Structural Conformation of Bovine Serum Albumin Layers at the Air–Water Interface Studied by Neutron Reflection

Journal of Colloid and Interface Science 213, 426 – 437 (1999) Article ID jcis.1999.6157, available online at http://www.idealibrary.com on

Structural Conformation of Bovine Serum Albumin Layers at the Air–Water Interface Studied by Neutron Reflection J. R. Lu,* ,1 T. J. Su,* and R. K. Thomas† *Department of Chemistry, University of Surrey, Guildford GU2 5XH, United Kingdom; and †Physical and Theoretical Chemistry Laboratory, South Parks Road, Oxford OX1 3QZ, United Kingdom Received September 8, 1998; accepted February 17, 1999

the consequence that some parts of the surface are more hydrophilic than others. Adsorption at the air–water interface is driven by the tendency for the more hydrophobic regions on the outer surface to minimize their exposure to the aqueous environment. At the air–water interface the hydrophobic portion of the protein is therefore expected to be exposed to air while the hydrophilic portion is submerged in the aqueous subphase. The structural detail of the adsorbed layer is then determined by preferential orientation and packing of protein molecules in the interface which is, in turn, associated with electrostatic, hydrophobic, and entropic interactions. The combination of all the interactions may result in structural rearrangements of the protein and may even lead to denaturation. Adsorption of proteins at the air–water interface is generally accepted to occur in two main steps (1– 4). In the first step protein molecules diffuse to the air–water interface and the resulting adsorption is marked by an increase in surface pressure and an increase in surface excess. In the slower second step there may be structural rearrangement of the protein molecules within the adsorbed layer with partial unfolding of the globular assembly. Subsequent multilayer formation may then also occur, in which there is reversible association of nondenatured protein molecules on the underside of the primary surface layer at relatively high protein concentrations. The physical state of protein molecules within the layer may in principle be deduced from changes in the structure of the adsorbed protein layer relative to that of the stable globular structure in solution. Previous investigations on adsorbed layers of globular proteins have relied on surface tension measurements, radiolabeling techniques, ellipsometry, infrared spectroscopy, the use of proteolytic enzymes to detect accessibility of hydrolysis-sensitive bonds in the adsorbed layer, and enzymatic cleavage (5–11). Although these methods may be sensitive to changes of the protein with time and concentration, they are not sufficiently accurate at measuring either the adsorbed amount or the structural characteristics of adsorbed layers, both of which are key quantities in the determination of the changes induced by adsorption. The main difficulty preventing derivation of accurate surface excesses from surface tension data is the unknown activity coefficient of the protein

The adsorption of bovine serum albumin (BSA) at the air–water interface has been studied by specular neutron reflection. The variation of the adsorbed amount and the total thickness of the BSA layer with respect to bulk BSA concentration was determined at pH 5, close to its isoelectric point (IP). While the surface excess showed a steady increase with bulk concentration the thickness of the protein layer was found to be close to the short axial length of 40 Å of the globular solution structure of BSA at concentrations below 0.1 g dm 23, suggesting that BSA molecules adsorb with their long axes parallel to the surface of water. At 1 g dm 23 the adsorbed layer can be modeled as an upper layer of 40 Å with a volume fraction of 0.4 and a sublayer of 30 Å underneath the top main layer with a volume fraction of 0.12. The results suggest that, although there is some structural deformation accompanying adsorption, there is no denaturation. The extent of immersion of the BSA in water was determined by performing the measurements in D 2O and in a mixture of H 2O and D 2O whose contrast matches that of BSA. The signal is then only from the part of the layer out of water. At pH 5 this layer was about 10 6 5 Å at a bulk concentration of 5 3 10 24 g dm 23 and decreased to 5 6 3 Å at 1 g dm 23. The fraction of the BSA layer immersed in water therefore varies from about 70 to over 90%. The effect of pH on the adsorption was examined at two BSA concentrations. While pH had little effect on the adsorption at a low BSA concentration of 5 3 10 23 g dm 23, both surface excess and layer thickness showed pronounced peaks at pH 5 at the higher concentration of 1 g dm 23. The increased adsorption at pH 5 is attributed to the reduced lateral electrostatic repulsion around the IP. This adsorption pattern became less pronounced when the total ionic strength was increased from 0.02 to 1 M, indicating that the electrolyte screens the electrostatic repulsions within the adsorbed layer. © 1999 Academic Press

Key Words: BSA adsorption; protein adsorption; protein conformation; neutron reflection; adsorption.

INTRODUCTION

The outer surface of a globular protein contains polar and charged amino acid groups. The distribution of these functional groups on the hydrophilic outer shell is usually uneven, with 1

To whom correspondence should be addressed.

0021-9797/99 $30.00 Copyright © 1999 by Academic Press All rights of reproduction in any form reserved.

426

427

BSA ADSORPTION ON THE SURFACE OF WATER

molecule in bulk solution, with the further uncertainty as to whether true equilibrium has been established between adsorbed layer and bulk solution. Although several authors have shown that surface excesses can be derived indirectly using semi-empirical equations the inconsistency in the surface excesses and layer thicknesses obtained through these indirect approaches has caused much confusion in the literature. Neutron reflection has recently been used to study the adsorption of b-casein and b-lactoglobulin at the air–water interface (12, 13). The resolution of neutron reflection is sufficient to allow the direct determination of the protein distribution along the surface normal direction. A comparison of the thickness of the layer with the known dimensions of b-lactoglobulin leads to the conclusion that surface adsorption results in the breakdown of the globular structure. On the other hand, the framework of the more robust globular protein, lysozyme, was shown to remain intact after adsorption at the air–water interface (14, 15). In the present paper we apply neutron reflection to the adsorption of bovine serum albumin (BSA) at the air–water interface. In comparison with lysozyme and b-lactoglobulin BSA is less rigid and therefore more likely to denature at the surface. Its structural flexibility arises because it consists of three principal domains rather loosely joined together. The structure within each subdomain is maintained in a reasonably ordered state in aqueous solution by a large number of disulfide bridges. On the other hand, the interactions between the three domains are noncovalent. It is, therefore, interesting to examine how the physical state of the adsorbed BSA is affected by surface adsorption under different solution conditions. EXPERIMENTAL

Neutron specular reflection measurements were made on the white beam reflectometer SURF at ISIS, Didcot, United Kingdom (16) using a range of incident wavelengths from 0.5 to 6.5 Å. The protein solutions were held in teflon troughs with a positive meniscus to prevent obstruction of the incoming and reflected beams by the teflon. The troughs were mounted on an active antivibration table to prevent surface vibrations. Samples were accurately aligned using a laser beam which follows the same path as the neutron beam. The neutron beam intensity was calibrated with respect to the reflectivity of clean D 2O. A flat background determined by extrapolation to high values of momentum transfer, k (k 5 (4p sin u)/l where l is the wavelength and u is the glancing angle of incidence), was subtracted. The reflectivity was measured at three different angles of incidence, 0.5°, 0.8°, and 1.8°, in order to ensure a range of k sufficient to determine the thickness of the protein layer. Fatty acid free BSA was used as supplied (Sigma, Cat. No. A0281, Lot No. 10H9304). The molecular weight of BSA is 66700 Daltons and its isoelectric point (IP) is 4.7– 4.8 [22]. The solution pH was controlled by using phosphate buffer and the

pH varied by changing the ratio of Na 2HPO 4, NaH 2PO 4, and H 3PO 4, keeping the total ionic strength fixed at 0.02 M. There were small differences in pH between H 2O and D 2O but this was controlled to within 0.2 pH units. D 2O was purchased from Fluorochem (99.9% D) and its surface tension was typically over 71 mN m 21 at 25°C, indicating the absence of any surface active impurity. H 2O was processed through an Elgastat ultra pure water system (UHQ) and its surface tension at 25°C was constant at 71.5 mN m 21. The glassware and Teflon troughs for the reflection measurements were cleaned using alkaline detergent (Decon 90) followed by repeated washing in UHQ water. All the experiments were performed at 25°C. NEUTRON REFLECTION

The specular neutron reflectivity is related to the variation of scattering length density r ( z) along the surface normal direction by the approximate relation (17, 18) R~ k ! 5

16 p 2 u rˆ ~ k !u 2 , k2

[1]

where R is the reflectivity, k is the momentum transfer (see above), and rˆ (k) is the one-dimensional Fourier transform of r ( z)

rˆ ~ k ! 5

E

`

exp~2i k z! r ~ z!dz.

[2]

2`

The scattering length density is related to the chemical composition across the surface by

r ~ z! 5 Sn i ~ z!b i,

[3]

where n i ( z) is the number density of element i and b i its scattering amplitude (scattering length). The three equations given above demonstrate the direct relationship between neutron reflectivity and interfacial composition. Values of b i vary from isotope to isotope and isotopic substitution can therefore be employed to vary r ( z) as can be seen from Eq. [3]. This, in turn, alters the reflectivity R( k ) through Eqs. [2] and [1], although the density distribution n i ( z) and chemical nature of the surface are unaltered. Since the scattering length of hydrogen is 23.7 3 10 25 Å, that of deuterium 6.7 3 10 25 Å and that of oxygen 5.8 3 10 25 Å, the scattering lengths for H 2O and D 2O are of opposite sign. An important consequence of this is that water with the approximate ratio of H 2O to D 2O of 11 to 1 has a scattering length density of zero. Since this is identical with that of air, such water is null reflecting (NRW). When a protein layer is adsorbed at the air–water interface of NRW the protein in the layer is the only species that contributes to the specular reflec-

428

LU, SU, AND THOMAS

TABLE 1 Physical Constants for BSA a

tivity. Under this condition, if the adsorbed layer is uniform, the area per molecule, A, is given by Sm i b i A5 , rt

[4]

where m i is the number of element i with scattering length b i . The surface excess G is related to A by G5

1 , Na A

[5]

where N a is Avogadro’s constant. Neutron reflectivity profiles are usually analyzed by means of the optical matrix formalism, described by Born and Wolf (19) and Lekner (20). The procedure for extracting structural information from experimental profiles is straightforward. A structural model is assumed and the exact reflectivity calculated using the optical matrix method. The calculated reflectivity is compared with the measured one and the structural parameters modified in a least-squares iteration. The parameters used in the calculation are the thicknesses of the layers, t, and the corresponding scattering length densities, r, which are related to the composition of the layers by Eq. [3]. In fitting the reflectivity profiles with a uniform layer model, it is usually found that t and r can be varied over a limited range, but that their variations cancel in their contribution to A in such a way that A is independent of the initial assumption that the layer is uniform (17). When the scattering length density of the water is different from zero the reflectivity will contain a contribution from the underlying water and hence from any disruption of its distribution across the interface by an adsorbed protein layer. To a first approximation the protein layer can be regarded as consisting of two layers, an upper one in air and a lower one fully immersed in the water. The following equations then describe the distribution of scattering length density of the two layers,

Water contrast

NRW

CM2.5 d

D 2O

Scattering length 3 10 25/Å b BSA volume c/Å 3 Scattering length density 3 10 26/Å 22

15860 79110 2.0

21850 79110 2.5

25730 79110 3.3

a The scattering length densities are slightly pH dependent but the variation is within 1 3 10 27/Å 22. b The amino acid sequence for BSA was obtained from Voet et al. (32) and its scattering length calculated using data from Sears (33). c The molecular volume of BSA was calculated from Chalikian et al. (34) and Van Krevelen (35). CM2.5 is the mixed water of H 2O and D 2O with the volume fraction of D 2O equal to 0.44.

r 2 5 f pr p 1 ~1 2 f p ! r w ,

where f p is the volume fraction of the protein in the layer, and r p and r w are the respective scattering length densities for protein and water. The number of water molecules in the layer can be estimated by assuming 100% filling of the volume of the second layer. If V t is the total volume for the second layer, V p the protein volume excluding hydrated water molecules, and V w the molecular volume for water, then V t 5 A t 2 5 V p 1 n wV w

b p ~1 2 f ! t 1A

r2 5

b p f 1 n wb w , t 2A

[6]

where b p and b w are the scattering lengths for protein and water, n w is the number of water associated with each protein molecule, and f is the fraction of the protein molecule immersed in water. For the second layer the scattering length density comprises a contribution from the adsorbed protein itself and one from the water forming the rest of the layer. Alternatively the scattering length density of the second layer may be written in terms of the composition of the layer by

[8]

and

fp 5

Vp . At2

[9]

The values of A and n w can be evaluated from Eqs. [8] and [9]: nw 5 A5

r1 5

[7]

~1 2 f p !Sb i r 2 V w 2 b w ~1 2 f p !

[10]

n wV w , t ~1 2 f p !

[11]

where Sb i is the scattering length for the fraction of the protein molecule immersed in water. BSA contains a large number of labile hydrogens which will exchange with the D 2O. On the timescale of the neutron reflection experiment exchange will be complete for the labile hydrogens on the amino acid side chains and for most of those labile hydrogens on the peptide backbone that are easily accessible to bulk water. However, exchange may be slow for labile protons on the peptide chains encapsulated in hydrophobic portions, although deformation of the globular structure following surface adsorption may cause sufficient perturbation to facilitate hydrogen exchange. The variation of the scattering length density of BSA with H/D content of the water is given in Table 1 together with the other physical

BSA ADSORPTION ON THE SURFACE OF WATER

FIG. 1. Plots of log(reflectivity) against k for the adsorption of BSA on the surface of null reflecting water at pH 5 with bulk concentrations of 5 3 10 24 g dm 23 (x), 5 3 10 23 g dm 23 (‚), and 5 3 10 22 g dm 23. The total ionic strength was kept at 0.02 M. The continuous lines were calculated using the structural parameters listed in Table 2.

constants used in the data analysis. The effect of incomplete exchange of the labile hydrogens varies according to the contrast conditions and will therefore be discussed later in conjunction with the structural determinations. However, at this stage we assume complete exchange. RESULTS AND DISCUSSION

(A) Variation of the Layer Structure with Bulk Concentration The surface coverage of BSA was first measured at pH 5, close to its isoelectric point. In NRW the specular reflectivity results entirely from the adsorbed protein layer and the measurements were therefore made in NRW at different BSA concentrations. Figure 1 shows the variation of the neutron reflectivity with bulk BSA concentration. The results are plotted in terms of log(reflectivity) against k. Most of the variation in reflectivity occurs at values of k below 0.1 Å 21 and above this all the reflectivity curves fall to the level of the flat background. The level of reflectivity is an indicator of the surface coverage and its slope is related to the thickness of the adsorbed layer. BSA adsorption at the air–water interface was found to be time dependent. The results shown in Fig. 1 were recorded after allowing sufficient time for equilibrium to be attained. One hour was typically required for equilibrium to be reached

429

at high BSA concentrations but some 10 h was required at low concentrations, around 5 3 10 24 g dm 23. The rate of equilibration is in broad agreement with results from the literature (1– 4). Many attempts have been made to relate variation of adsorption with time to the kinetic process of adsorption (1–3, 9, 10), but the time reproducibility is generally poor. We therefore confined our measurements to the equilibrium situation, which could easily be detected by following the change of neutron reflectivity with time. Quantitative information about the adsorbed layer structure was obtained by model fitting based on the optical matrix method as outlined above. For all concentrations below 1 g dm 23, the measured reflectivity profiles could be fitted using a model of a single uniform layer. The good fits obtained with this model, shown as continuous lines in Fig. 1, show that the adsorbed layers are reasonably uniform and therefore that the adsorbed BSA molecules probably retain their globular structure. Denaturing of the protein usually generates layers of different density and a more complex layer structure is then required to fit the data (see, for example, (21)). The fits of the single uniform layer model give the thickness and scattering length density of the layer and the volume fraction of protein is derived from the latter. At the two lowest concentrations the thickness of the BSA layer was found to be 32 6 3 Å and the area per molecule to decrease from 15000 6 300 Å to 7400 6 200 Å 2 at the higher of the two concentrations. The change in reflectivity results entirely from the increase in surface concentration; the layer thickness is constant. Since the globular structure of BSA in aqueous solution is approximately cylindrical with dimensions 40 3 40 3 140 Å 3 (22, 23), the constant thickness, and its value, suggest that BSA molecules adopt a sideways-on orientation in this low surface concentration range; that is, the molecules adsorb with their long axes parallel to the surface. The fact that the thickness of the layer is less than the dimension of the short axis of the cylindrical structure of BSA suggests that there is some flattening of the molecules on adsorption. Alternatively, the effect might be attributed to the effect of the cylindrical shape of the protein. When a layer of cylinders is represented by a uniform layer the effective thickness will be less than the diameter of the cylinder. However, this effect is not easy to quantify when the molecules are not true cylinders. In lysozyme layers at the air–water interface the uniform layer thickness exactly matched the short and long axial lengths where sideways-on and longways-on adsorption occurred, respectively. In any case, the difference between 40 and 32 Å is somewhat too large to be explained in this way. When the BSA concentration is increased to 5 3 10 22 g dm 23 the area per molecule decreases to 5400 6 200 Å 2, which is close to the estimated limiting area per molecule of 5600 Å 2 for sideways-on adsorption. The result suggests that at this concentration BSA molecules within the adsorbed layer are starting to repel each other as a result of close packing. Although the net charge at this pH is almost zero, local charges

430

LU, SU, AND THOMAS

deformation there is no breakdown of the globular framework into peptide fragments. The variation of the total BSA surface excess and the thickness of the surface layer with bulk concentration is shown in Fig. 3. Since the main increase in surface excess occurs below 5 3 10 22 g dm 23, the results are plotted in terms of surface coverage (Fig. 3a) and thickness (Fig. 3b) against log[concentration] to highlight the changes over the low concentration

FIG. 2. Plot of log(reflectivity) against k for the adsorption of BSA on the surface of null reflecting water at pH 5 at 1 g dm 23. The total ionic strength was kept at 0.02 M. The continuous line was calculated using a two layer model with the structural parameters listed in Table 2. The dashed line was calculated by taking account of the top layer only.

are present on the surface of the globular framework and some weak electrostatic repulsion may still exist within the adsorbed layer. The increase of the layer thickness from 32 6 3 Å to 37 6 3 Å suggests that the globular framework of BSA is not at all rigid. A further increase in BSA concentration to 1 g dm 23 produces a layer thickness of some 40 Å, which is comparable with the full length of the short axis for the globular structure. The area per molecule at this concentration was found to be around 3800 Å 2 compared with the estimated limit of 5600 Å 2, indicating a greater degree of structural deformation. However, it can be seen from Fig. 2 that the uniform layer model does not fit the high k range of the measured profile. A better fit is achieved by incorporating a second layer beneath the main surface layer (24). In fitting this two layer model the upper layer was taken to be similar to that obtained in the single uniform layer fit. The continuous line in Fig. 2 is the best fit of this model and gives a thickness of 42 Å for the upper layer and 30 Å for the totally immersed lower layer with a packing corresponding to an area per molecule of 3800 6 200 Å 2 for the upper layer and 19,000 6 1000 Å 2 for the lower layer. This result suggests that as the top layer becomes more close-packed a further increase in bulk concentration leads to the adsorption of a secondary layer underneath the main layer. It was found to be impossible to fit alternative models with the less ordered fragment distributions more characteristic of denatured protein (21). Again this suggests that in spite of the high degree of

FIG. 3. Variation of surface excess (a) and total layer thickness (b) as a function of log[concentration] for the adsorption of BSA on the surface of water at pH 3 (E), 5 (F), and 7 (1). The total ionic strength was kept at 0.02 M.

431

BSA ADSORPTION ON THE SURFACE OF WATER

TABLE 2 Structural Parameters for the BSA Layer Adsorbed on the Surface of Water at pH 5 a C/g dm 23

t 1 6 3/Å

t 2 6 5/Å

A 1 /Å 2

A 2 /Å 2

G 6 0.2/mg dm 23

f1

f2

5 3 10 24 5 3 10 23 5 3 10 22 1

32 32 37 42

— — — 30

15000 6 2000 7400 6 500 5000 6 350 3800 6 300

— — — 19000 6 1000

0.74 1.5 2.1 2.9

0.16 0.33 0.4 0.4

— — — 0.12

When the bulk scattering length density is nonzero the first layer has to be split into two layers. The thickness for the layer above water is between 5 6 3 to 10 6 5 Å (see text for more details). a

region. The G and t vs log[concentration] plots show that while the surface excess increases steadily with bulk concentration the total thickness of the layers remains reasonably constant until the formation of the additional lower layer at 1 g dm 23. The structural parameters of the layer are given in Table 2. Similar measurements were made at pH 3 and 7 and the results are also plotted in Fig. 3 for comparison. Although a similar trend of surface excess with bulk BSA concentration is observed, the surface excess at pH values away from the IP is lower. The thickness of the layer is always less than 40 Å over the whole concentration range except where the sublayer is formed at pH 5. The thicknesses of layers measured at the air–water interface always contain a contribution from the thermal motion of the surface (capillary waves). The relationship between the total measured thickness t and capillary roughness w is given approximately by (17, 25)

t 2 5 s 2 1 w 2,

[12]

where s can be regarded as the intrinsic thickness of the layer in the absence of capillary waves. The roughness from the thermal motion in a pure liquid is inversely proportional to the square root of the surface tension. At 25°C, the surface tension of pure water is about 72 3 10 23 Nm 21 and the surface ˚ , which is equivalent to a value of w of 7 Å roughness is 2.8 A (the definition of roughness is not the same and there is a factor of 2.3 difference (25)). Since the surface tension of BSA at 1 g dm 23 is about 53 mN m 21 (10) the thermal roughness of the BSA solution at this concentration is estimated to be 9 Å. Taking the thickness of BSA at 1 g dm 23 to be 40 Å gives an intrinsic thickness of 39 Å. This correction is well within the experimental error and, unlike surfactant monolayers where roughness is an important structural feature (17, 25), thermal roughness does not make a measurable contribution to the mean thickness of the protein layer. The surface excess of BSA has been determined by several authors using surface tension, radiolabeling, and ellipsometry (1–3, 5). The most systematic study was made by Graham et al. (1–3) who examined the surface behavior of several model proteins at the air–water interface. Their BSA surface excesses are, however, substantially higher than our values over the

whole concentration range. At 5 3 10 24 g dm 23 their surface excess is 1.5 mg m 22 as compared with 0.75 mg m 22 obtained from neutron reflection. At 5 3 10 22 g dm 23 their surface excess is 3.5 mg m 22 compared with the neutron value of 2.1 mg m 22. As the concentration is further increased their values of the surface excess appear to increase exponentially. At 1 g dm 23 their surface excess is 10 mg m 22 compared with our value of 2.9 mg m 22. We cannot explain this discrepancy but we note that it is not easy to observe protein adsorption at the air–water interface with ellipsometry because of the lack of contrast in refractive index between the adsorbed layer and the water substrate. The difficulties at the higher concentration region may be caused by the inability of ellipsometry to make an unambiguous separation of coverage and the protein distribution profile along the surface normal direction (27). Not surprisingly, there is also a large discrepancy between their layer thicknesses and the neutron results. Below 0.1 g dm 23 they found a thickness of 90 Å to be compared with 35 Å from neutron reflection! The maximum error in the determination of the thickness by neutron reflection would occur if the H/D exchange is incomplete, which we discuss further below, but it is less than about 15%. The assumptions that have to be made to obtain a layer thickness by ellipsometry at the air–water interface are extremely uncertain. (B) The Effect of Solution pH and Salt Concentration In our previous study of the adsorption of BSA at the hydrophilic silicon oxide–water interface we showed that the BSA surface excess attains its maximum value at the isoelectric point (28). The major difference at the air–water interface is that the surface itself does not have an intrinsic charge which also varies with pH. There is no electrostatic interaction comparable with that between protein and solid substrate, although electrostatic forces still matter within the adsorbed layer. The hydrophobic effect and gains in entropy arising from protein dehydration and change in structural conformation are the main factors contributing to surface adsorption. The effect of pH on surface adsorption has been shown in the plots in Fig. 3. An alternative way of showing the effects of pH is to plot the change in surface excess and layer thickness against pH at a fixed protein concentration and this is done in Fig. 4. The measurements were made at two BSA concentra-

432

LU, SU, AND THOMAS

TABLE 3 Structural Parameters for the BSA Layer Adsorbed on the Surface of water at pH 3 a C/g dm 23

t 1 6 3/Å

A 1 /Å 2

G 6 0.2/mg dm 23

f1

5 3 10 23 1

28 37

10860 6 2000 6000 6 300

1.05 1.9

0.26 0.37

a When the bulk scattering length density is nonzero the first layer has to be split into two layers. The thickness for the layer above water is about 5 6 3 Å at both concentrations.

FIG. 4. Variation of surface excess (a) and total layer thickness (b) as a function of solution pH at the BSA concentrations of 5 3 10 23 g dm 23 (E) and 1 g dm 23 (F). The total ionic strength was kept at 0.02 M. The solid lines were drawn for a guide.

tions, 5 3 10 23 and 1 g dm 23. Figure 4a shows that the surface excess peaks at the IP at both BSA concentrations, although the effect is greater at high BSA concentration. The change in the thickness of the BSA layers is shown in Fig. 4b where it can be seen that the thickness reaches its maximum at the IP at 1 g dm 23 but is approximately constant at 5 3 10 23 g dm 23. Tables 2– 4 list the structural parameters obtained from the analysis at all three pH values. The variation of surface excess with pH

appears to correlate with the change in the extent of electrostatic interaction within the adsorbed BSA layer. At the IP the electrostatic repulsion should be at a minimum since the net charge within the protein is zero. Lateral repulsion starts to increase as the pH is shifted away from the IP. However, an increase in the number of net charges within the protein also increases its hydrophilicity, which also reduces its tendency to adsorb. The adsorption of BSA onto the hydrophilic silicon oxide– water interface has also been studied by neutron reflection (28). Since the measurements were all made under similar conditions it is useful to compare the two sets of results. The basic pattern of surface excess and thickness variation with respect to pH and bulk concentration is remarkably similar, that is, both surface excess and layer thickness peak at pH 5. It is interesting to find that for both interfaces the thickness over the whole pH and concentration ranges is below 40 Å at pH 3 and 7 and over most of the concentration range at pH 5. When the bulk concentration at pH 5 is above 0.5 g dm 23 an immersed sublayer forms at both air–water and solid–water interfaces. Even the absolute values of the surface excess at pH 5 are within 0.5 mg m 23. The values at the air–water interface at pH 3 and 7 are higher than at the solid–water interface but this might be the effect of pH charging both surface and protein in the case of the solid–water interface. The effects of electrostatic repulsion between surface and protein should then be correspondingly greater than at the air–water interface. All the results so far presented were measured at a total ionic strength of 0.02 M. We have also examined the effect of ionic strength by adding sodium chloride to bring the total ionic TABLE 4 Structural Parameters for the BSA Layer Adsorbed on the Surface of Water at pH 7.2 a C/g dm 23

t 1 6 3/Å

A 1 /Å 2

G 6 0.2/mg dm 23

f1

5 3 10 23 1

30 38

11000 6 2000 5580 6 300

1.0 2.0

0.24 0.37

a When the bulk scattering length density is nonzero the first layer has to be split into two layers. The thickness for the layer above water is about 5 6 3 Å at both concentrations.

BSA ADSORPTION ON THE SURFACE OF WATER

433

variation with respect to pH at the high salt concentration is probably caused by the screening of the electrostatic charges on the protein. The reduced adsorption may be caused by association of ions with the protein molecules, lowering its tendency to adsorb. The slight increase of surface excess with pH in Fig. 5a can be correlated with the lower solubility of BSA under these conditions. It was observed that at pH 3 and 5 3 10 23 g dm 23 the protein solution became slightly blue, suggesting that there was some degree of aggregation. As the concentration increased to 1 g dm 23 there was some precipitation, which made the measurements under these conditions unreliable. The surface excess here was found to be around 4 mg m 22 and the total layer thickness about 150 Å. Although the results broadly fit the trend of the effect of pH, they are significantly greater than values at other pH. (C) The Extent of Immersion of BSA Layer in Water

FIG. 5. Variation of surface excess (a) and total layer thickness (b) as a function of solution pH at the BSA concentrations of 5 3 10 23 g dm 23 (E) and 1 g dm 23 (F). The total ionic strength was kept at 1 M. The solid lines were drawn for a guide.

strength to 1 M. The resultant surface excess is shown in Fig. 5a and the thickness variation in Fig. 5b. The main change is that the pattern of maximum adsorption at the IP disappears. There is a significant reduction in surface excess at pH 5 and at 1 g dm 23. The maximum in the thickness at the same concentration also disappears. Salt addition also systematically reduces the gap in surface excess between the two BSA concentrations, over almost all the pH range. The less dramatic

It is usually assumed that the monolayers formed by globular proteins at the air–water interface are fully immersed in water (9). However, other amphiphilic species such as surfactants and many polymers usually have a significant fraction of their molecules completely out of the underlying aqueous subphase (17, 18). Where a protein has the flexibility to allow segregation of hydrophobic fragments, such as b-casein, these would be expected to project out of the water. The situation for globular proteins is less clear. A sharp segregation of hydrophobic and hydrophilic portions is obviously not possible when the globular structure is retained. The molecule may, however, gain some advantage by removing some portions of the globular assembly from the water. There are two ways of determining the extent of immersion of the protein layer. The most direct approach is to adjust the H/D ratio of the water (the contrast) so that its scattering length density is the same as that of the protein. The immersed part of the layer is then completely invisible to neutrons. The reflectivity profile at this contrast will be entirely determined by the part of the protein layer out of the water. The exact match contrast to BSA can be obtained by using a 1:1.27 mixture of D 2O and H 2O whose scattering length density is 2.5 3 10 26 Å 22 (assuming complete H/D equilibration). The volume fraction of BSA in the layer above the water surface can be calculated from the fitted scattering length density since r p is already known from the measurements previously described. Since f p is less than unity, the scattering length density for this “hydrophobic” layer should be quite different from the underlying solution. Figure 6 shows the reflectivities at this contrast at concentrations of 5 3 10 24 and 1 g dm 23 at pH 5 and ionic strength 0.02 M, together with the best fits of a uniform layer model with respective thicknesses of 10 6 5 and 5 6 3 Å. To demonstrate the sensitivity of the fits to the extent of immersion, reflectivity profiles corresponding to half and complete immersion are also shown for comparison. For the two intermediate concentrations the thickness of this layer was found to

434

LU, SU, AND THOMAS

FIG. 6. Reflectivity profiles measured from BSA layers adsorbed on the surface of the mixed H 2O and D 2O (molar ratio 5 1.27:1, r 5 2.49 3 10 26 Å 22) at (a) 5 3 10 24 g dm 23 and (b) 1 g dm 23. The continuous lines were calculated using a single layer model of 10 Å for (a) and 5 Å for (b). The dashed line was calculated assuming that the layer was fully immersed in water and the dash– dotted line assuming that there was 25 Å (a) and 15 Å (b) of the layer out of water, respectively. The solution pH was 5 and the total ionic strength was 0.02 M.

be 8 6 4 and 5 6 3 Å. Since the total thickness of the layer at each concentration has already been determined from the measurements in NRW, the thickness for the portion of the layer under water can be obtained. The fraction of the monolayer

immersed in water therefore varies from 0.65 6 0.2 to 0.85 6 0.1. The measurement of the reflectivity in D 2O yields data that are more sensitive to the extent of immersion of the BSA layer in water. In D 2O both regions above and below the water contribute to the reflectivity. However, once again, since the thickness and the composition of the whole layer are already known, the only unknown parameter in the fitting of such profiles is the fraction of the layer immersed in water. Figure 7 shows the reflectivities measured at the two extreme concentrations, 5 3 10 24 and 1 g dm 23, in D 2O. The continuous lines were calculated using the identical structural models used to account for Figs. 1 and 6, except to allow for the different scattering length density of the protein in D 2O. That the fits between the calculated and observed reflectivities, shown in Fig. 7, are good gives further support to the model deduced in the previous paragraph. The same approach was used to determine the extent of immersion in water at the other two pH values. In all cases the thickness for the layer above water was found to be 5 6 3 Å. Thus, there is apparently no particular trend of the variation of the extent of immersion with respect to concentration or pH. It is quite common in reflection experiments to include roughness as an additional fitting parameter. If the extent of immersion is almost complete it may then be difficult to differentiate between a protein layer that is not quite totally immersed and the roughness generated by the difference between the water and protein scattering. This is an issue that has not been well explored, although it only arises in the limit when immersion is almost complete. However, in the case where the water is matched to the protein this roughness effect cannot occur; only a protruding fragment can generate a nonzero signal. Thus, the experiments shown in Fig. 6 in principle can only be explained by incomplete immersion of the protein in the aqueous subphase. In practice, this experiment relies on a correct assessment of the extent of exchange of the labile protons. In fitting the reflectivity profiles we have throughout assumed complete exchange of the labile protons with the bulk water and no fractionation of isotopes between protein and water. That the reflectivity profiles at the three water contrasts (NRW, r 5 2.9, and D 2O) were always well fitted with the same structural parameters indicates that the assumption is probably correct. The exchange of labile hydrogens in BSA with D 2O has been systematically examined by Hvidt and Nielsen (29). Of the 1015 labile hydrogens in BSA some 750 of these exchange almost instantly at pH 7 and 0°C. A further 250 or so exchange over a period of 2 h. The remaining 30 to 50 appear not to exchange within a further 24 h. We have stated previously that all the reflectivity profiles reported in this work were measured when there was no longer any change in the reflectivity. Since it typically took several hours for this equilibrium to be reached, over 90% of the labile hydrogen must be exchanged before the start of the measurements. The

435

BSA ADSORPTION ON THE SURFACE OF WATER

structure to allow exchange of the labile protons in the hydrophobic regions. This would then leave a very small proportion of nonequilibrated protons, too small to have any significant effect on the analyses presented above. It is important to realize that the neutron reflectivity profile would be very sensitive to levels of exchange less than about 90% and, if these were still occurring during an experiment, there would be noticeable drifts in the reflected intensity. Part of the time dependent behavior at the beginning of the neutron experiment may include some H/D exchange as well as structural adjustments in the layer. That the reflectivity was always recorded when its value had stabilized signifies that exchange had reached a stable level. CONCLUSIONS

FIG. 7. Reflectivity profiles measured from BSA layers adsorbed on the surface of D 2O at (a) 5 3 10 24 g dm 23 and (b) 1 g dm 23. The continuous lines were calculated using a two layer model for (a) and a three layer model for (b). The structural parameters are given in Table 2. The dashed lines were calculated assuming that the layers are completely immersed in water. The solution pH was 5 and the total ionic strength was 0.02 M.

hydrogens that are difficult to exchange are probably those on the peptide backbone that are in a hydrophobic environment, although Radford et al. (30) have argued, using H/D exchange data for lysozyme, that two independent factors may lead to slow rates of exchange, strong hydrogen bonding of a given proton, or restricted access of the water to a given amide group. It is expected that any structural deformation accompanying adsorption may cause sufficient disturbance to the globular

The adsorption of BSA at the air–water interface is summarized schematically in Fig. 8. Neutron reflection has been used to make an accurate measurement of the distribution of protein along the surface normal direction. That the adsorbed layer can be modeled by a single uniform layer over almost the entire pH and concentration range, except at pH 5 1 g dm 23 and the total ionic strength of 0.02 M, suggests that the adsorbed molecules retain their globular framework. No indication of the disorder characteristic of denaturation was observed. Isotopic substitution has made it possible to estimate the extent of immersion of the adsorbed layer in water. In concentration regions where only a monolayer is formed the thickness of the layer is always less than the length of the short axial diameter, suggesting that the molecules are adsorbed with their long axis parallel to the surface. The slight increase of the layer thickness from 30 to 40 Å as the surface concentration increases indicates a structural deformation caused either by geometrical constraint or lateral electrostatic repulsion and is consistent with the flexible BSA framework. At pH 5 and at a high BSA bulk concentration a dilute sublayer was detected below the main layer. The pattern of BSA adsorption at the air–water interface is remarkably similar to that at the hydrophilic silica–water interface (28). At both interfaces adsorption is in the form of a sideways-on monolayer over a wide pH and concentration range. The adsorbed amount peaks at pH 5 and a dilute sublayer forms when bulk BSA concentration approaches 1 g dm 23. Although the two surfaces are quite different in nature the surface excess only differs by less than 0.5 mg m 22 over most of the conditions studied. It is also interesting that at both interfaces the thickness varies from about 30 Å at very low BSA concentration to about 40 Å at higher concentrations. This indicates that the structural deformation is dominated by lateral interactions within the adsorbed layer. The adsorption of lysozyme at the air–water interface has also recently been studied by neutron reflection (14, 15). In comparison with BSA adsorbed lysozyme shows greater structural variation. At pH 7 lysozyme forms a 30 Å thick layer at a concentration below 0.1 g dm 23. Since the globular dimen-

436

LU, SU, AND THOMAS

FIG. 8. Schematic representation of the change of surface coverage and conformation of BSA layers at the air–water interface. Since the measurement was not sensitive to the precise location of the immersed sublayer at 1 g dm 23 and pH 5, the schematic diagram only represents the average structural composition of the layer.

sion of lysozyme is 30 3 30 3 45 Å 3, the result indicates that lysozyme also adopts a sideways-on conformation. At 0.1 g dm 23 the thickness of the layer is about 35 Å, suggesting that the molecules within the adsorbed layer start to tilt. The area per molecule at this concentration is about 1400 Å 2, which is comparable to the limiting area per molecule of 1350 Å 2 (30Å 3 45Å) for sideways-on adsorption. The tilting is thus an indication of the increased geometrical constraint and electrostatic repulsion within the layer. At 1 g dm 23 the layer is about 47 Å thick, close to the length of the long axial diameter, indicating a vertical orientation of the molecule. The corresponding area per molecule is about 970 Å 2, which is also close to the limiting area of 900 Å 2 (30Å 3 30Å) for head-on adsorption. A further increase in the bulk concentration to 4 g dm 23 leads to even closer packing within the upper layer together with the formation of a less dense sublayer underneath the main layer. These observations are consistent with the much greater rigidity of the lysozyme. Finally, when the molecules are adsorbed in the sideways-on orientation the lysozyme molecules protrude some 30 to 50% into the air, while BSA is almost fully immersed in the water. This observation is also consistent with the view that lysozyme has a relatively more hydrophobic outer surface than many other globular proteins (31).

ACKNOWLEDGMENTS We thank the Biotechnology and Biological Sciences Research Council for support. We also thank Dr. Sean Langridge at the ISIS neutron facilities for technical support.

REFERENCES 1. Graham, D. E., and Phillips, M. C., J. Colloid Interface Sci. 70, 403 (1979). 2. Graham, D. E., and Phillips, M. C., J. Colloid Interface Sci. 70, 415 (1979). 3. Graham, D. E., and Phillips, M. C., J. Colloid Interface Sci. 70, 427 (1979). 4. Clark, D. C., Coke, M., Mackie, A. R., Pinder, A. C., and Wilson, D. R., J. Colloid Interface Sci. 138, 207 (1990). 5. Mo¨bius, D., and Miller, R., “Proteins at Liquid Interfaces,” p. 7, Elsevier, Amsterdam, 1998. 6. Leaver, J., and Dalgleish, D. G., Biochim. Biophys. Acta 1041, 217 (1990). 7. Suttiprasit, P., Krisdhasima, V., and Mcguire, J., J. Colloid Interface Sci. 154, 316 (1992). 8. Anand, K., and Damodaran, S., J. Colloid Interface Sci. 176, 63 (1995). 9. Wang, J., and McGuire, J., J. Colloid Interface Sci. 185, 317 (1997). 10. Flach, C. R., Brauner, J. W., Taylor, J. W., Baldwin, R. C., and Mendelsohn, R., Biophys. J. 67, 402 (1994). 11. Flach, C. R., Prendergast, F. G., and Mendelsohn, R., Biophys. J. 70, 539 (1996).

BSA ADSORPTION ON THE SURFACE OF WATER 12. Dickinson, E., Horne, D. S., Phipps, J. S., and Richardson, R. M., Langmuir 9, 242 (1993). 13. Atkinson, P. J., Dickinson, E., Horne, D. S., and Richardson, R. M., J. Chem. Soc. Faraday Trans. 91, 2847 (1995). 14. Lu, J. R., Su, T. J., Howlin, B. J., J. Phys. Chem. B, in press. 15. Lu, J. R., Su, T. J., Thomas, R. K., Penfold, J., and Webster, J., J. Chem. Soc. Faraday Trans. 94, 3279 (1998). 16. Penfold, J., Richardson, R. M., Zarbakhsh, A., Webster, J. W. P., Bucknall, D. G., Rennie, A. R., Jones, R. A. L., Cosgrove, T., Thomas, R. K., Higgins, J. S., Fletcher, P. D. I., Dickinson, E., Roser, S. J., McLure, I. A., Hillman, A. R., Richards, R. W., Staples, E. J., Burgess, A. N., Simister, E. A., and White, J. W., J. Chem. Soc. Far. Trans. 93, 3899 (1997). 17. Lu, J. R., Lee, E. M., and Thomas, R. K., Acta Cryst. A 52, 11 (1996). 18. Lu, J. R., and Thomas, R. K., J. Chem. Soc. Far. Trans. 94, 995 (1998). 19. Born, M., and Wolf, E., “Principles of Optics,” Pergamon, Oxford, 1970. 20. Lekner, J., “Theory of Reflection,” Nijhoff, Dordrecht, 1987. 21. Lu, J. R., Su, T. J., Thirtle, P. N., Thomas, R. K., Rennie, A. R., and Cubitt, R., J. Colloid Interface Sci. 206, 212 (1998). 22. Peters, T., Adv. Protein Chem. 37, 161 (1985). 23. Bendedouch, D., and Chen, S. H., J. Phys. Chem. 87, 1473 (1983).

437

24. Lu, J. R., Su, T. J., Thomas, R. K., Penfold, J., and Richards, R. W., Polymer 37, 109 (1996). 25. Lu, J. R., Simister, E. A., Thomas, R. K., and Penfold, J., J. Phys. Condens. Matter 6, A403 (1994). 26. Schwartz, D. K., Schlossman, M. L., Kawmoto, G. H., Kellogg, G. J., Pershan, P. S., and Ocko, B. M., Phys. Rev. A 41, 5687 (1990). 27. Paudler, M., Ruths, J., and Riegler, H., Langmuir 8, 184 (1992). 28. Su, T. J., Lu, J. R., Cui, Z. F., Thomas, R. K., and Penfold, J., J. Phys. Chem. B 102, 8100 (1998). 29. Hvidt, A., and Nielsen, S. O., Adv. Protein Chem. 21, 287 (1966). 30. Radford, S. E., Buck, M., Topping, K. D., Dobson, C. M., and Evans, P. A., Proteins: Structure, Function, and Genetics 14, 237 (1992). 31. Haynes, C. A., and Norde, W., Colloid Surf. B: Biointerface 2, 517 (1994). 32. Voet, D., and Voet, J. G., “Biochemistry,” Wiley, New York, 1990. 33. Sears, V. F., “Neutron Optics,” Oxford University Press, Oxford, 1989. 34. Chalikian, T., Totrov, M., Abagyan, R., and Breslauer, K., J. Mol. Biol. 260, 588 (1996). 35. Van Krevelen, D. W., “Properties of Polymers,” 3rd ed. Elsevier, New York, 1990.