubes were taken out of the bath, aliquots were taken and the tubes were returned into the bath. The aliquots were quenched in ice-cold water and diluted on basis of the initial β-lg monomer concentration to a final protein concentration of 0.1 wt %. The protein dilution was done to facilitate quantitative comparison of the results obtained from the experiments performed at different initial protein concentrations. Protein monomers and/or protein aggregates were adsorbed on silicon substrates. Before use, the silicon plates were cleaned in pure alcohol by ultrasound. Subsequently, they were dried and plasma cleaned. The cleaned silicon plates were dipped into the diluted β-lg samples for a period of 1 h and subsequently were quickly dried using pure and dry nitrogen. The tapping mode AFM observations were performed in air using these dried plates covered with β-lg.

1616

Biomacromolecules, Vol. 4, No. 6, 2003

Arnaudov et al.

Figure 1. Tapping mode AFM height images of linear aggregates of β-lg obtained at pH ) 2.0 and ionic strength I ) 0.013 M, by starting with three different protein concentrations [i.e., 1 wt % (top), 2 wt % (middle), and 3 wt % (bottom)] and using different lengths of heating time at 80 °C [i.e., 1 h (left), 5 h (middle), and 24 h (right)].

Light Scattering (LS). Light scattering experiments were carried out using an ALV-5000 multiple tau digital correlator equipped with an argon laser that operates at a wavelength of 514.5 nm. The scattering angle was 90° for all measurements. Prior to the experiments, sample solutions were filtered into the test tubes, by using 0.45 µm low-protein adsorbing filters. In all cases, dynamic light scattering (DLS) experiments were first carried out at ambient temperature to ensure that only monomers were present in these initial solutions. The aggregation process was then followed in situ, using two different thermal trajectories. The first trajectory involves gradual heating (∼2 °C /min) of the sample from 25 to 80 °C, keeping the sample at 80 °C for a period of time ranging from 1 to 20 h, and then gradually (∼2 °C /min) cooling the sample down to 25 °C. The second trajectory consisted of directly inserting the sample in the preheated sample holder of the LS setup which was at 80 °C, keeping it there for a period of time ranging from 1 to

48 h, while collecting scattering data, and subsequently quenching the sample in ice-cold water. Nuclear Magnetic Resonance (NMR). NMR experiments were carried out using a Bruker AMX 500 spectrometer operating at a 1H frequency of 500 MHz. Solutions of β-lactoglobulin (500 µL of 8.2, 4.1, 3.1, 2, and 1 wt % β-lg, respectively) were prepared from stock solutions including 10% (v/v) D2O, adjusted to pH 1.92 with HCl, and were transferred to Wilmad (Buena, NJ) PP-528 5 mm NMR tubes. The NMR tubes containing protein solution were introduced in the spectrometer either at 70 °C (at which no aggregation occurs), after which the temperature of the NMR probe interior was rapidly increased from 70 to 80 °C, or alternatively by quickly introducing the NMR tube in a preheated probe at the desired temperature of 80 °C. In both cases, data acquisition started after the sample temperature equilibrated for approximately 3-5 min. To follow aggregation of β-lactoglobulin in real time, many one-dimensional

Formation of β-Lactoglobulin Fibrils

Biomacromolecules, Vol. 4, No. 6, 2003 1617

proton NMR spectra were sequentially acquired over an extended period of time ranging from 19 to 64 h in total and stored into a 2D serial file. Depending on the protein concentration of the sample, 96-256 scans were collected per sequential proton NMR spectrum. The residual water resonance in the middle of the spectrum was suppressed by presaturation of the solvent signal during the relaxation delay period of 2 s. After data acquisition, the sequential FIDs were multiplied by a cosine bell window function and Fourier transformed. The transformed spectra were baseline corrected by a third-order polynomial function. All spectral processing was done with the Bruker (Rheinstetten, Germany) Xwinnmr Linux 3.1 software package. The same software was also used to integrate the remaining soluble proton NMR signals in the NMR spectra of β-lg that were acquired as a function of incubation time. Signals in different spectral regions were integrated, including the total spectral proton resonance range (-1 to +11 ppm), the region containing the combined amide and aromatic protons (5.5 to 11 ppm), and the region containing the nonexchangeable protons (-1 to +4 ppm). Results and Discussion

Figure 2. AFM height image of split fibrils observed for 2 wt % β-lg at pH ) 2.0 and ionic strength I ) 0.013 M, after 4 h of heating at 80° C (white arrows show the places where the fibrils are split).

Tapping Mode Atomic Force Microscopy. The main objective to use AFM measurements in the study of protein fibrils is to obtain information about the structural details of these fibrils. However, AFM can also be used to obtain qualitative information about the dependence of protein aggregation on the initial protein monomer concentration and on the length of the applied heating time. Figure 1 shows the tapping mode AFM height images of linear aggregates of β-lg obtained by starting with three different protein concentrations (i.e., 1, 2, and 3 wt %) and using three different lengths of heating times at 80 °C (i.e., 1, 5, and 24 h). It can be seen that the rate of protein aggregation is β-lg concentration dependent. A higher initial β-lg monomer concentration leads to increased formation of long aggregates in 24 h. The simultaneous presence of aggregates with a short length (∼100 nm) and a long length (>1 µm) is observed. It is not clear whether the shorter length aggregates are eventually converted into fibrils or whether they exist as a separate aggregation state. We should also mention that since the aliquots for the AFM samples were quenched, the pictures in Figure 1 represent a “frozen” state of the process of aggregation. Next the structural details that can be resolved by AFM measurements for individual fibrils are considered. The first observation concerns fibril splitting. Figure 2 shows the AFM height image of what is apparently a split fibril. This particular case was observed for 2 wt % β-lg after 4 h of heating at 80 °C. Fibril splitting is observed only occasionally. A further structural detail observed is a periodic height fluctuation along the contour of the fibrils. This is observed for all fibrils. A typical example is shown in Figure 3A. The figure shows a 3D height AFM image of a linear aggregate obtained from 0.5 wt % β-lg after 17 h of heating at 80 °C. The spectral period as obtained from a Fourier transform of the height profile along the length of this fibril (Figure 3 B)

is about 26 nm. The maximum height of the observed fibril is around 4 nm, which is close to the size of the native monomer. However, the sample may be compressed due to interaction with the AFM tip. Such an effect can lead to a 60% underestimation of the height of the observed object as is shown by Yang et al.32 If the latter is the case, then the height of the fibril can be comparable to the size of a β-lg dimer. The combination of the fibril height, the fibril height periodicity, the magnitude of the height variations (maximally about 2 nm), plus the occasionally observed fibril splitting strongly suggests that the β-lg fibrils have a helical structure. The addition of monomers to protofibrils, which subsequently twist around each other to form fibrils, may be the first possible mechanism of formation of the observed β-lg fibrils. The fibrils may also form via a second mechanism in which dimers are added to one another. The lengthening product could form a coiled, helical-like ribbon, with the periodicity observed by us. The first mechanism allows the observed fibril splitting to occur (Figure 2). The fact that no protofibrils are observed and the rare observation of fibril splitting may be, in the first case, explained by the much higher preference of the fibril formation, compared to the existence of separate protofilaments. The second mechanism explains why no protofilaments are observed, but cannot explain the occasional observation of fibril splitting. In short, our AFM observations are not yet sufficient to distinguish between different possible mechanisms of fibril formation. In the following, the AFM results obtained by us are compared with other AFM studies on linear protein aggregation. Aggregates formed by β-lg at pH ) 2.0 after heating at 80 °C were also observed by AFM by Ikeda and Morris20 but at an ionic strength of 0.1 M. Under this circumstance, the fibrils are much shorter and more flexible than the fibrils detected by us. The height of the aggregates observed in the work of Ikeda and Morris, 3.8 nm, is similar to what we

1618

Biomacromolecules, Vol. 4, No. 6, 2003

Arnaudov et al.

Figure 3. A. 3D projection of a height AFM image of a linear aggregate obtained from 0.5 wt % β-lg at pH ) 2.0 and ionic strength I ) 0.013 M, after 17 h of heating at 80° C. B. Height profile along the fibril shown in part A.

observe. By using AFM, Gosal et al.22 observed fibrils formed from β-lg solutions at pH ) 2.0 after heating at 80 °C, that have lengths between 0.1 and 2 µm and a height of 3.6 ( 0.5 nm in agreement with our observations. However, Gosal et al. observed fine filaments and globular species and could not observe any periodic structure in their β-lg linear aggregates obtained in water solution. The difference with our observations may be caused by Gosal et al. not dialysing the protein samples, in contrast to our experiments. In the case of β-lg in TFE-water mixtures, Gosal et al. observed a beaded appearance of the β-lg linear aggregates, with height fluctuations similar to the ones detected by us. By using AFM, Ikeda23 also observed fibrilar aggregates formed from β-lg solutions at pH ) 2 in the presence of 0.1 M NaCl after heating at 80 °C. The height of the fibrils was 4 nm, and Ikeda concluded that they consisted of strings of partially unfolded β-lg monomers. The similarities between β-lg fibril formation and amyloid protein aggregation have been pointed out by a number of authors.22,33 Amyloid fibrils obtained from TTR10-19, SH3 domain, and human wild-type lysozyme were studied by AFM by Chamberlain et al.35 The fine structures observed for the SH3 domain and lysozyme fibrils are indeed very similar to the fine structure of the β-lg fibrils reported here (Figure 3). In the case of SH3 domain and lysozyme fibrils, helical structures are formed from two to four filaments which are twisted around each other.35 A helical structure of fibrils was also observed in the AFM study of the linear aggregation of SMA, a variable domain of Ig light-chain known to form amyloids.36 Light Scattering. LS is extremely sensitive to the presence of even the smallest amount of protein aggregates. As soon as particles larger than monomers appear, their scattering prevails and the apparent hydrodynamic radius increases to values many times the hydrodynamic radius of the monomers. The AFM results discussed above show that the β-lg fibrils obtained by us are extremely thin and long and extremely stiff. Therefore, even if only a small fraction of the initially present β-lg molecules is converted into fibrils, the dynamic light scattering will be essentially that of a solution of entangled, semi-flexible polyelectrolytes. A quantitative interpretation of the filament dynamics in such a system is very difficult. We use DLS as a qualitative technique for the detection of protein aggregates. This enables the in situ investigation of the reversibility of aggregate

Figure 4. Scattered intensity of 8.2 wt % (circles), 4.1 wt % (squares), and 2.6 wt % (triangles) β-lg solutions at pH ) 2.0 and ionic strength I ) 0.013 M, as a function of heating time theat at 80 °C, as determined by LS. The standard deviations are included as vertical bars at the individual data points. The lines are drawn by means of linear regression and are given to facilitate comparison.

formation under different thermal circumstances. The intensity of the scattered light can also be used in the investigation of the process of protein aggregation because the change in the scattering intensity is a measure for the concentration of the aggregated species. The evolution of the DLS autocorrelation function of β-lg as a function of the heating time at 80 °C follows a distinct pattern. At ambient temperature, we observe a single DLS decay curve consistent with the presence of only monomers in the solution. Upon increasing the temperature to 80 °C, a “shoulder” develops on the DLS decay curve of β-lg that suggests a bimodal distribution of monomeric species and aggregates. Upon increasing the heating time at 80 °C, the DLS decay curve becomes monomodal because of the predominance of the aggregate scattering. Such single exponential behavior is generally observed once fibrils longer than several tens of nanometers are formed. A similar pattern for the evolution of the DLS autocorrelation function of β-lg at pH ) 2.0 as a function of the heating time at 80 °C is reported by Aymard et al.11 In Figure 4, the scattered intensity of 8.2 wt % (circles), 4.1 wt % (squares), and 2.6 wt % (triangles) β-lg solutions is plotted as a function of heating time at 80 °C. As soon as the temperature is equilibrated at 80 °C (with a typical duration of 5 min), the light scattering measurement is

Biomacromolecules, Vol. 4, No. 6, 2003 1619

Formation of β-Lactoglobulin Fibrils

Table 1. Summary of the Data on Reversibility of Aggregation as Studied by DLSa

Cβ-lg, [wt %]

tup, [min]

theat, [min]

tdown, [min]

reversibility

0.6 1.0 1.0 5.0

60.0 90.0 90.0 60.0

60.0 180.0 540.0 60.0

60.0 90.0 90.0 60.0

yes yes no no

a The three columns t , t up heat, and tdown define the thermal trajectory followed for the particular experiment. tup is the time to reach 80 °C starting from 25 °C, theat is the time during which the temperature was kept constant at 80 °C, and tdown is the time for the sample temperature to get down to 25 °C.

Figure 5. Mean hydrodynamic radius Rh (left ordinate) and sample temperature T (right ordinate) for 0.6 wt % β-lg at pH ) 2.0 and ionic strength I ) 0.013 M as a function of heating time at 80 °C during the in situ DLS experiment for a thermal trajectory consisting of a period of slow heating, followed by a period of constant temperature and subsequently a period of slow cooling. The standard deviations are included as vertical bars at the individual data points.

started. At that time, the apparent hydrodynamic radius of β-lg is already much larger than the one of the monomeric protein and continues to increase with heating time (data not shown). As can be seen in Figure 4, the intensity of the scattered light increases linearly with heating time and the slope of the lines is concentration dependent, increasing drastically at higher concentrations. The nucleation or lag period for β-lg fibril formation is thus short which is in contrast to what is usually observed for amyloid protein aggregation. Although the structures of the β-lg fibrils are remarkably similar to those observed for amyloid protein aggregates, some aspects of the kinetics of their formation are different. The influence on β-lg aggregate formation of slowly increasing the temperature and subsequent cooling of the β-lg sample is investigated (Figure 5). First, the temperature is slowly increased from room temperature to 80 °C and kept there for approximately 1 h. Subsequently, the temperature is lowered back to room temperature. At the start of the experiment at ambient temperature, only monomer particles are detected with a hydrodynamic radius of about 2 nm. As soon as the temperature rises the apparent hydrodynamic radius of β-lg increases significantly. As the temperature reaches a constant value of 80 °C, the size of the aggregates continues to grow. Upon gradually cooling the solution back to 25 °C, the mean hydrodynamic radius decreases, and at 25 °C, only monomers are detected. Values for the mean hydrodynamic radius are given only for the equilibrated temperature since during the heating or the cooling of the sample nonequilibrium effects such as convection in the sample can produce artificial results. We should also stress that the magnitude of the hydrodynamic radius given in Figure 5 is apparent but is useful to show the change in the decay rate with the change of the heating temperature and the heating time. A summary of observations for different β-lg concentrations and for different heating times is given in Table 1. For longer heating times the decrease in size upon slow cooling is still observed, but the effective hydrodynamic radius fails to drop back to the monomer size and rather stays large (apparent

hydrodynamic radius of several hundred nanometers depending on the heating time and initial monomer concentration). The effect of the protein concentration is similar, and at 5 wt % β-lg, we can no longer observe a decrease of the hydrodynamic radius of the protein aggregates upon cooling of the solution. We should also mention that whenever we observed reversibility, i.e., the hydrodynamic radius of the β-lg went back to the monomeric size, the intensity of the scattered light after the T ramp was equal to the initial intensity of the scattered light obtained from the nonheated protein solution. This is evidence that the decrease in the hydrodynamic radius is truly due to a disassembly of the β-lg aggregates upon cooling rather than to a partial precipitation of large aggregates. The DLS observations show reversibility of the earlier stages of the process of β-lg fibril formation. The reversibility decreases after extended periods of β-lg heating at 80 °C and can be caused by a slow process of “consolidation” of the aggregated monomers constituting the fibrils during which rearrangements of the remaining secondary and/or tertiary structure of the proteins in the fibrils may occur. This could be similar to the “dock-lock” mechanism that has been suggested for amyloid elongation.37 The mentioned decrease in the reversibility of β-lg fibril formation during extended periods of heating also implies that presumably no depolymerization of β-lg fibrils occurs; consequently, the observation that not all β-lg protein molecules are incorporated into the fibrils cannot be explained by invoking depolymerization, as in the Oosawa nucleation and growth model. Proton Nuclear Magnetic Resonance. We use proton NMR spectroscopy as a direct, noninvasive and in situ method to measure the amount of low molecular weight species in aggregating β-lg solutions. The intensities of the proton resonances in a protein NMR spectrum strongly depend on the overall rotational diffusion coefficient of the molecules involved.30,31 As soon as free monomers become part of an aggregate larger than ∼200 kDa, their contribution to the measured proton resonance intensity disappears in the baseline of the NMR spectrum. Therefore, the observed decrease of the total proton NMR signal intensity is proportional to the amount of protein incorporated into high molecular weight aggregates. To determine at which temperature the proton NMR spectrum of native β-lg starts to change, proton NMR experiments were carried out at 25, 35, 56, and at 80 °C at protein concentrations of 1 and 2 wt %, respectively. The spectra at 25, 35, and 56 °C are typical for native β-lg at pH

1620

Biomacromolecules, Vol. 4, No. 6, 2003

Arnaudov et al.

Figure 7. Change in the integral of the whole 500 MHz proton NMR spectrum of 8.2 wt % (circles), 4.1 wt % (squares), and 3.1 wt % (triangles) β-lg solutions at pH ) 2.0 and ionic strength I ) 0.013 M as a function of heating time at 80 °C.

Figure 6. 500 MHz proton NMR spectra of 8.2 wt % β-lg, pH ) 2.0, ionic strength I ) 0.013 M: (A) Spectrum of native β-lg at 35.2 °C. (B-D) spectra of β-lg at 80 °C for three different heating times, as indicated in the figure.

2 (similar to spectrum A) in Figure 6). The NMR spectrum acquired at 56 °C did not change during a 24 h period of time. As soon as the temperature is increased above 60 °C, the proton NMR spectrum becomes typical for that of a largely unfolded protein similar to the spectra seen in Figure 6B). Some traces of native β-lg can still be found in the initial proton NMR spectrum, but they disappear in about 3 h at 80 °C. Proton NMR signals are recorded at 80 °C of solutions containing increasing initial concentrations of β-lg. Virtually no change in the overall proton NMR signal intensity of the spectra is observed for 1 and 2 wt % β-lg solutions during a period of heating to 24 h. However, heating a solution of 8.2 wt % β-lg leads to a significant decrease in the overall proton NMR signal intensity in time (Figure 6B-D). A similar observation is made for β-lg solutions containing 4.1 and 3.1 wt % protein, although the overall intensity changes to different extents. The samples with a protein concentration higher than 2 wt % have a gellike appearance after being heated for more than 24 h at 80 °C in the spectrometer. To relate the information obtained from the NMR spectra to the kinetics of aggregation, one can integrate the proton signal intensities over the whole range of proton chemical shifts available or one can integrate proton signal intensities in specific regions of the NMR spectrum as a function of the heating time. Both approaches lead to similar results, and here we show the results obtained by integrating over

the whole range of chemical shifts (Figure 7). The proton NMR signal intensity drops quickly as a function of time for the 8.2 wt % β-lg sample (Figure 7 - circles) suggesting fast conversion of protein monomers to fibrils during the first 5 h of the experiment at 80 °C. Subsequently, the process continues even after 24 h of heating but at a lower rate. For the 4.1 wt % sample (Figure 7, squares), there is virtually no change in the proton NMR signal intensity during the first 5-6 h of the experiment at 80 °C. Then the signal starts to decrease with a similar slope as the one observed for the 8.2 wt % β-lg sample and continues to decrease until the experiment is stopped. The smallest change in the proton NMR signal intensity with time is observed for the 3.1 wt % β-lg sample at 80 °C (Figure 7, triangles). In summary, Figure 7 shows that many β-lg molecules do not form fibrils upon heating at 80 °C and remain in solution either in a monomeric state or as small aggregates (dimers, small oligomers). These results, together with the results from the experiments carried out with 1 and 2 wt % β-lg, suggest that there is an apparent critical aggregation concentration of approximately 2.5 wt % as is estimated from the time dependence of the proton NMR signal intensities at the protein concentrations studied. This apparent critical aggregation concentration should be considered as a concentration above which the β-lg monomers are predominantly converted into linear aggregates and not into other aggregated, though low molecular weight, species. At any given β-lg concentration at pH ) 2 and at low ionic strength, there may be linear aggregates formed into the solution but up to the apparent critical aggregation concentration their concentration is too low to be detected by proton NMR spectroscopy. This is different from the notion of the conventional equilibrium critical aggregation concentration only above which aggregates that are in equilibrium with the monomers in the solution are present. The estimated value for the critical aggregation concentration of about 2.5 wt % is in agreement with the static light scattering data (Figure 4). Close to the critical concentration, only a small fraction of the monomeric β-lg is converted into fibrils; hence, the change in the scattering intensity with time is small. At concentrations significantly larger than the critical aggregation concentration,

Formation of β-Lactoglobulin Fibrils

Figure 8. 500 MHz proton NMR spectra of 8.2 wt % β-lg at 35.2 °C, pH ) 2.0 and ionic strength I ) 0.013 M. (A) Spectrum of native β-lg; (B) Spectrum of β-lg which is heated for 38 h at 80 °C and subsequently cooled to 35.2 °C.

a large part of the monomeric β-lg is converted into fibrils, so the change in the scattering intensity with heating time is drastic. Previous kinetic aggregation experiments using the selective precipitation technique34 give qualitatively similar results to the ones presented here.10,11 The work of Aymard et al.11 shows that 50% of the protein molecules in a 0.5 wt % β-lg end up in β-lg aggregates. The work of Veerman et al.10 shows that approximately 60% of the protein molecules end up in aggregates for a range of concentrations between 0.5 and 4 wt % and ionic strengths from 0.01 to 0.08 M. Both results are, however, in quantitative disagreement with the NMR results presented here in which an apparent critical aggregation concentration around 2.5 wt % is observed. The selective precipitation method34 is based on a pH quench to 4.8 or 7.0 of a sample that has already undergone aggregation.10,11 This pH quench causes a fraction of the protein molecules to precipitate, and the authors, using this method, argue that solely nonaggregated protein monomers do not precipitate. The precipitate is separated from the protein solution by means of centrifugation and the remaining protein concentration in the supernatant is subsequently measured by UV spectrophotometry. In our opinion, this pH quench may induce precipitation not only of the aggregates but also of a fraction of denatured protein molecules. The proton NMR spectra obtained of samples of β-lg at high temperature after prolonged heating show the typical features of the presence of denatured protein, i.e., limited chemical shift dispersion of proton NMR resonances (Figures 6 and 8). The NMR spectrum does not change significantly even when the sample which is subjected to 38 h of heating at 80 °C is cooled to 35 °C (compare Figure 6D with Figure 8B). General Discussion The kinetics of the linear aggregation of β-lg follows a complex scheme. The DLS results presented suggest that the

Biomacromolecules, Vol. 4, No. 6, 2003 1621

β-lg aggregation becomes irreversible only after prolonged heating periods. Yet, the NMR data show that a significant fraction of the β-lg molecules is still present as low molecular weight complexes in the solution after heating the sample for more than 24 h at 80 °C. These experimental results, as well as our AFM observations, i.e., the observed fibril splitting, the fact that no protofilaments are observed, the periodic structure of the fibrils, the fact that short and long aggregates are simultaneously present into the β-lg solution after prolonged heating at 80 °C, suggest that multiple steps in the formation of β-lg fibrils occur when heating β-lg at pH ) 2.0 and low ionic strength at 80 °C. A possible explanation for the transition from reversible to irreversible β-lg aggregation upon increasing the heating time is that the process of fibril formation is at least a twostep process. The first step is the aggregation of β-lg, and the second step, which takes much more time, is the consolidation of the monomers into the already formed aggregates. At any time during the aggregation process, short and long fibrils (see AFM results, Figure 1) coexist in solution with low molecular weight species, which are monomers or very short oligomers (as detected by NMR, see Figures 6 and 7). At high protein concentrations, AFM results suggest that long fibrils dominate over short ones and DLS results show that initially a bimodal distribution of long fibrils and low molecular weight species exists. NMR spectroscopy shows that after prolonged heating periods at high temperature the low molecular weight species survives as the corresponding NMR proton signals are still observed (Figures 6 and 7). The low molecular weight species do not aggregate any further, and its corresponding conformation does not change in time even upon decreasing the temperature. The low molecular weight species observed after cooling an 8.2 wt % β-lg solution from 80 to 35.2 °C has the characteristics of an unfolded protein (Figure 8B), similar to the low molecular weight species observed in a β-lg solution at 80 °C (Figure 6). The corresponding NMR spectra (Figures 6 and 8) are characterized by limited chemical shift dispersion of the proton resonances of the amino acid residues of the same type in β-lg. This is characteristic of the NMR features of largely unfolded protein molecules. Taking all results together, the following scheme regarding β-lg fibril formation emerges. Upon heating, the β-lg molecules (partially) unfold and form an intermediate, which can follow basically two routes of aggregation. The first route involves at least two steps: the reversible formation of linear aggregates, followed by some kind of consolidation step, which leads to fibrils. The second route leads to low molecular weight “dead-end” species from which fibrils cannot be formed. The “dead-end” species can either be a denatured monomer or an oligomer that cannot aggregate any further. Such a scheme can explain the existence of a critical β-lg aggregation concentration, which is not caused by equilibrium between monomers and aggregates but rather by a competition between reactions of different order and with a different rate. At low protein concentration, the transformation of monomers into “dead-end” species dominates over the linear aggregation and as a result, only a small

1622

Biomacromolecules, Vol. 4, No. 6, 2003

amount of fibrils is observed (though this amount is still enough to be detected by DLS and AFM). Once the initial β-lg monomer concentration exceeds a critical value (in our case around 2.5 wt %) the reaction rate for fibril formation exceeds that for the formation of “dead-end” species, and the excess of β-lg monomers is converted into fibrils. Conclusions The linear aggregation of β-lactoglobulin upon prolonged heating at pH ) 2.0 and at low ionic strength at 80 °C appears to be a multistep process. Competing reactions lead to two products: long linear aggregates and some low molecular weight “dead end” species. The “dead end” species comprises small oligomers and/or monomeric non-native protein molecules and cannot form fibrils. Fibril formation involves at least two steps: the reversible formation of linear aggregates, followed by a slow process of “consolidation” after which the fibrils no longer disintegrate upon subsequent slow cooling. Atomic force microscopy results suggest that the fibrils formed are helical with a thickness of one or two protein monomers, and a periodicity of about 25 nm. Acknowledgment. This work was supported by The Netherlands Research Council for Chemical Sciences (CW) with financial aid from The Netherlands Organization for Scientific Research (NWO), in the context of the SOFTLINK program “Theoretical Biophysics of Proteins in Complex Fluids”. References and Notes (1) Langton, M.; Hermansson, A.-M. Food Hydrocolloids 1992, 5 (6), 523. (2) Stading, M.; Langton, M.; Hermansson, A.-M. Food Hydrocolloids 1992, 6 (5), 455. (3) Tobitani, A.; Ross-Murphy, S. B. Macromolecules 1997, 30, 4845. (4) Tobitani, A.; Ross-Murphy, S. B. Macromolecules 1997, 30, 4855. (5) Kavanagh, G. M.; Clark, A. H.; Ross-Murphy, S. B. Int. J. Biol. Macromol. 2000, 28, 41. (6) Kavanagh, G. M.; Clark, A. H.; Ross-Murphy, S. B. Langmuir 2000, 16, 9584.

Arnaudov et al. (7) Le Bon, Ch.; Nicolai, T.; Durand, D. Macromolecules 1999, 32, 6120. (8) 8. Lefevre, T.; Subirade, M. Biopolymers 2000, 54, 578. (9) Lefebvre, J.; Renard, D.; Sanches-Gimeno, A. C. Rheol. Acta 1998, 37, 345. (10) Veerman, C.; Ruis, H.; Sagis, L. M. C.; van der Linden, E. Biomacromolecules 2002, 3, 869. (11) Aymard, P.; Nicolai, T.; Durand, D.; Clark, A. Macromolecules 1999, 32, 2542. (12) Koike, A.; Nemoto, N.; Doi, E. Polymer 1996, 37 (4), 587. (13) Koike, A.; Takada, A.; Nemoto, N. Polym. Gels Networks 1998, 6, 257. (14) Schokker, E. P.; Singh, H.; Pinder, D. N.; Creamer, L. K. Int. Dairy J. 2000, 10, 233. (15) Nemoto, N.; Koike, A.; Osaki, K.; Koseki, T.; Doi, E. Biopolymers 1993, 33, 551. (16) Koseki, T.; Kitabatake, N.; Doi, E. Food Hydrocolloids 1989, 3 (2), 123. (17) Renard, D.; Lefebvre, J. Int. J. Biol. Macromol. 1992, 14, 287. (18) Harwalkar, V. R.; Kalab, M. Milchwissenschaft 1985, 40 (2), 65. (19) Gimel, J. C.; Durand, D.; Nicolai, T. Macromolecules 1994, 27, (2), 583. (20) Ikeda, S.; Morris, V. J. Biomacromolecules 2002, 3, 382. (21) Wong, D. W. S.; Camirand, W. M.; Pavalath, A. E. Crit. ReV. Food Sci. Nutr. 1996, 36 (8), 807. (22) Gosal, W. S.; Clark, A. H.; Pudney, P. D. A.; Ross-Murphy, S. B. Langmuir 2002, 18, 7174. (23) Ikeda, S. Spectrosc.sInt. J. 2003, 17, 195. (24) Roefs, P. F. M.; de Kruif, K. G. Eur. J. Biochem. 1994, 226, 883. (25) Dickinson, E. J. Chem. Soc., Faraday Trans. 1994, 90 (1), 173. (26) Bijsterbosch, B. H.; Bos, M. T. A.; Dickinson, E.; van Opheusden, J. H. J.; Walstra, P. Faraday Discuss. 1995, 101, 51. (27) Witten, T. A., Jr.; Sander, L. M. Phys. ReV. Lett. 1981, 47, 1400. (28) Clark A. H.; Kavanagh, S. B.; Ross-Murphy, S. B. Food Hydrocolloids 2001, 15, 383. (29) Oosawa, F.; Asakura, Sh. Thermodynamics of the Polymerization of Protein; Academic Press Inc.: New York, 1975. (30) Belloque, J.; Ramos, M. Trends Food Sci. Technol. 1999, 10, 313. (31) Zurdo, J.; Guijarro, J. I.; Jimenez, J. L.; Saibil, H. R.; Dobson, Ch. M. J. Mol. Biol. 2001, 311, 325. (32) Yang, J.; Mou, J.; Shao, Z. Biochim. Biophys. Acta 1994, 1199, 105. (33) Hamada, D.; Dobson, C. M. Protein Sci. 2002, 11, 2417. (34) Harwalkar, V. R.; Kalab, M. Milchwissenschaft 1985, 40 (11), 665. (35) Chamberlain, A. K.; Cait, E. M.; Zurdo, J, Morozova-Roche, L. A.; Allen, H.; Hill, O.; Dobson, Ch. M.; Davis, J. J. Biophys. J. 2000, 79, 3282-3293. (36) Ionescu-Zanetti, C.; Khurana, R.; Gillespie, J. R.; Petrick, J. S.; Trabachino, L. C.; Minert, L. J.; Carter, S. A.; Fink, A. L. PNAS 1999, 96, 23, 13175-13179. (37) Massi, F.; Straub, J. E. Proteins 2001, 42, 217.

BM034096B